xref: /dflybsd-src/sys/kern/init_main.c (revision 97478218adb62ce7c64b01ab9fa63bcddcd602ab)
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 1.22 2003/08/03 10:07:41 hmp 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 
67 #include <machine/cpu.h>
68 
69 #include <vm/vm.h>
70 #include <vm/vm_param.h>
71 #include <sys/lock.h>
72 #include <vm/pmap.h>
73 #include <vm/vm_map.h>
74 #include <sys/user.h>
75 #include <sys/copyright.h>
76 
77 extern struct linker_set	sysinit_set;	/* XXX */
78 
79 void mi_startup(void);				/* Should be elsewhere */
80 
81 /* Components of the first process -- never freed. */
82 static struct session session0;
83 static struct pgrp pgrp0;
84 static struct procsig procsig0;
85 static struct filedesc0 filedesc0;
86 static struct plimit limit0;
87 static struct vmspace vmspace0;
88 struct proc *initproc;
89 struct proc proc0;
90 struct thread thread0;
91 
92 int cmask = CMASK;
93 extern	struct user *proc0paddr;
94 extern int fallback_elf_brand;
95 
96 struct	vnode *rootvp;
97 int	boothowto = 0;		/* initialized so that it can be patched */
98 SYSCTL_INT(_debug, OID_AUTO, boothowto, CTLFLAG_RD, &boothowto, 0, "");
99 
100 /*
101  * This ensures that there is at least one entry so that the sysinit_set
102  * symbol is not undefined.  A sybsystem ID of SI_SUB_DUMMY is never
103  * executed.
104  */
105 SYSINIT(placeholder, SI_SUB_DUMMY, SI_ORDER_ANY, NULL, NULL)
106 
107 /*
108  * The sysinit table itself.  Items are checked off as the are run.
109  * If we want to register new sysinit types, add them to newsysinit.
110  */
111 struct sysinit **sysinit = (struct sysinit **)sysinit_set.ls_items;
112 struct sysinit **newsysinit;
113 
114 /*
115  * Merge a new sysinit set into the current set, reallocating it if
116  * necessary.  This can only be called after malloc is running.
117  */
118 void
119 sysinit_add(struct sysinit **set)
120 {
121 	struct sysinit **newset;
122 	struct sysinit **sipp;
123 	struct sysinit **xipp;
124 	int count = 0;
125 
126 	if (newsysinit)
127 		for (sipp = newsysinit; *sipp; sipp++)
128 			count++;
129 	else
130 		for (sipp = sysinit; *sipp; sipp++)
131 			count++;
132 	for (sipp = set; *sipp; sipp++)
133 		count++;
134 	count++;		/* Trailing NULL */
135 	newset = malloc(count * sizeof(*sipp), M_TEMP, M_NOWAIT);
136 	if (newset == NULL)
137 		panic("cannot malloc for sysinit");
138 	xipp = newset;
139 	if (newsysinit)
140 		for (sipp = newsysinit; *sipp; sipp++)
141 			*xipp++ = *sipp;
142 	else
143 		for (sipp = sysinit; *sipp; sipp++)
144 			*xipp++ = *sipp;
145 	for (sipp = set; *sipp; sipp++)
146 		*xipp++ = *sipp;
147 	*xipp = NULL;
148 	if (newsysinit)
149 		free(newsysinit, M_TEMP);
150 	newsysinit = newset;
151 }
152 
153 /*
154  * System startup; initialize the world, create process 0, mount root
155  * filesystem, and fork to create init and pagedaemon.  Most of the
156  * hard work is done in the lower-level initialization routines including
157  * startup(), which does memory initialization and autoconfiguration.
158  *
159  * This allows simple addition of new kernel subsystems that require
160  * boot time initialization.  It also allows substitution of subsystem
161  * (for instance, a scheduler, kernel profiler, or VM system) by object
162  * module.  Finally, it allows for optional "kernel threads".
163  */
164 void
165 mi_startup(void)
166 {
167 	struct sysinit *sip;		/* system initialization*/
168 	struct sysinit **sipp;		/* system initialization*/
169 	struct sysinit **xipp;		/* interior loop of sort*/
170 	struct sysinit *save;		/* bubble*/
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; sipp++) {
178 		for (xipp = sipp + 1; *xipp; 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; (sip = *sipp) != NULL; sipp++) {
197 		if (sip->subsystem == SI_SUB_DUMMY)
198 			continue;	/* skip dummy task(s)*/
199 
200 		if (sip->subsystem == SI_SUB_DONE)
201 			continue;
202 
203 		/* Call function */
204 		(*(sip->func))(sip->udata);
205 
206 		/* Check off the one we're just done */
207 		sip->subsystem = SI_SUB_DONE;
208 
209 		/* Check if we've installed more sysinit items via KLD */
210 		if (newsysinit != NULL) {
211 			if (sysinit != (struct sysinit **)sysinit_set.ls_items)
212 				free(sysinit, M_TEMP);
213 			sysinit = newsysinit;
214 			newsysinit = NULL;
215 			goto restart;
216 		}
217 	}
218 
219 	panic("Shouldn't get here!");
220 	/* NOTREACHED*/
221 }
222 
223 
224 /*
225  ***************************************************************************
226  ****
227  **** The following SYSINIT's belong elsewhere, but have not yet
228  **** been moved.
229  ****
230  ***************************************************************************
231  */
232 static void
233 print_caddr_t(void *data __unused)
234 {
235 	printf("%s", (char *)data);
236 }
237 SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)
238 
239 
240 /*
241  ***************************************************************************
242  ****
243  **** The two following SYSINT's are proc0 specific glue code.  I am not
244  **** convinced that they can not be safely combined, but their order of
245  **** operation has been maintained as the same as the original init_main.c
246  **** for right now.
247  ****
248  **** These probably belong in init_proc.c or kern_proc.c, since they
249  **** deal with proc0 (the fork template process).
250  ****
251  ***************************************************************************
252  */
253 /* ARGSUSED*/
254 static void
255 proc0_init(void *dummy __unused)
256 {
257 	struct proc		*p;
258 	struct filedesc0	*fdp;
259 	unsigned i;
260 
261 	p = &proc0;
262 
263 	/*
264 	 * Initialize process and pgrp structures.
265 	 */
266 	procinit();
267 
268 	/*
269 	 * Initialize sleep queue hash table
270 	 */
271 	sleepinit();
272 
273 	/*
274 	 * additional VM structures
275 	 */
276 	vm_init2();
277 
278 	/*
279 	 * Create process 0 (the swapper).
280 	 */
281 	LIST_INSERT_HEAD(&allproc, p, p_list);
282 	p->p_pgrp = &pgrp0;
283 	LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
284 	LIST_INIT(&pgrp0.pg_members);
285 	LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist);
286 
287 	pgrp0.pg_session = &session0;
288 	session0.s_count = 1;
289 	session0.s_leader = p;
290 
291 	p->p_sysent = &aout_sysvec;
292 
293 	p->p_flag = P_INMEM | P_SYSTEM | P_CP_RELEASED;
294 	p->p_stat = SRUN;
295 	p->p_nice = NZERO;
296 	p->p_rtprio.type = RTP_PRIO_NORMAL;
297 	p->p_rtprio.prio = 0;
298 
299 	p->p_peers = 0;
300 	p->p_leader = p;
301 
302 	bcopy("swapper", p->p_comm, sizeof ("swapper"));
303 
304 	/* Create credentials. */
305 	p->p_ucred = crget();
306 	p->p_ucred->cr_ruidinfo = uifind(0);
307 	p->p_ucred->cr_ngroups = 1;	/* group 0 */
308 	p->p_ucred->cr_uidinfo = uifind(0);
309 
310 	/* Don't jail it */
311 	p->p_ucred->cr_prison = NULL;
312 
313 	/* Create procsig. */
314 	p->p_procsig = &procsig0;
315 	p->p_procsig->ps_refcnt = 1;
316 
317 	/* Initialize signal state for process 0. */
318 	siginit(&proc0);
319 
320 	/* Create the file descriptor table. */
321 	fdp = &filedesc0;
322 	p->p_fd = &fdp->fd_fd;
323 	p->p_fdtol = NULL;
324 	fdp->fd_fd.fd_refcnt = 1;
325 	fdp->fd_fd.fd_cmask = cmask;
326 	fdp->fd_fd.fd_ofiles = fdp->fd_dfiles;
327 	fdp->fd_fd.fd_ofileflags = fdp->fd_dfileflags;
328 	fdp->fd_fd.fd_nfiles = NDFILE;
329 
330 	/* Create the limits structures. */
331 	p->p_limit = &limit0;
332 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
333 		limit0.pl_rlimit[i].rlim_cur =
334 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
335 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
336 	    limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
337 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
338 	    limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
339 	i = ptoa(vmstats.v_free_count);
340 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = i;
341 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = i;
342 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = i / 3;
343 	limit0.p_cpulimit = RLIM_INFINITY;
344 	limit0.p_refcnt = 1;
345 
346 	/* Allocate a prototype map so we have something to fork. */
347 	pmap_pinit0(vmspace_pmap(&vmspace0));
348 	p->p_vmspace = &vmspace0;
349 	vmspace0.vm_refcnt = 1;
350 	vm_map_init(&vmspace0.vm_map, round_page(VM_MIN_ADDRESS),
351 	    trunc_page(VM_MAXUSER_ADDRESS));
352 	vmspace0.vm_map.pmap = vmspace_pmap(&vmspace0);
353 
354 	/*
355 	 * We continue to place resource usage info and signal
356 	 * actions in the user struct so they're pageable.
357 	 */
358 	p->p_stats = &p->p_addr->u_stats;
359 	p->p_sigacts = &p->p_addr->u_sigacts;
360 
361 	/*
362 	 * Charge root for one process.
363 	 */
364 	(void)chgproccnt(p->p_ucred->cr_uidinfo, 1, 0);
365 
366 }
367 SYSINIT(p0init, SI_SUB_INTRINSIC, SI_ORDER_FIRST, proc0_init, NULL)
368 
369 /* ARGSUSED*/
370 static void
371 proc0_post(void *dummy __unused)
372 {
373 	struct timespec ts;
374 	struct proc *p;
375 
376 	/*
377 	 * Now we can look at the time, having had a chance to verify the
378 	 * time from the file system.  Pretend that proc0 started now.
379 	 */
380 	FOREACH_PROC_IN_SYSTEM(p) {
381 		microtime(&p->p_stats->p_start);
382 	}
383 
384 	/*
385 	 * Give the ``random'' number generator a thump.
386 	 * XXX: Does read_random() contain enough bits to be used here ?
387 	 */
388 	nanotime(&ts);
389 	srandom(ts.tv_sec ^ ts.tv_nsec);
390 }
391 SYSINIT(p0post, SI_SUB_INTRINSIC_POST, SI_ORDER_FIRST, proc0_post, NULL)
392 
393 /*
394  ***************************************************************************
395  ****
396  **** The following SYSINIT's and glue code should be moved to the
397  **** respective files on a per subsystem basis.
398  ****
399  ***************************************************************************
400  */
401 
402 
403 /*
404  ***************************************************************************
405  ****
406  **** The following code probably belongs in another file, like
407  **** kern/init_init.c.
408  ****
409  ***************************************************************************
410  */
411 
412 /*
413  * List of paths to try when searching for "init".
414  */
415 static char init_path[MAXPATHLEN] =
416 #ifdef	INIT_PATH
417     __XSTRING(INIT_PATH);
418 #else
419     "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
420 #endif
421 SYSCTL_STRING(_kern, OID_AUTO, init_path, CTLFLAG_RD, init_path, 0, "");
422 
423 /*
424  * Start the initial user process; try exec'ing each pathname in init_path.
425  * The program is invoked with one argument containing the boot flags.
426  *
427  * The MP lock is held on entry.
428  */
429 static void
430 start_init(void *dummy)
431 {
432 	vm_offset_t addr;
433 	struct execve_args args;
434 	int options, error;
435 	char *var, *path, *next, *s;
436 	char *ucp, **uap, *arg0, *arg1;
437 	struct proc *p;
438 
439 	p = curproc;
440 
441 	/* Get the vnode for '/'.  Set p->p_fd->fd_cdir to reference it. */
442 	if (VFS_ROOT(TAILQ_FIRST(&mountlist), &rootvnode))
443 		panic("cannot find root vnode");
444 	p->p_fd->fd_cdir = rootvnode;
445 	VREF(p->p_fd->fd_cdir);
446 	p->p_fd->fd_rdir = rootvnode;
447 	VREF(p->p_fd->fd_rdir);
448 	VOP_UNLOCK(rootvnode, 0, curthread);
449 
450 	/*
451 	 * Need just enough stack to hold the faked-up "execve()" arguments.
452 	 */
453 	addr = trunc_page(USRSTACK - PAGE_SIZE);
454 	if (vm_map_find(&p->p_vmspace->vm_map, NULL, 0, &addr, PAGE_SIZE,
455 			FALSE, VM_PROT_ALL, VM_PROT_ALL, 0) != 0)
456 		panic("init: couldn't allocate argument space");
457 	p->p_vmspace->vm_maxsaddr = (caddr_t)addr;
458 	p->p_vmspace->vm_ssize = 1;
459 
460 	if ((var = getenv("init_path")) != NULL) {
461 		strncpy(init_path, var, sizeof init_path);
462 		init_path[sizeof init_path - 1] = 0;
463 	}
464 	if ((var = getenv("kern.fallback_elf_brand")) != NULL)
465 		fallback_elf_brand = strtol(var, NULL, 0);
466 
467 	for (path = init_path; *path != '\0'; path = next) {
468 		while (*path == ':')
469 			path++;
470 		if (*path == '\0')
471 			break;
472 		for (next = path; *next != '\0' && *next != ':'; next++)
473 			/* nothing */ ;
474 		if (bootverbose)
475 			printf("start_init: trying %.*s\n", (int)(next - path),
476 			    path);
477 
478 		/*
479 		 * Move out the boot flag argument.
480 		 */
481 		options = 0;
482 		ucp = (char *)USRSTACK;
483 		(void)subyte(--ucp, 0);		/* trailing zero */
484 		if (boothowto & RB_SINGLE) {
485 			(void)subyte(--ucp, 's');
486 			options = 1;
487 		}
488 #ifdef notyet
489                 if (boothowto & RB_FASTBOOT) {
490 			(void)subyte(--ucp, 'f');
491 			options = 1;
492 		}
493 #endif
494 
495 #ifdef BOOTCDROM
496 		(void)subyte(--ucp, 'C');
497 		options = 1;
498 #endif
499 		if (options == 0)
500 			(void)subyte(--ucp, '-');
501 		(void)subyte(--ucp, '-');		/* leading hyphen */
502 		arg1 = ucp;
503 
504 		/*
505 		 * Move out the file name (also arg 0).
506 		 */
507 		(void)subyte(--ucp, 0);
508 		for (s = next - 1; s >= path; s--)
509 			(void)subyte(--ucp, *s);
510 		arg0 = ucp;
511 
512 		/*
513 		 * Move out the arg pointers.
514 		 */
515 		uap = (char **)((intptr_t)ucp & ~(sizeof(intptr_t)-1));
516 		(void)suword((caddr_t)--uap, (long)0);	/* terminator */
517 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg1);
518 		(void)suword((caddr_t)--uap, (long)(intptr_t)arg0);
519 
520 		/*
521 		 * Point at the arguments.
522 		 */
523 		args.fname = arg0;
524 		args.argv = uap;
525 		args.envv = NULL;
526 
527 		/*
528 		 * Now try to exec the program.  If can't for any reason
529 		 * other than it doesn't exist, complain.
530 		 *
531 		 * Otherwise, return via fork_trampoline() all the way
532 		 * to user mode as init!
533 		 *
534 		 * WARNING!  We may have been moved to another cpu after
535 		 * acquiring P_CURPROC.  The MP lock will migrate with us
536 		 * though so we still have to release it.
537 		 */
538 		if ((error = execve(&args)) == 0) {
539 			acquire_curproc(p);
540 			rel_mplock();
541 			return;
542 		}
543 		if (error != ENOENT)
544 			printf("exec %.*s: error %d\n", (int)(next - path),
545 			    path, error);
546 	}
547 	printf("init: not found in path %s\n", init_path);
548 	panic("no init");
549 }
550 
551 /*
552  * Like kthread_create(), but runs in it's own address space.
553  * We do this early to reserve pid 1.
554  *
555  * Note special case - do not make it runnable yet.  Other work
556  * in progress will change this more.
557  */
558 static void
559 create_init(const void *udata __unused)
560 {
561 	int error;
562 	int s;
563 
564 	s = splhigh();
565 	error = fork1(&proc0, RFFDG | RFPROC, &initproc);
566 	if (error)
567 		panic("cannot fork init: %d\n", error);
568 	initproc->p_flag |= P_INMEM | P_SYSTEM;
569 	cpu_set_fork_handler(initproc, start_init, NULL);
570 	splx(s);
571 }
572 SYSINIT(init,SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)
573 
574 /*
575  * Make it runnable now.
576  */
577 static void
578 kick_init(const void *udata __unused)
579 {
580 	start_forked_proc(&proc0, initproc);
581 }
582 SYSINIT(kickinit,SI_SUB_KTHREAD_INIT, SI_ORDER_FIRST, kick_init, NULL)
583 
584 /*
585  * Machine independant globaldata initialization
586  *
587  * WARNING!  Called from early boot, 'mycpu' may not work yet.
588  */
589 void
590 mi_gdinit(struct globaldata *gd, int cpuid)
591 {
592 	TAILQ_INIT(&gd->gd_tdfreeq);    /* for pmap_{new,dispose}_thread() */
593 	gd->gd_cpuid = cpuid;
594 	lwkt_gdinit(gd);
595 }
596 
597 
598