xref: /dflybsd-src/sys/kern/init_main.c (revision 9dd6ab89a53571a63351083f86cb7bf1662fe356)
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 1.26 2003/11/10 06:12:13 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 void mi_startup(void);				/* Should be elsewhere */
79 
80 /* Components of the first process -- never freed. */
81 static struct session session0;
82 static struct pgrp pgrp0;
83 static struct procsig procsig0;
84 static struct filedesc0 filedesc0;
85 static struct plimit limit0;
86 static struct vmspace vmspace0;
87 struct proc *initproc;
88 struct proc proc0;
89 struct thread thread0;
90 
91 int cmask = CMASK;
92 extern	struct user *proc0paddr;
93 extern int fallback_elf_brand;
94 
95 struct	vnode *rootvp;
96 int	boothowto = 0;		/* initialized so that it can be patched */
97 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
98 
99 /*
100  * This ensures that there is at least one entry so that the sysinit_set
101  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
102  * executed.
103  */
104 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
105 
106 /*
107  * The sysinit table itself.  Items are checked off as the are run.
108  * If we want to register new sysinit types, add them to newsysinit.
109  */
110 SET_DECLARE(sysinit_set, struct sysinit);
111 struct sysinit **sysinit, **sysinit_end;
112 struct sysinit **newsysinit, **newsysinit_end;
113 
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, struct sysinit **set_end)
121 {
122 	struct sysinit **newset;
123 	struct sysinit **sipp;
124 	struct sysinit **xipp;
125 	int count;
126 
127 	count = set_end - set;
128 	if (newsysinit)
129 		count += newsysinit_end - newsysinit;
130 	else
131 		count += sysinit_end - sysinit;
132 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
133 	if (newset == NULL)
134 		panic("cannot malloc for sysinit");
135 	xipp = newset;
136 	if (newsysinit) {
137 		for (sipp = newsysinit; sipp < newsysinit_end; sipp++)
138 			*xipp++ = *sipp;
139 	} else {
140 		for (sipp = sysinit; sipp < sysinit_end; sipp++)
141 			*xipp++ = *sipp;
142 	}
143 	for (sipp = set; sipp < set_end; sipp++)
144 		*xipp++ = *sipp;
145 	if (newsysinit)
146 		free(newsysinit, M_TEMP);
147 	newsysinit = newset;
148 	newsysinit_end = newset + count;
149 }
150 
151 /*
152  * System startup; initialize the world, create process 0, mount root
153  * filesystem, and fork to create init and pagedaemon.  Most of the
154  * hard work is done in the lower-level initialization routines including
155  * startup(), which does memory initialization and autoconfiguration.
156  *
157  * This allows simple addition of new kernel subsystems that require
158  * boot time initialization.  It also allows substitution of subsystem
159  * (for instance, a scheduler, kernel profiler, or VM system) by object
160  * module.  Finally, it allows for optional "kernel threads".
161  */
162 void
163 mi_startup(void)
164 {
165 	struct sysinit *sip;		/* system initialization*/
166 	struct sysinit **sipp;		/* system initialization*/
167 	struct sysinit **xipp;		/* interior loop of sort*/
168 	struct sysinit *save;		/* bubble*/
169 
170 	if (sysinit == NULL) {
171 		sysinit = SET_BEGIN(sysinit_set);
172 		sysinit_end = SET_LIMIT(sysinit_set);
173 	}
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 < sysinit_end; sipp++) {
181 		for (xipp = sipp + 1; xipp < sysinit_end; 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 < sysinit_end; sipp++) {
200 		sip = *sipp;
201 		if (sip->subsystem == SI_SUB_DUMMY)
202 			continue;	/* skip dummy task(s)*/
203 
204 		if (sip->subsystem == SI_SUB_DONE)
205 			continue;
206 
207 		/* Call function */
208 		(*(sip->func))(sip->udata);
209 
210 		/* Check off the one we're just done */
211 		sip->subsystem = SI_SUB_DONE;
212 
213 		/* Check if we've installed more sysinit items via KLD */
214 		if (newsysinit != NULL) {
215 			if (sysinit != SET_BEGIN(sysinit_set))
216 				free(sysinit, M_TEMP);
217 			sysinit = newsysinit;
218 			sysinit_end = newsysinit_end;
219 			newsysinit = NULL;
220 			newsysinit_end = NULL;
221 			goto restart;
222 		}
223 	}
224 
225 	panic("Shouldn't get here!");
226 	/* NOTREACHED*/
227 }
228 
229 
230 /*
231  ***************************************************************************
232  ****
233  **** The following SYSINIT's belong elsewhere, but have not yet
234  **** been moved.
235  ****
236  ***************************************************************************
237  */
238 static void
239 print_caddr_t(void *data __unused)
240 {
241 	printf("%s", (char *)data);
242 }
243 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
244 
245 /*
246  * Leave the critical section that protected us from spurious interrupts
247  * so device probes work.
248  */
249 static void
250 leavecrit(void *dummy __unused)
251 {
252 	crit_exit();
253 	KKASSERT(!IN_CRITICAL_SECT(curthread));
254 	if (bootverbose)
255 		printf("Leaving critical section, allowing interrupts\n");
256 }
257 SYSINIT(leavecrit, SI_SUB_LEAVE_CRIT, SI_ORDER_ANY, leavecrit, NULL)
258 
259 /*
260  ***************************************************************************
261  ****
262  **** The two following SYSINT's are proc0 specific glue code.  I am not
263  **** convinced that they can not be safely combined, but their order of
264  **** operation has been maintained as the same as the original init_main.c
265  **** for right now.
266  ****
267  **** These probably belong in init_proc.c or kern_proc.c, since they
268  **** deal with proc0 (the fork template process).
269  ****
270  ***************************************************************************
271  */
272 /* ARGSUSED*/
273 static void
274 proc0_init(void *dummy __unused)
275 {
276 	struct proc		*p;
277 	struct filedesc0	*fdp;
278 	unsigned i;
279 
280 	p = &proc0;
281 
282 	/*
283 	 * Initialize process and pgrp structures.
284 	 */
285 	procinit();
286 
287 	/*
288 	 * Initialize sleep queue hash table
289 	 */
290 	sleepinit();
291 
292 	/*
293 	 * additional VM structures
294 	 */
295 	vm_init2();
296 
297 	/*
298 	 * Create process 0 (the swapper).
299 	 */
300 	LIST_INSERT_HEAD(&allproc, p, p_list);
301 	p->p_pgrp = &pgrp0;
302 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
303 	LIST_INIT(&pgrp0.pg_members);
304 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
305 
306 	pgrp0.pg_session = &session0;
307 	session0.s_count = 1;
308 	session0.s_leader = p;
309 
310 	p->p_sysent = &aout_sysvec;
311 
312 	p->p_flag = P_INMEM | P_SYSTEM | P_CP_RELEASED;
313 	p->p_stat = SRUN;
314 	p->p_nice = NZERO;
315 	p->p_rtprio.type = RTP_PRIO_NORMAL;
316 	p->p_rtprio.prio = 0;
317 
318 	p->p_peers = 0;
319 	p->p_leader = p;
320 
321 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
322 
323 	/* Create credentials. */
324 	p->p_ucred = crget();
325 	p->p_ucred->cr_ruidinfo = uifind(0);
326 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
327 	p->p_ucred->cr_uidinfo = uifind(0);
328 
329 	/* Don't jail it */
330 	p->p_ucred->cr_prison = NULL;
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 	p->p_fdtol = NULL;
343 	fdp->fd_fd.fd_refcnt = 1;
344 	fdp->fd_fd.fd_cmask = cmask;
345 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
346 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
347 	fdp->fd_fd.fd_nfiles = NDFILE;
348 
349 	/* Create the limits structures. */
350 	p->p_limit = &limit0;
351 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
352 		limit0.pl_rlimit[i].rlim_cur =
353 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
354 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
355 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
356 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
357 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
358 	i = ptoa(vmstats.v_free_count);
359 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
360 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
361 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
362 	limit0.p_cpulimit = RLIM_INFINITY;
363 	limit0.p_refcnt = 1;
364 
365 	/* Allocate a prototype map so we have something to fork. */
366 	pmap_pinit0(vmspace_pmap(&vmspace0));
367 	p->p_vmspace = &vmspace0;
368 	vmspace0.vm_refcnt = 1;
369 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
370 	    trunc_page(VM_MAXUSER_ADDRESS));
371 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
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(p->p_ucred->cr_uidinfo, 1, 0);
384 
385 }
386 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
387 
388 /* ARGSUSED*/
389 static void
390 proc0_post(void *dummy __unused)
391 {
392 	struct timespec ts;
393 	struct proc *p;
394 
395 	/*
396 	 * Now we can look at the time, having had a chance to verify the
397 	 * time from the file system.  Pretend that proc0 started now.
398 	 */
399 	FOREACH_PROC_IN_SYSTEM(p) {
400 		microtime(&p->p_stats->p_start);
401 	}
402 
403 	/*
404 	 * Give the ``random'' number generator a thump.
405 	 * XXX: Does read_random() contain enough bits to be used here ?
406 	 */
407 	nanotime(&ts);
408 	srandom(ts.tv_sec ^ ts.tv_nsec);
409 }
410 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
411 
412 /*
413  ***************************************************************************
414  ****
415  **** The following SYSINIT's and glue code should be moved to the
416  **** respective files on a per subsystem basis.
417  ****
418  ***************************************************************************
419  */
420 
421 
422 /*
423  ***************************************************************************
424  ****
425  **** The following code probably belongs in another file, like
426  **** kern/init_init.c.
427  ****
428  ***************************************************************************
429  */
430 
431 /*
432  * List of paths to try when searching for "init".
433  */
434 static char init_path[MAXPATHLEN] =
435 #ifdef	INIT_PATH
436     __XSTRING(INIT_PATH);
437 #else
438     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
439 #endif
440 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
441 
442 /*
443  * Start the initial user process; try exec'ing each pathname in init_path.
444  * The program is invoked with one argument containing the boot flags.
445  *
446  * The MP lock is held on entry.
447  */
448 static void
449 start_init(void *dummy)
450 {
451 	vm_offset_t addr;
452 	struct execve_args args;
453 	int options, error;
454 	char *var, *path, *next, *s;
455 	char *ucp, **uap, *arg0, *arg1;
456 	struct proc *p;
457 
458 	p = curproc;
459 
460 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
461 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
462 		panic("cannot find root vnode");
463 	p->p_fd->fd_cdir = rootvnode;
464 	VREF(p->p_fd->fd_cdir);
465 	p->p_fd->fd_rdir = rootvnode;
466 	VREF(p->p_fd->fd_rdir);
467 	vfs_cache_setroot(rootvnode);
468 	VOP_UNLOCK(rootvnode, 0, curthread);
469 
470 	/*
471 	 * Need just enough stack to hold the faked-up "execve()" arguments.
472 	 */
473 	addr = trunc_page(USRSTACK - PAGE_SIZE);
474 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
475 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
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 = getenv("init_path")) != NULL) {
481 		strncpy(init_path, var, sizeof init_path);
482 		init_path[sizeof init_path - 1] = 0;
483 	}
484 	if ((var = getenv("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 P_CURPROC.  The MP lock will migrate with us
556 		 * though so we still have to release it.
557 		 */
558 		if ((error = execve(&args)) == 0) {
559 			acquire_curproc(p);
560 			rel_mplock();
561 			return;
562 		}
563 		if (error != ENOENT)
564 			printf("exec %.*s: error %d\n", (int)(next - path),
565 			    path, error);
566 	}
567 	printf("init: not found in path %s\n", init_path);
568 	panic("no init");
569 }
570 
571 /*
572  * Like kthread_create(), but runs in it's own address space.
573  * We do this early to reserve pid 1.
574  *
575  * Note special case - do not make it runnable yet.  Other work
576  * in progress will change this more.
577  */
578 static void
579 create_init(const void *udata __unused)
580 {
581 	int error;
582 	int s;
583 
584 	s = splhigh();
585 	error = fork1(&proc0, RFFDG | RFPROC, &initproc);
586 	if (error)
587 		panic("cannot fork init: %d\n", error);
588 	initproc->p_flag |= P_INMEM | P_SYSTEM;
589 	cpu_set_fork_handler(initproc, start_init, NULL);
590 	splx(s);
591 }
592 SYSINIT(init,SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
593 
594 /*
595  * Make it runnable now.
596  */
597 static void
598 kick_init(const void *udata __unused)
599 {
600 	start_forked_proc(&proc0, initproc);
601 }
602 SYSINIT(kickinit,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
603 
604 /*
605  * Machine independant globaldata initialization
606  *
607  * WARNING!  Called from early boot, 'mycpu' may not work yet.
608  */
609 void
610 mi_gdinit(struct globaldata *gd, int cpuid)
611 {
612 	TAILQ_INIT(&gd->gd_tdfreeq);    /* for pmap_{new,dispose}_thread() */
613 	gd->gd_cpuid = cpuid;
614 	lwkt_gdinit(gd);
615 }
616 
617 
618