xref: /dflybsd-src/sys/kern/kern_sysctl.c (revision 6b08710ee11d58fba7e8982f046f86f888c48dfa)
1 /*-
2  * Copyright (c) 1982, 1986, 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Mike Karels at Berkeley Software Design, Inc.
7  *
8  * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
9  * project, to make these variables more userfriendly.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the University of
22  *	California, Berkeley and its contributors.
23  * 4. Neither the name of the University nor the names of its contributors
24  *    may be used to endorse or promote products derived from this software
25  *    without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37  * SUCH DAMAGE.
38  *
39  *	@(#)kern_sysctl.c	8.4 (Berkeley) 4/14/94
40  * $FreeBSD: src/sys/kern/kern_sysctl.c,v 1.92.2.9 2003/05/01 22:48:09 trhodes Exp $
41  * $DragonFly: src/sys/kern/kern_sysctl.c,v 1.13 2003/11/14 02:54:52 daver Exp $
42  */
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/buf.h>
48 #include <sys/sysctl.h>
49 #include <sys/malloc.h>
50 #include <sys/proc.h>
51 #include <sys/sysproto.h>
52 #include <vm/vm.h>
53 #include <vm/vm_extern.h>
54 
55 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
56 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
57 
58 /*
59  * Locking and stats
60  */
61 static struct sysctl_lock {
62 	int	sl_lock;
63 	int	sl_want;
64 	int	sl_locked;
65 } memlock;
66 
67 static int sysctl_root(SYSCTL_HANDLER_ARGS);
68 
69 struct sysctl_oid_list sysctl__children; /* root list */
70 
71 static struct sysctl_oid *
72 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
73 {
74 	struct sysctl_oid *oidp;
75 
76 	SLIST_FOREACH(oidp, list, oid_link) {
77 		if (strcmp(oidp->oid_name, name) == 0) {
78 			return (oidp);
79 		}
80 	}
81 	return (NULL);
82 }
83 
84 /*
85  * Initialization of the MIB tree.
86  *
87  * Order by number in each list.
88  */
89 
90 void sysctl_register_oid(struct sysctl_oid *oidp)
91 {
92 	struct sysctl_oid_list *parent = oidp->oid_parent;
93 	struct sysctl_oid *p;
94 	struct sysctl_oid *q;
95 
96 	/*
97 	 * First check if another oid with the same name already
98 	 * exists in the parent's list.
99 	 */
100 	p = sysctl_find_oidname(oidp->oid_name, parent);
101 	if (p != NULL) {
102 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
103 			p->oid_refcnt++;
104 			return;
105 		} else {
106 			printf("can't re-use a leaf (%s)!\n", p->oid_name);
107 			return;
108 		}
109 	}
110 	/*
111 	 * If this oid has a number OID_AUTO, give it a number which
112 	 * is greater than any current oid.  Make sure it is at least
113 	 * 100 to leave space for pre-assigned oid numbers.
114 	 */
115 	if (oidp->oid_number == OID_AUTO) {
116 		static int newoid = 100;
117 		oidp->oid_number = newoid++;
118 		if (newoid == 0x7fffffff)
119 			panic("out of oids");
120 	}
121 
122 	/*
123 	 * Insert the oid into the parent's list in order.
124 	 */
125 	q = NULL;
126 	SLIST_FOREACH(p, parent, oid_link) {
127 		if (oidp->oid_number < p->oid_number)
128 			break;
129 		q = p;
130 	}
131 	if (q)
132 		SLIST_INSERT_AFTER(q, oidp, oid_link);
133 	else
134 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
135 }
136 
137 void sysctl_unregister_oid(struct sysctl_oid *oidp)
138 {
139 	struct sysctl_oid *p;
140 	int error;
141 
142 	error = ENOENT;
143 	if (oidp->oid_number == OID_AUTO) {
144 		error = EINVAL;
145 	} else {
146 		SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
147 			if (p == oidp) {
148 				SLIST_REMOVE(oidp->oid_parent, oidp,
149 						sysctl_oid, oid_link);
150 				error = 0;
151 				break;
152 			}
153 		}
154 	}
155 
156 	/*
157 	 * This can happen when a module fails to register and is
158 	 * being unloaded afterwards.  It should not be a panic()
159 	 * for normal use.
160 	 */
161 	if (error)
162 		printf("%s: failed to unregister sysctl\n", __func__);
163 }
164 
165 /* Initialize a new context to keep track of dynamically added sysctls. */
166 int
167 sysctl_ctx_init(struct sysctl_ctx_list *c)
168 {
169 
170 	if (c == NULL) {
171 		return (EINVAL);
172 	}
173 	TAILQ_INIT(c);
174 	return (0);
175 }
176 
177 /* Free the context, and destroy all dynamic oids registered in this context */
178 int
179 sysctl_ctx_free(struct sysctl_ctx_list *clist)
180 {
181 	struct sysctl_ctx_entry *e, *e1;
182 	int error;
183 
184 	error = 0;
185 	/*
186 	 * First perform a "dry run" to check if it's ok to remove oids.
187 	 * XXX FIXME
188 	 * XXX This algorithm is a hack. But I don't know any
189 	 * XXX better solution for now...
190 	 */
191 	TAILQ_FOREACH(e, clist, link) {
192 		error = sysctl_remove_oid(e->entry, 0, 0);
193 		if (error)
194 			break;
195 	}
196 	/*
197 	 * Restore deregistered entries, either from the end,
198 	 * or from the place where error occured.
199 	 * e contains the entry that was not unregistered
200 	 */
201 	if (error)
202 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
203 	else
204 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
205 	while (e1 != NULL) {
206 		sysctl_register_oid(e1->entry);
207 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
208 	}
209 	if (error)
210 		return(EBUSY);
211 	/* Now really delete the entries */
212 	e = TAILQ_FIRST(clist);
213 	while (e != NULL) {
214 		e1 = TAILQ_NEXT(e, link);
215 		error = sysctl_remove_oid(e->entry, 1, 0);
216 		if (error)
217 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
218 			    e->entry->oid_name);
219 		free(e, M_SYSCTLOID);
220 		e = e1;
221 	}
222 	return (error);
223 }
224 
225 /* Add an entry to the context */
226 struct sysctl_ctx_entry *
227 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
228 {
229 	struct sysctl_ctx_entry *e;
230 
231 	if (clist == NULL || oidp == NULL)
232 		return(NULL);
233 	e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
234 	e->entry = oidp;
235 	TAILQ_INSERT_HEAD(clist, e, link);
236 	return (e);
237 }
238 
239 /* Find an entry in the context */
240 struct sysctl_ctx_entry *
241 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
242 {
243 	struct sysctl_ctx_entry *e;
244 
245 	if (clist == NULL || oidp == NULL)
246 		return(NULL);
247 	for (e = TAILQ_FIRST(clist); e != NULL; e = TAILQ_NEXT(e, link)) {
248 		if(e->entry == oidp)
249 			return(e);
250 	}
251 	return (e);
252 }
253 
254 /*
255  * Delete an entry from the context.
256  * NOTE: this function doesn't free oidp! You have to remove it
257  * with sysctl_remove_oid().
258  */
259 int
260 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
261 {
262 	struct sysctl_ctx_entry *e;
263 
264 	if (clist == NULL || oidp == NULL)
265 		return (EINVAL);
266 	e = sysctl_ctx_entry_find(clist, oidp);
267 	if (e != NULL) {
268 		TAILQ_REMOVE(clist, e, link);
269 		free(e, M_SYSCTLOID);
270 		return (0);
271 	} else
272 		return (ENOENT);
273 }
274 
275 /*
276  * Remove dynamically created sysctl trees.
277  * oidp - top of the tree to be removed
278  * del - if 0 - just deregister, otherwise free up entries as well
279  * recurse - if != 0 traverse the subtree to be deleted
280  */
281 int
282 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
283 {
284 	struct sysctl_oid *p;
285 	int error;
286 
287 	if (oidp == NULL)
288 		return(EINVAL);
289 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
290 		printf("can't remove non-dynamic nodes!\n");
291 		return (EINVAL);
292 	}
293 	/*
294 	 * WARNING: normal method to do this should be through
295 	 * sysctl_ctx_free(). Use recursing as the last resort
296 	 * method to purge your sysctl tree of leftovers...
297 	 * However, if some other code still references these nodes,
298 	 * it will panic.
299 	 */
300 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
301 		if (oidp->oid_refcnt == 1) {
302 			SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
303 				if (!recurse)
304 					return (ENOTEMPTY);
305 				error = sysctl_remove_oid(p, del, recurse);
306 				if (error)
307 					return (error);
308 			}
309 			if (del)
310 				free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
311 		}
312 	}
313 	if (oidp->oid_refcnt > 1 ) {
314 		oidp->oid_refcnt--;
315 	} else {
316 		if (oidp->oid_refcnt == 0) {
317 			printf("Warning: bad oid_refcnt=%u (%s)!\n",
318 				oidp->oid_refcnt, oidp->oid_name);
319 			return (EINVAL);
320 		}
321 		sysctl_unregister_oid(oidp);
322 		if (del) {
323 			if (oidp->oid_descr)
324 				free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
325 			free((void *)(uintptr_t)(const void *)oidp->oid_name,
326 			     M_SYSCTLOID);
327 			free(oidp, M_SYSCTLOID);
328 		}
329 	}
330 	return (0);
331 }
332 
333 /*
334  * Create new sysctls at run time.
335  * clist may point to a valid context initialized with sysctl_ctx_init().
336  */
337 struct sysctl_oid *
338 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
339 	int number, const char *name, int kind, void *arg1, int arg2,
340 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
341 {
342 	struct sysctl_oid *oidp;
343 	ssize_t len;
344 	char *newname;
345 
346 	/* You have to hook up somewhere.. */
347 	if (parent == NULL)
348 		return(NULL);
349 	/* Check if the node already exists, otherwise create it */
350 	oidp = sysctl_find_oidname(name, parent);
351 	if (oidp != NULL) {
352 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
353 			oidp->oid_refcnt++;
354 			/* Update the context */
355 			if (clist != NULL)
356 				sysctl_ctx_entry_add(clist, oidp);
357 			return (oidp);
358 		} else {
359 			printf("can't re-use a leaf (%s)!\n", name);
360 			return (NULL);
361 		}
362 	}
363 	oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK);
364 	bzero(oidp, sizeof(struct sysctl_oid));
365 	oidp->oid_parent = parent;
366 	SLIST_NEXT(oidp, oid_link) = NULL;
367 	oidp->oid_number = number;
368 	oidp->oid_refcnt = 1;
369 	len = strlen(name);
370 	newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
371 	bcopy(name, newname, len + 1);
372 	newname[len] = '\0';
373 	oidp->oid_name = newname;
374 	oidp->oid_handler = handler;
375 	oidp->oid_kind = CTLFLAG_DYN | kind;
376 	if ((kind & CTLTYPE) == CTLTYPE_NODE) {
377 		/* Allocate space for children */
378 		SYSCTL_CHILDREN(oidp) = malloc(sizeof(struct sysctl_oid_list),
379 		    M_SYSCTLOID, M_WAITOK);
380 		SLIST_INIT(SYSCTL_CHILDREN(oidp));
381 	} else {
382 		oidp->oid_arg1 = arg1;
383 		oidp->oid_arg2 = arg2;
384 	}
385 	oidp->oid_fmt = fmt;
386 	if (descr) {
387 		int len = strlen(descr) + 1;
388 		oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
389 		if (oidp->oid_descr)
390 			strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
391 	};
392 	/* Update the context, if used */
393 	if (clist != NULL)
394 		sysctl_ctx_entry_add(clist, oidp);
395 	/* Register this oid */
396 	sysctl_register_oid(oidp);
397 	return (oidp);
398 }
399 
400 /*
401  * Register the kernel's oids on startup.
402  */
403 SET_DECLARE(sysctl_set, struct sysctl_oid);
404 
405 static void sysctl_register_all(void *arg)
406 {
407 	struct sysctl_oid **oidp;
408 
409 	SET_FOREACH(oidp, sysctl_set)
410 		sysctl_register_oid(*oidp);
411 }
412 
413 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
414 
415 /*
416  * "Staff-functions"
417  *
418  * These functions implement a presently undocumented interface
419  * used by the sysctl program to walk the tree, and get the type
420  * so it can print the value.
421  * This interface is under work and consideration, and should probably
422  * be killed with a big axe by the first person who can find the time.
423  * (be aware though, that the proper interface isn't as obvious as it
424  * may seem, there are various conflicting requirements.
425  *
426  * {0,0}	printf the entire MIB-tree.
427  * {0,1,...}	return the name of the "..." OID.
428  * {0,2,...}	return the next OID.
429  * {0,3}	return the OID of the name in "new"
430  * {0,4,...}	return the kind & format info for the "..." OID.
431  */
432 
433 static void
434 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
435 {
436 	int k;
437 	struct sysctl_oid *oidp;
438 
439 	SLIST_FOREACH(oidp, l, oid_link) {
440 
441 		for (k=0; k<i; k++)
442 			printf(" ");
443 
444 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
445 
446 		printf("%c%c",
447 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
448 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
449 
450 		if (oidp->oid_handler)
451 			printf(" *Handler");
452 
453 		switch (oidp->oid_kind & CTLTYPE) {
454 			case CTLTYPE_NODE:
455 				printf(" Node\n");
456 				if (!oidp->oid_handler) {
457 					sysctl_sysctl_debug_dump_node(
458 						oidp->oid_arg1, i+2);
459 				}
460 				break;
461 			case CTLTYPE_INT:    printf(" Int\n"); break;
462 			case CTLTYPE_STRING: printf(" String\n"); break;
463 			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
464 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
465 			default:	     printf("\n");
466 		}
467 
468 	}
469 }
470 
471 static int
472 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
473 {
474 	int error;
475 
476 	error = suser(req->td);
477 	if (error)
478 		return error;
479 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
480 	return ENOENT;
481 }
482 
483 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
484 	0, 0, sysctl_sysctl_debug, "-", "");
485 
486 static int
487 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
488 {
489 	int *name = (int *) arg1;
490 	u_int namelen = arg2;
491 	int error = 0;
492 	struct sysctl_oid *oid;
493 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
494 	char buf[10];
495 
496 	while (namelen) {
497 		if (!lsp) {
498 			snprintf(buf,sizeof(buf),"%d",*name);
499 			if (req->oldidx)
500 				error = SYSCTL_OUT(req, ".", 1);
501 			if (!error)
502 				error = SYSCTL_OUT(req, buf, strlen(buf));
503 			if (error)
504 				return (error);
505 			namelen--;
506 			name++;
507 			continue;
508 		}
509 		lsp2 = 0;
510 		SLIST_FOREACH(oid, lsp, oid_link) {
511 			if (oid->oid_number != *name)
512 				continue;
513 
514 			if (req->oldidx)
515 				error = SYSCTL_OUT(req, ".", 1);
516 			if (!error)
517 				error = SYSCTL_OUT(req, oid->oid_name,
518 					strlen(oid->oid_name));
519 			if (error)
520 				return (error);
521 
522 			namelen--;
523 			name++;
524 
525 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
526 				break;
527 
528 			if (oid->oid_handler)
529 				break;
530 
531 			lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
532 			break;
533 		}
534 		lsp = lsp2;
535 	}
536 	return (SYSCTL_OUT(req, "", 1));
537 }
538 
539 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
540 
541 static int
542 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
543 	int *next, int *len, int level, struct sysctl_oid **oidpp)
544 {
545 	struct sysctl_oid *oidp;
546 
547 	*len = level;
548 	SLIST_FOREACH(oidp, lsp, oid_link) {
549 		*next = oidp->oid_number;
550 		*oidpp = oidp;
551 
552 		if (!namelen) {
553 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
554 				return 0;
555 			if (oidp->oid_handler)
556 				/* We really should call the handler here...*/
557 				return 0;
558 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
559 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
560 				len, level+1, oidpp))
561 				return 0;
562 			goto emptynode;
563 		}
564 
565 		if (oidp->oid_number < *name)
566 			continue;
567 
568 		if (oidp->oid_number > *name) {
569 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
570 				return 0;
571 			if (oidp->oid_handler)
572 				return 0;
573 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
574 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
575 				next+1, len, level+1, oidpp))
576 				return (0);
577 			goto next;
578 		}
579 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
580 			continue;
581 
582 		if (oidp->oid_handler)
583 			continue;
584 
585 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
586 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
587 			len, level+1, oidpp))
588 			return (0);
589 	next:
590 		namelen = 1;
591 		*len = level;
592 	emptynode:
593 		*len = level;
594 	}
595 	return 1;
596 }
597 
598 static int
599 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
600 {
601 	int *name = (int *) arg1;
602 	u_int namelen = arg2;
603 	int i, j, error;
604 	struct sysctl_oid *oid;
605 	struct sysctl_oid_list *lsp = &sysctl__children;
606 	int newoid[CTL_MAXNAME];
607 
608 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
609 	if (i)
610 		return ENOENT;
611 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
612 	return (error);
613 }
614 
615 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
616 
617 static int
618 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
619 {
620 	int i;
621 	struct sysctl_oid *oidp;
622 	struct sysctl_oid_list *lsp = &sysctl__children;
623 	char *p;
624 
625 	if (!*name)
626 		return ENOENT;
627 
628 	p = name + strlen(name) - 1 ;
629 	if (*p == '.')
630 		*p = '\0';
631 
632 	*len = 0;
633 
634 	for (p = name; *p && *p != '.'; p++)
635 		;
636 	i = *p;
637 	if (i == '.')
638 		*p = '\0';
639 
640 	oidp = SLIST_FIRST(lsp);
641 
642 	while (oidp && *len < CTL_MAXNAME) {
643 		if (strcmp(name, oidp->oid_name)) {
644 			oidp = SLIST_NEXT(oidp, oid_link);
645 			continue;
646 		}
647 		*oid++ = oidp->oid_number;
648 		(*len)++;
649 
650 		if (!i) {
651 			if (oidpp)
652 				*oidpp = oidp;
653 			return (0);
654 		}
655 
656 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
657 			break;
658 
659 		if (oidp->oid_handler)
660 			break;
661 
662 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
663 		oidp = SLIST_FIRST(lsp);
664 		name = p+1;
665 		for (p = name; *p && *p != '.'; p++)
666 				;
667 		i = *p;
668 		if (i == '.')
669 			*p = '\0';
670 	}
671 	return ENOENT;
672 }
673 
674 static int
675 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
676 {
677 	char *p;
678 	int error, oid[CTL_MAXNAME], len;
679 	struct sysctl_oid *op = 0;
680 
681 	if (!req->newlen)
682 		return ENOENT;
683 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
684 		return (ENAMETOOLONG);
685 
686 	p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
687 
688 	error = SYSCTL_IN(req, p, req->newlen);
689 	if (error) {
690 		free(p, M_SYSCTL);
691 		return (error);
692 	}
693 
694 	p [req->newlen] = '\0';
695 
696 	error = name2oid(p, oid, &len, &op);
697 
698 	free(p, M_SYSCTL);
699 
700 	if (error)
701 		return (error);
702 
703 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
704 	return (error);
705 }
706 
707 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
708 	sysctl_sysctl_name2oid, "I", "");
709 
710 static int
711 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
712 {
713 	struct sysctl_oid *oid;
714 	int error;
715 
716 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
717 	if (error)
718 		return (error);
719 
720 	if (!oid->oid_fmt)
721 		return (ENOENT);
722 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
723 	if (error)
724 		return (error);
725 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
726 	return (error);
727 }
728 
729 
730 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
731 
732 static int
733 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
734 {
735 	struct sysctl_oid *oid;
736 	int error;
737 
738 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
739 	if (error)
740 		return (error);
741 
742 	if (!oid->oid_descr)
743 		return (ENOENT);
744 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
745 	return (error);
746 }
747 
748 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
749 
750 /*
751  * Default "handler" functions.
752  */
753 
754 /*
755  * Handle an int, signed or unsigned.
756  * Two cases:
757  *     a variable:  point arg1 at it.
758  *     a constant:  pass it in arg2.
759  */
760 
761 int
762 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
763 {
764 	int error = 0;
765 
766 	if (arg1)
767 		error = SYSCTL_OUT(req, arg1, sizeof(int));
768 	else
769 		error = SYSCTL_OUT(req, &arg2, sizeof(int));
770 
771 	if (error || !req->newptr)
772 		return (error);
773 
774 	if (!arg1)
775 		error = EPERM;
776 	else
777 		error = SYSCTL_IN(req, arg1, sizeof(int));
778 	return (error);
779 }
780 
781 /*
782  * Handle a long, signed or unsigned.  arg1 points to it.
783  */
784 
785 int
786 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
787 {
788 	int error = 0;
789 
790 	if (!arg1)
791 		return (EINVAL);
792 	error = SYSCTL_OUT(req, arg1, sizeof(long));
793 
794 	if (error || !req->newptr)
795 		return (error);
796 
797 	error = SYSCTL_IN(req, arg1, sizeof(long));
798 	return (error);
799 }
800 
801 /*
802  * Handle a quad, signed or unsigned.  arg1 points to it.
803  */
804 
805 int
806 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
807 {
808 	int error = 0;
809 
810 	if (!arg1)
811 		return (EINVAL);
812 	error = SYSCTL_OUT(req, arg1, sizeof(quad_t));
813 
814 	if (error || !req->newptr)
815 		return (error);
816 
817 	error = SYSCTL_IN(req, arg1, sizeof(quad_t));
818 	return (error);
819 }
820 
821 /*
822  * Handle our generic '\0' terminated 'C' string.
823  * Two cases:
824  * 	a variable string:  point arg1 at it, arg2 is max length.
825  * 	a constant string:  point arg1 at it, arg2 is zero.
826  */
827 
828 int
829 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
830 {
831 	int error=0;
832 
833 	error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
834 
835 	if (error || !req->newptr)
836 		return (error);
837 
838 	if ((req->newlen - req->newidx) >= arg2) {
839 		error = EINVAL;
840 	} else {
841 		arg2 = (req->newlen - req->newidx);
842 		error = SYSCTL_IN(req, arg1, arg2);
843 		((char *)arg1)[arg2] = '\0';
844 	}
845 
846 	return (error);
847 }
848 
849 /*
850  * Handle any kind of opaque data.
851  * arg1 points to it, arg2 is the size.
852  */
853 
854 int
855 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
856 {
857 	int error;
858 
859 	error = SYSCTL_OUT(req, arg1, arg2);
860 
861 	if (error || !req->newptr)
862 		return (error);
863 
864 	error = SYSCTL_IN(req, arg1, arg2);
865 
866 	return (error);
867 }
868 
869 /*
870  * Transfer functions to/from kernel space.
871  * XXX: rather untested at this point
872  */
873 static int
874 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
875 {
876 	size_t i = 0;
877 
878 	if (req->oldptr) {
879 		i = l;
880 		if (i > req->oldlen - req->oldidx)
881 			i = req->oldlen - req->oldidx;
882 		if (i > 0)
883 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
884 	}
885 	req->oldidx += l;
886 	if (req->oldptr && i != l)
887 		return (ENOMEM);
888 	return (0);
889 }
890 
891 static int
892 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
893 {
894 
895 	if (!req->newptr)
896 		return 0;
897 	if (req->newlen - req->newidx < l)
898 		return (EINVAL);
899 	bcopy((char *)req->newptr + req->newidx, p, l);
900 	req->newidx += l;
901 	return (0);
902 }
903 
904 int
905 kernel_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval)
906 {
907 	int error = 0;
908 	struct sysctl_req req;
909 
910 	bzero(&req, sizeof req);
911 
912 	req.td = curthread;
913 
914 	if (oldlenp) {
915 		req.oldlen = *oldlenp;
916 	}
917 
918 	if (old) {
919 		req.oldptr = old;
920 	}
921 
922 	if (new != NULL) {
923 		req.newlen = newlen;
924 		req.newptr = new;
925 	}
926 
927 	req.oldfunc = sysctl_old_kernel;
928 	req.newfunc = sysctl_new_kernel;
929 	req.lock = 1;
930 
931 	/* XXX this should probably be done in a general way */
932 	while (memlock.sl_lock) {
933 		memlock.sl_want = 1;
934 		(void) tsleep((caddr_t)&memlock, 0, "sysctl", 0);
935 		memlock.sl_locked++;
936 	}
937 	memlock.sl_lock = 1;
938 
939 	error = sysctl_root(0, name, namelen, &req);
940 
941 	if (req.lock == 2)
942 		vsunlock(req.oldptr, req.oldlen);
943 
944 	memlock.sl_lock = 0;
945 
946 	if (memlock.sl_want) {
947 		memlock.sl_want = 0;
948 		wakeup((caddr_t)&memlock);
949 	}
950 
951 	if (error && error != ENOMEM)
952 		return (error);
953 
954 	if (retval) {
955 		if (req.oldptr && req.oldidx > req.oldlen)
956 			*retval = req.oldlen;
957 		else
958 			*retval = req.oldidx;
959 	}
960 	return (error);
961 }
962 
963 int
964 kernel_sysctlbyname(char *name, void *old, size_t *oldlenp,
965     void *new, size_t newlen, size_t *retval)
966 {
967         int oid[CTL_MAXNAME];
968         size_t oidlen, plen;
969 	int error;
970 
971 	oid[0] = 0;		/* sysctl internal magic */
972 	oid[1] = 3;		/* name2oid */
973 	oidlen = sizeof(oid);
974 
975 	error = kernel_sysctl(oid, 2, oid, &oidlen, (void *)name,
976 		    strlen(name), &plen);
977 	if (error)
978 		return (error);
979 
980 	error = kernel_sysctl(oid, plen / sizeof(int), old, oldlenp,
981 	    new, newlen, retval);
982 	return (error);
983 }
984 
985 /*
986  * Transfer function to/from user space.
987  */
988 static int
989 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
990 {
991 	int error = 0;
992 	size_t i = 0;
993 
994 	if (req->lock == 1 && req->oldptr) {
995 		vslock(req->oldptr, req->oldlen);
996 		req->lock = 2;
997 	}
998 	if (req->oldptr) {
999 		i = l;
1000 		if (i > req->oldlen - req->oldidx)
1001 			i = req->oldlen - req->oldidx;
1002 		if (i > 0)
1003 			error = copyout(p, (char *)req->oldptr + req->oldidx,
1004 					i);
1005 	}
1006 	req->oldidx += l;
1007 	if (error)
1008 		return (error);
1009 	if (req->oldptr && i < l)
1010 		return (ENOMEM);
1011 	return (0);
1012 }
1013 
1014 static int
1015 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1016 {
1017 	int error;
1018 
1019 	if (!req->newptr)
1020 		return 0;
1021 	if (req->newlen - req->newidx < l)
1022 		return (EINVAL);
1023 	error = copyin((char *)req->newptr + req->newidx, p, l);
1024 	req->newidx += l;
1025 	return (error);
1026 }
1027 
1028 int
1029 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1030     int *nindx, struct sysctl_req *req)
1031 {
1032 	struct sysctl_oid *oid;
1033 	int indx;
1034 
1035 	oid = SLIST_FIRST(&sysctl__children);
1036 	indx = 0;
1037 	while (oid && indx < CTL_MAXNAME) {
1038 		if (oid->oid_number == name[indx]) {
1039 			indx++;
1040 			if (oid->oid_kind & CTLFLAG_NOLOCK)
1041 				req->lock = 0;
1042 			if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1043 				if (oid->oid_handler != NULL ||
1044 				    indx == namelen) {
1045 					*noid = oid;
1046 					if (nindx != NULL)
1047 						*nindx = indx;
1048 					return (0);
1049 				}
1050 				oid = SLIST_FIRST(
1051 				    (struct sysctl_oid_list *)oid->oid_arg1);
1052 			} else if (indx == namelen) {
1053 				*noid = oid;
1054 				if (nindx != NULL)
1055 					*nindx = indx;
1056 				return (0);
1057 			} else {
1058 				return (ENOTDIR);
1059 			}
1060 		} else {
1061 			oid = SLIST_NEXT(oid, oid_link);
1062 		}
1063 	}
1064 	return (ENOENT);
1065 }
1066 
1067 /*
1068  * Traverse our tree, and find the right node, execute whatever it points
1069  * to, and return the resulting error code.
1070  */
1071 
1072 int
1073 sysctl_root(SYSCTL_HANDLER_ARGS)
1074 {
1075 	struct thread *td = req->td;
1076 	struct proc *p = td ? td->td_proc : NULL;
1077 	struct sysctl_oid *oid;
1078 	int error, indx;
1079 
1080 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1081 	if (error)
1082 		return (error);
1083 
1084 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1085 		/*
1086 		 * You can't call a sysctl when it's a node, but has
1087 		 * no handler.  Inform the user that it's a node.
1088 		 * The indx may or may not be the same as namelen.
1089 		 */
1090 		if (oid->oid_handler == NULL)
1091 			return (EISDIR);
1092 	}
1093 
1094 	/* If writing isn't allowed */
1095 	if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) ||
1096 	    ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0)))
1097 		return (EPERM);
1098 
1099 	/* Most likely only root can write */
1100 	if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p &&
1101 	    (error = suser_cred(p->p_ucred,
1102 	     (oid->oid_kind & CTLFLAG_PRISON) ? PRISON_ROOT : 0)))
1103 		return (error);
1104 
1105 	if (!oid->oid_handler)
1106 		return EINVAL;
1107 
1108 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE)
1109 		error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx,
1110 		    req);
1111 	else
1112 		error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2,
1113 		    req);
1114 	return (error);
1115 }
1116 
1117 int
1118 __sysctl(struct sysctl_args *uap)
1119 {
1120 	int error, i, name[CTL_MAXNAME];
1121 	size_t j;
1122 
1123 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1124 		return (EINVAL);
1125 
1126  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1127  	if (error)
1128 		return (error);
1129 
1130 	error = userland_sysctl(name, uap->namelen,
1131 		uap->old, uap->oldlenp, 0,
1132 		uap->new, uap->newlen, &j);
1133 	if (error && error != ENOMEM)
1134 		return (error);
1135 	if (uap->oldlenp) {
1136 		i = copyout(&j, uap->oldlenp, sizeof(j));
1137 		if (i)
1138 			return (i);
1139 	}
1140 	return (error);
1141 }
1142 
1143 /*
1144  * This is used from various compatibility syscalls too.  That's why name
1145  * must be in kernel space.
1146  */
1147 int
1148 userland_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval)
1149 {
1150 	int error = 0;
1151 	struct sysctl_req req, req2;
1152 
1153 	bzero(&req, sizeof req);
1154 
1155 	if (oldlenp) {
1156 		if (inkernel) {
1157 			req.oldlen = *oldlenp;
1158 		} else {
1159 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1160 			if (error)
1161 				return (error);
1162 		}
1163 	}
1164 
1165 	if (old) {
1166 		if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1167 			return (EFAULT);
1168 		req.oldptr= old;
1169 	}
1170 
1171 	if (new != NULL) {
1172 		if (!useracc(new, req.newlen, VM_PROT_READ))
1173 			return (EFAULT);
1174 		req.newlen = newlen;
1175 		req.newptr = new;
1176 	}
1177 
1178 	req.oldfunc = sysctl_old_user;
1179 	req.newfunc = sysctl_new_user;
1180 	req.lock = 1;
1181 	req.td = curthread;
1182 
1183 	/* XXX this should probably be done in a general way */
1184 	while (memlock.sl_lock) {
1185 		memlock.sl_want = 1;
1186 		(void) tsleep((caddr_t)&memlock, 0, "sysctl", 0);
1187 		memlock.sl_locked++;
1188 	}
1189 	memlock.sl_lock = 1;
1190 
1191 	do {
1192 	    req2 = req;
1193 	    error = sysctl_root(0, name, namelen, &req2);
1194 	} while (error == EAGAIN);
1195 
1196 	req = req2;
1197 	if (req.lock == 2)
1198 		vsunlock(req.oldptr, req.oldlen);
1199 
1200 	memlock.sl_lock = 0;
1201 
1202 	if (memlock.sl_want) {
1203 		memlock.sl_want = 0;
1204 		wakeup((caddr_t)&memlock);
1205 	}
1206 
1207 	if (error && error != ENOMEM)
1208 		return (error);
1209 
1210 	if (retval) {
1211 		if (req.oldptr && req.oldidx > req.oldlen)
1212 			*retval = req.oldlen;
1213 		else
1214 			*retval = req.oldidx;
1215 	}
1216 	return (error);
1217 }
1218