xref: /dragonfly/sys/kern/kern_sysctl.c (revision 86fe9e07)
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 1.17 2004/05/10 10:51:31 hmp 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 <sys/lock.h>
53 #include <vm/vm.h>
54 #include <vm/vm_extern.h>
55 
56 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
57 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
58 
59 static struct lock sysctl_lkp;
60 static struct lock sysctl_ctx_lkp;
61 
62 static void	sysctl_lock(int type);
63 static void	sysctl_unlock(void);
64 static void	sysctl_ctx_lock(int type);
65 static void	sysctl_ctx_unlock(void);
66 
67 static int	sysctl_root(SYSCTL_HANDLER_ARGS);
68 static void	sysctl_register_oid_int(struct sysctl_oid *oipd);
69 static void	sysctl_unregister_oid_int(struct sysctl_oid *oipd);
70 static struct sysctl_ctx_entry* sysctl_ctx_entry_find_int
71 	(struct sysctl_ctx_list *, struct sysctl_oid *oidp);
72 
73 struct sysctl_oid_list sysctl__children; /* root list */
74 
75 static struct sysctl_oid *
76 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list, int lock)
77 {
78 	struct sysctl_oid *oidp;
79 
80 	SLIST_FOREACH(oidp, list, oid_link) {
81 		if (strcmp(oidp->oid_name, name) == 0) {
82 			break;
83 		}
84 	}
85 	return (oidp);
86 }
87 
88 /*
89  * Initialization of the MIB tree.
90  *
91  * Order by number in each list.
92  */
93 
94 void
95 sysctl_register_oid(struct sysctl_oid *oidp)
96 {
97 	sysctl_lock(LK_EXCLUSIVE);
98 	sysctl_register_oid_int(oidp);
99 	sysctl_unlock();
100 }
101 
102 static void
103 sysctl_register_oid_int(struct sysctl_oid *oidp)
104 {
105 	struct sysctl_oid_list *parent = oidp->oid_parent;
106 	struct sysctl_oid *p;
107 	struct sysctl_oid *q;
108 
109 	/*
110 	 * First check if another oid with the same name already
111 	 * exists in the parent's list.
112 	 */
113 	p = sysctl_find_oidname(oidp->oid_name, parent, 0);
114 	if (p != NULL) {
115 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE)
116 			p->oid_refcnt++;
117 		else
118 			printf("can't re-use a leaf (%s)!\n", p->oid_name);
119 		return;
120 	}
121 
122 	/*
123 	 * If this oid has a number OID_AUTO, give it a number which
124 	 * is greater than any current oid.  Make sure it is at least
125 	 * 256 to leave space for pre-assigned oid numbers.
126 	 */
127 	if (oidp->oid_number == OID_AUTO) {
128 		int newoid = 0x100;	/* minimum AUTO oid */
129 
130 		/*
131 		 * Adjust based on highest oid in parent list
132 		 */
133 		SLIST_FOREACH(p, parent, oid_link) {
134 			if (newoid <= p->oid_number)
135 				newoid = p->oid_number + 1;
136 		}
137 		oidp->oid_number = newoid;
138 	}
139 
140 	/*
141 	 * Insert the oid into the parent's list in order.
142 	 */
143 	q = NULL;
144 	SLIST_FOREACH(p, parent, oid_link) {
145 		if (oidp->oid_number < p->oid_number)
146 			break;
147 		q = p;
148 	}
149 	if (q)
150 		SLIST_INSERT_AFTER(q, oidp, oid_link);
151 	else
152 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
153 }
154 
155 void
156 sysctl_unregister_oid(struct sysctl_oid *oidp)
157 {
158 	sysctl_lock(LK_EXCLUSIVE);
159 	sysctl_unregister_oid_int(oidp);
160 	sysctl_unlock();
161 }
162 
163 static void
164 sysctl_unregister_oid_int(struct sysctl_oid *oidp)
165 {
166 	struct sysctl_oid *p;
167 
168 	if (oidp->oid_number == OID_AUTO)
169 		panic("Trying to unregister OID_AUTO entry: %p", oidp);
170 
171 	SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
172 		if (p != oidp)
173 			continue;
174 		SLIST_REMOVE(oidp->oid_parent, oidp, sysctl_oid, oid_link);
175 		return;
176 	}
177 
178 	/*
179 	 * This can happen when a module fails to register and is
180 	 * being unloaded afterwards.  It should not be a panic()
181 	 * for normal use.
182 	 */
183 	printf("%s: failed to unregister sysctl\n", __func__);
184 }
185 
186 /* Initialize a new context to keep track of dynamically added sysctls. */
187 int
188 sysctl_ctx_init(struct sysctl_ctx_list *c)
189 {
190 	if (c == NULL)
191 		return(EINVAL);
192 	TAILQ_INIT(c);
193 	return(0);
194 }
195 
196 /* Free the context, and destroy all dynamic oids registered in this context */
197 int
198 sysctl_ctx_free(struct sysctl_ctx_list *clist)
199 {
200 	struct sysctl_ctx_entry *e, *e1;
201 	int error;
202 
203 	error = 0;
204 	sysctl_ctx_lock(LK_EXCLUSIVE);
205 	/*
206 	 * First perform a "dry run" to check if it's ok to remove oids.
207 	 * XXX FIXME
208 	 * XXX This algorithm is a hack. But I don't know any
209 	 * XXX better solution for now...
210 	 */
211 	TAILQ_FOREACH(e, clist, link) {
212 		error = sysctl_remove_oid(e->entry, 0, 0);
213 		if (error)
214 			break;
215 	}
216 	/*
217 	 * Restore deregistered entries, either from the end,
218 	 * or from the place where error occured.
219 	 * e contains the entry that was not unregistered
220 	 */
221 	if (error)
222 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
223 	else
224 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
225 	while (e1 != NULL) {
226 		sysctl_register_oid(e1->entry);
227 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
228 	}
229 	if (error) {
230 		sysctl_ctx_unlock();
231 		return(EBUSY);
232 	}
233 	/* Now really delete the entries */
234 	e = TAILQ_FIRST(clist);
235 	while (e != NULL) {
236 		e1 = TAILQ_NEXT(e, link);
237 		error = sysctl_remove_oid(e->entry, 1, 0);
238 		if (error)
239 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
240 			    e->entry->oid_name);
241 		free(e, M_SYSCTLOID);
242 		e = e1;
243 	}
244 	sysctl_ctx_unlock();
245 	return (error);
246 }
247 
248 /* Add an entry to the context */
249 struct sysctl_ctx_entry *
250 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
251 {
252 	struct sysctl_ctx_entry *e;
253 
254 	if (clist == NULL || oidp == NULL)
255 		return(NULL);
256 	e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
257 	e->entry = oidp;
258 	sysctl_ctx_lock(LK_EXCLUSIVE);
259 	TAILQ_INSERT_HEAD(clist, e, link);
260 	sysctl_ctx_unlock();
261 	return (e);
262 }
263 
264 /* Find an entry in the context */
265 struct sysctl_ctx_entry *
266 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
267 {
268 	struct sysctl_ctx_entry *e;
269 
270 	if (clist == NULL || oidp == NULL)
271 		return(NULL);
272 
273 	sysctl_ctx_lock(LK_SHARED);
274 	e = sysctl_ctx_entry_find_int(clist, oidp);
275 	sysctl_ctx_unlock();
276 
277 	return(e);
278 }
279 
280 struct sysctl_ctx_entry *
281 sysctl_ctx_entry_find_int(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
282 {
283 	struct sysctl_ctx_entry *e;
284 
285 	KKASSERT(clist != NULL && oidp != NULL);
286 
287 	for (e = TAILQ_FIRST(clist); e != NULL; e = TAILQ_NEXT(e, link)) {
288 		if(e->entry == oidp)
289 			break;
290 	}
291 
292 	return (e);
293 }
294 
295 /*
296  * Delete an entry from the context.
297  * NOTE: this function doesn't free oidp! You have to remove it
298  * with sysctl_remove_oid().
299  */
300 int
301 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
302 {
303 	struct sysctl_ctx_entry *e;
304 
305 	if (clist == NULL || oidp == NULL)
306 		return (EINVAL);
307 
308 	sysctl_ctx_lock(LK_EXCLUSIVE);
309 	e = sysctl_ctx_entry_find_int(clist, oidp);
310 	if (e == NULL) {
311 		sysctl_ctx_unlock();
312 		return (ENOENT);
313 	}
314 	TAILQ_REMOVE(clist, e, link);
315 	free(e, M_SYSCTLOID);
316 	sysctl_ctx_unlock();
317 
318 	return(0);
319 }
320 
321 /*
322  * Remove dynamically created sysctl trees.
323  * oidp - top of the tree to be removed
324  * del - if 0 - just deregister, otherwise free up entries as well
325  * recurse - if != 0 traverse the subtree to be deleted
326  */
327 int
328 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
329 {
330 	struct sysctl_oid *p;
331 	int error;
332 
333 	if (oidp == NULL)
334 		return(EINVAL);
335 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
336 		printf("can't remove non-dynamic nodes!\n");
337 		return (EINVAL);
338 	}
339 	sysctl_lock(LK_EXCLUSIVE | LK_CANRECURSE);
340 	/*
341 	 * WARNING: normal method to do this should be through
342 	 * sysctl_ctx_free(). Use recursing as the last resort
343 	 * method to purge your sysctl tree of leftovers...
344 	 * However, if some other code still references these nodes,
345 	 * it will panic.
346 	 */
347 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
348 		if (oidp->oid_refcnt == 1) {
349 			SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
350 				if (!recurse) {
351 					sysctl_unlock();
352 					return(ENOTEMPTY);
353 				}
354 				error = sysctl_remove_oid(p, del, recurse);
355 				if (error) {
356 					sysctl_unlock();
357 					return(error);
358 				}
359 			}
360 			if (del)
361 				free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
362 		}
363 	}
364 	if (oidp->oid_refcnt > 1 ) {
365 		oidp->oid_refcnt--;
366 	} else {
367 		if (oidp->oid_refcnt == 0) {
368 			printf("Warning: bad oid_refcnt=%u (%s)!\n",
369 			       oidp->oid_refcnt, oidp->oid_name);
370 			sysctl_unlock();
371 			return(EINVAL);
372 		}
373 		sysctl_unregister_oid_int(oidp);
374 		if (del) {
375 			if (oidp->oid_descr)
376 				free(__DECONST(char *,oidp->oid_descr),
377 				     M_SYSCTLOID);
378 			free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
379 			free(oidp, M_SYSCTLOID);
380 		}
381 	}
382 	sysctl_unlock();
383 	return(0);
384 }
385 
386 /*
387  * Create new sysctls at run time.
388  * clist may point to a valid context initialized with sysctl_ctx_init().
389  */
390 struct sysctl_oid *
391 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
392 	int number, const char *name, int kind, void *arg1, int arg2,
393 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
394 {
395 	struct sysctl_oid *oidp;
396 	ssize_t len;
397 	char *newname;
398 
399 	/* You have to hook up somewhere.. */
400 	if (parent == NULL)
401 		return(NULL);
402 	sysctl_lock(LK_EXCLUSIVE);
403 	/* Check if the node already exists, otherwise create it */
404 	oidp = sysctl_find_oidname(name, parent, 0);
405 	if (oidp != NULL) {
406 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
407 			oidp->oid_refcnt++;
408 			/* Update the context */
409 			if (clist != NULL)
410 				sysctl_ctx_entry_add(clist, oidp);
411 			sysctl_unlock();
412 			return (oidp);
413 		} else {
414 			printf("can't re-use a leaf (%s)!\n", name);
415 			sysctl_unlock();
416 			return (NULL);
417 		}
418 	}
419 	oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK | M_ZERO);
420 	oidp->oid_parent = parent;
421 	SLIST_NEXT(oidp, oid_link) = NULL;
422 	oidp->oid_number = number;
423 	oidp->oid_refcnt = 1;
424 	len = strlen(name);
425 	newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
426 	bcopy(name, newname, len + 1);
427 	newname[len] = '\0';
428 	oidp->oid_name = newname;
429 	oidp->oid_handler = handler;
430 	oidp->oid_kind = CTLFLAG_DYN | kind;
431 	if ((kind & CTLTYPE) == CTLTYPE_NODE) {
432 		/* Allocate space for children */
433 		SYSCTL_CHILDREN(oidp) = malloc(sizeof(struct sysctl_oid_list),
434 		    M_SYSCTLOID, M_WAITOK);
435 		SLIST_INIT(SYSCTL_CHILDREN(oidp));
436 	} else {
437 		oidp->oid_arg1 = arg1;
438 		oidp->oid_arg2 = arg2;
439 	}
440 	oidp->oid_fmt = fmt;
441 	if (descr) {
442 		int len = strlen(descr) + 1;
443 		oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
444 		if (oidp->oid_descr)
445 			strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
446 	};
447 	/* Update the context, if used */
448 	if (clist != NULL)
449 		sysctl_ctx_entry_add(clist, oidp);
450 	/* Register this oid */
451 	sysctl_register_oid_int(oidp);
452 	sysctl_unlock();
453 	return (oidp);
454 }
455 
456 /*
457  * Register the kernel's oids on startup.
458  */
459 SET_DECLARE(sysctl_set, struct sysctl_oid);
460 
461 static void sysctl_register_all(void *arg)
462 {
463 	struct sysctl_oid **oidp;
464 
465 	lockinit(&sysctl_lkp, 0, "sysctl", 0, 0);
466 	lockinit(&sysctl_ctx_lkp, 0, "sysctl ctx", 0, 0);
467 	SET_FOREACH(oidp, sysctl_set)
468 		sysctl_register_oid_int(*oidp);
469 }
470 
471 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
472 
473 /*
474  * "Staff-functions"
475  *
476  * These functions implement a presently undocumented interface
477  * used by the sysctl program to walk the tree, and get the type
478  * so it can print the value.
479  * This interface is under work and consideration, and should probably
480  * be killed with a big axe by the first person who can find the time.
481  * (be aware though, that the proper interface isn't as obvious as it
482  * may seem, there are various conflicting requirements.
483  *
484  * {0,0}	printf the entire MIB-tree.
485  * {0,1,...}	return the name of the "..." OID.
486  * {0,2,...}	return the next OID.
487  * {0,3}	return the OID of the name in "new"
488  * {0,4,...}	return the kind & format info for the "..." OID.
489  */
490 
491 static void
492 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
493 {
494 	int k;
495 	struct sysctl_oid *oidp;
496 
497 	sysctl_lock(LK_SHARED);
498 	SLIST_FOREACH(oidp, l, oid_link) {
499 
500 		for (k=0; k<i; k++)
501 			printf(" ");
502 
503 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
504 
505 		printf("%c%c",
506 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
507 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
508 
509 		if (oidp->oid_handler)
510 			printf(" *Handler");
511 
512 		switch (oidp->oid_kind & CTLTYPE) {
513 			case CTLTYPE_NODE:
514 				printf(" Node\n");
515 				if (!oidp->oid_handler) {
516 					sysctl_sysctl_debug_dump_node(
517 						oidp->oid_arg1, i+2);
518 				}
519 				break;
520 			case CTLTYPE_INT:    printf(" Int\n"); break;
521 			case CTLTYPE_STRING: printf(" String\n"); break;
522 			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
523 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
524 			default:	     printf("\n");
525 		}
526 
527 	}
528 	sysctl_unlock();
529 }
530 
531 static int
532 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
533 {
534 	int error;
535 
536 	error = suser(req->td);
537 	if (error)
538 		return error;
539 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
540 	return ENOENT;
541 }
542 
543 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
544 	0, 0, sysctl_sysctl_debug, "-", "");
545 
546 static int
547 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
548 {
549 	int *name = (int *) arg1;
550 	u_int namelen = arg2;
551 	int error = 0;
552 	struct sysctl_oid *oid;
553 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
554 	char buf[10];
555 
556 	sysctl_lock(LK_SHARED);
557 	while (namelen) {
558 		if (!lsp) {
559 			snprintf(buf,sizeof(buf),"%d",*name);
560 			if (req->oldidx)
561 				error = SYSCTL_OUT(req, ".", 1);
562 			if (!error)
563 				error = SYSCTL_OUT(req, buf, strlen(buf));
564 			if (error) {
565 				sysctl_unlock();
566 				return (error);
567 			}
568 			namelen--;
569 			name++;
570 			continue;
571 		}
572 		lsp2 = 0;
573 		SLIST_FOREACH(oid, lsp, oid_link) {
574 			if (oid->oid_number != *name)
575 				continue;
576 
577 			if (req->oldidx)
578 				error = SYSCTL_OUT(req, ".", 1);
579 			if (!error)
580 				error = SYSCTL_OUT(req, oid->oid_name,
581 					strlen(oid->oid_name));
582 			if (error) {
583 				sysctl_unlock();
584 				return (error);
585 			}
586 
587 			namelen--;
588 			name++;
589 
590 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
591 				break;
592 
593 			if (oid->oid_handler)
594 				break;
595 
596 			lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
597 			break;
598 		}
599 		lsp = lsp2;
600 	}
601 	sysctl_unlock();
602 	return (SYSCTL_OUT(req, "", 1));
603 }
604 
605 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
606 
607 static int
608 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
609 	int *next, int *len, int level, struct sysctl_oid **oidpp)
610 {
611 	struct sysctl_oid *oidp;
612 
613 	*len = level;
614 	sysctl_lock(LK_SHARED);
615 	SLIST_FOREACH(oidp, lsp, oid_link) {
616 		*next = oidp->oid_number;
617 		*oidpp = oidp;
618 
619 		if (!namelen) {
620 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) {
621 				sysctl_unlock();
622 				return 0;
623 			}
624 			if (oidp->oid_handler) {
625 				/* We really should call the handler here...*/
626 				sysctl_unlock();
627 				return 0;
628 			}
629 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
630 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
631 				len, level+1, oidpp)) {
632 				sysctl_unlock();
633 				return 0;
634 			}
635 			goto emptynode;
636 		}
637 
638 		if (oidp->oid_number < *name)
639 			continue;
640 
641 		if (oidp->oid_number > *name) {
642 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) {
643 				sysctl_unlock();
644 				return 0;
645 			}
646 			if (oidp->oid_handler) {
647 				sysctl_unlock();
648 				return 0;
649 			}
650 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
651 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
652 				next+1, len, level+1, oidpp)) {
653 				sysctl_unlock();
654 				return (0);
655 			}
656 			goto next;
657 		}
658 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
659 			continue;
660 
661 		if (oidp->oid_handler)
662 			continue;
663 
664 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
665 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
666 			len, level+1, oidpp)) {
667 			sysctl_unlock();
668 			return (0);
669 		}
670 	next:
671 		namelen = 1;
672 		*len = level;
673 	emptynode:
674 		*len = level;
675 	}
676 	sysctl_unlock();
677 	return 1;
678 }
679 
680 static int
681 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
682 {
683 	int *name = (int *) arg1;
684 	u_int namelen = arg2;
685 	int i, j, error;
686 	struct sysctl_oid *oid;
687 	struct sysctl_oid_list *lsp = &sysctl__children;
688 	int newoid[CTL_MAXNAME];
689 
690 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
691 	if (i)
692 		return ENOENT;
693 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
694 	return (error);
695 }
696 
697 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
698 
699 static int
700 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
701 {
702 	int i;
703 	struct sysctl_oid *oidp;
704 	struct sysctl_oid_list *lsp = &sysctl__children;
705 	char *p;
706 
707 	if (!*name)
708 		return ENOENT;
709 
710 	p = name + strlen(name) - 1 ;
711 	if (*p == '.')
712 		*p = '\0';
713 
714 	*len = 0;
715 
716 	for (p = name; *p && *p != '.'; p++)
717 		;
718 	i = *p;
719 	if (i == '.')
720 		*p = '\0';
721 
722 	sysctl_lock(LK_SHARED);
723 	oidp = SLIST_FIRST(lsp);
724 
725 	while (oidp && *len < CTL_MAXNAME) {
726 		if (strcmp(name, oidp->oid_name)) {
727 			oidp = SLIST_NEXT(oidp, oid_link);
728 			continue;
729 		}
730 		*oid++ = oidp->oid_number;
731 		(*len)++;
732 
733 		if (!i) {
734 			if (oidpp)
735 				*oidpp = oidp;
736 			sysctl_unlock();
737 			return (0);
738 		}
739 
740 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
741 			break;
742 
743 		if (oidp->oid_handler)
744 			break;
745 
746 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
747 		oidp = SLIST_FIRST(lsp);
748 		name = p+1;
749 		for (p = name; *p && *p != '.'; p++)
750 				;
751 		i = *p;
752 		if (i == '.')
753 			*p = '\0';
754 	}
755 	sysctl_unlock();
756 	return ENOENT;
757 }
758 
759 static int
760 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
761 {
762 	char *p;
763 	int error, oid[CTL_MAXNAME], len;
764 	struct sysctl_oid *op = 0;
765 
766 	if (!req->newlen)
767 		return ENOENT;
768 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
769 		return (ENAMETOOLONG);
770 
771 	p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
772 
773 	error = SYSCTL_IN(req, p, req->newlen);
774 	if (error) {
775 		free(p, M_SYSCTL);
776 		return (error);
777 	}
778 
779 	p [req->newlen] = '\0';
780 
781 	error = name2oid(p, oid, &len, &op);
782 
783 	free(p, M_SYSCTL);
784 
785 	if (error)
786 		return (error);
787 
788 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
789 	return (error);
790 }
791 
792 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
793 	sysctl_sysctl_name2oid, "I", "");
794 
795 static int
796 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
797 {
798 	struct sysctl_oid *oid;
799 	int error;
800 
801 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
802 	if (error)
803 		return (error);
804 
805 	if (!oid->oid_fmt)
806 		return (ENOENT);
807 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
808 	if (error)
809 		return (error);
810 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
811 	return (error);
812 }
813 
814 
815 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
816 
817 static int
818 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
819 {
820 	struct sysctl_oid *oid;
821 	int error;
822 
823 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
824 	if (error)
825 		return (error);
826 
827 	if (!oid->oid_descr)
828 		return (ENOENT);
829 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
830 	return (error);
831 }
832 
833 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
834 
835 /*
836  * Default "handler" functions.
837  */
838 
839 /*
840  * Handle an int, signed or unsigned.
841  * Two cases:
842  *     a variable:  point arg1 at it.
843  *     a constant:  pass it in arg2.
844  */
845 
846 int
847 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
848 {
849 	int error = 0;
850 
851 	if (arg1)
852 		error = SYSCTL_OUT(req, arg1, sizeof(int));
853 	else
854 		error = SYSCTL_OUT(req, &arg2, sizeof(int));
855 
856 	if (error || !req->newptr)
857 		return (error);
858 
859 	if (!arg1)
860 		error = EPERM;
861 	else
862 		error = SYSCTL_IN(req, arg1, sizeof(int));
863 	return (error);
864 }
865 
866 /*
867  * Handle a long, signed or unsigned.  arg1 points to it.
868  */
869 
870 int
871 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
872 {
873 	int error = 0;
874 
875 	if (!arg1)
876 		return (EINVAL);
877 	error = SYSCTL_OUT(req, arg1, sizeof(long));
878 
879 	if (error || !req->newptr)
880 		return (error);
881 
882 	error = SYSCTL_IN(req, arg1, sizeof(long));
883 	return (error);
884 }
885 
886 /*
887  * Handle a quad, signed or unsigned.  arg1 points to it.
888  */
889 
890 int
891 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
892 {
893 	int error = 0;
894 
895 	if (!arg1)
896 		return (EINVAL);
897 	error = SYSCTL_OUT(req, arg1, sizeof(quad_t));
898 
899 	if (error || !req->newptr)
900 		return (error);
901 
902 	error = SYSCTL_IN(req, arg1, sizeof(quad_t));
903 	return (error);
904 }
905 
906 /*
907  * Handle our generic '\0' terminated 'C' string.
908  * Two cases:
909  * 	a variable string:  point arg1 at it, arg2 is max length.
910  * 	a constant string:  point arg1 at it, arg2 is zero.
911  */
912 
913 int
914 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
915 {
916 	int error=0;
917 
918 	error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
919 
920 	if (error || !req->newptr)
921 		return (error);
922 
923 	if ((req->newlen - req->newidx) >= arg2) {
924 		error = EINVAL;
925 	} else {
926 		arg2 = (req->newlen - req->newidx);
927 		error = SYSCTL_IN(req, arg1, arg2);
928 		((char *)arg1)[arg2] = '\0';
929 	}
930 
931 	return (error);
932 }
933 
934 /*
935  * Handle any kind of opaque data.
936  * arg1 points to it, arg2 is the size.
937  */
938 
939 int
940 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
941 {
942 	int error;
943 
944 	error = SYSCTL_OUT(req, arg1, arg2);
945 
946 	if (error || !req->newptr)
947 		return (error);
948 
949 	error = SYSCTL_IN(req, arg1, arg2);
950 
951 	return (error);
952 }
953 
954 /*
955  * Transfer functions to/from kernel space.
956  * XXX: rather untested at this point
957  */
958 static int
959 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
960 {
961 	size_t i = 0;
962 
963 	if (req->oldptr) {
964 		i = l;
965 		if (i > req->oldlen - req->oldidx)
966 			i = req->oldlen - req->oldidx;
967 		if (i > 0)
968 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
969 	}
970 	req->oldidx += l;
971 	if (req->oldptr && i != l)
972 		return (ENOMEM);
973 	return (0);
974 }
975 
976 static int
977 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
978 {
979 
980 	if (!req->newptr)
981 		return 0;
982 	if (req->newlen - req->newidx < l)
983 		return (EINVAL);
984 	bcopy((char *)req->newptr + req->newidx, p, l);
985 	req->newidx += l;
986 	return (0);
987 }
988 
989 int
990 kernel_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval)
991 {
992 	int error = 0;
993 	struct sysctl_req req;
994 
995 	bzero(&req, sizeof req);
996 
997 	req.td = curthread;
998 
999 	if (oldlenp) {
1000 		req.oldlen = *oldlenp;
1001 	}
1002 
1003 	if (old) {
1004 		req.oldptr = old;
1005 	}
1006 
1007 	if (new != NULL) {
1008 		req.newlen = newlen;
1009 		req.newptr = new;
1010 	}
1011 
1012 	req.oldfunc = sysctl_old_kernel;
1013 	req.newfunc = sysctl_new_kernel;
1014 	req.lock = 1;
1015 
1016 	sysctl_lock(LK_SHARED);
1017 
1018 	error = sysctl_root(0, name, namelen, &req);
1019 
1020 	if (req.lock == 2)
1021 		vsunlock(req.oldptr, req.oldlen);
1022 
1023 	sysctl_unlock();
1024 
1025 	if (error && error != ENOMEM)
1026 		return (error);
1027 
1028 	if (retval) {
1029 		if (req.oldptr && req.oldidx > req.oldlen)
1030 			*retval = req.oldlen;
1031 		else
1032 			*retval = req.oldidx;
1033 	}
1034 	return (error);
1035 }
1036 
1037 int
1038 kernel_sysctlbyname(char *name, void *old, size_t *oldlenp,
1039     void *new, size_t newlen, size_t *retval)
1040 {
1041         int oid[CTL_MAXNAME];
1042         size_t oidlen, plen;
1043 	int error;
1044 
1045 	oid[0] = 0;		/* sysctl internal magic */
1046 	oid[1] = 3;		/* name2oid */
1047 	oidlen = sizeof(oid);
1048 
1049 	error = kernel_sysctl(oid, 2, oid, &oidlen, (void *)name,
1050 		    strlen(name), &plen);
1051 	if (error)
1052 		return (error);
1053 
1054 	error = kernel_sysctl(oid, plen / sizeof(int), old, oldlenp,
1055 	    new, newlen, retval);
1056 	return (error);
1057 }
1058 
1059 /*
1060  * Transfer function to/from user space.
1061  */
1062 static int
1063 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1064 {
1065 	int error = 0;
1066 	size_t i = 0;
1067 
1068 	if (req->lock == 1 && req->oldptr) {
1069 		vslock(req->oldptr, req->oldlen);
1070 		req->lock = 2;
1071 	}
1072 	if (req->oldptr) {
1073 		i = l;
1074 		if (i > req->oldlen - req->oldidx)
1075 			i = req->oldlen - req->oldidx;
1076 		if (i > 0)
1077 			error = copyout(p, (char *)req->oldptr + req->oldidx,
1078 					i);
1079 	}
1080 	req->oldidx += l;
1081 	if (error)
1082 		return (error);
1083 	if (req->oldptr && i < l)
1084 		return (ENOMEM);
1085 	return (0);
1086 }
1087 
1088 static int
1089 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1090 {
1091 	int error;
1092 
1093 	if (!req->newptr)
1094 		return 0;
1095 	if (req->newlen - req->newidx < l)
1096 		return (EINVAL);
1097 	error = copyin((char *)req->newptr + req->newidx, p, l);
1098 	req->newidx += l;
1099 	return (error);
1100 }
1101 
1102 int
1103 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1104     int *nindx, struct sysctl_req *req)
1105 {
1106 	struct sysctl_oid *oid;
1107 	int indx;
1108 
1109 	sysctl_lock(LK_SHARED);
1110 	oid = SLIST_FIRST(&sysctl__children);
1111 	indx = 0;
1112 	while (oid && indx < CTL_MAXNAME) {
1113 		if (oid->oid_number == name[indx]) {
1114 			indx++;
1115 			if (oid->oid_kind & CTLFLAG_NOLOCK)
1116 				req->lock = 0;
1117 			if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1118 				if (oid->oid_handler != NULL ||
1119 				    indx == namelen) {
1120 					*noid = oid;
1121 					if (nindx != NULL)
1122 						*nindx = indx;
1123 					sysctl_unlock();
1124 					return (0);
1125 				}
1126 				oid = SLIST_FIRST(
1127 				    (struct sysctl_oid_list *)oid->oid_arg1);
1128 			} else if (indx == namelen) {
1129 				*noid = oid;
1130 				if (nindx != NULL)
1131 					*nindx = indx;
1132 				sysctl_unlock();
1133 				return (0);
1134 			} else {
1135 				sysctl_unlock();
1136 				return (ENOTDIR);
1137 			}
1138 		} else {
1139 			oid = SLIST_NEXT(oid, oid_link);
1140 		}
1141 	}
1142 	sysctl_unlock();
1143 	return (ENOENT);
1144 }
1145 
1146 /*
1147  * Traverse our tree, and find the right node, execute whatever it points
1148  * to, and return the resulting error code.
1149  */
1150 
1151 int
1152 sysctl_root(SYSCTL_HANDLER_ARGS)
1153 {
1154 	struct thread *td = req->td;
1155 	struct proc *p = td ? td->td_proc : NULL;
1156 	struct sysctl_oid *oid;
1157 	int error, indx;
1158 
1159 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1160 	if (error)
1161 		return (error);
1162 
1163 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1164 		/*
1165 		 * You can't call a sysctl when it's a node, but has
1166 		 * no handler.  Inform the user that it's a node.
1167 		 * The indx may or may not be the same as namelen.
1168 		 */
1169 		if (oid->oid_handler == NULL)
1170 			return (EISDIR);
1171 	}
1172 
1173 	/* If writing isn't allowed */
1174 	if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) ||
1175 	    ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0)))
1176 		return (EPERM);
1177 
1178 	/* Most likely only root can write */
1179 	if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p &&
1180 	    (error = suser_cred(p->p_ucred,
1181 	     (oid->oid_kind & CTLFLAG_PRISON) ? PRISON_ROOT : 0)))
1182 		return (error);
1183 
1184 	if (!oid->oid_handler)
1185 		return EINVAL;
1186 
1187 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE)
1188 		error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx,
1189 		    req);
1190 	else
1191 		error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2,
1192 		    req);
1193 	return (error);
1194 }
1195 
1196 int
1197 __sysctl(struct sysctl_args *uap)
1198 {
1199 	int error, i, name[CTL_MAXNAME];
1200 	size_t j;
1201 
1202 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1203 		return (EINVAL);
1204 
1205  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1206  	if (error)
1207 		return (error);
1208 
1209 	error = userland_sysctl(name, uap->namelen,
1210 		uap->old, uap->oldlenp, 0,
1211 		uap->new, uap->newlen, &j);
1212 	if (error && error != ENOMEM)
1213 		return (error);
1214 	if (uap->oldlenp) {
1215 		i = copyout(&j, uap->oldlenp, sizeof(j));
1216 		if (i)
1217 			return (i);
1218 	}
1219 	return (error);
1220 }
1221 
1222 /*
1223  * This is used from various compatibility syscalls too.  That's why name
1224  * must be in kernel space.
1225  */
1226 int
1227 userland_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval)
1228 {
1229 	int error = 0;
1230 	struct sysctl_req req, req2;
1231 
1232 	bzero(&req, sizeof req);
1233 
1234 	if (oldlenp) {
1235 		if (inkernel) {
1236 			req.oldlen = *oldlenp;
1237 		} else {
1238 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1239 			if (error)
1240 				return (error);
1241 		}
1242 	}
1243 
1244 	if (old) {
1245 		if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1246 			return (EFAULT);
1247 		req.oldptr= old;
1248 	}
1249 
1250 	if (new != NULL) {
1251 		if (!useracc(new, req.newlen, VM_PROT_READ))
1252 			return (EFAULT);
1253 		req.newlen = newlen;
1254 		req.newptr = new;
1255 	}
1256 
1257 	req.oldfunc = sysctl_old_user;
1258 	req.newfunc = sysctl_new_user;
1259 	req.lock = 1;
1260 	req.td = curthread;
1261 
1262 	sysctl_lock(LK_SHARED);
1263 
1264 	do {
1265 	    req2 = req;
1266 	    error = sysctl_root(0, name, namelen, &req2);
1267 	} while (error == EAGAIN);
1268 
1269 	req = req2;
1270 	if (req.lock == 2)
1271 		vsunlock(req.oldptr, req.oldlen);
1272 
1273 	sysctl_unlock();
1274 
1275 	if (error && error != ENOMEM)
1276 		return (error);
1277 
1278 	if (retval) {
1279 		if (req.oldptr && req.oldidx > req.oldlen)
1280 			*retval = req.oldlen;
1281 		else
1282 			*retval = req.oldidx;
1283 	}
1284 	return (error);
1285 }
1286 
1287 static void
1288 sysctl_lock(int flag)
1289 {
1290 	lockmgr(&sysctl_lkp, flag, NULL, curthread);
1291 }
1292 
1293 static void
1294 sysctl_unlock(void)
1295 {
1296 	lockmgr(&sysctl_lkp, LK_RELEASE, NULL, curthread);
1297 }
1298 
1299 static void
1300 sysctl_ctx_lock(int flag)
1301 {
1302 	lockmgr(&sysctl_ctx_lkp, flag, NULL, curthread);
1303 }
1304 
1305 static void
1306 sysctl_ctx_unlock(void)
1307 {
1308 	lockmgr(&sysctl_ctx_lkp, LK_RELEASE, NULL, curthread);
1309 }
1310