xref: /dflybsd-src/sys/kern/init_main.c (revision f8c7a42d831dbc6ff46a5448591fd28de34b649b)
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: src/sys/kern/init_main.c,v 1.134.2.8 2003/06/06 20:21:32 tegge Exp $
43  * $DragonFly: src/sys/kern/init_main.c,v 1.67 2006/12/04 18:03:26 dillon Exp $
44  */
45 
46 #include "opt_init_path.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 #include <sys/unistd.h>
64 #include <sys/malloc.h>
65 #include <sys/file2.h>
66 #include <sys/thread2.h>
67 
68 #include <machine/cpu.h>
69 
70 #include <vm/vm.h>
71 #include <vm/vm_param.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 /* Components of the first process -- never freed. */
79 static struct session session0;
80 static struct pgrp pgrp0;
81 static struct procsig procsig0;
82 static struct filedesc filedesc0;
83 static struct plimit limit0;
84 static struct vmspace vmspace0;
85 struct proc *initproc;
86 struct proc proc0;
87 struct thread thread0;
88 
89 int cmask = CMASK;
90 extern	struct user *proc0paddr;
91 extern int fallback_elf_brand;
92 
93 int	boothowto = 0;		/* initialized so that it can be patched */
94 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
95 
96 /*
97  * This ensures that there is at least one entry so that the sysinit_set
98  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
99  * executed.
100  */
101 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
102 
103 /*
104  * The sysinit table itself.  Items are checked off as the are run.
105  * If we want to register new sysinit types, add them to newsysinit.
106  */
107 SET_DECLARE(sysinit_set, struct sysinit);
108 struct sysinit **sysinit, **sysinit_end;
109 struct sysinit **newsysinit, **newsysinit_end;
110 
111 
112 /*
113  * Merge a new sysinit set into the current set, reallocating it if
114  * necessary.  This can only be called after malloc is running.
115  */
116 void
117 sysinit_add(struct sysinit **set, struct sysinit **set_end)
118 {
119 	struct sysinit **newset;
120 	struct sysinit **sipp;
121 	struct sysinit **xipp;
122 	int count;
123 
124 	count = set_end - set;
125 	if (newsysinit)
126 		count += newsysinit_end - newsysinit;
127 	else
128 		count += sysinit_end - sysinit;
129 	newset = kmalloc(count * sizeof(*sipp), M_TEMP, M_WAITOK);
130 	if (newset == NULL)
131 		panic("cannot malloc for sysinit");
132 	xipp = newset;
133 	if (newsysinit) {
134 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
135 			*xipp++ = *sipp;
136 	} else {
137 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
138 			*xipp++ = *sipp;
139 	}
140 	for (sipp = set; sipp < set_end; sipp++)
141 		*xipp++ = *sipp;
142 	if (newsysinit)
143 		kfree(newsysinit, M_TEMP);
144 	newsysinit = newset;
145 	newsysinit_end = newset + count;
146 }
147 
148 /*
149  * System startup; initialize the world, create process 0, mount root
150  * filesystem, and fork to create init and pagedaemon.  Most of the
151  * hard work is done in the lower-level initialization routines including
152  * startup(), which does memory initialization and autoconfiguration.
153  *
154  * This allows simple addition of new kernel subsystems that require
155  * boot time initialization.  It also allows substitution of subsystem
156  * (for instance, a scheduler, kernel profiler, or VM system) by object
157  * module.  Finally, it allows for optional "kernel threads".
158  */
159 void
160 mi_startup(void)
161 {
162 	struct sysinit *sip;		/* system initialization*/
163 	struct sysinit **sipp;		/* system initialization*/
164 	struct sysinit **xipp;		/* interior loop of sort*/
165 	struct sysinit *save;		/* bubble*/
166 
167 	if (sysinit == NULL) {
168 		sysinit = SET_BEGIN(sysinit_set);
169 		sysinit_end = SET_LIMIT(sysinit_set);
170 	}
171 
172 restart:
173 	/*
174 	 * Perform a bubble sort of the system initialization objects by
175 	 * their subsystem (primary key) and order (secondary key).
176 	 */
177 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
178 		for (xipp = sipp + 1; xipp < sysinit_end; xipp++) {
179 			if ((*sipp)->subsystem < (*xipp)->subsystem ||
180 			     ((*sipp)->subsystem == (*xipp)->subsystem &&
181 			      (*sipp)->order <= (*xipp)->order))
182 				continue;	/* skip*/
183 			save = *sipp;
184 			*sipp = *xipp;
185 			*xipp = save;
186 		}
187 	}
188 
189 	/*
190 	 * Traverse the (now) ordered list of system initialization tasks.
191 	 * Perform each task, and continue on to the next task.
192 	 *
193 	 * The last item on the list is expected to be the scheduler,
194 	 * which will not return.
195 	 */
196 	for (sipp = sysinit; sipp < sysinit_end; sipp++) {
197 		sip = *sipp;
198 		if (sip->subsystem == SI_SUB_DUMMY)
199 			continue;	/* skip dummy task(s)*/
200 
201 		if (sip->subsystem == SI_SUB_DONE)
202 			continue;
203 
204 		/* Call function */
205 		(*(sip->func))(sip->udata);
206 
207 		/* Check off the one we're just done */
208 		sip->subsystem = SI_SUB_DONE;
209 
210 		/* Check if we've installed more sysinit items via KLD */
211 		if (newsysinit != NULL) {
212 			if (sysinit != SET_BEGIN(sysinit_set))
213 				kfree(sysinit, M_TEMP);
214 			sysinit = newsysinit;
215 			sysinit_end = newsysinit_end;
216 			newsysinit = NULL;
217 			newsysinit_end = NULL;
218 			goto restart;
219 		}
220 	}
221 
222 	panic("Shouldn't get here!");
223 	/* NOTREACHED*/
224 }
225 
226 
227 /*
228  ***************************************************************************
229  ****
230  **** The following SYSINIT's belong elsewhere, but have not yet
231  **** been moved.
232  ****
233  ***************************************************************************
234  */
235 static void
236 print_caddr_t(void *data __unused)
237 {
238 	printf("%s", (char *)data);
239 }
240 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
241 
242 /*
243  * Leave the critical section that protected us from spurious interrupts
244  * so device probes work.
245  */
246 static void
247 leavecrit(void *dummy __unused)
248 {
249 	crit_exit();
250 	KKASSERT(!IN_CRITICAL_SECT(curthread));
251 	if (bootverbose)
252 		printf("Leaving critical section, allowing interrupts\n");
253 }
254 SYSINIT(leavecrit, SI_SUB_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
255 
256 /*
257  ***************************************************************************
258  ****
259  **** The two following SYSINT's are proc0 specific glue code.  I am not
260  **** convinced that they can not be safely combined, but their order of
261  **** operation has been maintained as the same as the original init_main.c
262  **** for right now.
263  ****
264  **** These probably belong in init_proc.c or kern_proc.c, since they
265  **** deal with proc0 (the fork template process).
266  ****
267  ***************************************************************************
268  */
269 /* ARGSUSED*/
270 static void
271 proc0_init(void *dummy __unused)
272 {
273 	struct proc *p;
274 	struct lwp *lp;
275 
276 	p = &proc0;
277 	lp = &proc0.p_lwp;	/* XXX lwp to be: lwp0 */
278 
279 	/*
280 	 * Initialize process and pgrp structures.
281 	 */
282 	procinit();
283 
284 	/*
285 	 * additional VM structures
286 	 */
287 	vm_init2();
288 
289 	/*
290 	 * Create process 0 (the swapper).
291 	 */
292 	LIST_INSERT_HEAD(&allproc, p, p_list);
293 	p->p_pgrp = &pgrp0;
294 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
295 	LIST_INIT(&pgrp0.pg_members);
296 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
297 
298 	pgrp0.pg_session = &session0;
299 	session0.s_count = 1;
300 	session0.s_leader = p;
301 
302 	p->p_sysent = &aout_sysvec;
303 
304 	p->p_flag = 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 	p->p_lwp.lwp_rtprio = p->p_rtprio;
310 
311 	p->p_peers = 0;
312 	p->p_leader = p;
313 
314 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
315 	bcopy("swapper", thread0.td_comm, sizeof ("swapper"));
316 
317 	/* Create credentials. */
318 	p->p_ucred = crget();
319 	p->p_ucred->cr_ruidinfo = uifind(0);
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_ucred->cr_prison = NULL;
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(p);
332 
333 	/* Create the file descriptor table. */
334 	fdinit_bootstrap(p, &filedesc0, cmask);
335 
336 	/* Create the limits structures. */
337 	plimit_init0(&limit0);
338 	p->p_limit = &limit0;
339 
340 	/* Allocate a prototype map so we have something to fork. */
341 	pmap_pinit0(vmspace_pmap(&vmspace0));
342 	p->p_vmspace = &vmspace0;
343 	vmspace0.vm_refcnt = 1;
344 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_USER_ADDRESS),
345 	    trunc_page(VM_MAX_USER_ADDRESS));
346 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
347 
348 	/*
349 	 * We continue to place resource usage info and signal
350 	 * actions in the user struct so they're pageable.
351 	 */
352 	p->p_stats = &p->p_addr->u_stats;
353 	p->p_sigacts = &p->p_addr->u_sigacts;
354 
355 	/*
356 	 * Charge root for one process.
357 	 */
358 	(void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
359 
360 }
361 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
362 
363 static int proc0_post_callback(struct proc *p, void *data __unused);
364 
365 /* ARGSUSED*/
366 static void
367 proc0_post(void *dummy __unused)
368 {
369 	struct timespec ts;
370 
371 	/*
372 	 * Now we can look at the time, having had a chance to verify the
373 	 * time from the file system.  Pretend that proc0 started now.
374 	 */
375 	allproc_scan(proc0_post_callback, NULL);
376 
377 	/*
378 	 * Give the ``random'' number generator a thump.
379 	 * XXX: Does read_random() contain enough bits to be used here ?
380 	 */
381 	nanotime(&ts);
382 	skrandom(ts.tv_sec ^ ts.tv_nsec);
383 }
384 
385 static int
386 proc0_post_callback(struct proc *p, void *data __unused)
387 {
388 	microtime(&p->p_start);
389 	return(0);
390 }
391 
392 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
393 
394 /*
395  ***************************************************************************
396  ****
397  **** The following SYSINIT's and glue code should be moved to the
398  **** respective files on a per subsystem basis.
399  ****
400  ***************************************************************************
401  */
402 
403 
404 /*
405  ***************************************************************************
406  ****
407  **** The following code probably belongs in another file, like
408  **** kern/init_init.c.
409  ****
410  ***************************************************************************
411  */
412 
413 /*
414  * List of paths to try when searching for "init".
415  */
416 static char init_path[MAXPATHLEN] =
417 #ifdef	INIT_PATH
418     __XSTRING(INIT_PATH);
419 #else
420     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
421 #endif
422 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
423 
424 /*
425  * Start the initial user process; try exec'ing each pathname in init_path.
426  * The program is invoked with one argument containing the boot flags.
427  *
428  * The MP lock is held on entry.
429  */
430 static void
431 start_init(void *dummy)
432 {
433 	vm_offset_t addr;
434 	struct execve_args args;
435 	int options, error;
436 	char *var, *path, *next, *s;
437 	char *ucp, **uap, *arg0, *arg1;
438 	struct proc *p;
439 	struct lwp *lp;
440 	struct mount *mp;
441 	struct vnode *vp;
442 
443 	p = curproc;
444 
445 	KKASSERT(p->p_nthreads == 1);
446 
447 	lp = LIST_FIRST(&p->p_lwps);
448 
449 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
450 	mp = mountlist_boot_getfirst();
451 	if (VFS_ROOT(mp, &vp))
452 		panic("cannot find root vnode");
453 	if (mp->mnt_ncmountpt.ncp == NULL) {
454 		cache_allocroot(&mp->mnt_ncmountpt, mp, vp);
455 		cache_unlock(&mp->mnt_ncmountpt);	/* leave ref intact */
456 	}
457 	p->p_fd->fd_cdir = vp;
458 	vref(p->p_fd->fd_cdir);
459 	p->p_fd->fd_rdir = vp;
460 	vref(p->p_fd->fd_rdir);
461 	vfs_cache_setroot(vp, cache_hold(&mp->mnt_ncmountpt));
462 	vn_unlock(vp);			/* leave ref intact */
463 	cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_ncdir);
464 	cache_copy(&mp->mnt_ncmountpt, &p->p_fd->fd_nrdir);
465 
466 	/*
467 	 * Need just enough stack to hold the faked-up "execve()" arguments.
468 	 */
469 	addr = trunc_page(USRSTACK - PAGE_SIZE);
470 	error = vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
471 			    FALSE,
472 			    VM_MAPTYPE_NORMAL,
473 			    VM_PROT_ALL, VM_PROT_ALL,
474 			    0);
475 	if (error)
476 		panic("init: couldn't allocate argument space");
477 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
478 	p->p_vmspace->vm_ssize = 1;
479 
480 	if ((var = kgetenv("init_path")) != NULL) {
481 		strncpy(init_path, var, sizeof init_path);
482 		init_path[sizeof init_path - 1] = 0;
483 	}
484 	if ((var = kgetenv("kern.fallback_elf_brand")) != NULL)
485 		fallback_elf_brand = strtol(var, NULL, 0);
486 
487 	for (path = init_path; *path != '\0'; path = next) {
488 		while (*path == ':')
489 			path++;
490 		if (*path == '\0')
491 			break;
492 		for (next = path; *next != '\0' && *next != ':'; next++)
493 			/* nothing */ ;
494 		if (bootverbose)
495 			printf("start_init: trying %.*s\n", (int)(next - path),
496 			    path);
497 
498 		/*
499 		 * Move out the boot flag argument.
500 		 */
501 		options = 0;
502 		ucp = (char *)USRSTACK;
503 		(void)subyte(--ucp, 0);		/* trailing zero */
504 		if (boothowto & RB_SINGLE) {
505 			(void)subyte(--ucp, 's');
506 			options = 1;
507 		}
508 #ifdef notyet
509                 if (boothowto & RB_FASTBOOT) {
510 			(void)subyte(--ucp, 'f');
511 			options = 1;
512 		}
513 #endif
514 
515 #ifdef BOOTCDROM
516 		(void)subyte(--ucp, 'C');
517 		options = 1;
518 #endif
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 		 * WARNING!  We may have been moved to another cpu after
555 		 * acquiring the current user process designation.  The
556 		 * MP lock will migrate with us though so we still have to
557 		 * release it.
558 		 */
559 		if ((error = sys_execve(&args)) == 0) {
560 			rel_mplock();
561 			lp->lwp_proc->p_usched->acquire_curproc(lp);
562 			return;
563 		}
564 		if (error != ENOENT)
565 			printf("exec %.*s: error %d\n", (int)(next - path),
566 			    path, error);
567 	}
568 	printf("init: not found in path %s\n", init_path);
569 	panic("no init");
570 }
571 
572 /*
573  * Like kthread_create(), but runs in it's own address space.
574  * We do this early to reserve pid 1.
575  *
576  * Note special case - do not make it runnable yet.  Other work
577  * in progress will change this more.
578  */
579 static void
580 create_init(const void *udata __unused)
581 {
582 	int error;
583 	struct lwp *lp;
584 
585 	crit_enter();
586 	error = fork1(&proc0.p_lwp, RFFDG | RFPROC, &initproc);
587 	if (error)
588 		panic("cannot fork init: %d", error);
589 	initproc->p_flag |= P_SYSTEM;
590 	lp = LIST_FIRST(&initproc->p_lwps);
591 	cpu_set_fork_handler(lp, start_init, NULL);
592 	crit_exit();
593 }
594 SYSINIT(init,SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
595 
596 /*
597  * Make it runnable now.
598  */
599 static void
600 kick_init(const void *udata __unused)
601 {
602 	start_forked_proc(&proc0.p_lwp, initproc);
603 }
604 SYSINIT(kickinit,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
605 
606 /*
607  * Machine independant globaldata initialization
608  *
609  * WARNING!  Called from early boot, 'mycpu' may not work yet.
610  */
611 void
612 mi_gdinit(struct globaldata *gd, int cpuid)
613 {
614 	TAILQ_INIT(&gd->gd_tdfreeq);    /* for pmap_{new,dispose}_thread() */
615 	TAILQ_INIT(&gd->gd_systimerq);
616 	gd->gd_cpuid = cpuid;
617 	gd->gd_cpumask = (cpumask_t)1 << cpuid;
618 	lwkt_gdinit(gd);
619 	vm_map_entry_reserve_cpu_init(gd);
620 	sleep_gdinit(gd);
621 }
622 
623 
624