xref: /freebsd/sys/kern/kern_sysctl.c (revision c0020399a650364d0134f79f3fa319f84064372d)
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  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)kern_sysctl.c	8.4 (Berkeley) 4/14/94
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_compat.h"
42 #include "opt_ktrace.h"
43 #include "opt_mac.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
48 #include <sys/sysctl.h>
49 #include <sys/malloc.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/lock.h>
53 #include <sys/mutex.h>
54 #include <sys/sx.h>
55 #include <sys/sysproto.h>
56 #include <sys/uio.h>
57 #include <sys/vimage.h>
58 #ifdef KTRACE
59 #include <sys/ktrace.h>
60 #endif
61 
62 #include <security/mac/mac_framework.h>
63 
64 #include <vm/vm.h>
65 #include <vm/vm_extern.h>
66 
67 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
68 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
69 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
70 
71 /*
72  * The sysctllock protects the MIB tree.  It also protects sysctl
73  * contexts used with dynamic sysctls.  The sysctl_register_oid() and
74  * sysctl_unregister_oid() routines require the sysctllock to already
75  * be held, so the sysctl_lock() and sysctl_unlock() routines are
76  * provided for the few places in the kernel which need to use that
77  * API rather than using the dynamic API.  Use of the dynamic API is
78  * strongly encouraged for most code.
79  *
80  * This lock is also used to serialize userland sysctl requests.  Some
81  * sysctls wire user memory, and serializing the requests limits the
82  * amount of wired user memory in use.
83  */
84 static struct sx sysctllock;
85 
86 #define	SYSCTL_SLOCK()		sx_slock(&sysctllock)
87 #define	SYSCTL_SUNLOCK()	sx_sunlock(&sysctllock)
88 #define	SYSCTL_XLOCK()		sx_xlock(&sysctllock)
89 #define	SYSCTL_XUNLOCK()	sx_xunlock(&sysctllock)
90 #define	SYSCTL_ASSERT_XLOCKED()	sx_assert(&sysctllock, SA_XLOCKED)
91 #define	SYSCTL_ASSERT_LOCKED()	sx_assert(&sysctllock, SA_LOCKED)
92 #define	SYSCTL_INIT()		sx_init(&sysctllock, "sysctl lock")
93 
94 static int sysctl_root(SYSCTL_HANDLER_ARGS);
95 
96 struct sysctl_oid_list sysctl__children; /* root list */
97 
98 static int	sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
99 		    int recurse);
100 
101 static struct sysctl_oid *
102 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
103 {
104 	struct sysctl_oid *oidp;
105 
106 	SYSCTL_ASSERT_LOCKED();
107 	SLIST_FOREACH(oidp, list, oid_link) {
108 		if (strcmp(oidp->oid_name, name) == 0) {
109 			return (oidp);
110 		}
111 	}
112 	return (NULL);
113 }
114 
115 /*
116  * Initialization of the MIB tree.
117  *
118  * Order by number in each list.
119  */
120 void
121 sysctl_lock(void)
122 {
123 
124 	SYSCTL_XLOCK();
125 }
126 
127 void
128 sysctl_unlock(void)
129 {
130 
131 	SYSCTL_XUNLOCK();
132 }
133 
134 void
135 sysctl_register_oid(struct sysctl_oid *oidp)
136 {
137 	struct sysctl_oid_list *parent = oidp->oid_parent;
138 	struct sysctl_oid *p;
139 	struct sysctl_oid *q;
140 
141 	/*
142 	 * First check if another oid with the same name already
143 	 * exists in the parent's list.
144 	 */
145 	SYSCTL_ASSERT_XLOCKED();
146 	p = sysctl_find_oidname(oidp->oid_name, parent);
147 	if (p != NULL) {
148 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
149 			p->oid_refcnt++;
150 			return;
151 		} else {
152 			printf("can't re-use a leaf (%s)!\n", p->oid_name);
153 			return;
154 		}
155 	}
156 	/*
157 	 * If this oid has a number OID_AUTO, give it a number which
158 	 * is greater than any current oid.
159 	 * NOTE: DO NOT change the starting value here, change it in
160 	 * <sys/sysctl.h>, and make sure it is at least 256 to
161 	 * accomodate e.g. net.inet.raw as a static sysctl node.
162 	 */
163 	if (oidp->oid_number == OID_AUTO) {
164 		static int newoid = CTL_AUTO_START;
165 
166 		oidp->oid_number = newoid++;
167 		if (newoid == 0x7fffffff)
168 			panic("out of oids");
169 	}
170 #if 0
171 	else if (oidp->oid_number >= CTL_AUTO_START) {
172 		/* do not panic; this happens when unregistering sysctl sets */
173 		printf("static sysctl oid too high: %d", oidp->oid_number);
174 	}
175 #endif
176 
177 	/*
178 	 * Insert the oid into the parent's list in order.
179 	 */
180 	q = NULL;
181 	SLIST_FOREACH(p, parent, oid_link) {
182 		if (oidp->oid_number < p->oid_number)
183 			break;
184 		q = p;
185 	}
186 	if (q)
187 		SLIST_INSERT_AFTER(q, oidp, oid_link);
188 	else
189 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
190 }
191 
192 void
193 sysctl_unregister_oid(struct sysctl_oid *oidp)
194 {
195 	struct sysctl_oid *p;
196 	int error;
197 
198 	SYSCTL_ASSERT_XLOCKED();
199 	error = ENOENT;
200 	if (oidp->oid_number == OID_AUTO) {
201 		error = EINVAL;
202 	} else {
203 		SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
204 			if (p == oidp) {
205 				SLIST_REMOVE(oidp->oid_parent, oidp,
206 				    sysctl_oid, oid_link);
207 				error = 0;
208 				break;
209 			}
210 		}
211 	}
212 
213 	/*
214 	 * This can happen when a module fails to register and is
215 	 * being unloaded afterwards.  It should not be a panic()
216 	 * for normal use.
217 	 */
218 	if (error)
219 		printf("%s: failed to unregister sysctl\n", __func__);
220 }
221 
222 /* Initialize a new context to keep track of dynamically added sysctls. */
223 int
224 sysctl_ctx_init(struct sysctl_ctx_list *c)
225 {
226 
227 	if (c == NULL) {
228 		return (EINVAL);
229 	}
230 
231 	/*
232 	 * No locking here, the caller is responsible for not adding
233 	 * new nodes to a context until after this function has
234 	 * returned.
235 	 */
236 	TAILQ_INIT(c);
237 	return (0);
238 }
239 
240 /* Free the context, and destroy all dynamic oids registered in this context */
241 int
242 sysctl_ctx_free(struct sysctl_ctx_list *clist)
243 {
244 	struct sysctl_ctx_entry *e, *e1;
245 	int error;
246 
247 	error = 0;
248 	/*
249 	 * First perform a "dry run" to check if it's ok to remove oids.
250 	 * XXX FIXME
251 	 * XXX This algorithm is a hack. But I don't know any
252 	 * XXX better solution for now...
253 	 */
254 	SYSCTL_XLOCK();
255 	TAILQ_FOREACH(e, clist, link) {
256 		error = sysctl_remove_oid_locked(e->entry, 0, 0);
257 		if (error)
258 			break;
259 	}
260 	/*
261 	 * Restore deregistered entries, either from the end,
262 	 * or from the place where error occured.
263 	 * e contains the entry that was not unregistered
264 	 */
265 	if (error)
266 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
267 	else
268 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
269 	while (e1 != NULL) {
270 		sysctl_register_oid(e1->entry);
271 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
272 	}
273 	if (error) {
274 		SYSCTL_XUNLOCK();
275 		return(EBUSY);
276 	}
277 	/* Now really delete the entries */
278 	e = TAILQ_FIRST(clist);
279 	while (e != NULL) {
280 		e1 = TAILQ_NEXT(e, link);
281 		error = sysctl_remove_oid_locked(e->entry, 1, 0);
282 		if (error)
283 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
284 			    e->entry->oid_name);
285 		free(e, M_SYSCTLOID);
286 		e = e1;
287 	}
288 	SYSCTL_XUNLOCK();
289 	return (error);
290 }
291 
292 /* Add an entry to the context */
293 struct sysctl_ctx_entry *
294 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
295 {
296 	struct sysctl_ctx_entry *e;
297 
298 	SYSCTL_ASSERT_XLOCKED();
299 	if (clist == NULL || oidp == NULL)
300 		return(NULL);
301 	e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
302 	e->entry = oidp;
303 	TAILQ_INSERT_HEAD(clist, e, link);
304 	return (e);
305 }
306 
307 /* Find an entry in the context */
308 struct sysctl_ctx_entry *
309 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
310 {
311 	struct sysctl_ctx_entry *e;
312 
313 	SYSCTL_ASSERT_LOCKED();
314 	if (clist == NULL || oidp == NULL)
315 		return(NULL);
316 	TAILQ_FOREACH(e, clist, link) {
317 		if(e->entry == oidp)
318 			return(e);
319 	}
320 	return (e);
321 }
322 
323 /*
324  * Delete an entry from the context.
325  * NOTE: this function doesn't free oidp! You have to remove it
326  * with sysctl_remove_oid().
327  */
328 int
329 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
330 {
331 	struct sysctl_ctx_entry *e;
332 
333 	if (clist == NULL || oidp == NULL)
334 		return (EINVAL);
335 	SYSCTL_XLOCK();
336 	e = sysctl_ctx_entry_find(clist, oidp);
337 	if (e != NULL) {
338 		TAILQ_REMOVE(clist, e, link);
339 		SYSCTL_XUNLOCK();
340 		free(e, M_SYSCTLOID);
341 		return (0);
342 	} else {
343 		SYSCTL_XUNLOCK();
344 		return (ENOENT);
345 	}
346 }
347 
348 /*
349  * Remove dynamically created sysctl trees.
350  * oidp - top of the tree to be removed
351  * del - if 0 - just deregister, otherwise free up entries as well
352  * recurse - if != 0 traverse the subtree to be deleted
353  */
354 int
355 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
356 {
357 	int error;
358 
359 	SYSCTL_XLOCK();
360 	error = sysctl_remove_oid_locked(oidp, del, recurse);
361 	SYSCTL_XUNLOCK();
362 	return (error);
363 }
364 
365 static int
366 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
367 {
368 	struct sysctl_oid *p;
369 	int error;
370 
371 	SYSCTL_ASSERT_XLOCKED();
372 	if (oidp == NULL)
373 		return(EINVAL);
374 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
375 		printf("can't remove non-dynamic nodes!\n");
376 		return (EINVAL);
377 	}
378 	/*
379 	 * WARNING: normal method to do this should be through
380 	 * sysctl_ctx_free(). Use recursing as the last resort
381 	 * method to purge your sysctl tree of leftovers...
382 	 * However, if some other code still references these nodes,
383 	 * it will panic.
384 	 */
385 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
386 		if (oidp->oid_refcnt == 1) {
387 			SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
388 				if (!recurse)
389 					return (ENOTEMPTY);
390 				error = sysctl_remove_oid_locked(p, del,
391 				    recurse);
392 				if (error)
393 					return (error);
394 			}
395 			if (del)
396 				free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
397 		}
398 	}
399 	if (oidp->oid_refcnt > 1 ) {
400 		oidp->oid_refcnt--;
401 	} else {
402 		if (oidp->oid_refcnt == 0) {
403 			printf("Warning: bad oid_refcnt=%u (%s)!\n",
404 				oidp->oid_refcnt, oidp->oid_name);
405 			return (EINVAL);
406 		}
407 		sysctl_unregister_oid(oidp);
408 		if (del) {
409 			if (oidp->oid_descr)
410 				free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
411 			free((void *)(uintptr_t)(const void *)oidp->oid_name,
412 			     M_SYSCTLOID);
413 			free(oidp, M_SYSCTLOID);
414 		}
415 	}
416 	return (0);
417 }
418 
419 /*
420  * Create new sysctls at run time.
421  * clist may point to a valid context initialized with sysctl_ctx_init().
422  */
423 struct sysctl_oid *
424 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
425 	int number, const char *name, int kind, void *arg1, int arg2,
426 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
427 {
428 	struct sysctl_oid *oidp;
429 	ssize_t len;
430 	char *newname;
431 
432 	/* You have to hook up somewhere.. */
433 	if (parent == NULL)
434 		return(NULL);
435 	/* Check if the node already exists, otherwise create it */
436 	SYSCTL_XLOCK();
437 	oidp = sysctl_find_oidname(name, parent);
438 	if (oidp != NULL) {
439 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
440 			oidp->oid_refcnt++;
441 			/* Update the context */
442 			if (clist != NULL)
443 				sysctl_ctx_entry_add(clist, oidp);
444 			SYSCTL_XUNLOCK();
445 			return (oidp);
446 		} else {
447 			SYSCTL_XUNLOCK();
448 			printf("can't re-use a leaf (%s)!\n", name);
449 			return (NULL);
450 		}
451 	}
452 	oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
453 	oidp->oid_parent = parent;
454 	SLIST_NEXT(oidp, oid_link) = NULL;
455 	oidp->oid_number = number;
456 	oidp->oid_refcnt = 1;
457 	len = strlen(name);
458 	newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
459 	bcopy(name, newname, len + 1);
460 	newname[len] = '\0';
461 	oidp->oid_name = newname;
462 	oidp->oid_handler = handler;
463 	oidp->oid_kind = CTLFLAG_DYN | kind;
464 	if ((kind & CTLTYPE) == CTLTYPE_NODE) {
465 		/* Allocate space for children */
466 		SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
467 		    M_SYSCTLOID, M_WAITOK));
468 		SLIST_INIT(SYSCTL_CHILDREN(oidp));
469 	} else {
470 		oidp->oid_arg1 = arg1;
471 		oidp->oid_arg2 = arg2;
472 	}
473 	oidp->oid_fmt = fmt;
474 	if (descr) {
475 		int len = strlen(descr) + 1;
476 		oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
477 		if (oidp->oid_descr)
478 			strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
479 	}
480 	/* Update the context, if used */
481 	if (clist != NULL)
482 		sysctl_ctx_entry_add(clist, oidp);
483 	/* Register this oid */
484 	sysctl_register_oid(oidp);
485 	SYSCTL_XUNLOCK();
486 	return (oidp);
487 }
488 
489 /*
490  * Rename an existing oid.
491  */
492 void
493 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
494 {
495 	ssize_t len;
496 	char *newname;
497 	void *oldname;
498 
499 	len = strlen(name);
500 	newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
501 	bcopy(name, newname, len + 1);
502 	newname[len] = '\0';
503 	SYSCTL_XLOCK();
504 	oldname = (void *)(uintptr_t)(const void *)oidp->oid_name;
505 	oidp->oid_name = newname;
506 	SYSCTL_XUNLOCK();
507 	free(oldname, M_SYSCTLOID);
508 }
509 
510 /*
511  * Reparent an existing oid.
512  */
513 int
514 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
515 {
516 	struct sysctl_oid *oidp;
517 
518 	SYSCTL_XLOCK();
519 	if (oid->oid_parent == parent) {
520 		SYSCTL_XUNLOCK();
521 		return (0);
522 	}
523 	oidp = sysctl_find_oidname(oid->oid_name, parent);
524 	if (oidp != NULL) {
525 		SYSCTL_XUNLOCK();
526 		return (EEXIST);
527 	}
528 	sysctl_unregister_oid(oid);
529 	oid->oid_parent = parent;
530 	oid->oid_number = OID_AUTO;
531 	sysctl_register_oid(oid);
532 	SYSCTL_XUNLOCK();
533 	return (0);
534 }
535 
536 /*
537  * Register the kernel's oids on startup.
538  */
539 SET_DECLARE(sysctl_set, struct sysctl_oid);
540 
541 static void
542 sysctl_register_all(void *arg)
543 {
544 	struct sysctl_oid **oidp;
545 
546 	SYSCTL_INIT();
547 	SYSCTL_XLOCK();
548 	SET_FOREACH(oidp, sysctl_set)
549 		sysctl_register_oid(*oidp);
550 	SYSCTL_XUNLOCK();
551 }
552 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
553 
554 /*
555  * "Staff-functions"
556  *
557  * These functions implement a presently undocumented interface
558  * used by the sysctl program to walk the tree, and get the type
559  * so it can print the value.
560  * This interface is under work and consideration, and should probably
561  * be killed with a big axe by the first person who can find the time.
562  * (be aware though, that the proper interface isn't as obvious as it
563  * may seem, there are various conflicting requirements.
564  *
565  * {0,0}	printf the entire MIB-tree.
566  * {0,1,...}	return the name of the "..." OID.
567  * {0,2,...}	return the next OID.
568  * {0,3}	return the OID of the name in "new"
569  * {0,4,...}	return the kind & format info for the "..." OID.
570  * {0,5,...}	return the description the "..." OID.
571  */
572 
573 #ifdef SYSCTL_DEBUG
574 static void
575 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
576 {
577 	int k;
578 	struct sysctl_oid *oidp;
579 
580 	SYSCTL_ASSERT_LOCKED();
581 	SLIST_FOREACH(oidp, l, oid_link) {
582 
583 		for (k=0; k<i; k++)
584 			printf(" ");
585 
586 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
587 
588 		printf("%c%c",
589 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
590 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
591 
592 		if (oidp->oid_handler)
593 			printf(" *Handler");
594 
595 		switch (oidp->oid_kind & CTLTYPE) {
596 			case CTLTYPE_NODE:
597 				printf(" Node\n");
598 				if (!oidp->oid_handler) {
599 					sysctl_sysctl_debug_dump_node(
600 						oidp->oid_arg1, i+2);
601 				}
602 				break;
603 			case CTLTYPE_INT:    printf(" Int\n"); break;
604 			case CTLTYPE_STRING: printf(" String\n"); break;
605 			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
606 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
607 			default:	     printf("\n");
608 		}
609 
610 	}
611 }
612 
613 static int
614 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
615 {
616 	int error;
617 
618 	error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
619 	if (error)
620 		return (error);
621 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
622 	return (ENOENT);
623 }
624 
625 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
626 	0, 0, sysctl_sysctl_debug, "-", "");
627 #endif
628 
629 static int
630 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
631 {
632 	int *name = (int *) arg1;
633 	u_int namelen = arg2;
634 	int error = 0;
635 	struct sysctl_oid *oid;
636 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
637 	char buf[10];
638 
639 	SYSCTL_ASSERT_LOCKED();
640 	while (namelen) {
641 		if (!lsp) {
642 			snprintf(buf,sizeof(buf),"%d",*name);
643 			if (req->oldidx)
644 				error = SYSCTL_OUT(req, ".", 1);
645 			if (!error)
646 				error = SYSCTL_OUT(req, buf, strlen(buf));
647 			if (error)
648 				return (error);
649 			namelen--;
650 			name++;
651 			continue;
652 		}
653 		lsp2 = 0;
654 		SLIST_FOREACH(oid, lsp, oid_link) {
655 			if (oid->oid_number != *name)
656 				continue;
657 
658 			if (req->oldidx)
659 				error = SYSCTL_OUT(req, ".", 1);
660 			if (!error)
661 				error = SYSCTL_OUT(req, oid->oid_name,
662 					strlen(oid->oid_name));
663 			if (error)
664 				return (error);
665 
666 			namelen--;
667 			name++;
668 
669 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
670 				break;
671 
672 			if (oid->oid_handler)
673 				break;
674 
675 			lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
676 			break;
677 		}
678 		lsp = lsp2;
679 	}
680 	return (SYSCTL_OUT(req, "", 1));
681 }
682 
683 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
684 
685 static int
686 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
687 	int *next, int *len, int level, struct sysctl_oid **oidpp)
688 {
689 	struct sysctl_oid *oidp;
690 
691 	SYSCTL_ASSERT_LOCKED();
692 	*len = level;
693 	SLIST_FOREACH(oidp, lsp, oid_link) {
694 		*next = oidp->oid_number;
695 		*oidpp = oidp;
696 
697 		if (oidp->oid_kind & CTLFLAG_SKIP)
698 			continue;
699 
700 		if (!namelen) {
701 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
702 				return (0);
703 			if (oidp->oid_handler)
704 				/* We really should call the handler here...*/
705 				return (0);
706 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
707 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
708 				len, level+1, oidpp))
709 				return (0);
710 			goto emptynode;
711 		}
712 
713 		if (oidp->oid_number < *name)
714 			continue;
715 
716 		if (oidp->oid_number > *name) {
717 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
718 				return (0);
719 			if (oidp->oid_handler)
720 				return (0);
721 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
722 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
723 				next+1, len, level+1, oidpp))
724 				return (0);
725 			goto next;
726 		}
727 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
728 			continue;
729 
730 		if (oidp->oid_handler)
731 			continue;
732 
733 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
734 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
735 			len, level+1, oidpp))
736 			return (0);
737 	next:
738 		namelen = 1;
739 	emptynode:
740 		*len = level;
741 	}
742 	return (1);
743 }
744 
745 static int
746 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
747 {
748 	int *name = (int *) arg1;
749 	u_int namelen = arg2;
750 	int i, j, error;
751 	struct sysctl_oid *oid;
752 	struct sysctl_oid_list *lsp = &sysctl__children;
753 	int newoid[CTL_MAXNAME];
754 
755 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
756 	if (i)
757 		return (ENOENT);
758 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
759 	return (error);
760 }
761 
762 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
763 
764 static int
765 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
766 {
767 	int i;
768 	struct sysctl_oid *oidp;
769 	struct sysctl_oid_list *lsp = &sysctl__children;
770 	char *p;
771 
772 	SYSCTL_ASSERT_LOCKED();
773 
774 	if (!*name)
775 		return (ENOENT);
776 
777 	p = name + strlen(name) - 1 ;
778 	if (*p == '.')
779 		*p = '\0';
780 
781 	*len = 0;
782 
783 	for (p = name; *p && *p != '.'; p++)
784 		;
785 	i = *p;
786 	if (i == '.')
787 		*p = '\0';
788 
789 	oidp = SLIST_FIRST(lsp);
790 
791 	while (oidp && *len < CTL_MAXNAME) {
792 		if (strcmp(name, oidp->oid_name)) {
793 			oidp = SLIST_NEXT(oidp, oid_link);
794 			continue;
795 		}
796 		*oid++ = oidp->oid_number;
797 		(*len)++;
798 
799 		if (!i) {
800 			if (oidpp)
801 				*oidpp = oidp;
802 			return (0);
803 		}
804 
805 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
806 			break;
807 
808 		if (oidp->oid_handler)
809 			break;
810 
811 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
812 		oidp = SLIST_FIRST(lsp);
813 		name = p+1;
814 		for (p = name; *p && *p != '.'; p++)
815 				;
816 		i = *p;
817 		if (i == '.')
818 			*p = '\0';
819 	}
820 	return (ENOENT);
821 }
822 
823 static int
824 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
825 {
826 	char *p;
827 	int error, oid[CTL_MAXNAME], len;
828 	struct sysctl_oid *op = 0;
829 
830 	SYSCTL_ASSERT_LOCKED();
831 
832 	if (!req->newlen)
833 		return (ENOENT);
834 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
835 		return (ENAMETOOLONG);
836 
837 	p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
838 
839 	error = SYSCTL_IN(req, p, req->newlen);
840 	if (error) {
841 		free(p, M_SYSCTL);
842 		return (error);
843 	}
844 
845 	p [req->newlen] = '\0';
846 
847 	error = name2oid(p, oid, &len, &op);
848 
849 	free(p, M_SYSCTL);
850 
851 	if (error)
852 		return (error);
853 
854 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
855 	return (error);
856 }
857 
858 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_MPSAFE,
859     0, 0, sysctl_sysctl_name2oid, "I", "");
860 
861 static int
862 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
863 {
864 	struct sysctl_oid *oid;
865 	int error;
866 
867 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
868 	if (error)
869 		return (error);
870 
871 	if (!oid->oid_fmt)
872 		return (ENOENT);
873 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
874 	if (error)
875 		return (error);
876 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
877 	return (error);
878 }
879 
880 
881 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE,
882     sysctl_sysctl_oidfmt, "");
883 
884 static int
885 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
886 {
887 	struct sysctl_oid *oid;
888 	int error;
889 
890 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
891 	if (error)
892 		return (error);
893 
894 	if (!oid->oid_descr)
895 		return (ENOENT);
896 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
897 	return (error);
898 }
899 
900 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
901 
902 /*
903  * Default "handler" functions.
904  */
905 
906 /*
907  * Handle an int, signed or unsigned.
908  * Two cases:
909  *     a variable:  point arg1 at it.
910  *     a constant:  pass it in arg2.
911  */
912 
913 int
914 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
915 {
916 	int tmpout, error = 0;
917 
918 	/*
919 	 * Attempt to get a coherent snapshot by making a copy of the data.
920 	 */
921 	if (arg1)
922 		tmpout = *(int *)arg1;
923 	else
924 		tmpout = arg2;
925 	error = SYSCTL_OUT(req, &tmpout, sizeof(int));
926 
927 	if (error || !req->newptr)
928 		return (error);
929 
930 	if (!arg1)
931 		error = EPERM;
932 	else
933 		error = SYSCTL_IN(req, arg1, sizeof(int));
934 	return (error);
935 }
936 
937 
938 /*
939  * Based on on sysctl_handle_int() convert milliseconds into ticks.
940  */
941 
942 int
943 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
944 {
945 	int error, s, tt;
946 
947 	tt = *(int *)oidp->oid_arg1;
948 	s = (int)((int64_t)tt * 1000 / hz);
949 
950 	error = sysctl_handle_int(oidp, &s, 0, req);
951 	if (error || !req->newptr)
952 		return (error);
953 
954 	tt = (int)((int64_t)s * hz / 1000);
955 	if (tt < 1)
956 		return (EINVAL);
957 
958 	*(int *)oidp->oid_arg1 = tt;
959 	return (0);
960 }
961 
962 
963 /*
964  * Handle a long, signed or unsigned.  arg1 points to it.
965  */
966 
967 int
968 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
969 {
970 	int error = 0;
971 	long tmplong;
972 #ifdef SCTL_MASK32
973 	int tmpint;
974 #endif
975 
976 	/*
977 	 * Attempt to get a coherent snapshot by making a copy of the data.
978 	 */
979 	if (!arg1)
980 		return (EINVAL);
981 	tmplong = *(long *)arg1;
982 #ifdef SCTL_MASK32
983 	if (req->flags & SCTL_MASK32) {
984 		tmpint = tmplong;
985 		error = SYSCTL_OUT(req, &tmpint, sizeof(int));
986 	} else
987 #endif
988 		error = SYSCTL_OUT(req, &tmplong, sizeof(long));
989 
990 	if (error || !req->newptr)
991 		return (error);
992 
993 #ifdef SCTL_MASK32
994 	if (req->flags & SCTL_MASK32) {
995 		error = SYSCTL_IN(req, &tmpint, sizeof(int));
996 		*(long *)arg1 = (long)tmpint;
997 	} else
998 #endif
999 		error = SYSCTL_IN(req, arg1, sizeof(long));
1000 	return (error);
1001 }
1002 
1003 /*
1004  * Handle a 64 bit int, signed or unsigned.  arg1 points to it.
1005  */
1006 
1007 int
1008 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
1009 {
1010 	int error = 0;
1011 	uint64_t tmpout;
1012 
1013 	/*
1014 	 * Attempt to get a coherent snapshot by making a copy of the data.
1015 	 */
1016 	if (!arg1)
1017 		return (EINVAL);
1018 	tmpout = *(uint64_t *)arg1;
1019 	error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1020 
1021 	if (error || !req->newptr)
1022 		return (error);
1023 
1024 	error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1025 	return (error);
1026 }
1027 
1028 /*
1029  * Handle our generic '\0' terminated 'C' string.
1030  * Two cases:
1031  * 	a variable string:  point arg1 at it, arg2 is max length.
1032  * 	a constant string:  point arg1 at it, arg2 is zero.
1033  */
1034 
1035 int
1036 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1037 {
1038 	int error=0;
1039 	char *tmparg;
1040 	size_t outlen;
1041 
1042 	/*
1043 	 * Attempt to get a coherent snapshot by copying to a
1044 	 * temporary kernel buffer.
1045 	 */
1046 retry:
1047 	outlen = strlen((char *)arg1)+1;
1048 	tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
1049 
1050 	if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
1051 		free(tmparg, M_SYSCTLTMP);
1052 		goto retry;
1053 	}
1054 
1055 	error = SYSCTL_OUT(req, tmparg, outlen);
1056 	free(tmparg, M_SYSCTLTMP);
1057 
1058 	if (error || !req->newptr)
1059 		return (error);
1060 
1061 	if ((req->newlen - req->newidx) >= arg2) {
1062 		error = EINVAL;
1063 	} else {
1064 		arg2 = (req->newlen - req->newidx);
1065 		error = SYSCTL_IN(req, arg1, arg2);
1066 		((char *)arg1)[arg2] = '\0';
1067 	}
1068 
1069 	return (error);
1070 }
1071 
1072 /*
1073  * Handle any kind of opaque data.
1074  * arg1 points to it, arg2 is the size.
1075  */
1076 
1077 int
1078 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1079 {
1080 	int error, tries;
1081 	u_int generation;
1082 	struct sysctl_req req2;
1083 
1084 	/*
1085 	 * Attempt to get a coherent snapshot, by using the thread
1086 	 * pre-emption counter updated from within mi_switch() to
1087 	 * determine if we were pre-empted during a bcopy() or
1088 	 * copyout(). Make 3 attempts at doing this before giving up.
1089 	 * If we encounter an error, stop immediately.
1090 	 */
1091 	tries = 0;
1092 	req2 = *req;
1093 retry:
1094 	generation = curthread->td_generation;
1095 	error = SYSCTL_OUT(req, arg1, arg2);
1096 	if (error)
1097 		return (error);
1098 	tries++;
1099 	if (generation != curthread->td_generation && tries < 3) {
1100 		*req = req2;
1101 		goto retry;
1102 	}
1103 
1104 	error = SYSCTL_IN(req, arg1, arg2);
1105 
1106 	return (error);
1107 }
1108 
1109 /*
1110  * Transfer functions to/from kernel space.
1111  * XXX: rather untested at this point
1112  */
1113 static int
1114 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1115 {
1116 	size_t i = 0;
1117 
1118 	if (req->oldptr) {
1119 		i = l;
1120 		if (req->oldlen <= req->oldidx)
1121 			i = 0;
1122 		else
1123 			if (i > req->oldlen - req->oldidx)
1124 				i = req->oldlen - req->oldidx;
1125 		if (i > 0)
1126 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
1127 	}
1128 	req->oldidx += l;
1129 	if (req->oldptr && i != l)
1130 		return (ENOMEM);
1131 	return (0);
1132 }
1133 
1134 static int
1135 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1136 {
1137 	if (!req->newptr)
1138 		return (0);
1139 	if (req->newlen - req->newidx < l)
1140 		return (EINVAL);
1141 	bcopy((char *)req->newptr + req->newidx, p, l);
1142 	req->newidx += l;
1143 	return (0);
1144 }
1145 
1146 int
1147 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1148     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1149 {
1150 	int error = 0;
1151 	struct sysctl_req req;
1152 
1153 	bzero(&req, sizeof req);
1154 
1155 	req.td = td;
1156 	req.flags = flags;
1157 
1158 	if (oldlenp) {
1159 		req.oldlen = *oldlenp;
1160 	}
1161 	req.validlen = req.oldlen;
1162 
1163 	if (old) {
1164 		req.oldptr= old;
1165 	}
1166 
1167 	if (new != NULL) {
1168 		req.newlen = newlen;
1169 		req.newptr = new;
1170 	}
1171 
1172 	req.oldfunc = sysctl_old_kernel;
1173 	req.newfunc = sysctl_new_kernel;
1174 	req.lock = REQ_LOCKED;
1175 
1176 	SYSCTL_SLOCK();
1177 	error = sysctl_root(0, name, namelen, &req);
1178 	SYSCTL_SUNLOCK();
1179 
1180 	if (req.lock == REQ_WIRED && req.validlen > 0)
1181 		vsunlock(req.oldptr, req.validlen);
1182 
1183 	if (error && error != ENOMEM)
1184 		return (error);
1185 
1186 	if (retval) {
1187 		if (req.oldptr && req.oldidx > req.validlen)
1188 			*retval = req.validlen;
1189 		else
1190 			*retval = req.oldidx;
1191 	}
1192 	return (error);
1193 }
1194 
1195 int
1196 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1197     void *new, size_t newlen, size_t *retval, int flags)
1198 {
1199         int oid[CTL_MAXNAME];
1200         size_t oidlen, plen;
1201 	int error;
1202 
1203 	oid[0] = 0;		/* sysctl internal magic */
1204 	oid[1] = 3;		/* name2oid */
1205 	oidlen = sizeof(oid);
1206 
1207 	error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1208 	    (void *)name, strlen(name), &plen, flags);
1209 	if (error)
1210 		return (error);
1211 
1212 	error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1213 	    new, newlen, retval, flags);
1214 	return (error);
1215 }
1216 
1217 /*
1218  * Transfer function to/from user space.
1219  */
1220 static int
1221 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1222 {
1223 	int error = 0;
1224 	size_t i, len, origidx;
1225 
1226 	origidx = req->oldidx;
1227 	req->oldidx += l;
1228 	if (req->oldptr == NULL)
1229 		return (0);
1230 	/*
1231 	 * If we have not wired the user supplied buffer and we are currently
1232 	 * holding locks, drop a witness warning, as it's possible that
1233 	 * write operations to the user page can sleep.
1234 	 */
1235 	if (req->lock != REQ_WIRED)
1236 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1237 		    "sysctl_old_user()");
1238 	i = l;
1239 	len = req->validlen;
1240 	if (len <= origidx)
1241 		i = 0;
1242 	else {
1243 		if (i > len - origidx)
1244 			i = len - origidx;
1245 		error = copyout(p, (char *)req->oldptr + origidx, i);
1246 	}
1247 	if (error)
1248 		return (error);
1249 	if (i < l)
1250 		return (ENOMEM);
1251 	return (0);
1252 }
1253 
1254 static int
1255 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1256 {
1257 	int error;
1258 
1259 	if (!req->newptr)
1260 		return (0);
1261 	if (req->newlen - req->newidx < l)
1262 		return (EINVAL);
1263 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1264 	    "sysctl_new_user()");
1265 	error = copyin((char *)req->newptr + req->newidx, p, l);
1266 	req->newidx += l;
1267 	return (error);
1268 }
1269 
1270 /*
1271  * Wire the user space destination buffer.  If set to a value greater than
1272  * zero, the len parameter limits the maximum amount of wired memory.
1273  */
1274 int
1275 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1276 {
1277 	int ret;
1278 	size_t i, wiredlen;
1279 	char *cp, dummy;
1280 
1281 	wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1282 	ret = 0;
1283 	if (req->lock == REQ_LOCKED && req->oldptr &&
1284 	    req->oldfunc == sysctl_old_user) {
1285 		if (wiredlen != 0) {
1286 			ret = vslock(req->oldptr, wiredlen);
1287 			if (ret != 0) {
1288 				if (ret != ENOMEM)
1289 					return (ret);
1290 				wiredlen = 0;
1291 			}
1292 			/*
1293 			 * Touch all the wired pages to avoid PTE modified
1294 			 * bit emulation traps on Alpha while holding locks
1295 			 * in the sysctl handler.
1296 			 */
1297 			for (i = (wiredlen + PAGE_SIZE - 1) / PAGE_SIZE,
1298 			    cp = req->oldptr; i > 0; i--, cp += PAGE_SIZE) {
1299 				copyin(cp, &dummy, 1);
1300 				copyout(&dummy, cp, 1);
1301 			}
1302 		}
1303 		req->lock = REQ_WIRED;
1304 		req->validlen = wiredlen;
1305 	}
1306 	return (0);
1307 }
1308 
1309 int
1310 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1311     int *nindx, struct sysctl_req *req)
1312 {
1313 	struct sysctl_oid *oid;
1314 	int indx;
1315 
1316 	SYSCTL_ASSERT_LOCKED();
1317 	oid = SLIST_FIRST(&sysctl__children);
1318 	indx = 0;
1319 	while (oid && indx < CTL_MAXNAME) {
1320 		if (oid->oid_number == name[indx]) {
1321 			indx++;
1322 			if (oid->oid_kind & CTLFLAG_NOLOCK)
1323 				req->lock = REQ_UNLOCKED;
1324 			if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1325 				if (oid->oid_handler != NULL ||
1326 				    indx == namelen) {
1327 					*noid = oid;
1328 					if (nindx != NULL)
1329 						*nindx = indx;
1330 					return (0);
1331 				}
1332 				oid = SLIST_FIRST(
1333 				    (struct sysctl_oid_list *)oid->oid_arg1);
1334 			} else if (indx == namelen) {
1335 				*noid = oid;
1336 				if (nindx != NULL)
1337 					*nindx = indx;
1338 				return (0);
1339 			} else {
1340 				return (ENOTDIR);
1341 			}
1342 		} else {
1343 			oid = SLIST_NEXT(oid, oid_link);
1344 		}
1345 	}
1346 	return (ENOENT);
1347 }
1348 
1349 /*
1350  * Traverse our tree, and find the right node, execute whatever it points
1351  * to, and return the resulting error code.
1352  */
1353 
1354 static int
1355 sysctl_root(SYSCTL_HANDLER_ARGS)
1356 {
1357 	struct sysctl_oid *oid;
1358 	int error, indx, lvl;
1359 
1360 	SYSCTL_ASSERT_LOCKED();
1361 
1362 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1363 	if (error)
1364 		return (error);
1365 
1366 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1367 		/*
1368 		 * You can't call a sysctl when it's a node, but has
1369 		 * no handler.  Inform the user that it's a node.
1370 		 * The indx may or may not be the same as namelen.
1371 		 */
1372 		if (oid->oid_handler == NULL)
1373 			return (EISDIR);
1374 	}
1375 
1376 	/* Is this sysctl writable? */
1377 	if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1378 		return (EPERM);
1379 
1380 	KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1381 
1382 	/* Is this sysctl sensitive to securelevels? */
1383 	if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1384 		lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1385 		error = securelevel_gt(req->td->td_ucred, lvl);
1386 		if (error)
1387 			return (error);
1388 	}
1389 
1390 	/* Is this sysctl writable by only privileged users? */
1391 	if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1392 		if (oid->oid_kind & CTLFLAG_PRISON)
1393 			error = priv_check(req->td, PRIV_SYSCTL_WRITEJAIL);
1394 		else
1395 			error = priv_check(req->td, PRIV_SYSCTL_WRITE);
1396 		if (error)
1397 			return (error);
1398 	}
1399 
1400 	if (!oid->oid_handler)
1401 		return (EINVAL);
1402 
1403 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1404 		arg1 = (int *)arg1 + indx;
1405 		arg2 -= indx;
1406 	} else {
1407 		arg1 = oid->oid_arg1;
1408 		arg2 = oid->oid_arg2;
1409 	}
1410 #ifdef MAC
1411 	error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1412 	    req);
1413 	if (error != 0)
1414 		return (error);
1415 #endif
1416 	if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1417 		mtx_lock(&Giant);
1418 	error = oid->oid_handler(oid, arg1, arg2, req);
1419 	if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1420 		mtx_unlock(&Giant);
1421 
1422 	return (error);
1423 }
1424 
1425 #ifndef _SYS_SYSPROTO_H_
1426 struct sysctl_args {
1427 	int	*name;
1428 	u_int	namelen;
1429 	void	*old;
1430 	size_t	*oldlenp;
1431 	void	*new;
1432 	size_t	newlen;
1433 };
1434 #endif
1435 int
1436 __sysctl(struct thread *td, struct sysctl_args *uap)
1437 {
1438 	int error, i, name[CTL_MAXNAME];
1439 	size_t j;
1440 
1441 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1442 		return (EINVAL);
1443 
1444  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1445  	if (error)
1446 		return (error);
1447 
1448 	error = userland_sysctl(td, name, uap->namelen,
1449 		uap->old, uap->oldlenp, 0,
1450 		uap->new, uap->newlen, &j, 0);
1451 	if (error && error != ENOMEM)
1452 		return (error);
1453 	if (uap->oldlenp) {
1454 		i = copyout(&j, uap->oldlenp, sizeof(j));
1455 		if (i)
1456 			return (i);
1457 	}
1458 	return (error);
1459 }
1460 
1461 /*
1462  * This is used from various compatibility syscalls too.  That's why name
1463  * must be in kernel space.
1464  */
1465 int
1466 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1467     size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1468     int flags)
1469 {
1470 	int error = 0;
1471 	struct sysctl_req req;
1472 
1473 	bzero(&req, sizeof req);
1474 
1475 	req.td = td;
1476 	req.flags = flags;
1477 
1478 	if (oldlenp) {
1479 		if (inkernel) {
1480 			req.oldlen = *oldlenp;
1481 		} else {
1482 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1483 			if (error)
1484 				return (error);
1485 		}
1486 	}
1487 	req.validlen = req.oldlen;
1488 
1489 	if (old) {
1490 		if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1491 			return (EFAULT);
1492 		req.oldptr= old;
1493 	}
1494 
1495 	if (new != NULL) {
1496 		if (!useracc(new, newlen, VM_PROT_READ))
1497 			return (EFAULT);
1498 		req.newlen = newlen;
1499 		req.newptr = new;
1500 	}
1501 
1502 	req.oldfunc = sysctl_old_user;
1503 	req.newfunc = sysctl_new_user;
1504 	req.lock = REQ_LOCKED;
1505 
1506 #ifdef KTRACE
1507 	if (KTRPOINT(curthread, KTR_SYSCTL))
1508 		ktrsysctl(name, namelen);
1509 #endif
1510 
1511 	SYSCTL_XLOCK();
1512 	CURVNET_SET(TD_TO_VNET(curthread));
1513 
1514 	for (;;) {
1515 		req.oldidx = 0;
1516 		req.newidx = 0;
1517 		error = sysctl_root(0, name, namelen, &req);
1518 		if (error != EAGAIN)
1519 			break;
1520 		uio_yield();
1521 	}
1522 
1523 	CURVNET_RESTORE();
1524 
1525 	if (req.lock == REQ_WIRED && req.validlen > 0)
1526 		vsunlock(req.oldptr, req.validlen);
1527 	SYSCTL_XUNLOCK();
1528 
1529 	if (error && error != ENOMEM)
1530 		return (error);
1531 
1532 	if (retval) {
1533 		if (req.oldptr && req.oldidx > req.validlen)
1534 			*retval = req.validlen;
1535 		else
1536 			*retval = req.oldidx;
1537 	}
1538 	return (error);
1539 }
1540