xref: /dragonfly/sys/kern/kern_sysctl.c (revision 768af85b)
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 1.30 2008/08/03 11:00:32 sephe 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/priv.h>
52 #include <sys/sysproto.h>
53 #include <sys/lock.h>
54 
55 #include <sys/mplock2.h>
56 
57 #include <vm/vm.h>
58 #include <vm/vm_extern.h>
59 
60 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
61 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
62 
63 static struct lock sysctl_lkp;
64 static struct lock sysctl_ctx_lkp;
65 
66 static void	sysctl_lock(int type);
67 static void	sysctl_unlock(void);
68 static void	sysctl_ctx_lock(int type);
69 static void	sysctl_ctx_unlock(void);
70 
71 static int	sysctl_root(SYSCTL_HANDLER_ARGS);
72 static void	sysctl_register_oid_int(struct sysctl_oid *oipd);
73 static void	sysctl_unregister_oid_int(struct sysctl_oid *oipd);
74 static struct sysctl_ctx_entry* sysctl_ctx_entry_find_int
75 	(struct sysctl_ctx_list *, struct sysctl_oid *oidp);
76 
77 struct sysctl_oid_list sysctl__children; /* root list */
78 
79 static struct sysctl_oid *
80 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list, int lock)
81 {
82 	struct sysctl_oid *oidp;
83 
84 	SLIST_FOREACH(oidp, list, oid_link) {
85 		if (strcmp(oidp->oid_name, name) == 0) {
86 			break;
87 		}
88 	}
89 	return (oidp);
90 }
91 
92 /*
93  * Initialization of the MIB tree.
94  *
95  * Order by number in each list.
96  */
97 
98 void
99 sysctl_register_oid(struct sysctl_oid *oidp)
100 {
101 	sysctl_lock(LK_EXCLUSIVE);
102 	sysctl_register_oid_int(oidp);
103 	sysctl_unlock();
104 }
105 
106 static void
107 sysctl_register_oid_int(struct sysctl_oid *oidp)
108 {
109 	struct sysctl_oid_list *parent = oidp->oid_parent;
110 	struct sysctl_oid *p;
111 	struct sysctl_oid *q;
112 
113 	/*
114 	 * First check if another oid with the same name already
115 	 * exists in the parent's list.
116 	 */
117 	p = sysctl_find_oidname(oidp->oid_name, parent, 0);
118 	if (p != NULL) {
119 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE)
120 			p->oid_refcnt++;
121 		else
122 			kprintf("can't re-use a leaf (%s)!\n", p->oid_name);
123 		return;
124 	}
125 
126 	/*
127 	 * If this oid has a number OID_AUTO, give it a number which
128 	 * is greater than any current oid.  Make sure it is at least
129 	 * 256 to leave space for pre-assigned oid numbers.
130 	 */
131 	if (oidp->oid_number == OID_AUTO) {
132 		int newoid = 0x100;	/* minimum AUTO oid */
133 
134 		/*
135 		 * Adjust based on highest oid in parent list
136 		 */
137 		SLIST_FOREACH(p, parent, oid_link) {
138 			if (newoid <= p->oid_number)
139 				newoid = p->oid_number + 1;
140 		}
141 		oidp->oid_number = newoid;
142 	}
143 
144 	/*
145 	 * Insert the oid into the parent's list in order.
146 	 */
147 	q = NULL;
148 	SLIST_FOREACH(p, parent, oid_link) {
149 		if (oidp->oid_number < p->oid_number)
150 			break;
151 		q = p;
152 	}
153 	if (q)
154 		SLIST_INSERT_AFTER(q, oidp, oid_link);
155 	else
156 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
157 }
158 
159 void
160 sysctl_unregister_oid(struct sysctl_oid *oidp)
161 {
162 	sysctl_lock(LK_EXCLUSIVE);
163 	sysctl_unregister_oid_int(oidp);
164 	sysctl_unlock();
165 }
166 
167 static void
168 sysctl_unregister_oid_int(struct sysctl_oid *oidp)
169 {
170 	struct sysctl_oid *p;
171 
172 	if (oidp->oid_number == OID_AUTO)
173 		panic("Trying to unregister OID_AUTO entry: %p", oidp);
174 
175 	SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
176 		if (p != oidp)
177 			continue;
178 		SLIST_REMOVE(oidp->oid_parent, oidp, sysctl_oid, oid_link);
179 		return;
180 	}
181 
182 	/*
183 	 * This can happen when a module fails to register and is
184 	 * being unloaded afterwards.  It should not be a panic()
185 	 * for normal use.
186 	 */
187 	kprintf("%s: failed to unregister sysctl\n", __func__);
188 }
189 
190 /* Initialize a new context to keep track of dynamically added sysctls. */
191 int
192 sysctl_ctx_init(struct sysctl_ctx_list *c)
193 {
194 	if (c == NULL)
195 		return(EINVAL);
196 	TAILQ_INIT(c);
197 	return(0);
198 }
199 
200 /* Free the context, and destroy all dynamic oids registered in this context */
201 int
202 sysctl_ctx_free(struct sysctl_ctx_list *clist)
203 {
204 	struct sysctl_ctx_entry *e, *e1;
205 	int error;
206 
207 	error = 0;
208 	sysctl_ctx_lock(LK_EXCLUSIVE);
209 	/*
210 	 * First perform a "dry run" to check if it's ok to remove oids.
211 	 * XXX FIXME
212 	 * XXX This algorithm is a hack. But I don't know any
213 	 * XXX better solution for now...
214 	 */
215 	TAILQ_FOREACH(e, clist, link) {
216 		error = sysctl_remove_oid(e->entry, 0, 0);
217 		if (error)
218 			break;
219 	}
220 	/*
221 	 * Restore deregistered entries, either from the end,
222 	 * or from the place where error occured.
223 	 * e contains the entry that was not unregistered
224 	 */
225 	if (error)
226 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
227 	else
228 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
229 	while (e1 != NULL) {
230 		sysctl_register_oid(e1->entry);
231 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
232 	}
233 	if (error) {
234 		sysctl_ctx_unlock();
235 		return(EBUSY);
236 	}
237 	/* Now really delete the entries */
238 	e = TAILQ_FIRST(clist);
239 	while (e != NULL) {
240 		e1 = TAILQ_NEXT(e, link);
241 		error = sysctl_remove_oid(e->entry, 1, 0);
242 		if (error)
243 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
244 			    e->entry->oid_name);
245 		kfree(e, M_SYSCTLOID);
246 		e = e1;
247 	}
248 	sysctl_ctx_unlock();
249 	return (error);
250 }
251 
252 /* Add an entry to the context */
253 struct sysctl_ctx_entry *
254 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
255 {
256 	struct sysctl_ctx_entry *e;
257 
258 	if (clist == NULL || oidp == NULL)
259 		return(NULL);
260 	e = kmalloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
261 	e->entry = oidp;
262 	sysctl_ctx_lock(LK_EXCLUSIVE);
263 	TAILQ_INSERT_HEAD(clist, e, link);
264 	sysctl_ctx_unlock();
265 	return (e);
266 }
267 
268 /* Find an entry in the context */
269 struct sysctl_ctx_entry *
270 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
271 {
272 	struct sysctl_ctx_entry *e;
273 
274 	if (clist == NULL || oidp == NULL)
275 		return(NULL);
276 
277 	sysctl_ctx_lock(LK_SHARED);
278 	e = sysctl_ctx_entry_find_int(clist, oidp);
279 	sysctl_ctx_unlock();
280 
281 	return(e);
282 }
283 
284 struct sysctl_ctx_entry *
285 sysctl_ctx_entry_find_int(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
286 {
287 	struct sysctl_ctx_entry *e;
288 
289 	KKASSERT(clist != NULL && oidp != NULL);
290 
291 	for (e = TAILQ_FIRST(clist); e != NULL; e = TAILQ_NEXT(e, link)) {
292 		if(e->entry == oidp)
293 			break;
294 	}
295 
296 	return (e);
297 }
298 
299 /*
300  * Delete an entry from the context.
301  * NOTE: this function doesn't free oidp! You have to remove it
302  * with sysctl_remove_oid().
303  */
304 int
305 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
306 {
307 	struct sysctl_ctx_entry *e;
308 
309 	if (clist == NULL || oidp == NULL)
310 		return (EINVAL);
311 
312 	sysctl_ctx_lock(LK_EXCLUSIVE);
313 	e = sysctl_ctx_entry_find_int(clist, oidp);
314 	if (e == NULL) {
315 		sysctl_ctx_unlock();
316 		return (ENOENT);
317 	}
318 	TAILQ_REMOVE(clist, e, link);
319 	kfree(e, M_SYSCTLOID);
320 	sysctl_ctx_unlock();
321 
322 	return(0);
323 }
324 
325 /*
326  * Remove dynamically created sysctl trees.
327  * oidp - top of the tree to be removed
328  * del - if 0 - just deregister, otherwise free up entries as well
329  * recurse - if != 0 traverse the subtree to be deleted
330  */
331 int
332 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
333 {
334 	struct sysctl_oid *p;
335 	int error;
336 
337 	if (oidp == NULL)
338 		return(EINVAL);
339 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
340 		kprintf("can't remove non-dynamic nodes!\n");
341 		return (EINVAL);
342 	}
343 	sysctl_lock(LK_EXCLUSIVE | LK_CANRECURSE);
344 	/*
345 	 * WARNING: normal method to do this should be through
346 	 * sysctl_ctx_free(). Use recursing as the last resort
347 	 * method to purge your sysctl tree of leftovers...
348 	 * However, if some other code still references these nodes,
349 	 * it will panic.
350 	 */
351 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
352 		if (oidp->oid_refcnt == 1) {
353 			SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
354 				if (!recurse) {
355 					sysctl_unlock();
356 					return(ENOTEMPTY);
357 				}
358 				error = sysctl_remove_oid(p, del, recurse);
359 				if (error) {
360 					sysctl_unlock();
361 					return(error);
362 				}
363 			}
364 			if (del)
365 				kfree(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
366 		}
367 	}
368 	if (oidp->oid_refcnt > 1 ) {
369 		oidp->oid_refcnt--;
370 	} else {
371 		if (oidp->oid_refcnt == 0) {
372 			kprintf("Warning: bad oid_refcnt=%u (%s)!\n",
373 			       oidp->oid_refcnt, oidp->oid_name);
374 			sysctl_unlock();
375 			return(EINVAL);
376 		}
377 		sysctl_unregister_oid_int(oidp);
378 		if (del) {
379 			if (oidp->oid_descr)
380 				kfree(__DECONST(char *,oidp->oid_descr),
381 				     M_SYSCTLOID);
382 			kfree(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
383 			kfree(oidp, M_SYSCTLOID);
384 		}
385 	}
386 	sysctl_unlock();
387 	return(0);
388 }
389 
390 /*
391  * Create new sysctls at run time.
392  * clist may point to a valid context initialized with sysctl_ctx_init().
393  */
394 struct sysctl_oid *
395 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
396 	int number, const char *name, int kind, void *arg1, int arg2,
397 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
398 {
399 	struct sysctl_oid *oidp;
400 	ssize_t len;
401 	char *newname;
402 
403 	/* You have to hook up somewhere.. */
404 	if (parent == NULL)
405 		return(NULL);
406 	sysctl_lock(LK_EXCLUSIVE);
407 	/* Check if the node already exists, otherwise create it */
408 	oidp = sysctl_find_oidname(name, parent, 0);
409 	if (oidp != NULL) {
410 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
411 			oidp->oid_refcnt++;
412 			/* Update the context */
413 			if (clist != NULL)
414 				sysctl_ctx_entry_add(clist, oidp);
415 			sysctl_unlock();
416 			return (oidp);
417 		} else {
418 			kprintf("can't re-use a leaf (%s)!\n", name);
419 			sysctl_unlock();
420 			return (NULL);
421 		}
422 	}
423 	oidp = kmalloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK | M_ZERO);
424 	oidp->oid_parent = parent;
425 	SLIST_NEXT(oidp, oid_link) = NULL;
426 	oidp->oid_number = number;
427 	oidp->oid_refcnt = 1;
428 	len = strlen(name);
429 	newname = kmalloc(len + 1, M_SYSCTLOID, M_WAITOK);
430 	bcopy(name, newname, len + 1);
431 	newname[len] = '\0';
432 	oidp->oid_name = newname;
433 	oidp->oid_handler = handler;
434 	oidp->oid_kind = CTLFLAG_DYN | kind;
435 	if ((kind & CTLTYPE) == CTLTYPE_NODE) {
436 		struct sysctl_oid_list *children;
437 
438 		/* Allocate space for children */
439 		children = kmalloc(sizeof(*children), M_SYSCTLOID, M_WAITOK);
440 		SYSCTL_SET_CHILDREN(oidp, children);
441 		SLIST_INIT(children);
442 	} else {
443 		oidp->oid_arg1 = arg1;
444 		oidp->oid_arg2 = arg2;
445 	}
446 	oidp->oid_fmt = fmt;
447 	if (descr) {
448 		int len = strlen(descr) + 1;
449 		oidp->oid_descr = kmalloc(len, M_SYSCTLOID, M_WAITOK);
450 		strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
451 	};
452 	/* Update the context, if used */
453 	if (clist != NULL)
454 		sysctl_ctx_entry_add(clist, oidp);
455 	/* Register this oid */
456 	sysctl_register_oid_int(oidp);
457 	sysctl_unlock();
458 	return (oidp);
459 }
460 
461 /*
462  * Register the kernel's oids on startup.
463  */
464 SET_DECLARE(sysctl_set, struct sysctl_oid);
465 
466 static void
467 sysctl_register_all(void *arg)
468 {
469 	struct sysctl_oid **oidp;
470 
471 	lockinit(&sysctl_lkp, "sysctl", 0, 0);
472 	lockinit(&sysctl_ctx_lkp, "sysctl ctx", 0, 0);
473 	SET_FOREACH(oidp, sysctl_set)
474 		sysctl_register_oid_int(*oidp);
475 }
476 
477 SYSINIT(sysctl, SI_BOOT1_POST, SI_ORDER_ANY, sysctl_register_all, 0);
478 
479 /*
480  * "Staff-functions"
481  *
482  * These functions implement a presently undocumented interface
483  * used by the sysctl program to walk the tree, and get the type
484  * so it can print the value.
485  * This interface is under work and consideration, and should probably
486  * be killed with a big axe by the first person who can find the time.
487  * (be aware though, that the proper interface isn't as obvious as it
488  * may seem, there are various conflicting requirements.
489  *
490  * {0,0}	kprintf the entire MIB-tree.
491  * {0,1,...}	return the name of the "..." OID.
492  * {0,2,...}	return the next OID.
493  * {0,3}	return the OID of the name in "new"
494  * {0,4,...}	return the kind & format info for the "..." OID.
495  */
496 
497 static void
498 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
499 {
500 	int k;
501 	struct sysctl_oid *oidp;
502 
503 	sysctl_lock(LK_SHARED);
504 	SLIST_FOREACH(oidp, l, oid_link) {
505 
506 		for (k=0; k<i; k++)
507 			kprintf(" ");
508 
509 		kprintf("%d %s ", oidp->oid_number, oidp->oid_name);
510 
511 		kprintf("%c%c",
512 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
513 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
514 
515 		if (oidp->oid_handler)
516 			kprintf(" *Handler");
517 
518 		switch (oidp->oid_kind & CTLTYPE) {
519 			case CTLTYPE_NODE:
520 				kprintf(" Node\n");
521 				if (!oidp->oid_handler) {
522 					sysctl_sysctl_debug_dump_node(
523 						oidp->oid_arg1, i+2);
524 				}
525 				break;
526 			case CTLTYPE_INT:    kprintf(" Int\n"); break;
527 			case CTLTYPE_STRING: kprintf(" String\n"); break;
528 			case CTLTYPE_QUAD:   kprintf(" Quad\n"); break;
529 			case CTLTYPE_OPAQUE: kprintf(" Opaque/struct\n"); break;
530 			default:	     kprintf("\n");
531 		}
532 
533 	}
534 	sysctl_unlock();
535 }
536 
537 static int
538 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
539 {
540 	int error;
541 
542 	error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
543 	if (error)
544 		return error;
545 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
546 	return ENOENT;
547 }
548 
549 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
550 	0, 0, sysctl_sysctl_debug, "-", "");
551 
552 static int
553 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
554 {
555 	int *name = (int *) arg1;
556 	u_int namelen = arg2;
557 	int error = 0;
558 	struct sysctl_oid *oid;
559 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
560 	char buf[10];
561 
562 	sysctl_lock(LK_SHARED);
563 	while (namelen) {
564 		if (!lsp) {
565 			ksnprintf(buf,sizeof(buf),"%d",*name);
566 			if (req->oldidx)
567 				error = SYSCTL_OUT(req, ".", 1);
568 			if (!error)
569 				error = SYSCTL_OUT(req, buf, strlen(buf));
570 			if (error) {
571 				sysctl_unlock();
572 				return (error);
573 			}
574 			namelen--;
575 			name++;
576 			continue;
577 		}
578 		lsp2 = 0;
579 		SLIST_FOREACH(oid, lsp, oid_link) {
580 			if (oid->oid_number != *name)
581 				continue;
582 
583 			if (req->oldidx)
584 				error = SYSCTL_OUT(req, ".", 1);
585 			if (!error)
586 				error = SYSCTL_OUT(req, oid->oid_name,
587 					strlen(oid->oid_name));
588 			if (error) {
589 				sysctl_unlock();
590 				return (error);
591 			}
592 
593 			namelen--;
594 			name++;
595 
596 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
597 				break;
598 
599 			if (oid->oid_handler)
600 				break;
601 
602 			lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
603 			break;
604 		}
605 		lsp = lsp2;
606 	}
607 	sysctl_unlock();
608 	return (SYSCTL_OUT(req, "", 1));
609 }
610 
611 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
612 
613 static int
614 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
615 	int *next, int *len, int level, struct sysctl_oid **oidpp)
616 {
617 	struct sysctl_oid *oidp;
618 
619 	*len = level;
620 	sysctl_lock(LK_SHARED);
621 	SLIST_FOREACH(oidp, lsp, oid_link) {
622 		*next = oidp->oid_number;
623 		*oidpp = oidp;
624 
625 		if (!namelen) {
626 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) {
627 				sysctl_unlock();
628 				return 0;
629 			}
630 			if (oidp->oid_handler) {
631 				/* We really should call the handler here...*/
632 				sysctl_unlock();
633 				return 0;
634 			}
635 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
636 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
637 				len, level+1, oidpp)) {
638 				sysctl_unlock();
639 				return 0;
640 			}
641 			goto emptynode;
642 		}
643 
644 		if (oidp->oid_number < *name)
645 			continue;
646 
647 		if (oidp->oid_number > *name) {
648 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) {
649 				sysctl_unlock();
650 				return 0;
651 			}
652 			if (oidp->oid_handler) {
653 				sysctl_unlock();
654 				return 0;
655 			}
656 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
657 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
658 				next+1, len, level+1, oidpp)) {
659 				sysctl_unlock();
660 				return (0);
661 			}
662 			goto next;
663 		}
664 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
665 			continue;
666 
667 		if (oidp->oid_handler)
668 			continue;
669 
670 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
671 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
672 			len, level+1, oidpp)) {
673 			sysctl_unlock();
674 			return (0);
675 		}
676 	next:
677 		namelen = 1;
678 		*len = level;
679 	emptynode:
680 		*len = level;
681 	}
682 	sysctl_unlock();
683 	return 1;
684 }
685 
686 static int
687 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
688 {
689 	int *name = (int *) arg1;
690 	u_int namelen = arg2;
691 	int i, j, error;
692 	struct sysctl_oid *oid;
693 	struct sysctl_oid_list *lsp = &sysctl__children;
694 	int newoid[CTL_MAXNAME];
695 
696 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
697 	if (i)
698 		return ENOENT;
699 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
700 	return (error);
701 }
702 
703 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
704 
705 static int
706 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
707 {
708 	int i;
709 	struct sysctl_oid *oidp;
710 	struct sysctl_oid_list *lsp = &sysctl__children;
711 	char *p;
712 
713 	if (!*name)
714 		return ENOENT;
715 
716 	p = name + strlen(name) - 1 ;
717 	if (*p == '.')
718 		*p = '\0';
719 
720 	*len = 0;
721 
722 	for (p = name; *p && *p != '.'; p++)
723 		;
724 	i = *p;
725 	if (i == '.')
726 		*p = '\0';
727 
728 	sysctl_lock(LK_SHARED);
729 	oidp = SLIST_FIRST(lsp);
730 
731 	while (oidp && *len < CTL_MAXNAME) {
732 		if (strcmp(name, oidp->oid_name)) {
733 			oidp = SLIST_NEXT(oidp, oid_link);
734 			continue;
735 		}
736 		*oid++ = oidp->oid_number;
737 		(*len)++;
738 
739 		if (!i) {
740 			if (oidpp)
741 				*oidpp = oidp;
742 			sysctl_unlock();
743 			return (0);
744 		}
745 
746 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
747 			break;
748 
749 		if (oidp->oid_handler)
750 			break;
751 
752 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
753 		oidp = SLIST_FIRST(lsp);
754 		name = p+1;
755 		for (p = name; *p && *p != '.'; p++)
756 				;
757 		i = *p;
758 		if (i == '.')
759 			*p = '\0';
760 	}
761 	sysctl_unlock();
762 	return ENOENT;
763 }
764 
765 static int
766 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
767 {
768 	char *p;
769 	int error, oid[CTL_MAXNAME], len;
770 	struct sysctl_oid *op = 0;
771 
772 	if (!req->newlen)
773 		return ENOENT;
774 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
775 		return (ENAMETOOLONG);
776 
777 	p = kmalloc(req->newlen+1, M_SYSCTL, M_WAITOK);
778 
779 	error = SYSCTL_IN(req, p, req->newlen);
780 	if (error) {
781 		kfree(p, M_SYSCTL);
782 		return (error);
783 	}
784 
785 	p [req->newlen] = '\0';
786 
787 	error = name2oid(p, oid, &len, &op);
788 
789 	kfree(p, M_SYSCTL);
790 
791 	if (error)
792 		return (error);
793 
794 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
795 	return (error);
796 }
797 
798 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
799 	sysctl_sysctl_name2oid, "I", "");
800 
801 static int
802 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
803 {
804 	struct sysctl_oid *oid;
805 	int error;
806 
807 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
808 	if (error)
809 		return (error);
810 
811 	if (!oid->oid_fmt)
812 		return (ENOENT);
813 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
814 	if (error)
815 		return (error);
816 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
817 	return (error);
818 }
819 
820 
821 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
822 
823 static int
824 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
825 {
826 	struct sysctl_oid *oid;
827 	int error;
828 
829 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
830 	if (error)
831 		return (error);
832 
833 	if (!oid->oid_descr)
834 		return (ENOENT);
835 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
836 	return (error);
837 }
838 
839 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
840 
841 /*
842  * Default "handler" functions.
843  */
844 
845 /*
846  * Handle an int, signed or unsigned.
847  * Two cases:
848  *     a variable:  point arg1 at it.
849  *     a constant:  pass it in arg2.
850  */
851 
852 int
853 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
854 {
855 	int error = 0;
856 
857 	if (arg1)
858 		error = SYSCTL_OUT(req, arg1, sizeof(int));
859 	else
860 		error = SYSCTL_OUT(req, &arg2, sizeof(int));
861 
862 	if (error || !req->newptr)
863 		return (error);
864 
865 	if (!arg1)
866 		error = EPERM;
867 	else
868 		error = SYSCTL_IN(req, arg1, sizeof(int));
869 	return (error);
870 }
871 
872 /*
873  * Handle a long, signed or unsigned.  arg1 points to it.
874  */
875 
876 int
877 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
878 {
879 	int error = 0;
880 
881 	if (!arg1)
882 		return (EINVAL);
883 	error = SYSCTL_OUT(req, arg1, sizeof(long));
884 
885 	if (error || !req->newptr)
886 		return (error);
887 
888 	error = SYSCTL_IN(req, arg1, sizeof(long));
889 	return (error);
890 }
891 
892 /*
893  * Handle a quad, signed or unsigned.  arg1 points to it.
894  */
895 
896 int
897 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
898 {
899 	int error = 0;
900 
901 	if (!arg1)
902 		return (EINVAL);
903 	error = SYSCTL_OUT(req, arg1, sizeof(quad_t));
904 
905 	if (error || !req->newptr)
906 		return (error);
907 
908 	error = SYSCTL_IN(req, arg1, sizeof(quad_t));
909 	return (error);
910 }
911 
912 /*
913  * Handle our generic '\0' terminated 'C' string.
914  * Two cases:
915  * 	a variable string:  point arg1 at it, arg2 is max length.
916  * 	a constant string:  point arg1 at it, arg2 is zero.
917  */
918 
919 int
920 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
921 {
922 	int error=0;
923 
924 	error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
925 
926 	if (error || !req->newptr)
927 		return (error);
928 
929 	if ((req->newlen - req->newidx) >= arg2) {
930 		error = EINVAL;
931 	} else {
932 		arg2 = (req->newlen - req->newidx);
933 		error = SYSCTL_IN(req, arg1, arg2);
934 		((char *)arg1)[arg2] = '\0';
935 	}
936 
937 	return (error);
938 }
939 
940 /*
941  * Handle any kind of opaque data.
942  * arg1 points to it, arg2 is the size.
943  */
944 
945 int
946 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
947 {
948 	int error;
949 
950 	error = SYSCTL_OUT(req, arg1, arg2);
951 
952 	if (error || !req->newptr)
953 		return (error);
954 
955 	error = SYSCTL_IN(req, arg1, arg2);
956 
957 	return (error);
958 }
959 
960 /*
961  * Transfer functions to/from kernel space.
962  * XXX: rather untested at this point
963  */
964 static int
965 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
966 {
967 	size_t i = 0;
968 
969 	if (req->oldptr) {
970 		i = l;
971 		if (i > req->oldlen - req->oldidx)
972 			i = req->oldlen - req->oldidx;
973 		if (i > 0)
974 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
975 	}
976 	req->oldidx += l;
977 	if (req->oldptr && i != l)
978 		return (ENOMEM);
979 	return (0);
980 }
981 
982 static int
983 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
984 {
985 
986 	if (!req->newptr)
987 		return 0;
988 	if (req->newlen - req->newidx < l)
989 		return (EINVAL);
990 	bcopy((char *)req->newptr + req->newidx, p, l);
991 	req->newidx += l;
992 	return (0);
993 }
994 
995 int
996 kernel_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval)
997 {
998 	int error = 0;
999 	struct sysctl_req req;
1000 
1001 	bzero(&req, sizeof req);
1002 
1003 	req.td = curthread;
1004 
1005 	if (oldlenp) {
1006 		req.oldlen = *oldlenp;
1007 	}
1008 
1009 	if (old) {
1010 		req.oldptr = old;
1011 	}
1012 
1013 	if (new != NULL) {
1014 		req.newlen = newlen;
1015 		req.newptr = new;
1016 	}
1017 
1018 	req.oldfunc = sysctl_old_kernel;
1019 	req.newfunc = sysctl_new_kernel;
1020 #if 0
1021 	req.lock = 1;
1022 #endif
1023 
1024 	sysctl_lock(LK_SHARED);
1025 
1026 	error = sysctl_root(0, name, namelen, &req);
1027 
1028 #if 0
1029 	if (req.lock == 2)
1030 		vsunlock(req.oldptr, req.oldlen);
1031 #endif
1032 
1033 	sysctl_unlock();
1034 
1035 	if (error && error != ENOMEM)
1036 		return (error);
1037 
1038 	if (retval) {
1039 		if (req.oldptr && req.oldidx > req.oldlen)
1040 			*retval = req.oldlen;
1041 		else
1042 			*retval = req.oldidx;
1043 	}
1044 	return (error);
1045 }
1046 
1047 int
1048 kernel_sysctlbyname(char *name, void *old, size_t *oldlenp,
1049     void *new, size_t newlen, size_t *retval)
1050 {
1051         int oid[CTL_MAXNAME];
1052         size_t oidlen, plen;
1053 	int error;
1054 
1055 	oid[0] = 0;		/* sysctl internal magic */
1056 	oid[1] = 3;		/* name2oid */
1057 	oidlen = sizeof(oid);
1058 
1059 	error = kernel_sysctl(oid, 2, oid, &oidlen, (void *)name,
1060 		    strlen(name), &plen);
1061 	if (error)
1062 		return (error);
1063 
1064 	error = kernel_sysctl(oid, plen / sizeof(int), old, oldlenp,
1065 	    new, newlen, retval);
1066 	return (error);
1067 }
1068 
1069 /*
1070  * Transfer function to/from user space.
1071  */
1072 static int
1073 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1074 {
1075 	int error = 0;
1076 	size_t i = 0;
1077 
1078 #if 0
1079 	if (req->lock == 1 && req->oldptr) {
1080 		vslock(req->oldptr, req->oldlen);
1081 		req->lock = 2;
1082 	}
1083 #endif
1084 	if (req->oldptr) {
1085 		i = l;
1086 		if (i > req->oldlen - req->oldidx)
1087 			i = req->oldlen - req->oldidx;
1088 		if (i > 0)
1089 			error = copyout(p, (char *)req->oldptr + req->oldidx,
1090 					i);
1091 	}
1092 	req->oldidx += l;
1093 	if (error)
1094 		return (error);
1095 	if (req->oldptr && i < l)
1096 		return (ENOMEM);
1097 	return (0);
1098 }
1099 
1100 static int
1101 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1102 {
1103 	int error;
1104 
1105 	if (!req->newptr)
1106 		return 0;
1107 	if (req->newlen - req->newidx < l)
1108 		return (EINVAL);
1109 	error = copyin((char *)req->newptr + req->newidx, p, l);
1110 	req->newidx += l;
1111 	return (error);
1112 }
1113 
1114 int
1115 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1116     int *nindx, struct sysctl_req *req)
1117 {
1118 	struct sysctl_oid *oid;
1119 	int indx;
1120 
1121 	sysctl_lock(LK_SHARED);
1122 	oid = SLIST_FIRST(&sysctl__children);
1123 	indx = 0;
1124 	while (oid && indx < CTL_MAXNAME) {
1125 		if (oid->oid_number == name[indx]) {
1126 			indx++;
1127 			if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1128 				if (oid->oid_handler != NULL ||
1129 				    indx == namelen) {
1130 					*noid = oid;
1131 					if (nindx != NULL)
1132 						*nindx = indx;
1133 					sysctl_unlock();
1134 					return (0);
1135 				}
1136 				oid = SLIST_FIRST(
1137 				    (struct sysctl_oid_list *)oid->oid_arg1);
1138 			} else if (indx == namelen) {
1139 				*noid = oid;
1140 				if (nindx != NULL)
1141 					*nindx = indx;
1142 				sysctl_unlock();
1143 				return (0);
1144 			} else {
1145 				sysctl_unlock();
1146 				return (ENOTDIR);
1147 			}
1148 		} else {
1149 			oid = SLIST_NEXT(oid, oid_link);
1150 		}
1151 	}
1152 	sysctl_unlock();
1153 	return (ENOENT);
1154 }
1155 
1156 /*
1157  * Traverse our tree, and find the right node, execute whatever it points
1158  * to, and return the resulting error code.
1159  */
1160 
1161 int
1162 sysctl_root(SYSCTL_HANDLER_ARGS)
1163 {
1164 	struct thread *td = req->td;
1165 	struct proc *p = td ? td->td_proc : NULL;
1166 	struct sysctl_oid *oid;
1167 	int error, indx;
1168 
1169 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1170 	if (error)
1171 		return (error);
1172 
1173 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1174 		/*
1175 		 * You can't call a sysctl when it's a node, but has
1176 		 * no handler.  Inform the user that it's a node.
1177 		 * The indx may or may not be the same as namelen.
1178 		 */
1179 		if (oid->oid_handler == NULL)
1180 			return (EISDIR);
1181 	}
1182 
1183 	/* If writing isn't allowed */
1184 	if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) ||
1185 	    ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0)))
1186 		return (EPERM);
1187 
1188 	/* Most likely only root can write */
1189 	if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p &&
1190 	    (error = priv_check_cred(td->td_ucred,
1191 	     (oid->oid_kind & CTLFLAG_PRISON) ? PRIV_SYSCTL_WRITEJAIL :
1192 	                                        PRIV_SYSCTL_WRITE, 0)))
1193 		return (error);
1194 
1195 	if (!oid->oid_handler)
1196 		return EINVAL;
1197 
1198 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE)
1199 		error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx,
1200 		    req);
1201 	else
1202 		error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2,
1203 		    req);
1204 	return (error);
1205 }
1206 
1207 /*
1208  * MPALMOSTSAFE
1209  */
1210 int
1211 sys___sysctl(struct sysctl_args *uap)
1212 {
1213 	int error, i, name[CTL_MAXNAME];
1214 	size_t j;
1215 
1216 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1217 		return (EINVAL);
1218 
1219  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1220  	if (error)
1221 		return (error);
1222 
1223 	get_mplock();
1224 	error = userland_sysctl(name, uap->namelen,
1225 		uap->old, uap->oldlenp, 0,
1226 		uap->new, uap->newlen, &j);
1227 	rel_mplock();
1228 	if (error && error != ENOMEM)
1229 		return (error);
1230 	if (uap->oldlenp) {
1231 		i = copyout(&j, uap->oldlenp, sizeof(j));
1232 		if (i)
1233 			return (i);
1234 	}
1235 	return (error);
1236 }
1237 
1238 /*
1239  * This is used from various compatibility syscalls too.  That's why name
1240  * must be in kernel space.
1241  */
1242 int
1243 userland_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval)
1244 {
1245 	int error = 0;
1246 	struct sysctl_req req, req2;
1247 
1248 	bzero(&req, sizeof req);
1249 
1250 	if (oldlenp) {
1251 		if (inkernel) {
1252 			req.oldlen = *oldlenp;
1253 		} else {
1254 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1255 			if (error)
1256 				return (error);
1257 		}
1258 	}
1259 
1260 	if (old) {
1261 		if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1262 			return (EFAULT);
1263 		req.oldptr= old;
1264 	}
1265 
1266 	if (new != NULL) {
1267 		if (!useracc(new, newlen, VM_PROT_READ))
1268 			return (EFAULT);
1269 		req.newlen = newlen;
1270 		req.newptr = new;
1271 	}
1272 
1273 	req.oldfunc = sysctl_old_user;
1274 	req.newfunc = sysctl_new_user;
1275 #if 0
1276 	req.lock = 1;
1277 #endif
1278 	req.td = curthread;
1279 
1280 	sysctl_lock(LK_SHARED);
1281 
1282 	do {
1283 	    req2 = req;
1284 	    error = sysctl_root(0, name, namelen, &req2);
1285 	} while (error == EAGAIN);
1286 
1287 	req = req2;
1288 #if 0
1289 	if (req.lock == 2)
1290 		vsunlock(req.oldptr, req.oldlen);
1291 #endif
1292 
1293 	sysctl_unlock();
1294 
1295 	if (error && error != ENOMEM)
1296 		return (error);
1297 
1298 	if (retval) {
1299 		if (req.oldptr && req.oldidx > req.oldlen)
1300 			*retval = req.oldlen;
1301 		else
1302 			*retval = req.oldidx;
1303 	}
1304 	return (error);
1305 }
1306 
1307 static void
1308 sysctl_lock(int flag)
1309 {
1310 	lockmgr(&sysctl_lkp, flag);
1311 }
1312 
1313 static void
1314 sysctl_unlock(void)
1315 {
1316 	lockmgr(&sysctl_lkp, LK_RELEASE);
1317 }
1318 
1319 static void
1320 sysctl_ctx_lock(int flag)
1321 {
1322 	lockmgr(&sysctl_ctx_lkp, flag);
1323 }
1324 
1325 static void
1326 sysctl_ctx_unlock(void)
1327 {
1328 	lockmgr(&sysctl_ctx_lkp, LK_RELEASE);
1329 }
1330 
1331 int
1332 sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high)
1333 {
1334 	int error, value;
1335 
1336 	value = *(int *)arg1;
1337 	error = sysctl_handle_int(oidp, &value, 0, req);
1338 	if (error || !req->newptr)
1339 		return (error);
1340 	if (value < low || value > high)
1341 		return (EINVAL);
1342 	*(int *)arg1 = value;
1343 	return (0);
1344 }
1345