xref: /dragonfly/sys/kern/kern_sysctl.c (revision 3bafb5c1)
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  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/buf.h>
47 #include <sys/sysctl.h>
48 #include <sys/malloc.h>
49 #include <sys/proc.h>
50 #include <sys/priv.h>
51 #include <sys/sysproto.h>
52 #include <sys/lock.h>
53 
54 #include <sys/mplock2.h>
55 
56 #include <vm/vm.h>
57 #include <vm/vm_extern.h>
58 
59 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
60 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
61 
62 static struct lock sysctl_lkp;
63 static struct lock sysctl_ctx_lkp;
64 
65 static void	sysctl_lock(int type);
66 static void	sysctl_unlock(void);
67 static void	sysctl_ctx_lock(int type);
68 static void	sysctl_ctx_unlock(void);
69 
70 static int	sysctl_root(SYSCTL_HANDLER_ARGS);
71 static void	sysctl_register_oid_int(struct sysctl_oid *oipd);
72 static void	sysctl_unregister_oid_int(struct sysctl_oid *oipd);
73 static struct sysctl_ctx_entry* sysctl_ctx_entry_find_int
74 	(struct sysctl_ctx_list *, struct sysctl_oid *oidp);
75 
76 struct sysctl_oid_list sysctl__children; /* root list */
77 
78 static struct sysctl_oid *
79 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list, int lock)
80 {
81 	struct sysctl_oid *oidp;
82 
83 	SLIST_FOREACH(oidp, list, oid_link) {
84 		if (strcmp(oidp->oid_name, name) == 0) {
85 			break;
86 		}
87 	}
88 	return (oidp);
89 }
90 
91 /*
92  * Initialization of the MIB tree.
93  *
94  * Order by number in each list.
95  */
96 
97 void
98 sysctl_register_oid(struct sysctl_oid *oidp)
99 {
100 	sysctl_lock(LK_EXCLUSIVE);
101 	sysctl_register_oid_int(oidp);
102 	sysctl_unlock();
103 }
104 
105 static void
106 sysctl_register_oid_int(struct sysctl_oid *oidp)
107 {
108 	struct sysctl_oid_list *parent = oidp->oid_parent;
109 	struct sysctl_oid *p;
110 	struct sysctl_oid *q;
111 
112 	/*
113 	 * First check if another oid with the same name already
114 	 * exists in the parent's list.
115 	 */
116 	p = sysctl_find_oidname(oidp->oid_name, parent, 0);
117 	if (p != NULL) {
118 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE)
119 			p->oid_refcnt++;
120 		else
121 			kprintf("can't re-use a leaf (%s)!\n", p->oid_name);
122 		return;
123 	}
124 
125 	/*
126 	 * If this oid has a number OID_AUTO, give it a number which
127 	 * is greater than any current oid.  Make sure it is at least
128 	 * 256 to leave space for pre-assigned oid numbers.
129 	 */
130 	if (oidp->oid_number == OID_AUTO) {
131 		int newoid = 0x100;	/* minimum AUTO oid */
132 
133 		/*
134 		 * Adjust based on highest oid in parent list
135 		 */
136 		SLIST_FOREACH(p, parent, oid_link) {
137 			if (newoid <= p->oid_number)
138 				newoid = p->oid_number + 1;
139 		}
140 		oidp->oid_number = newoid;
141 	}
142 
143 	/*
144 	 * Insert the oid into the parent's list in order.
145 	 */
146 	q = NULL;
147 	SLIST_FOREACH(p, parent, oid_link) {
148 		if (oidp->oid_number < p->oid_number)
149 			break;
150 		q = p;
151 	}
152 	if (q)
153 		SLIST_INSERT_AFTER(q, oidp, oid_link);
154 	else
155 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
156 }
157 
158 void
159 sysctl_unregister_oid(struct sysctl_oid *oidp)
160 {
161 	sysctl_lock(LK_EXCLUSIVE);
162 	sysctl_unregister_oid_int(oidp);
163 	sysctl_unlock();
164 }
165 
166 static void
167 sysctl_unregister_oid_int(struct sysctl_oid *oidp)
168 {
169 	struct sysctl_oid *p;
170 
171 	if (oidp->oid_number == OID_AUTO)
172 		panic("Trying to unregister OID_AUTO entry: %p", oidp);
173 
174 	SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
175 		if (p != oidp)
176 			continue;
177 		SLIST_REMOVE(oidp->oid_parent, oidp, sysctl_oid, oid_link);
178 		return;
179 	}
180 
181 	/*
182 	 * This can happen when a module fails to register and is
183 	 * being unloaded afterwards.  It should not be a panic()
184 	 * for normal use.
185 	 */
186 	kprintf("%s: failed to unregister sysctl\n", __func__);
187 }
188 
189 /* Initialize a new context to keep track of dynamically added sysctls. */
190 int
191 sysctl_ctx_init(struct sysctl_ctx_list *c)
192 {
193 	if (c == NULL)
194 		return(EINVAL);
195 	TAILQ_INIT(c);
196 	return(0);
197 }
198 
199 /* Free the context, and destroy all dynamic oids registered in this context */
200 int
201 sysctl_ctx_free(struct sysctl_ctx_list *clist)
202 {
203 	struct sysctl_ctx_entry *e, *e1;
204 	int error;
205 
206 	error = 0;
207 	sysctl_ctx_lock(LK_EXCLUSIVE);
208 	/*
209 	 * First perform a "dry run" to check if it's ok to remove oids.
210 	 * XXX FIXME
211 	 * XXX This algorithm is a hack. But I don't know any
212 	 * XXX better solution for now...
213 	 */
214 	TAILQ_FOREACH(e, clist, link) {
215 		error = sysctl_remove_oid(e->entry, 0, 0);
216 		if (error)
217 			break;
218 	}
219 	/*
220 	 * Restore deregistered entries, either from the end,
221 	 * or from the place where error occured.
222 	 * e contains the entry that was not unregistered
223 	 */
224 	if (error)
225 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
226 	else
227 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
228 	while (e1 != NULL) {
229 		sysctl_register_oid(e1->entry);
230 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
231 	}
232 	if (error) {
233 		sysctl_ctx_unlock();
234 		return(EBUSY);
235 	}
236 	/* Now really delete the entries */
237 	e = TAILQ_FIRST(clist);
238 	while (e != NULL) {
239 		e1 = TAILQ_NEXT(e, link);
240 		error = sysctl_remove_oid(e->entry, 1, 0);
241 		if (error)
242 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
243 			    e->entry->oid_name);
244 		kfree(e, M_SYSCTLOID);
245 		e = e1;
246 	}
247 	sysctl_ctx_unlock();
248 	return (error);
249 }
250 
251 /* Add an entry to the context */
252 struct sysctl_ctx_entry *
253 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
254 {
255 	struct sysctl_ctx_entry *e;
256 
257 	if (clist == NULL || oidp == NULL)
258 		return(NULL);
259 	e = kmalloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
260 	e->entry = oidp;
261 	sysctl_ctx_lock(LK_EXCLUSIVE);
262 	TAILQ_INSERT_HEAD(clist, e, link);
263 	sysctl_ctx_unlock();
264 	return (e);
265 }
266 
267 /* Find an entry in the context */
268 struct sysctl_ctx_entry *
269 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
270 {
271 	struct sysctl_ctx_entry *e;
272 
273 	if (clist == NULL || oidp == NULL)
274 		return(NULL);
275 
276 	sysctl_ctx_lock(LK_SHARED);
277 	e = sysctl_ctx_entry_find_int(clist, oidp);
278 	sysctl_ctx_unlock();
279 
280 	return(e);
281 }
282 
283 struct sysctl_ctx_entry *
284 sysctl_ctx_entry_find_int(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
285 {
286 	struct sysctl_ctx_entry *e;
287 
288 	KKASSERT(clist != NULL && oidp != NULL);
289 
290 	for (e = TAILQ_FIRST(clist); e != NULL; e = TAILQ_NEXT(e, link)) {
291 		if(e->entry == oidp)
292 			break;
293 	}
294 
295 	return (e);
296 }
297 
298 /*
299  * Delete an entry from the context.
300  * NOTE: this function doesn't free oidp! You have to remove it
301  * with sysctl_remove_oid().
302  */
303 int
304 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
305 {
306 	struct sysctl_ctx_entry *e;
307 
308 	if (clist == NULL || oidp == NULL)
309 		return (EINVAL);
310 
311 	sysctl_ctx_lock(LK_EXCLUSIVE);
312 	e = sysctl_ctx_entry_find_int(clist, oidp);
313 	if (e == NULL) {
314 		sysctl_ctx_unlock();
315 		return (ENOENT);
316 	}
317 	TAILQ_REMOVE(clist, e, link);
318 	kfree(e, M_SYSCTLOID);
319 	sysctl_ctx_unlock();
320 
321 	return(0);
322 }
323 
324 /*
325  * Remove dynamically created sysctl trees.
326  * oidp - top of the tree to be removed
327  * del - if 0 - just deregister, otherwise free up entries as well
328  * recurse - if != 0 traverse the subtree to be deleted
329  */
330 int
331 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
332 {
333 	struct sysctl_oid *p;
334 	int error;
335 
336 	if (oidp == NULL)
337 		return(EINVAL);
338 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
339 		kprintf("can't remove non-dynamic nodes!\n");
340 		return (EINVAL);
341 	}
342 	sysctl_lock(LK_EXCLUSIVE | LK_CANRECURSE);
343 	/*
344 	 * WARNING: normal method to do this should be through
345 	 * sysctl_ctx_free(). Use recursing as the last resort
346 	 * method to purge your sysctl tree of leftovers...
347 	 * However, if some other code still references these nodes,
348 	 * it will panic.
349 	 */
350 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
351 		if (oidp->oid_refcnt == 1) {
352 			SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
353 				if (!recurse) {
354 					sysctl_unlock();
355 					return(ENOTEMPTY);
356 				}
357 				error = sysctl_remove_oid(p, del, recurse);
358 				if (error) {
359 					sysctl_unlock();
360 					return(error);
361 				}
362 			}
363 			if (del)
364 				kfree(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
365 		}
366 	}
367 	if (oidp->oid_refcnt > 1 ) {
368 		oidp->oid_refcnt--;
369 	} else {
370 		if (oidp->oid_refcnt == 0) {
371 			kprintf("Warning: bad oid_refcnt=%u (%s)!\n",
372 			       oidp->oid_refcnt, oidp->oid_name);
373 			sysctl_unlock();
374 			return(EINVAL);
375 		}
376 		sysctl_unregister_oid_int(oidp);
377 		if (del) {
378 			if (oidp->oid_descr)
379 				kfree(__DECONST(char *,oidp->oid_descr),
380 				     M_SYSCTLOID);
381 			kfree(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
382 			kfree(oidp, M_SYSCTLOID);
383 		}
384 	}
385 	sysctl_unlock();
386 	return(0);
387 }
388 
389 /*
390  * Create new sysctls at run time.
391  * clist may point to a valid context initialized with sysctl_ctx_init().
392  */
393 struct sysctl_oid *
394 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
395 	int number, const char *name, int kind, void *arg1, int arg2,
396 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
397 {
398 	struct sysctl_oid *oidp;
399 	ssize_t len;
400 	char *newname;
401 
402 	/* You have to hook up somewhere.. */
403 	if (parent == NULL)
404 		return(NULL);
405 	sysctl_lock(LK_EXCLUSIVE);
406 	/* Check if the node already exists, otherwise create it */
407 	oidp = sysctl_find_oidname(name, parent, 0);
408 	if (oidp != NULL) {
409 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
410 			oidp->oid_refcnt++;
411 			/* Update the context */
412 			if (clist != NULL)
413 				sysctl_ctx_entry_add(clist, oidp);
414 			sysctl_unlock();
415 			return (oidp);
416 		} else {
417 			kprintf("can't re-use a leaf (%s)!\n", name);
418 			sysctl_unlock();
419 			return (NULL);
420 		}
421 	}
422 	oidp = kmalloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK | M_ZERO);
423 	oidp->oid_parent = parent;
424 	SLIST_NEXT(oidp, oid_link) = NULL;
425 	oidp->oid_number = number;
426 	oidp->oid_refcnt = 1;
427 	len = strlen(name);
428 	newname = kmalloc(len + 1, M_SYSCTLOID, M_WAITOK);
429 	bcopy(name, newname, len + 1);
430 	newname[len] = '\0';
431 	oidp->oid_name = newname;
432 	oidp->oid_handler = handler;
433 	oidp->oid_kind = CTLFLAG_DYN | kind;
434 	if ((kind & CTLTYPE) == CTLTYPE_NODE) {
435 		struct sysctl_oid_list *children;
436 
437 		/* Allocate space for children */
438 		children = kmalloc(sizeof(*children), M_SYSCTLOID, M_WAITOK);
439 		SYSCTL_SET_CHILDREN(oidp, children);
440 		SLIST_INIT(children);
441 	} else {
442 		oidp->oid_arg1 = arg1;
443 		oidp->oid_arg2 = arg2;
444 	}
445 	oidp->oid_fmt = fmt;
446 	if (descr) {
447 		int len = strlen(descr) + 1;
448 		oidp->oid_descr = kmalloc(len, M_SYSCTLOID, M_WAITOK);
449 		strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
450 	};
451 	/* Update the context, if used */
452 	if (clist != NULL)
453 		sysctl_ctx_entry_add(clist, oidp);
454 	/* Register this oid */
455 	sysctl_register_oid_int(oidp);
456 	sysctl_unlock();
457 	return (oidp);
458 }
459 
460 /*
461  * Register the kernel's oids on startup.
462  */
463 SET_DECLARE(sysctl_set, struct sysctl_oid);
464 
465 static void
466 sysctl_register_all(void *arg)
467 {
468 	struct sysctl_oid **oidp;
469 
470 	lockinit(&sysctl_lkp, "sysctl", 0, 0);
471 	lockinit(&sysctl_ctx_lkp, "sysctl ctx", 0, 0);
472 	SET_FOREACH(oidp, sysctl_set)
473 		sysctl_register_oid_int(*oidp);
474 }
475 
476 SYSINIT(sysctl, SI_BOOT1_POST, SI_ORDER_ANY, sysctl_register_all, 0);
477 
478 /*
479  * "Staff-functions"
480  *
481  * These functions implement a presently undocumented interface
482  * used by the sysctl program to walk the tree, and get the type
483  * so it can print the value.
484  * This interface is under work and consideration, and should probably
485  * be killed with a big axe by the first person who can find the time.
486  * (be aware though, that the proper interface isn't as obvious as it
487  * may seem, there are various conflicting requirements.
488  *
489  * {0,0}	kprintf the entire MIB-tree.
490  * {0,1,...}	return the name of the "..." OID.
491  * {0,2,...}	return the next OID.
492  * {0,3}	return the OID of the name in "new"
493  * {0,4,...}	return the kind & format info for the "..." OID.
494  */
495 
496 static void
497 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
498 {
499 	int k;
500 	struct sysctl_oid *oidp;
501 
502 	sysctl_lock(LK_SHARED);
503 	SLIST_FOREACH(oidp, l, oid_link) {
504 
505 		for (k=0; k<i; k++)
506 			kprintf(" ");
507 
508 		kprintf("%d %s ", oidp->oid_number, oidp->oid_name);
509 
510 		kprintf("%c%c",
511 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
512 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
513 
514 		if (oidp->oid_handler)
515 			kprintf(" *Handler");
516 
517 		switch (oidp->oid_kind & CTLTYPE) {
518 			case CTLTYPE_NODE:
519 				kprintf(" Node\n");
520 				if (!oidp->oid_handler) {
521 					sysctl_sysctl_debug_dump_node(
522 						oidp->oid_arg1, i+2);
523 				}
524 				break;
525 			case CTLTYPE_INT:    kprintf(" Int\n"); break;
526 			case CTLTYPE_STRING: kprintf(" String\n"); break;
527 			case CTLTYPE_QUAD:   kprintf(" Quad\n"); break;
528 			case CTLTYPE_OPAQUE: kprintf(" Opaque/struct\n"); break;
529 			default:	     kprintf("\n");
530 		}
531 
532 	}
533 	sysctl_unlock();
534 }
535 
536 static int
537 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
538 {
539 	int error;
540 
541 	error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
542 	if (error)
543 		return error;
544 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
545 	return ENOENT;
546 }
547 
548 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
549 	0, 0, sysctl_sysctl_debug, "-", "");
550 
551 static int
552 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
553 {
554 	int *name = (int *) arg1;
555 	u_int namelen = arg2;
556 	int error = 0;
557 	struct sysctl_oid *oid;
558 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
559 	char buf[16];
560 
561 	sysctl_lock(LK_SHARED);
562 	while (namelen) {
563 		if (!lsp) {
564 			ksnprintf(buf, sizeof(buf), "%d", *name);
565 			if (req->oldidx)
566 				error = SYSCTL_OUT(req, ".", 1);
567 			if (!error)
568 				error = SYSCTL_OUT(req, buf, strlen(buf));
569 			if (error) {
570 				sysctl_unlock();
571 				return (error);
572 			}
573 			namelen--;
574 			name++;
575 			continue;
576 		}
577 		lsp2 = NULL;
578 		SLIST_FOREACH(oid, lsp, oid_link) {
579 			if (oid->oid_number != *name)
580 				continue;
581 
582 			if (req->oldidx)
583 				error = SYSCTL_OUT(req, ".", 1);
584 			if (!error)
585 				error = SYSCTL_OUT(req, oid->oid_name,
586 					strlen(oid->oid_name));
587 			if (error) {
588 				sysctl_unlock();
589 				return (error);
590 			}
591 
592 			namelen--;
593 			name++;
594 
595 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
596 				break;
597 
598 			if (oid->oid_handler)
599 				break;
600 
601 			lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
602 			break;
603 		}
604 		lsp = lsp2;
605 	}
606 	sysctl_unlock();
607 	return (SYSCTL_OUT(req, "", 1));
608 }
609 
610 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
611 
612 static int
613 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
614 	int *next, int *len, int level, struct sysctl_oid **oidpp)
615 {
616 	struct sysctl_oid *oidp;
617 
618 	*len = level;
619 	sysctl_lock(LK_SHARED);
620 	SLIST_FOREACH(oidp, lsp, oid_link) {
621 		*next = oidp->oid_number;
622 		*oidpp = oidp;
623 
624 		if (!namelen) {
625 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) {
626 				sysctl_unlock();
627 				return 0;
628 			}
629 			if (oidp->oid_handler) {
630 				/* We really should call the handler here...*/
631 				sysctl_unlock();
632 				return 0;
633 			}
634 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
635 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
636 				len, level+1, oidpp)) {
637 				sysctl_unlock();
638 				return 0;
639 			}
640 			goto emptynode;
641 		}
642 
643 		if (oidp->oid_number < *name)
644 			continue;
645 
646 		if (oidp->oid_number > *name) {
647 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) {
648 				sysctl_unlock();
649 				return 0;
650 			}
651 			if (oidp->oid_handler) {
652 				sysctl_unlock();
653 				return 0;
654 			}
655 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
656 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
657 				next+1, len, level+1, oidpp)) {
658 				sysctl_unlock();
659 				return (0);
660 			}
661 			goto next;
662 		}
663 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
664 			continue;
665 
666 		if (oidp->oid_handler)
667 			continue;
668 
669 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
670 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
671 			len, level+1, oidpp)) {
672 			sysctl_unlock();
673 			return (0);
674 		}
675 	next:
676 		namelen = 1;
677 		*len = level;
678 	emptynode:
679 		*len = level;
680 	}
681 	sysctl_unlock();
682 	return 1;
683 }
684 
685 static int
686 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
687 {
688 	int *name = (int *) arg1;
689 	u_int namelen = arg2;
690 	int i, j, error;
691 	struct sysctl_oid *oid;
692 	struct sysctl_oid_list *lsp = &sysctl__children;
693 	int newoid[CTL_MAXNAME];
694 
695 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
696 	if (i)
697 		return ENOENT;
698 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
699 	return (error);
700 }
701 
702 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
703 
704 static int
705 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
706 {
707 	int i;
708 	struct sysctl_oid *oidp;
709 	struct sysctl_oid_list *lsp = &sysctl__children;
710 	char *p;
711 
712 	if (!*name)
713 		return ENOENT;
714 
715 	p = name + strlen(name) - 1 ;
716 	if (*p == '.')
717 		*p = '\0';
718 
719 	*len = 0;
720 
721 	for (p = name; *p && *p != '.'; p++)
722 		;
723 	i = *p;
724 	if (i == '.')
725 		*p = '\0';
726 
727 	sysctl_lock(LK_SHARED);
728 	oidp = SLIST_FIRST(lsp);
729 
730 	while (oidp && *len < CTL_MAXNAME) {
731 		if (strcmp(name, oidp->oid_name)) {
732 			oidp = SLIST_NEXT(oidp, oid_link);
733 			continue;
734 		}
735 		*oid++ = oidp->oid_number;
736 		(*len)++;
737 
738 		if (!i) {
739 			if (oidpp)
740 				*oidpp = oidp;
741 			sysctl_unlock();
742 			return (0);
743 		}
744 
745 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
746 			break;
747 
748 		if (oidp->oid_handler)
749 			break;
750 
751 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
752 		oidp = SLIST_FIRST(lsp);
753 		name = p+1;
754 		for (p = name; *p && *p != '.'; p++)
755 				;
756 		i = *p;
757 		if (i == '.')
758 			*p = '\0';
759 	}
760 	sysctl_unlock();
761 	return ENOENT;
762 }
763 
764 static int
765 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
766 {
767 	char *p;
768 	int error, oid[CTL_MAXNAME], len;
769 	struct sysctl_oid *op = NULL;
770 
771 	if (!req->newlen)
772 		return ENOENT;
773 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
774 		return (ENAMETOOLONG);
775 
776 	p = kmalloc(req->newlen+1, M_SYSCTL, M_WAITOK);
777 
778 	error = SYSCTL_IN(req, p, req->newlen);
779 	if (error) {
780 		kfree(p, M_SYSCTL);
781 		return (error);
782 	}
783 
784 	p [req->newlen] = '\0';
785 
786 	error = name2oid(p, oid, &len, &op);
787 
788 	kfree(p, M_SYSCTL);
789 
790 	if (error)
791 		return (error);
792 
793 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
794 	return (error);
795 }
796 
797 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
798 	sysctl_sysctl_name2oid, "I", "");
799 
800 static int
801 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
802 {
803 	struct sysctl_oid *oid;
804 	int error;
805 
806 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
807 	if (error)
808 		return (error);
809 
810 	if (!oid->oid_fmt)
811 		return (ENOENT);
812 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
813 	if (error)
814 		return (error);
815 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
816 	return (error);
817 }
818 
819 
820 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
821 
822 static int
823 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
824 {
825 	struct sysctl_oid *oid;
826 	int error;
827 
828 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
829 	if (error)
830 		return (error);
831 
832 	if (!oid->oid_descr)
833 		return (ENOENT);
834 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
835 	return (error);
836 }
837 
838 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
839 
840 /*
841  * Default "handler" functions.
842  */
843 
844 /*
845  * Handle an int, signed or unsigned.
846  * Two cases:
847  *     a variable:  point arg1 at it.
848  *     a constant:  pass it in arg2.
849  */
850 
851 int
852 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
853 {
854 	int error = 0;
855 
856 	if (arg1)
857 		error = SYSCTL_OUT(req, arg1, sizeof(int));
858 	else
859 		error = SYSCTL_OUT(req, &arg2, sizeof(int));
860 
861 	if (error || !req->newptr)
862 		return (error);
863 
864 	if (!arg1)
865 		error = EPERM;
866 	else
867 		error = SYSCTL_IN(req, arg1, sizeof(int));
868 	return (error);
869 }
870 
871 /*
872  * Handle a long, signed or unsigned.  arg1 points to it.
873  */
874 
875 int
876 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
877 {
878 	int error = 0;
879 
880 	if (!arg1)
881 		return (EINVAL);
882 	error = SYSCTL_OUT(req, arg1, sizeof(long));
883 
884 	if (error || !req->newptr)
885 		return (error);
886 
887 	error = SYSCTL_IN(req, arg1, sizeof(long));
888 	return (error);
889 }
890 
891 /*
892  * Handle a quad, signed or unsigned.  arg1 points to it.
893  */
894 
895 int
896 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
897 {
898 	int error = 0;
899 
900 	if (!arg1)
901 		return (EINVAL);
902 	error = SYSCTL_OUT(req, arg1, sizeof(quad_t));
903 
904 	if (error || !req->newptr)
905 		return (error);
906 
907 	error = SYSCTL_IN(req, arg1, sizeof(quad_t));
908 	return (error);
909 }
910 
911 /*
912  * Handle our generic '\0' terminated 'C' string.
913  * Two cases:
914  * 	a variable string:  point arg1 at it, arg2 is max length.
915  * 	a constant string:  point arg1 at it, arg2 is zero.
916  */
917 
918 int
919 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
920 {
921 	int error=0;
922 
923 	error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
924 
925 	if (error || !req->newptr)
926 		return (error);
927 
928 	if ((req->newlen - req->newidx) >= arg2) {
929 		error = EINVAL;
930 	} else {
931 		arg2 = (req->newlen - req->newidx);
932 		error = SYSCTL_IN(req, arg1, arg2);
933 		((char *)arg1)[arg2] = '\0';
934 	}
935 
936 	return (error);
937 }
938 
939 /*
940  * Handle any kind of opaque data.
941  * arg1 points to it, arg2 is the size.
942  */
943 
944 int
945 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
946 {
947 	int error;
948 
949 	error = SYSCTL_OUT(req, arg1, arg2);
950 
951 	if (error || !req->newptr)
952 		return (error);
953 
954 	error = SYSCTL_IN(req, arg1, arg2);
955 
956 	return (error);
957 }
958 
959 /*
960  * Transfer functions to/from kernel space.
961  * XXX: rather untested at this point
962  */
963 static int
964 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
965 {
966 	size_t i = 0;
967 
968 	if (req->oldptr) {
969 		i = l;
970 		if (i > req->oldlen - req->oldidx)
971 			i = req->oldlen - req->oldidx;
972 		if (i > 0)
973 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
974 	}
975 	req->oldidx += l;
976 	if (req->oldptr && i != l)
977 		return (ENOMEM);
978 	return (0);
979 }
980 
981 static int
982 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
983 {
984 
985 	if (!req->newptr)
986 		return 0;
987 	if (req->newlen - req->newidx < l)
988 		return (EINVAL);
989 	bcopy((char *)req->newptr + req->newidx, p, l);
990 	req->newidx += l;
991 	return (0);
992 }
993 
994 int
995 kernel_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval)
996 {
997 	int error = 0;
998 	struct sysctl_req req;
999 
1000 	bzero(&req, sizeof req);
1001 
1002 	req.td = curthread;
1003 
1004 	if (oldlenp) {
1005 		req.oldlen = *oldlenp;
1006 	}
1007 
1008 	if (old) {
1009 		req.oldptr = old;
1010 	}
1011 
1012 	if (new != NULL) {
1013 		req.newlen = newlen;
1014 		req.newptr = new;
1015 	}
1016 
1017 	req.oldfunc = sysctl_old_kernel;
1018 	req.newfunc = sysctl_new_kernel;
1019 #if 0
1020 	req.lock = 1;
1021 #endif
1022 
1023 	sysctl_lock(LK_SHARED);
1024 
1025 	error = sysctl_root(0, name, namelen, &req);
1026 
1027 #if 0
1028 	if (req.lock == 2)
1029 		vsunlock(req.oldptr, req.oldlen);
1030 #endif
1031 
1032 	sysctl_unlock();
1033 
1034 	if (error && error != ENOMEM)
1035 		return (error);
1036 
1037 	if (retval) {
1038 		if (req.oldptr && req.oldidx > req.oldlen)
1039 			*retval = req.oldlen;
1040 		else
1041 			*retval = req.oldidx;
1042 	}
1043 	return (error);
1044 }
1045 
1046 int
1047 kernel_sysctlbyname(char *name, void *old, size_t *oldlenp,
1048     void *new, size_t newlen, size_t *retval)
1049 {
1050         int oid[CTL_MAXNAME];
1051         size_t oidlen, plen;
1052 	int error;
1053 
1054 	oid[0] = 0;		/* sysctl internal magic */
1055 	oid[1] = 3;		/* name2oid */
1056 	oidlen = sizeof(oid);
1057 
1058 	error = kernel_sysctl(oid, 2, oid, &oidlen, (void *)name,
1059 		    strlen(name), &plen);
1060 	if (error)
1061 		return (error);
1062 
1063 	error = kernel_sysctl(oid, plen / sizeof(int), old, oldlenp,
1064 	    new, newlen, retval);
1065 	return (error);
1066 }
1067 
1068 /*
1069  * Transfer function to/from user space.
1070  */
1071 static int
1072 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1073 {
1074 	int error = 0;
1075 	size_t i = 0;
1076 
1077 #if 0
1078 	if (req->lock == 1 && req->oldptr) {
1079 		vslock(req->oldptr, req->oldlen);
1080 		req->lock = 2;
1081 	}
1082 #endif
1083 	if (req->oldptr) {
1084 		i = l;
1085 		if (i > req->oldlen - req->oldidx)
1086 			i = req->oldlen - req->oldidx;
1087 		if (i > 0)
1088 			error = copyout(p, (char *)req->oldptr + req->oldidx,
1089 					i);
1090 	}
1091 	req->oldidx += l;
1092 	if (error)
1093 		return (error);
1094 	if (req->oldptr && i < l)
1095 		return (ENOMEM);
1096 	return (0);
1097 }
1098 
1099 static int
1100 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1101 {
1102 	int error;
1103 
1104 	if (!req->newptr)
1105 		return 0;
1106 	if (req->newlen - req->newidx < l)
1107 		return (EINVAL);
1108 	error = copyin((char *)req->newptr + req->newidx, p, l);
1109 	req->newidx += l;
1110 	return (error);
1111 }
1112 
1113 int
1114 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1115     int *nindx, struct sysctl_req *req)
1116 {
1117 	struct sysctl_oid *oid;
1118 	int indx;
1119 
1120 	sysctl_lock(LK_SHARED);
1121 	oid = SLIST_FIRST(&sysctl__children);
1122 	indx = 0;
1123 	while (oid && indx < CTL_MAXNAME) {
1124 		if (oid->oid_number == name[indx]) {
1125 			indx++;
1126 			if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1127 				if (oid->oid_handler != NULL ||
1128 				    indx == namelen) {
1129 					*noid = oid;
1130 					if (nindx != NULL)
1131 						*nindx = indx;
1132 					sysctl_unlock();
1133 					return (0);
1134 				}
1135 				oid = SLIST_FIRST(
1136 				    (struct sysctl_oid_list *)oid->oid_arg1);
1137 			} else if (indx == namelen) {
1138 				*noid = oid;
1139 				if (nindx != NULL)
1140 					*nindx = indx;
1141 				sysctl_unlock();
1142 				return (0);
1143 			} else {
1144 				sysctl_unlock();
1145 				return (ENOTDIR);
1146 			}
1147 		} else {
1148 			oid = SLIST_NEXT(oid, oid_link);
1149 		}
1150 	}
1151 	sysctl_unlock();
1152 	return (ENOENT);
1153 }
1154 
1155 /*
1156  * Traverse our tree, and find the right node, execute whatever it points
1157  * to, and return the resulting error code.
1158  */
1159 
1160 int
1161 sysctl_root(SYSCTL_HANDLER_ARGS)
1162 {
1163 	struct thread *td = req->td;
1164 	struct proc *p = td ? td->td_proc : NULL;
1165 	struct sysctl_oid *oid;
1166 	int error, indx;
1167 
1168 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1169 	if (error)
1170 		return (error);
1171 
1172 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1173 		/*
1174 		 * You can't call a sysctl when it's a node, but has
1175 		 * no handler.  Inform the user that it's a node.
1176 		 * The indx may or may not be the same as namelen.
1177 		 */
1178 		if (oid->oid_handler == NULL)
1179 			return (EISDIR);
1180 	}
1181 
1182 	/* If writing isn't allowed */
1183 	if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) ||
1184 	    ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0)))
1185 		return (EPERM);
1186 
1187 	/* Most likely only root can write */
1188 	if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p &&
1189 	    (error = priv_check_cred(td->td_ucred,
1190 	     (oid->oid_kind & CTLFLAG_PRISON) ? PRIV_SYSCTL_WRITEJAIL :
1191 	                                        PRIV_SYSCTL_WRITE, 0)))
1192 		return (error);
1193 
1194 	if (!oid->oid_handler)
1195 		return EINVAL;
1196 
1197 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE)
1198 		error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx,
1199 		    req);
1200 	else
1201 		error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2,
1202 		    req);
1203 	return (error);
1204 }
1205 
1206 /*
1207  * MPALMOSTSAFE
1208  */
1209 int
1210 sys___sysctl(struct sysctl_args *uap)
1211 {
1212 	int error, i, name[CTL_MAXNAME];
1213 	size_t j;
1214 
1215 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1216 		return (EINVAL);
1217 
1218 	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1219  	if (error)
1220 		return (error);
1221 
1222 	get_mplock();
1223 	error = userland_sysctl(name, uap->namelen,
1224 		uap->old, uap->oldlenp, 0,
1225 		uap->new, uap->newlen, &j);
1226 	rel_mplock();
1227 	if (error && error != ENOMEM)
1228 		return (error);
1229 	if (uap->oldlenp) {
1230 		i = copyout(&j, uap->oldlenp, sizeof(j));
1231 		if (i)
1232 			return (i);
1233 	}
1234 	return (error);
1235 }
1236 
1237 /*
1238  * This is used from various compatibility syscalls too.  That's why name
1239  * must be in kernel space.
1240  */
1241 int
1242 userland_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval)
1243 {
1244 	int error = 0;
1245 	struct sysctl_req req, req2;
1246 
1247 	bzero(&req, sizeof req);
1248 
1249 	if (oldlenp) {
1250 		if (inkernel) {
1251 			req.oldlen = *oldlenp;
1252 		} else {
1253 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1254 			if (error)
1255 				return (error);
1256 		}
1257 	}
1258 
1259 	if (old) {
1260 		if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1261 			return (EFAULT);
1262 		req.oldptr= old;
1263 	}
1264 
1265 	if (new != NULL) {
1266 		if (!useracc(new, newlen, VM_PROT_READ))
1267 			return (EFAULT);
1268 		req.newlen = newlen;
1269 		req.newptr = new;
1270 	}
1271 
1272 	req.oldfunc = sysctl_old_user;
1273 	req.newfunc = sysctl_new_user;
1274 #if 0
1275 	req.lock = 1;
1276 #endif
1277 	req.td = curthread;
1278 
1279 	sysctl_lock(LK_SHARED);
1280 
1281 	do {
1282 	    req2 = req;
1283 	    error = sysctl_root(0, name, namelen, &req2);
1284 	} while (error == EAGAIN);
1285 
1286 	req = req2;
1287 #if 0
1288 	if (req.lock == 2)
1289 		vsunlock(req.oldptr, req.oldlen);
1290 #endif
1291 
1292 	sysctl_unlock();
1293 
1294 	if (error && error != ENOMEM)
1295 		return (error);
1296 
1297 	if (retval) {
1298 		if (req.oldptr && req.oldidx > req.oldlen)
1299 			*retval = req.oldlen;
1300 		else
1301 			*retval = req.oldidx;
1302 	}
1303 	return (error);
1304 }
1305 
1306 static void
1307 sysctl_lock(int flag)
1308 {
1309 	lockmgr(&sysctl_lkp, flag);
1310 }
1311 
1312 static void
1313 sysctl_unlock(void)
1314 {
1315 	lockmgr(&sysctl_lkp, LK_RELEASE);
1316 }
1317 
1318 static void
1319 sysctl_ctx_lock(int flag)
1320 {
1321 	lockmgr(&sysctl_ctx_lkp, flag);
1322 }
1323 
1324 static void
1325 sysctl_ctx_unlock(void)
1326 {
1327 	lockmgr(&sysctl_ctx_lkp, LK_RELEASE);
1328 }
1329 
1330 int
1331 sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high)
1332 {
1333 	int error, value;
1334 
1335 	value = *(int *)arg1;
1336 	error = sysctl_handle_int(oidp, &value, 0, req);
1337 	if (error || !req->newptr)
1338 		return (error);
1339 	if (value < low || value > high)
1340 		return (EINVAL);
1341 	*(int *)arg1 = value;
1342 	return (0);
1343 }
1344