xref: /freebsd/sys/kern/kern_sysctl.c (revision df347c8a2e8ac08df4c1a6058c12b9f01c289cff)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Mike Karels at Berkeley Software Design, Inc.
9  *
10  * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
11  * project, to make these variables more userfriendly.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)kern_sysctl.c	8.4 (Berkeley) 4/14/94
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_capsicum.h"
44 #include "opt_ddb.h"
45 #include "opt_ktrace.h"
46 
47 #include <sys/param.h>
48 #include <sys/fail.h>
49 #include <sys/systm.h>
50 #include <sys/capsicum.h>
51 #include <sys/kernel.h>
52 #include <sys/limits.h>
53 #include <sys/sysctl.h>
54 #include <sys/malloc.h>
55 #include <sys/priv.h>
56 #include <sys/proc.h>
57 #include <sys/jail.h>
58 #include <sys/kdb.h>
59 #include <sys/lock.h>
60 #include <sys/mutex.h>
61 #include <sys/rmlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sx.h>
64 #include <sys/sysproto.h>
65 #include <sys/uio.h>
66 #ifdef KTRACE
67 #include <sys/ktrace.h>
68 #endif
69 
70 #ifdef DDB
71 #include <ddb/ddb.h>
72 #include <ddb/db_lex.h>
73 #endif
74 
75 #include <net/vnet.h>
76 
77 #include <security/mac/mac_framework.h>
78 
79 #include <vm/vm.h>
80 #include <vm/vm_extern.h>
81 
82 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
83 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
84 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
85 
86 /*
87  * The sysctllock protects the MIB tree.  It also protects sysctl
88  * contexts used with dynamic sysctls.  The sysctl_register_oid() and
89  * sysctl_unregister_oid() routines require the sysctllock to already
90  * be held, so the sysctl_wlock() and sysctl_wunlock() routines are
91  * provided for the few places in the kernel which need to use that
92  * API rather than using the dynamic API.  Use of the dynamic API is
93  * strongly encouraged for most code.
94  *
95  * The sysctlmemlock is used to limit the amount of user memory wired for
96  * sysctl requests.  This is implemented by serializing any userland
97  * sysctl requests larger than a single page via an exclusive lock.
98  */
99 static struct rmlock sysctllock;
100 static struct sx __exclusive_cache_line sysctlmemlock;
101 
102 #define	SYSCTL_WLOCK()		rm_wlock(&sysctllock)
103 #define	SYSCTL_WUNLOCK()	rm_wunlock(&sysctllock)
104 #define	SYSCTL_RLOCK(tracker)	rm_rlock(&sysctllock, (tracker))
105 #define	SYSCTL_RUNLOCK(tracker)	rm_runlock(&sysctllock, (tracker))
106 #define	SYSCTL_WLOCKED()	rm_wowned(&sysctllock)
107 #define	SYSCTL_ASSERT_LOCKED()	rm_assert(&sysctllock, RA_LOCKED)
108 #define	SYSCTL_ASSERT_WLOCKED()	rm_assert(&sysctllock, RA_WLOCKED)
109 #define	SYSCTL_ASSERT_RLOCKED()	rm_assert(&sysctllock, RA_RLOCKED)
110 #define	SYSCTL_INIT()		rm_init_flags(&sysctllock, "sysctl lock", \
111 				    RM_SLEEPABLE)
112 #define	SYSCTL_SLEEP(ch, wmesg, timo)					\
113 				rm_sleep(ch, &sysctllock, 0, wmesg, timo)
114 
115 static int sysctl_root(SYSCTL_HANDLER_ARGS);
116 
117 /* Root list */
118 struct sysctl_oid_list sysctl__children = SLIST_HEAD_INITIALIZER(&sysctl__children);
119 
120 static int	sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
121 		    int recurse);
122 static int	sysctl_old_kernel(struct sysctl_req *, const void *, size_t);
123 static int	sysctl_new_kernel(struct sysctl_req *, void *, size_t);
124 
125 static struct sysctl_oid *
126 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
127 {
128 	struct sysctl_oid *oidp;
129 
130 	SYSCTL_ASSERT_LOCKED();
131 	SLIST_FOREACH(oidp, list, oid_link) {
132 		if (strcmp(oidp->oid_name, name) == 0) {
133 			return (oidp);
134 		}
135 	}
136 	return (NULL);
137 }
138 
139 /*
140  * Initialization of the MIB tree.
141  *
142  * Order by number in each list.
143  */
144 void
145 sysctl_wlock(void)
146 {
147 
148 	SYSCTL_WLOCK();
149 }
150 
151 void
152 sysctl_wunlock(void)
153 {
154 
155 	SYSCTL_WUNLOCK();
156 }
157 
158 static int
159 sysctl_root_handler_locked(struct sysctl_oid *oid, void *arg1, intmax_t arg2,
160     struct sysctl_req *req, struct rm_priotracker *tracker)
161 {
162 	int error;
163 
164 	if (oid->oid_kind & CTLFLAG_DYN)
165 		atomic_add_int(&oid->oid_running, 1);
166 
167 	if (tracker != NULL)
168 		SYSCTL_RUNLOCK(tracker);
169 	else
170 		SYSCTL_WUNLOCK();
171 
172 	if (!(oid->oid_kind & CTLFLAG_MPSAFE))
173 		mtx_lock(&Giant);
174 	error = oid->oid_handler(oid, arg1, arg2, req);
175 	if (!(oid->oid_kind & CTLFLAG_MPSAFE))
176 		mtx_unlock(&Giant);
177 
178 	KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
179 
180 	if (tracker != NULL)
181 		SYSCTL_RLOCK(tracker);
182 	else
183 		SYSCTL_WLOCK();
184 
185 	if (oid->oid_kind & CTLFLAG_DYN) {
186 		if (atomic_fetchadd_int(&oid->oid_running, -1) == 1 &&
187 		    (oid->oid_kind & CTLFLAG_DYING) != 0)
188 			wakeup(&oid->oid_running);
189 	}
190 
191 	return (error);
192 }
193 
194 static void
195 sysctl_load_tunable_by_oid_locked(struct sysctl_oid *oidp)
196 {
197 	struct sysctl_req req;
198 	struct sysctl_oid *curr;
199 	char *penv = NULL;
200 	char path[96];
201 	ssize_t rem = sizeof(path);
202 	ssize_t len;
203 	uint8_t data[512] __aligned(sizeof(uint64_t));
204 	int size;
205 	int error;
206 
207 	path[--rem] = 0;
208 
209 	for (curr = oidp; curr != NULL; curr = SYSCTL_PARENT(curr)) {
210 		len = strlen(curr->oid_name);
211 		rem -= len;
212 		if (curr != oidp)
213 			rem -= 1;
214 		if (rem < 0) {
215 			printf("OID path exceeds %d bytes\n", (int)sizeof(path));
216 			return;
217 		}
218 		memcpy(path + rem, curr->oid_name, len);
219 		if (curr != oidp)
220 			path[rem + len] = '.';
221 	}
222 
223 	memset(&req, 0, sizeof(req));
224 
225 	req.td = curthread;
226 	req.oldfunc = sysctl_old_kernel;
227 	req.newfunc = sysctl_new_kernel;
228 	req.lock = REQ_UNWIRED;
229 
230 	switch (oidp->oid_kind & CTLTYPE) {
231 	case CTLTYPE_INT:
232 		if (getenv_array(path + rem, data, sizeof(data), &size,
233 		    sizeof(int), GETENV_SIGNED) == 0)
234 			return;
235 		req.newlen = size;
236 		req.newptr = data;
237 		break;
238 	case CTLTYPE_UINT:
239 		if (getenv_array(path + rem, data, sizeof(data), &size,
240 		    sizeof(int), GETENV_UNSIGNED) == 0)
241 			return;
242 		req.newlen = size;
243 		req.newptr = data;
244 		break;
245 	case CTLTYPE_LONG:
246 		if (getenv_array(path + rem, data, sizeof(data), &size,
247 		    sizeof(long), GETENV_SIGNED) == 0)
248 			return;
249 		req.newlen = size;
250 		req.newptr = data;
251 		break;
252 	case CTLTYPE_ULONG:
253 		if (getenv_array(path + rem, data, sizeof(data), &size,
254 		    sizeof(long), GETENV_UNSIGNED) == 0)
255 			return;
256 		req.newlen = size;
257 		req.newptr = data;
258 		break;
259 	case CTLTYPE_S8:
260 		if (getenv_array(path + rem, data, sizeof(data), &size,
261 		    sizeof(int8_t), GETENV_SIGNED) == 0)
262 			return;
263 		req.newlen = size;
264 		req.newptr = data;
265 		break;
266 	case CTLTYPE_S16:
267 		if (getenv_array(path + rem, data, sizeof(data), &size,
268 		    sizeof(int16_t), GETENV_SIGNED) == 0)
269 			return;
270 		req.newlen = size;
271 		req.newptr = data;
272 		break;
273 	case CTLTYPE_S32:
274 		if (getenv_array(path + rem, data, sizeof(data), &size,
275 		    sizeof(int32_t), GETENV_SIGNED) == 0)
276 			return;
277 		req.newlen = size;
278 		req.newptr = data;
279 		break;
280 	case CTLTYPE_S64:
281 		if (getenv_array(path + rem, data, sizeof(data), &size,
282 		    sizeof(int64_t), GETENV_SIGNED) == 0)
283 			return;
284 		req.newlen = size;
285 		req.newptr = data;
286 		break;
287 	case CTLTYPE_U8:
288 		if (getenv_array(path + rem, data, sizeof(data), &size,
289 		    sizeof(uint8_t), GETENV_UNSIGNED) == 0)
290 			return;
291 		req.newlen = size;
292 		req.newptr = data;
293 		break;
294 	case CTLTYPE_U16:
295 		if (getenv_array(path + rem, data, sizeof(data), &size,
296 		    sizeof(uint16_t), GETENV_UNSIGNED) == 0)
297 			return;
298 		req.newlen = size;
299 		req.newptr = data;
300 		break;
301 	case CTLTYPE_U32:
302 		if (getenv_array(path + rem, data, sizeof(data), &size,
303 		    sizeof(uint32_t), GETENV_UNSIGNED) == 0)
304 			return;
305 		req.newlen = size;
306 		req.newptr = data;
307 		break;
308 	case CTLTYPE_U64:
309 		if (getenv_array(path + rem, data, sizeof(data), &size,
310 		    sizeof(uint64_t), GETENV_UNSIGNED) == 0)
311 			return;
312 		req.newlen = size;
313 		req.newptr = data;
314 		break;
315 	case CTLTYPE_STRING:
316 		penv = kern_getenv(path + rem);
317 		if (penv == NULL)
318 			return;
319 		req.newlen = strlen(penv);
320 		req.newptr = penv;
321 		break;
322 	default:
323 		return;
324 	}
325 	error = sysctl_root_handler_locked(oidp, oidp->oid_arg1,
326 	    oidp->oid_arg2, &req, NULL);
327 	if (error != 0)
328 		printf("Setting sysctl %s failed: %d\n", path + rem, error);
329 	if (penv != NULL)
330 		freeenv(penv);
331 }
332 
333 /*
334  * Locate the path to a given oid.  Returns the length of the resulting path,
335  * or -1 if the oid was not found.  nodes must have room for CTL_MAXNAME
336  * elements and be NULL initialized.
337  */
338 static int
339 sysctl_search_oid(struct sysctl_oid **nodes, struct sysctl_oid *needle)
340 {
341 	int indx;
342 
343 	SYSCTL_ASSERT_LOCKED();
344 	indx = 0;
345 	while (indx < CTL_MAXNAME && indx >= 0) {
346 		if (nodes[indx] == NULL && indx == 0)
347 			nodes[indx] = SLIST_FIRST(&sysctl__children);
348 		else if (nodes[indx] == NULL)
349 			nodes[indx] = SLIST_FIRST(&nodes[indx - 1]->oid_children);
350 		else
351 			nodes[indx] = SLIST_NEXT(nodes[indx], oid_link);
352 
353 		if (nodes[indx] == needle)
354 			return (indx + 1);
355 
356 		if (nodes[indx] == NULL) {
357 			indx--;
358 			continue;
359 		}
360 
361 		if ((nodes[indx]->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
362 			indx++;
363 			continue;
364 		}
365 	}
366 	return (-1);
367 }
368 
369 static void
370 sysctl_warn_reuse(const char *func, struct sysctl_oid *leaf)
371 {
372 	struct sysctl_oid *nodes[CTL_MAXNAME];
373 	char buf[128];
374 	struct sbuf sb;
375 	int rc, i;
376 
377 	(void)sbuf_new(&sb, buf, sizeof(buf), SBUF_FIXEDLEN | SBUF_INCLUDENUL);
378 	sbuf_set_drain(&sb, sbuf_printf_drain, NULL);
379 
380 	sbuf_printf(&sb, "%s: can't re-use a leaf (", __func__);
381 
382 	memset(nodes, 0, sizeof(nodes));
383 	rc = sysctl_search_oid(nodes, leaf);
384 	if (rc > 0) {
385 		for (i = 0; i < rc; i++)
386 			sbuf_printf(&sb, "%s%.*s", nodes[i]->oid_name,
387 			    i != (rc - 1), ".");
388 	} else {
389 		sbuf_printf(&sb, "%s", leaf->oid_name);
390 	}
391 	sbuf_printf(&sb, ")!\n");
392 
393 	(void)sbuf_finish(&sb);
394 }
395 
396 #ifdef SYSCTL_DEBUG
397 static int
398 sysctl_reuse_test(SYSCTL_HANDLER_ARGS)
399 {
400 	struct rm_priotracker tracker;
401 
402 	SYSCTL_RLOCK(&tracker);
403 	sysctl_warn_reuse(__func__, oidp);
404 	SYSCTL_RUNLOCK(&tracker);
405 	return (0);
406 }
407 SYSCTL_PROC(_sysctl, 0, reuse_test, CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE,
408 	0, 0, sysctl_reuse_test, "-", "");
409 #endif
410 
411 void
412 sysctl_register_oid(struct sysctl_oid *oidp)
413 {
414 	struct sysctl_oid_list *parent = oidp->oid_parent;
415 	struct sysctl_oid *p;
416 	struct sysctl_oid *q;
417 	int oid_number;
418 	int timeout = 2;
419 
420 	/*
421 	 * First check if another oid with the same name already
422 	 * exists in the parent's list.
423 	 */
424 	SYSCTL_ASSERT_WLOCKED();
425 	p = sysctl_find_oidname(oidp->oid_name, parent);
426 	if (p != NULL) {
427 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
428 			p->oid_refcnt++;
429 			return;
430 		} else {
431 			sysctl_warn_reuse(__func__, p);
432 			return;
433 		}
434 	}
435 	/* get current OID number */
436 	oid_number = oidp->oid_number;
437 
438 #if (OID_AUTO >= 0)
439 #error "OID_AUTO is expected to be a negative value"
440 #endif
441 	/*
442 	 * Any negative OID number qualifies as OID_AUTO. Valid OID
443 	 * numbers should always be positive.
444 	 *
445 	 * NOTE: DO NOT change the starting value here, change it in
446 	 * <sys/sysctl.h>, and make sure it is at least 256 to
447 	 * accommodate e.g. net.inet.raw as a static sysctl node.
448 	 */
449 	if (oid_number < 0) {
450 		static int newoid;
451 
452 		/*
453 		 * By decrementing the next OID number we spend less
454 		 * time inserting the OIDs into a sorted list.
455 		 */
456 		if (--newoid < CTL_AUTO_START)
457 			newoid = 0x7fffffff;
458 
459 		oid_number = newoid;
460 	}
461 
462 	/*
463 	 * Insert the OID into the parent's list sorted by OID number.
464 	 */
465 retry:
466 	q = NULL;
467 	SLIST_FOREACH(p, parent, oid_link) {
468 		/* check if the current OID number is in use */
469 		if (oid_number == p->oid_number) {
470 			/* get the next valid OID number */
471 			if (oid_number < CTL_AUTO_START ||
472 			    oid_number == 0x7fffffff) {
473 				/* wraparound - restart */
474 				oid_number = CTL_AUTO_START;
475 				/* don't loop forever */
476 				if (!timeout--)
477 					panic("sysctl: Out of OID numbers\n");
478 				goto retry;
479 			} else {
480 				oid_number++;
481 			}
482 		} else if (oid_number < p->oid_number)
483 			break;
484 		q = p;
485 	}
486 	/* check for non-auto OID number collision */
487 	if (oidp->oid_number >= 0 && oidp->oid_number < CTL_AUTO_START &&
488 	    oid_number >= CTL_AUTO_START) {
489 		printf("sysctl: OID number(%d) is already in use for '%s'\n",
490 		    oidp->oid_number, oidp->oid_name);
491 	}
492 	/* update the OID number, if any */
493 	oidp->oid_number = oid_number;
494 	if (q != NULL)
495 		SLIST_INSERT_AFTER(q, oidp, oid_link);
496 	else
497 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
498 
499 	if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE &&
500 #ifdef VIMAGE
501 	    (oidp->oid_kind & CTLFLAG_VNET) == 0 &&
502 #endif
503 	    (oidp->oid_kind & CTLFLAG_TUN) != 0 &&
504 	    (oidp->oid_kind & CTLFLAG_NOFETCH) == 0) {
505 		/* only fetch value once */
506 		oidp->oid_kind |= CTLFLAG_NOFETCH;
507 		/* try to fetch value from kernel environment */
508 		sysctl_load_tunable_by_oid_locked(oidp);
509 	}
510 }
511 
512 void
513 sysctl_register_disabled_oid(struct sysctl_oid *oidp)
514 {
515 
516 	/*
517 	 * Mark the leaf as dormant if it's not to be immediately enabled.
518 	 * We do not disable nodes as they can be shared between modules
519 	 * and it is always safe to access a node.
520 	 */
521 	KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
522 	    ("internal flag is set in oid_kind"));
523 	if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
524 		oidp->oid_kind |= CTLFLAG_DORMANT;
525 	sysctl_register_oid(oidp);
526 }
527 
528 void
529 sysctl_enable_oid(struct sysctl_oid *oidp)
530 {
531 
532 	SYSCTL_ASSERT_WLOCKED();
533 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
534 		KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
535 		    ("sysctl node is marked as dormant"));
536 		return;
537 	}
538 	KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) != 0,
539 	    ("enabling already enabled sysctl oid"));
540 	oidp->oid_kind &= ~CTLFLAG_DORMANT;
541 }
542 
543 void
544 sysctl_unregister_oid(struct sysctl_oid *oidp)
545 {
546 	struct sysctl_oid *p;
547 	int error;
548 
549 	SYSCTL_ASSERT_WLOCKED();
550 	if (oidp->oid_number == OID_AUTO) {
551 		error = EINVAL;
552 	} else {
553 		error = ENOENT;
554 		SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
555 			if (p == oidp) {
556 				SLIST_REMOVE(oidp->oid_parent, oidp,
557 				    sysctl_oid, oid_link);
558 				error = 0;
559 				break;
560 			}
561 		}
562 	}
563 
564 	/*
565 	 * This can happen when a module fails to register and is
566 	 * being unloaded afterwards.  It should not be a panic()
567 	 * for normal use.
568 	 */
569 	if (error) {
570 		printf("%s: failed(%d) to unregister sysctl(%s)\n",
571 		    __func__, error, oidp->oid_name);
572 	}
573 }
574 
575 /* Initialize a new context to keep track of dynamically added sysctls. */
576 int
577 sysctl_ctx_init(struct sysctl_ctx_list *c)
578 {
579 
580 	if (c == NULL) {
581 		return (EINVAL);
582 	}
583 
584 	/*
585 	 * No locking here, the caller is responsible for not adding
586 	 * new nodes to a context until after this function has
587 	 * returned.
588 	 */
589 	TAILQ_INIT(c);
590 	return (0);
591 }
592 
593 /* Free the context, and destroy all dynamic oids registered in this context */
594 int
595 sysctl_ctx_free(struct sysctl_ctx_list *clist)
596 {
597 	struct sysctl_ctx_entry *e, *e1;
598 	int error;
599 
600 	error = 0;
601 	/*
602 	 * First perform a "dry run" to check if it's ok to remove oids.
603 	 * XXX FIXME
604 	 * XXX This algorithm is a hack. But I don't know any
605 	 * XXX better solution for now...
606 	 */
607 	SYSCTL_WLOCK();
608 	TAILQ_FOREACH(e, clist, link) {
609 		error = sysctl_remove_oid_locked(e->entry, 0, 0);
610 		if (error)
611 			break;
612 	}
613 	/*
614 	 * Restore deregistered entries, either from the end,
615 	 * or from the place where error occurred.
616 	 * e contains the entry that was not unregistered
617 	 */
618 	if (error)
619 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
620 	else
621 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
622 	while (e1 != NULL) {
623 		sysctl_register_oid(e1->entry);
624 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
625 	}
626 	if (error) {
627 		SYSCTL_WUNLOCK();
628 		return(EBUSY);
629 	}
630 	/* Now really delete the entries */
631 	e = TAILQ_FIRST(clist);
632 	while (e != NULL) {
633 		e1 = TAILQ_NEXT(e, link);
634 		error = sysctl_remove_oid_locked(e->entry, 1, 0);
635 		if (error)
636 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
637 			    e->entry->oid_name);
638 		free(e, M_SYSCTLOID);
639 		e = e1;
640 	}
641 	SYSCTL_WUNLOCK();
642 	return (error);
643 }
644 
645 /* Add an entry to the context */
646 struct sysctl_ctx_entry *
647 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
648 {
649 	struct sysctl_ctx_entry *e;
650 
651 	SYSCTL_ASSERT_WLOCKED();
652 	if (clist == NULL || oidp == NULL)
653 		return(NULL);
654 	e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
655 	e->entry = oidp;
656 	TAILQ_INSERT_HEAD(clist, e, link);
657 	return (e);
658 }
659 
660 /* Find an entry in the context */
661 struct sysctl_ctx_entry *
662 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
663 {
664 	struct sysctl_ctx_entry *e;
665 
666 	SYSCTL_ASSERT_WLOCKED();
667 	if (clist == NULL || oidp == NULL)
668 		return(NULL);
669 	TAILQ_FOREACH(e, clist, link) {
670 		if(e->entry == oidp)
671 			return(e);
672 	}
673 	return (e);
674 }
675 
676 /*
677  * Delete an entry from the context.
678  * NOTE: this function doesn't free oidp! You have to remove it
679  * with sysctl_remove_oid().
680  */
681 int
682 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
683 {
684 	struct sysctl_ctx_entry *e;
685 
686 	if (clist == NULL || oidp == NULL)
687 		return (EINVAL);
688 	SYSCTL_WLOCK();
689 	e = sysctl_ctx_entry_find(clist, oidp);
690 	if (e != NULL) {
691 		TAILQ_REMOVE(clist, e, link);
692 		SYSCTL_WUNLOCK();
693 		free(e, M_SYSCTLOID);
694 		return (0);
695 	} else {
696 		SYSCTL_WUNLOCK();
697 		return (ENOENT);
698 	}
699 }
700 
701 /*
702  * Remove dynamically created sysctl trees.
703  * oidp - top of the tree to be removed
704  * del - if 0 - just deregister, otherwise free up entries as well
705  * recurse - if != 0 traverse the subtree to be deleted
706  */
707 int
708 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
709 {
710 	int error;
711 
712 	SYSCTL_WLOCK();
713 	error = sysctl_remove_oid_locked(oidp, del, recurse);
714 	SYSCTL_WUNLOCK();
715 	return (error);
716 }
717 
718 int
719 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
720     int del, int recurse)
721 {
722 	struct sysctl_oid *p, *tmp;
723 	int error;
724 
725 	error = ENOENT;
726 	SYSCTL_WLOCK();
727 	SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
728 		if (strcmp(p->oid_name, name) == 0) {
729 			error = sysctl_remove_oid_locked(p, del, recurse);
730 			break;
731 		}
732 	}
733 	SYSCTL_WUNLOCK();
734 
735 	return (error);
736 }
737 
738 
739 static int
740 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
741 {
742 	struct sysctl_oid *p, *tmp;
743 	int error;
744 
745 	SYSCTL_ASSERT_WLOCKED();
746 	if (oidp == NULL)
747 		return(EINVAL);
748 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
749 		printf("Warning: can't remove non-dynamic nodes (%s)!\n",
750 		    oidp->oid_name);
751 		return (EINVAL);
752 	}
753 	/*
754 	 * WARNING: normal method to do this should be through
755 	 * sysctl_ctx_free(). Use recursing as the last resort
756 	 * method to purge your sysctl tree of leftovers...
757 	 * However, if some other code still references these nodes,
758 	 * it will panic.
759 	 */
760 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
761 		if (oidp->oid_refcnt == 1) {
762 			SLIST_FOREACH_SAFE(p,
763 			    SYSCTL_CHILDREN(oidp), oid_link, tmp) {
764 				if (!recurse) {
765 					printf("Warning: failed attempt to "
766 					    "remove oid %s with child %s\n",
767 					    oidp->oid_name, p->oid_name);
768 					return (ENOTEMPTY);
769 				}
770 				error = sysctl_remove_oid_locked(p, del,
771 				    recurse);
772 				if (error)
773 					return (error);
774 			}
775 		}
776 	}
777 	if (oidp->oid_refcnt > 1 ) {
778 		oidp->oid_refcnt--;
779 	} else {
780 		if (oidp->oid_refcnt == 0) {
781 			printf("Warning: bad oid_refcnt=%u (%s)!\n",
782 				oidp->oid_refcnt, oidp->oid_name);
783 			return (EINVAL);
784 		}
785 		sysctl_unregister_oid(oidp);
786 		if (del) {
787 			/*
788 			 * Wait for all threads running the handler to drain.
789 			 * This preserves the previous behavior when the
790 			 * sysctl lock was held across a handler invocation,
791 			 * and is necessary for module unload correctness.
792 			 */
793 			while (oidp->oid_running > 0) {
794 				oidp->oid_kind |= CTLFLAG_DYING;
795 				SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
796 			}
797 			if (oidp->oid_descr)
798 				free(__DECONST(char *, oidp->oid_descr),
799 				    M_SYSCTLOID);
800 			if (oidp->oid_label)
801 				free(__DECONST(char *, oidp->oid_label),
802 				    M_SYSCTLOID);
803 			free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
804 			free(oidp, M_SYSCTLOID);
805 		}
806 	}
807 	return (0);
808 }
809 /*
810  * Create new sysctls at run time.
811  * clist may point to a valid context initialized with sysctl_ctx_init().
812  */
813 struct sysctl_oid *
814 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
815 	int number, const char *name, int kind, void *arg1, intmax_t arg2,
816 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr,
817 	const char *label)
818 {
819 	struct sysctl_oid *oidp;
820 
821 	/* You have to hook up somewhere.. */
822 	if (parent == NULL)
823 		return(NULL);
824 	/* Check if the node already exists, otherwise create it */
825 	SYSCTL_WLOCK();
826 	oidp = sysctl_find_oidname(name, parent);
827 	if (oidp != NULL) {
828 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
829 			oidp->oid_refcnt++;
830 			/* Update the context */
831 			if (clist != NULL)
832 				sysctl_ctx_entry_add(clist, oidp);
833 			SYSCTL_WUNLOCK();
834 			return (oidp);
835 		} else {
836 			sysctl_warn_reuse(__func__, oidp);
837 			SYSCTL_WUNLOCK();
838 			return (NULL);
839 		}
840 	}
841 	oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
842 	oidp->oid_parent = parent;
843 	SLIST_INIT(&oidp->oid_children);
844 	oidp->oid_number = number;
845 	oidp->oid_refcnt = 1;
846 	oidp->oid_name = strdup(name, M_SYSCTLOID);
847 	oidp->oid_handler = handler;
848 	oidp->oid_kind = CTLFLAG_DYN | kind;
849 	oidp->oid_arg1 = arg1;
850 	oidp->oid_arg2 = arg2;
851 	oidp->oid_fmt = fmt;
852 	if (descr != NULL)
853 		oidp->oid_descr = strdup(descr, M_SYSCTLOID);
854 	if (label != NULL)
855 		oidp->oid_label = strdup(label, M_SYSCTLOID);
856 	/* Update the context, if used */
857 	if (clist != NULL)
858 		sysctl_ctx_entry_add(clist, oidp);
859 	/* Register this oid */
860 	sysctl_register_oid(oidp);
861 	SYSCTL_WUNLOCK();
862 	return (oidp);
863 }
864 
865 /*
866  * Rename an existing oid.
867  */
868 void
869 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
870 {
871 	char *newname;
872 	char *oldname;
873 
874 	newname = strdup(name, M_SYSCTLOID);
875 	SYSCTL_WLOCK();
876 	oldname = __DECONST(char *, oidp->oid_name);
877 	oidp->oid_name = newname;
878 	SYSCTL_WUNLOCK();
879 	free(oldname, M_SYSCTLOID);
880 }
881 
882 /*
883  * Reparent an existing oid.
884  */
885 int
886 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
887 {
888 	struct sysctl_oid *oidp;
889 
890 	SYSCTL_WLOCK();
891 	if (oid->oid_parent == parent) {
892 		SYSCTL_WUNLOCK();
893 		return (0);
894 	}
895 	oidp = sysctl_find_oidname(oid->oid_name, parent);
896 	if (oidp != NULL) {
897 		SYSCTL_WUNLOCK();
898 		return (EEXIST);
899 	}
900 	sysctl_unregister_oid(oid);
901 	oid->oid_parent = parent;
902 	oid->oid_number = OID_AUTO;
903 	sysctl_register_oid(oid);
904 	SYSCTL_WUNLOCK();
905 	return (0);
906 }
907 
908 /*
909  * Register the kernel's oids on startup.
910  */
911 SET_DECLARE(sysctl_set, struct sysctl_oid);
912 
913 static void
914 sysctl_register_all(void *arg)
915 {
916 	struct sysctl_oid **oidp;
917 
918 	sx_init(&sysctlmemlock, "sysctl mem");
919 	SYSCTL_INIT();
920 	SYSCTL_WLOCK();
921 	SET_FOREACH(oidp, sysctl_set)
922 		sysctl_register_oid(*oidp);
923 	SYSCTL_WUNLOCK();
924 }
925 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_FIRST, sysctl_register_all, NULL);
926 
927 /*
928  * "Staff-functions"
929  *
930  * These functions implement a presently undocumented interface
931  * used by the sysctl program to walk the tree, and get the type
932  * so it can print the value.
933  * This interface is under work and consideration, and should probably
934  * be killed with a big axe by the first person who can find the time.
935  * (be aware though, that the proper interface isn't as obvious as it
936  * may seem, there are various conflicting requirements.
937  *
938  * {0,0}	printf the entire MIB-tree.
939  * {0,1,...}	return the name of the "..." OID.
940  * {0,2,...}	return the next OID.
941  * {0,3}	return the OID of the name in "new"
942  * {0,4,...}	return the kind & format info for the "..." OID.
943  * {0,5,...}	return the description of the "..." OID.
944  * {0,6,...}	return the aggregation label of the "..." OID.
945  */
946 
947 #ifdef SYSCTL_DEBUG
948 static void
949 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
950 {
951 	int k;
952 	struct sysctl_oid *oidp;
953 
954 	SYSCTL_ASSERT_LOCKED();
955 	SLIST_FOREACH(oidp, l, oid_link) {
956 
957 		for (k=0; k<i; k++)
958 			printf(" ");
959 
960 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
961 
962 		printf("%c%c",
963 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
964 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
965 
966 		if (oidp->oid_handler)
967 			printf(" *Handler");
968 
969 		switch (oidp->oid_kind & CTLTYPE) {
970 			case CTLTYPE_NODE:
971 				printf(" Node\n");
972 				if (!oidp->oid_handler) {
973 					sysctl_sysctl_debug_dump_node(
974 					    SYSCTL_CHILDREN(oidp), i + 2);
975 				}
976 				break;
977 			case CTLTYPE_INT:    printf(" Int\n"); break;
978 			case CTLTYPE_UINT:   printf(" u_int\n"); break;
979 			case CTLTYPE_LONG:   printf(" Long\n"); break;
980 			case CTLTYPE_ULONG:  printf(" u_long\n"); break;
981 			case CTLTYPE_STRING: printf(" String\n"); break;
982 			case CTLTYPE_S8:     printf(" int8_t\n"); break;
983 			case CTLTYPE_S16:    printf(" int16_t\n"); break;
984 			case CTLTYPE_S32:    printf(" int32_t\n"); break;
985 			case CTLTYPE_S64:    printf(" int64_t\n"); break;
986 			case CTLTYPE_U8:     printf(" uint8_t\n"); break;
987 			case CTLTYPE_U16:    printf(" uint16_t\n"); break;
988 			case CTLTYPE_U32:    printf(" uint32_t\n"); break;
989 			case CTLTYPE_U64:    printf(" uint64_t\n"); break;
990 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
991 			default:	     printf("\n");
992 		}
993 
994 	}
995 }
996 
997 static int
998 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
999 {
1000 	struct rm_priotracker tracker;
1001 	int error;
1002 
1003 	error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
1004 	if (error)
1005 		return (error);
1006 	SYSCTL_RLOCK(&tracker);
1007 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
1008 	SYSCTL_RUNLOCK(&tracker);
1009 	return (ENOENT);
1010 }
1011 
1012 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE,
1013 	0, 0, sysctl_sysctl_debug, "-", "");
1014 #endif
1015 
1016 static int
1017 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
1018 {
1019 	int *name = (int *) arg1;
1020 	u_int namelen = arg2;
1021 	int error = 0;
1022 	struct sysctl_oid *oid;
1023 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
1024 	struct rm_priotracker tracker;
1025 	char buf[10];
1026 
1027 	SYSCTL_RLOCK(&tracker);
1028 	while (namelen) {
1029 		if (!lsp) {
1030 			snprintf(buf,sizeof(buf),"%d",*name);
1031 			if (req->oldidx)
1032 				error = SYSCTL_OUT(req, ".", 1);
1033 			if (!error)
1034 				error = SYSCTL_OUT(req, buf, strlen(buf));
1035 			if (error)
1036 				goto out;
1037 			namelen--;
1038 			name++;
1039 			continue;
1040 		}
1041 		lsp2 = NULL;
1042 		SLIST_FOREACH(oid, lsp, oid_link) {
1043 			if (oid->oid_number != *name)
1044 				continue;
1045 
1046 			if (req->oldidx)
1047 				error = SYSCTL_OUT(req, ".", 1);
1048 			if (!error)
1049 				error = SYSCTL_OUT(req, oid->oid_name,
1050 					strlen(oid->oid_name));
1051 			if (error)
1052 				goto out;
1053 
1054 			namelen--;
1055 			name++;
1056 
1057 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1058 				break;
1059 
1060 			if (oid->oid_handler)
1061 				break;
1062 
1063 			lsp2 = SYSCTL_CHILDREN(oid);
1064 			break;
1065 		}
1066 		lsp = lsp2;
1067 	}
1068 	error = SYSCTL_OUT(req, "", 1);
1069  out:
1070 	SYSCTL_RUNLOCK(&tracker);
1071 	return (error);
1072 }
1073 
1074 /*
1075  * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
1076  * capability mode.
1077  */
1078 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD,
1079     sysctl_sysctl_name, "");
1080 
1081 static int
1082 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
1083 	int *next, int *len, int level, struct sysctl_oid **oidpp)
1084 {
1085 	struct sysctl_oid *oidp;
1086 
1087 	SYSCTL_ASSERT_LOCKED();
1088 	*len = level;
1089 	SLIST_FOREACH(oidp, lsp, oid_link) {
1090 		*next = oidp->oid_number;
1091 		*oidpp = oidp;
1092 
1093 		if ((oidp->oid_kind & (CTLFLAG_SKIP | CTLFLAG_DORMANT)) != 0)
1094 			continue;
1095 
1096 		if (!namelen) {
1097 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1098 				return (0);
1099 			if (oidp->oid_handler)
1100 				/* We really should call the handler here...*/
1101 				return (0);
1102 			lsp = SYSCTL_CHILDREN(oidp);
1103 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
1104 				len, level+1, oidpp))
1105 				return (0);
1106 			goto emptynode;
1107 		}
1108 
1109 		if (oidp->oid_number < *name)
1110 			continue;
1111 
1112 		if (oidp->oid_number > *name) {
1113 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1114 				return (0);
1115 			if (oidp->oid_handler)
1116 				return (0);
1117 			lsp = SYSCTL_CHILDREN(oidp);
1118 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
1119 				next+1, len, level+1, oidpp))
1120 				return (0);
1121 			goto next;
1122 		}
1123 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1124 			continue;
1125 
1126 		if (oidp->oid_handler)
1127 			continue;
1128 
1129 		lsp = SYSCTL_CHILDREN(oidp);
1130 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
1131 			len, level+1, oidpp))
1132 			return (0);
1133 	next:
1134 		namelen = 1;
1135 	emptynode:
1136 		*len = level;
1137 	}
1138 	return (1);
1139 }
1140 
1141 static int
1142 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
1143 {
1144 	int *name = (int *) arg1;
1145 	u_int namelen = arg2;
1146 	int i, j, error;
1147 	struct sysctl_oid *oid;
1148 	struct sysctl_oid_list *lsp = &sysctl__children;
1149 	struct rm_priotracker tracker;
1150 	int newoid[CTL_MAXNAME];
1151 
1152 	SYSCTL_RLOCK(&tracker);
1153 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
1154 	SYSCTL_RUNLOCK(&tracker);
1155 	if (i)
1156 		return (ENOENT);
1157 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
1158 	return (error);
1159 }
1160 
1161 /*
1162  * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
1163  * capability mode.
1164  */
1165 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD,
1166     sysctl_sysctl_next, "");
1167 
1168 static int
1169 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
1170 {
1171 	struct sysctl_oid *oidp;
1172 	struct sysctl_oid_list *lsp = &sysctl__children;
1173 	char *p;
1174 
1175 	SYSCTL_ASSERT_LOCKED();
1176 
1177 	for (*len = 0; *len < CTL_MAXNAME;) {
1178 		p = strsep(&name, ".");
1179 
1180 		oidp = SLIST_FIRST(lsp);
1181 		for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
1182 			if (oidp == NULL)
1183 				return (ENOENT);
1184 			if (strcmp(p, oidp->oid_name) == 0)
1185 				break;
1186 		}
1187 		*oid++ = oidp->oid_number;
1188 		(*len)++;
1189 
1190 		if (name == NULL || *name == '\0') {
1191 			if (oidpp)
1192 				*oidpp = oidp;
1193 			return (0);
1194 		}
1195 
1196 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1197 			break;
1198 
1199 		if (oidp->oid_handler)
1200 			break;
1201 
1202 		lsp = SYSCTL_CHILDREN(oidp);
1203 	}
1204 	return (ENOENT);
1205 }
1206 
1207 static int
1208 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
1209 {
1210 	char *p;
1211 	int error, oid[CTL_MAXNAME], len = 0;
1212 	struct sysctl_oid *op = NULL;
1213 	struct rm_priotracker tracker;
1214 	char buf[32];
1215 
1216 	if (!req->newlen)
1217 		return (ENOENT);
1218 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
1219 		return (ENAMETOOLONG);
1220 
1221 	p = buf;
1222 	if (req->newlen >= sizeof(buf))
1223 		p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
1224 
1225 	error = SYSCTL_IN(req, p, req->newlen);
1226 	if (error) {
1227 		if (p != buf)
1228 			free(p, M_SYSCTL);
1229 		return (error);
1230 	}
1231 
1232 	p [req->newlen] = '\0';
1233 
1234 	SYSCTL_RLOCK(&tracker);
1235 	error = name2oid(p, oid, &len, &op);
1236 	SYSCTL_RUNLOCK(&tracker);
1237 
1238 	if (p != buf)
1239 		free(p, M_SYSCTL);
1240 
1241 	if (error)
1242 		return (error);
1243 
1244 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
1245 	return (error);
1246 }
1247 
1248 /*
1249  * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
1250  * capability mode.
1251  */
1252 SYSCTL_PROC(_sysctl, 3, name2oid,
1253     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE
1254     | CTLFLAG_CAPRW, 0, 0, sysctl_sysctl_name2oid, "I", "");
1255 
1256 static int
1257 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
1258 {
1259 	struct sysctl_oid *oid;
1260 	struct rm_priotracker tracker;
1261 	int error;
1262 
1263 	SYSCTL_RLOCK(&tracker);
1264 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1265 	if (error)
1266 		goto out;
1267 
1268 	if (oid->oid_fmt == NULL) {
1269 		error = ENOENT;
1270 		goto out;
1271 	}
1272 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
1273 	if (error)
1274 		goto out;
1275 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
1276  out:
1277 	SYSCTL_RUNLOCK(&tracker);
1278 	return (error);
1279 }
1280 
1281 
1282 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
1283     sysctl_sysctl_oidfmt, "");
1284 
1285 static int
1286 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
1287 {
1288 	struct sysctl_oid *oid;
1289 	struct rm_priotracker tracker;
1290 	int error;
1291 
1292 	SYSCTL_RLOCK(&tracker);
1293 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1294 	if (error)
1295 		goto out;
1296 
1297 	if (oid->oid_descr == NULL) {
1298 		error = ENOENT;
1299 		goto out;
1300 	}
1301 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
1302  out:
1303 	SYSCTL_RUNLOCK(&tracker);
1304 	return (error);
1305 }
1306 
1307 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
1308     sysctl_sysctl_oiddescr, "");
1309 
1310 static int
1311 sysctl_sysctl_oidlabel(SYSCTL_HANDLER_ARGS)
1312 {
1313 	struct sysctl_oid *oid;
1314 	struct rm_priotracker tracker;
1315 	int error;
1316 
1317 	SYSCTL_RLOCK(&tracker);
1318 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1319 	if (error)
1320 		goto out;
1321 
1322 	if (oid->oid_label == NULL) {
1323 		error = ENOENT;
1324 		goto out;
1325 	}
1326 	error = SYSCTL_OUT(req, oid->oid_label, strlen(oid->oid_label) + 1);
1327  out:
1328 	SYSCTL_RUNLOCK(&tracker);
1329 	return (error);
1330 }
1331 
1332 static SYSCTL_NODE(_sysctl, 6, oidlabel,
1333     CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidlabel, "");
1334 
1335 /*
1336  * Default "handler" functions.
1337  */
1338 
1339 /*
1340  * Handle a bool.
1341  * Two cases:
1342  *     a variable:  point arg1 at it.
1343  *     a constant:  pass it in arg2.
1344  */
1345 
1346 int
1347 sysctl_handle_bool(SYSCTL_HANDLER_ARGS)
1348 {
1349 	uint8_t temp;
1350 	int error;
1351 
1352 	/*
1353 	 * Attempt to get a coherent snapshot by making a copy of the data.
1354 	 */
1355 	if (arg1)
1356 		temp = *(bool *)arg1 ? 1 : 0;
1357 	else
1358 		temp = arg2 ? 1 : 0;
1359 
1360 	error = SYSCTL_OUT(req, &temp, sizeof(temp));
1361 	if (error || !req->newptr)
1362 		return (error);
1363 
1364 	if (!arg1)
1365 		error = EPERM;
1366 	else {
1367 		error = SYSCTL_IN(req, &temp, sizeof(temp));
1368 		if (!error)
1369 			*(bool *)arg1 = temp ? 1 : 0;
1370 	}
1371 	return (error);
1372 }
1373 
1374 /*
1375  * Handle an int8_t, signed or unsigned.
1376  * Two cases:
1377  *     a variable:  point arg1 at it.
1378  *     a constant:  pass it in arg2.
1379  */
1380 
1381 int
1382 sysctl_handle_8(SYSCTL_HANDLER_ARGS)
1383 {
1384 	int8_t tmpout;
1385 	int error = 0;
1386 
1387 	/*
1388 	 * Attempt to get a coherent snapshot by making a copy of the data.
1389 	 */
1390 	if (arg1)
1391 		tmpout = *(int8_t *)arg1;
1392 	else
1393 		tmpout = arg2;
1394 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1395 
1396 	if (error || !req->newptr)
1397 		return (error);
1398 
1399 	if (!arg1)
1400 		error = EPERM;
1401 	else
1402 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1403 	return (error);
1404 }
1405 
1406 /*
1407  * Handle an int16_t, signed or unsigned.
1408  * Two cases:
1409  *     a variable:  point arg1 at it.
1410  *     a constant:  pass it in arg2.
1411  */
1412 
1413 int
1414 sysctl_handle_16(SYSCTL_HANDLER_ARGS)
1415 {
1416 	int16_t tmpout;
1417 	int error = 0;
1418 
1419 	/*
1420 	 * Attempt to get a coherent snapshot by making a copy of the data.
1421 	 */
1422 	if (arg1)
1423 		tmpout = *(int16_t *)arg1;
1424 	else
1425 		tmpout = arg2;
1426 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1427 
1428 	if (error || !req->newptr)
1429 		return (error);
1430 
1431 	if (!arg1)
1432 		error = EPERM;
1433 	else
1434 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1435 	return (error);
1436 }
1437 
1438 /*
1439  * Handle an int32_t, signed or unsigned.
1440  * Two cases:
1441  *     a variable:  point arg1 at it.
1442  *     a constant:  pass it in arg2.
1443  */
1444 
1445 int
1446 sysctl_handle_32(SYSCTL_HANDLER_ARGS)
1447 {
1448 	int32_t tmpout;
1449 	int error = 0;
1450 
1451 	/*
1452 	 * Attempt to get a coherent snapshot by making a copy of the data.
1453 	 */
1454 	if (arg1)
1455 		tmpout = *(int32_t *)arg1;
1456 	else
1457 		tmpout = arg2;
1458 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1459 
1460 	if (error || !req->newptr)
1461 		return (error);
1462 
1463 	if (!arg1)
1464 		error = EPERM;
1465 	else
1466 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1467 	return (error);
1468 }
1469 
1470 /*
1471  * Handle an int, signed or unsigned.
1472  * Two cases:
1473  *     a variable:  point arg1 at it.
1474  *     a constant:  pass it in arg2.
1475  */
1476 
1477 int
1478 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
1479 {
1480 	int tmpout, error = 0;
1481 
1482 	/*
1483 	 * Attempt to get a coherent snapshot by making a copy of the data.
1484 	 */
1485 	if (arg1)
1486 		tmpout = *(int *)arg1;
1487 	else
1488 		tmpout = arg2;
1489 	error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1490 
1491 	if (error || !req->newptr)
1492 		return (error);
1493 
1494 	if (!arg1)
1495 		error = EPERM;
1496 	else
1497 		error = SYSCTL_IN(req, arg1, sizeof(int));
1498 	return (error);
1499 }
1500 
1501 /*
1502  * Based on on sysctl_handle_int() convert milliseconds into ticks.
1503  * Note: this is used by TCP.
1504  */
1505 
1506 int
1507 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1508 {
1509 	int error, s, tt;
1510 
1511 	tt = *(int *)arg1;
1512 	s = (int)((int64_t)tt * 1000 / hz);
1513 
1514 	error = sysctl_handle_int(oidp, &s, 0, req);
1515 	if (error || !req->newptr)
1516 		return (error);
1517 
1518 	tt = (int)((int64_t)s * hz / 1000);
1519 	if (tt < 1)
1520 		return (EINVAL);
1521 
1522 	*(int *)arg1 = tt;
1523 	return (0);
1524 }
1525 
1526 
1527 /*
1528  * Handle a long, signed or unsigned.
1529  * Two cases:
1530  *     a variable:  point arg1 at it.
1531  *     a constant:  pass it in arg2.
1532  */
1533 
1534 int
1535 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1536 {
1537 	int error = 0;
1538 	long tmplong;
1539 #ifdef SCTL_MASK32
1540 	int tmpint;
1541 #endif
1542 
1543 	/*
1544 	 * Attempt to get a coherent snapshot by making a copy of the data.
1545 	 */
1546 	if (arg1)
1547 		tmplong = *(long *)arg1;
1548 	else
1549 		tmplong = arg2;
1550 #ifdef SCTL_MASK32
1551 	if (req->flags & SCTL_MASK32) {
1552 		tmpint = tmplong;
1553 		error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1554 	} else
1555 #endif
1556 		error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1557 
1558 	if (error || !req->newptr)
1559 		return (error);
1560 
1561 	if (!arg1)
1562 		error = EPERM;
1563 #ifdef SCTL_MASK32
1564 	else if (req->flags & SCTL_MASK32) {
1565 		error = SYSCTL_IN(req, &tmpint, sizeof(int));
1566 		*(long *)arg1 = (long)tmpint;
1567 	}
1568 #endif
1569 	else
1570 		error = SYSCTL_IN(req, arg1, sizeof(long));
1571 	return (error);
1572 }
1573 
1574 /*
1575  * Handle a 64 bit int, signed or unsigned.
1576  * Two cases:
1577  *     a variable:  point arg1 at it.
1578  *     a constant:  pass it in arg2.
1579  */
1580 int
1581 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1582 {
1583 	int error = 0;
1584 	uint64_t tmpout;
1585 
1586 	/*
1587 	 * Attempt to get a coherent snapshot by making a copy of the data.
1588 	 */
1589 	if (arg1)
1590 		tmpout = *(uint64_t *)arg1;
1591 	else
1592 		tmpout = arg2;
1593 	error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1594 
1595 	if (error || !req->newptr)
1596 		return (error);
1597 
1598 	if (!arg1)
1599 		error = EPERM;
1600 	else
1601 		error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1602 	return (error);
1603 }
1604 
1605 /*
1606  * Handle our generic '\0' terminated 'C' string.
1607  * Two cases:
1608  * 	a variable string:  point arg1 at it, arg2 is max length.
1609  * 	a constant string:  point arg1 at it, arg2 is zero.
1610  */
1611 
1612 int
1613 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1614 {
1615 	size_t outlen;
1616 	int error = 0, ro_string = 0;
1617 
1618 	/*
1619 	 * A zero-length buffer indicates a fixed size read-only
1620 	 * string.  In ddb, don't worry about trying to make a malloced
1621 	 * snapshot.
1622 	 */
1623 	if (arg2 == 0 || kdb_active) {
1624 		arg2 = strlen((char *)arg1) + 1;
1625 		ro_string = 1;
1626 	}
1627 
1628 	if (req->oldptr != NULL) {
1629 		char *tmparg;
1630 
1631 		if (ro_string) {
1632 			tmparg = arg1;
1633 		} else {
1634 			/* try to make a coherent snapshot of the string */
1635 			tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK);
1636 			memcpy(tmparg, arg1, arg2);
1637 		}
1638 
1639 		outlen = strnlen(tmparg, arg2 - 1) + 1;
1640 		error = SYSCTL_OUT(req, tmparg, outlen);
1641 
1642 		if (!ro_string)
1643 			free(tmparg, M_SYSCTLTMP);
1644 	} else {
1645 		outlen = strnlen((char *)arg1, arg2 - 1) + 1;
1646 		error = SYSCTL_OUT(req, NULL, outlen);
1647 	}
1648 	if (error || !req->newptr)
1649 		return (error);
1650 
1651 	if ((req->newlen - req->newidx) >= arg2) {
1652 		error = EINVAL;
1653 	} else {
1654 		arg2 = (req->newlen - req->newidx);
1655 		error = SYSCTL_IN(req, arg1, arg2);
1656 		((char *)arg1)[arg2] = '\0';
1657 	}
1658 	return (error);
1659 }
1660 
1661 /*
1662  * Handle any kind of opaque data.
1663  * arg1 points to it, arg2 is the size.
1664  */
1665 
1666 int
1667 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1668 {
1669 	int error, tries;
1670 	u_int generation;
1671 	struct sysctl_req req2;
1672 
1673 	/*
1674 	 * Attempt to get a coherent snapshot, by using the thread
1675 	 * pre-emption counter updated from within mi_switch() to
1676 	 * determine if we were pre-empted during a bcopy() or
1677 	 * copyout(). Make 3 attempts at doing this before giving up.
1678 	 * If we encounter an error, stop immediately.
1679 	 */
1680 	tries = 0;
1681 	req2 = *req;
1682 retry:
1683 	generation = curthread->td_generation;
1684 	error = SYSCTL_OUT(req, arg1, arg2);
1685 	if (error)
1686 		return (error);
1687 	tries++;
1688 	if (generation != curthread->td_generation && tries < 3) {
1689 		*req = req2;
1690 		goto retry;
1691 	}
1692 
1693 	error = SYSCTL_IN(req, arg1, arg2);
1694 
1695 	return (error);
1696 }
1697 
1698 /*
1699  * Based on on sysctl_handle_int() convert microseconds to a sbintime.
1700  */
1701 int
1702 sysctl_usec_to_sbintime(SYSCTL_HANDLER_ARGS)
1703 {
1704 	int error;
1705 	int64_t tt;
1706 	sbintime_t sb;
1707 
1708 	tt = *(int64_t *)arg1;
1709 	sb = sbttous(tt);
1710 
1711 	error = sysctl_handle_64(oidp, &sb, 0, req);
1712 	if (error || !req->newptr)
1713 		return (error);
1714 
1715 	tt = ustosbt(sb);
1716 	*(int64_t *)arg1 = tt;
1717 
1718 	return (0);
1719 }
1720 
1721 /*
1722  * Based on on sysctl_handle_int() convert milliseconds to a sbintime.
1723  */
1724 int
1725 sysctl_msec_to_sbintime(SYSCTL_HANDLER_ARGS)
1726 {
1727 	int error;
1728 	int64_t tt;
1729 	sbintime_t sb;
1730 
1731 	tt = *(int64_t *)arg1;
1732 	sb = sbttoms(tt);
1733 
1734 	error = sysctl_handle_64(oidp, &sb, 0, req);
1735 	if (error || !req->newptr)
1736 		return (error);
1737 
1738 	tt = mstosbt(sb);
1739 	*(int64_t *)arg1 = tt;
1740 
1741 	return (0);
1742 }
1743 
1744 /*
1745  * Convert seconds to a struct timeval.  Intended for use with
1746  * intervals and thus does not permit negative seconds.
1747  */
1748 int
1749 sysctl_sec_to_timeval(SYSCTL_HANDLER_ARGS)
1750 {
1751 	struct timeval *tv;
1752 	int error, secs;
1753 
1754 	tv = arg1;
1755 	secs = tv->tv_sec;
1756 
1757 	error = sysctl_handle_int(oidp, &secs, 0, req);
1758 	if (error || req->newptr == NULL)
1759 		return (error);
1760 
1761 	if (secs < 0)
1762 		return (EINVAL);
1763 	tv->tv_sec = secs;
1764 
1765 	return (0);
1766 }
1767 
1768 /*
1769  * Transfer functions to/from kernel space.
1770  * XXX: rather untested at this point
1771  */
1772 static int
1773 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1774 {
1775 	size_t i = 0;
1776 
1777 	if (req->oldptr) {
1778 		i = l;
1779 		if (req->oldlen <= req->oldidx)
1780 			i = 0;
1781 		else
1782 			if (i > req->oldlen - req->oldidx)
1783 				i = req->oldlen - req->oldidx;
1784 		if (i > 0)
1785 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
1786 	}
1787 	req->oldidx += l;
1788 	if (req->oldptr && i != l)
1789 		return (ENOMEM);
1790 	return (0);
1791 }
1792 
1793 static int
1794 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1795 {
1796 	if (!req->newptr)
1797 		return (0);
1798 	if (req->newlen - req->newidx < l)
1799 		return (EINVAL);
1800 	bcopy((const char *)req->newptr + req->newidx, p, l);
1801 	req->newidx += l;
1802 	return (0);
1803 }
1804 
1805 int
1806 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1807     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1808 {
1809 	int error = 0;
1810 	struct sysctl_req req;
1811 
1812 	bzero(&req, sizeof req);
1813 
1814 	req.td = td;
1815 	req.flags = flags;
1816 
1817 	if (oldlenp) {
1818 		req.oldlen = *oldlenp;
1819 	}
1820 	req.validlen = req.oldlen;
1821 
1822 	if (old) {
1823 		req.oldptr= old;
1824 	}
1825 
1826 	if (new != NULL) {
1827 		req.newlen = newlen;
1828 		req.newptr = new;
1829 	}
1830 
1831 	req.oldfunc = sysctl_old_kernel;
1832 	req.newfunc = sysctl_new_kernel;
1833 	req.lock = REQ_UNWIRED;
1834 
1835 	error = sysctl_root(0, name, namelen, &req);
1836 
1837 	if (req.lock == REQ_WIRED && req.validlen > 0)
1838 		vsunlock(req.oldptr, req.validlen);
1839 
1840 	if (error && error != ENOMEM)
1841 		return (error);
1842 
1843 	if (retval) {
1844 		if (req.oldptr && req.oldidx > req.validlen)
1845 			*retval = req.validlen;
1846 		else
1847 			*retval = req.oldidx;
1848 	}
1849 	return (error);
1850 }
1851 
1852 int
1853 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1854     void *new, size_t newlen, size_t *retval, int flags)
1855 {
1856         int oid[CTL_MAXNAME];
1857         size_t oidlen, plen;
1858 	int error;
1859 
1860 	oid[0] = 0;		/* sysctl internal magic */
1861 	oid[1] = 3;		/* name2oid */
1862 	oidlen = sizeof(oid);
1863 
1864 	error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1865 	    (void *)name, strlen(name), &plen, flags);
1866 	if (error)
1867 		return (error);
1868 
1869 	error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1870 	    new, newlen, retval, flags);
1871 	return (error);
1872 }
1873 
1874 /*
1875  * Transfer function to/from user space.
1876  */
1877 static int
1878 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1879 {
1880 	size_t i, len, origidx;
1881 	int error;
1882 
1883 	origidx = req->oldidx;
1884 	req->oldidx += l;
1885 	if (req->oldptr == NULL)
1886 		return (0);
1887 	/*
1888 	 * If we have not wired the user supplied buffer and we are currently
1889 	 * holding locks, drop a witness warning, as it's possible that
1890 	 * write operations to the user page can sleep.
1891 	 */
1892 	if (req->lock != REQ_WIRED)
1893 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1894 		    "sysctl_old_user()");
1895 	i = l;
1896 	len = req->validlen;
1897 	if (len <= origidx)
1898 		i = 0;
1899 	else {
1900 		if (i > len - origidx)
1901 			i = len - origidx;
1902 		if (req->lock == REQ_WIRED) {
1903 			error = copyout_nofault(p, (char *)req->oldptr +
1904 			    origidx, i);
1905 		} else
1906 			error = copyout(p, (char *)req->oldptr + origidx, i);
1907 		if (error != 0)
1908 			return (error);
1909 	}
1910 	if (i < l)
1911 		return (ENOMEM);
1912 	return (0);
1913 }
1914 
1915 static int
1916 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1917 {
1918 	int error;
1919 
1920 	if (!req->newptr)
1921 		return (0);
1922 	if (req->newlen - req->newidx < l)
1923 		return (EINVAL);
1924 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1925 	    "sysctl_new_user()");
1926 	error = copyin((const char *)req->newptr + req->newidx, p, l);
1927 	req->newidx += l;
1928 	return (error);
1929 }
1930 
1931 /*
1932  * Wire the user space destination buffer.  If set to a value greater than
1933  * zero, the len parameter limits the maximum amount of wired memory.
1934  */
1935 int
1936 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1937 {
1938 	int ret;
1939 	size_t wiredlen;
1940 
1941 	wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1942 	ret = 0;
1943 	if (req->lock != REQ_WIRED && req->oldptr &&
1944 	    req->oldfunc == sysctl_old_user) {
1945 		if (wiredlen != 0) {
1946 			ret = vslock(req->oldptr, wiredlen);
1947 			if (ret != 0) {
1948 				if (ret != ENOMEM)
1949 					return (ret);
1950 				wiredlen = 0;
1951 			}
1952 		}
1953 		req->lock = REQ_WIRED;
1954 		req->validlen = wiredlen;
1955 	}
1956 	return (0);
1957 }
1958 
1959 int
1960 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1961     int *nindx, struct sysctl_req *req)
1962 {
1963 	struct sysctl_oid_list *lsp;
1964 	struct sysctl_oid *oid;
1965 	int indx;
1966 
1967 	SYSCTL_ASSERT_LOCKED();
1968 	lsp = &sysctl__children;
1969 	indx = 0;
1970 	while (indx < CTL_MAXNAME) {
1971 		SLIST_FOREACH(oid, lsp, oid_link) {
1972 			if (oid->oid_number == name[indx])
1973 				break;
1974 		}
1975 		if (oid == NULL)
1976 			return (ENOENT);
1977 
1978 		indx++;
1979 		if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1980 			if (oid->oid_handler != NULL || indx == namelen) {
1981 				*noid = oid;
1982 				if (nindx != NULL)
1983 					*nindx = indx;
1984 				KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1985 				    ("%s found DYING node %p", __func__, oid));
1986 				return (0);
1987 			}
1988 			lsp = SYSCTL_CHILDREN(oid);
1989 		} else if (indx == namelen) {
1990 			if ((oid->oid_kind & CTLFLAG_DORMANT) != 0)
1991 				return (ENOENT);
1992 			*noid = oid;
1993 			if (nindx != NULL)
1994 				*nindx = indx;
1995 			KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1996 			    ("%s found DYING node %p", __func__, oid));
1997 			return (0);
1998 		} else {
1999 			return (ENOTDIR);
2000 		}
2001 	}
2002 	return (ENOENT);
2003 }
2004 
2005 /*
2006  * Traverse our tree, and find the right node, execute whatever it points
2007  * to, and return the resulting error code.
2008  */
2009 
2010 static int
2011 sysctl_root(SYSCTL_HANDLER_ARGS)
2012 {
2013 	struct sysctl_oid *oid;
2014 	struct rm_priotracker tracker;
2015 	int error, indx, lvl;
2016 
2017 	SYSCTL_RLOCK(&tracker);
2018 
2019 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
2020 	if (error)
2021 		goto out;
2022 
2023 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2024 		/*
2025 		 * You can't call a sysctl when it's a node, but has
2026 		 * no handler.  Inform the user that it's a node.
2027 		 * The indx may or may not be the same as namelen.
2028 		 */
2029 		if (oid->oid_handler == NULL) {
2030 			error = EISDIR;
2031 			goto out;
2032 		}
2033 	}
2034 
2035 	/* Is this sysctl writable? */
2036 	if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) {
2037 		error = EPERM;
2038 		goto out;
2039 	}
2040 
2041 	KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
2042 
2043 #ifdef CAPABILITY_MODE
2044 	/*
2045 	 * If the process is in capability mode, then don't permit reading or
2046 	 * writing unless specifically granted for the node.
2047 	 */
2048 	if (IN_CAPABILITY_MODE(req->td)) {
2049 		if ((req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) ||
2050 		    (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))) {
2051 			error = EPERM;
2052 			goto out;
2053 		}
2054 	}
2055 #endif
2056 
2057 	/* Is this sysctl sensitive to securelevels? */
2058 	if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
2059 		lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
2060 		error = securelevel_gt(req->td->td_ucred, lvl);
2061 		if (error)
2062 			goto out;
2063 	}
2064 
2065 	/* Is this sysctl writable by only privileged users? */
2066 	if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
2067 		int priv;
2068 
2069 		if (oid->oid_kind & CTLFLAG_PRISON)
2070 			priv = PRIV_SYSCTL_WRITEJAIL;
2071 #ifdef VIMAGE
2072 		else if ((oid->oid_kind & CTLFLAG_VNET) &&
2073 		     prison_owns_vnet(req->td->td_ucred))
2074 			priv = PRIV_SYSCTL_WRITEJAIL;
2075 #endif
2076 		else
2077 			priv = PRIV_SYSCTL_WRITE;
2078 		error = priv_check(req->td, priv);
2079 		if (error)
2080 			goto out;
2081 	}
2082 
2083 	if (!oid->oid_handler) {
2084 		error = EINVAL;
2085 		goto out;
2086 	}
2087 
2088 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2089 		arg1 = (int *)arg1 + indx;
2090 		arg2 -= indx;
2091 	} else {
2092 		arg1 = oid->oid_arg1;
2093 		arg2 = oid->oid_arg2;
2094 	}
2095 #ifdef MAC
2096 	error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
2097 	    req);
2098 	if (error != 0)
2099 		goto out;
2100 #endif
2101 #ifdef VIMAGE
2102 	if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
2103 		arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
2104 #endif
2105 	error = sysctl_root_handler_locked(oid, arg1, arg2, req, &tracker);
2106 
2107 out:
2108 	SYSCTL_RUNLOCK(&tracker);
2109 	return (error);
2110 }
2111 
2112 #ifndef _SYS_SYSPROTO_H_
2113 struct sysctl_args {
2114 	int	*name;
2115 	u_int	namelen;
2116 	void	*old;
2117 	size_t	*oldlenp;
2118 	void	*new;
2119 	size_t	newlen;
2120 };
2121 #endif
2122 int
2123 sys___sysctl(struct thread *td, struct sysctl_args *uap)
2124 {
2125 	int error, i, name[CTL_MAXNAME];
2126 	size_t j;
2127 
2128 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
2129 		return (EINVAL);
2130 
2131  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
2132  	if (error)
2133 		return (error);
2134 
2135 	error = userland_sysctl(td, name, uap->namelen,
2136 		uap->old, uap->oldlenp, 0,
2137 		uap->new, uap->newlen, &j, 0);
2138 	if (error && error != ENOMEM)
2139 		return (error);
2140 	if (uap->oldlenp) {
2141 		i = copyout(&j, uap->oldlenp, sizeof(j));
2142 		if (i)
2143 			return (i);
2144 	}
2145 	return (error);
2146 }
2147 
2148 int
2149 kern___sysctlbyname(struct thread *td, const char *oname, size_t namelen,
2150     void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval,
2151     int flags, bool inkernel)
2152 {
2153 	int oid[CTL_MAXNAME];
2154 	char namebuf[16];
2155 	char *name;
2156 	size_t oidlen;
2157 	int error;
2158 
2159 	if (namelen > MAXPATHLEN || namelen == 0)
2160 		return (EINVAL);
2161 	name = namebuf;
2162 	if (namelen > sizeof(namebuf))
2163 		name = malloc(namelen, M_SYSCTL, M_WAITOK);
2164 	error = copyin(oname, name, namelen);
2165 	if (error != 0)
2166 		goto out;
2167 
2168 	oid[0] = 0;
2169 	oid[1] = 3;
2170 	oidlen = sizeof(oid);
2171 	error = kernel_sysctl(td, oid, 2, oid, &oidlen, (void *)name, namelen,
2172 	    retval, flags);
2173 	if (error != 0)
2174 		goto out;
2175 	error = userland_sysctl(td, oid, *retval / sizeof(int), old, oldlenp,
2176 	    inkernel, new, newlen, retval, flags);
2177 
2178 out:
2179 	if (namelen > sizeof(namebuf))
2180 		free(name, M_SYSCTL);
2181 	return (error);
2182 }
2183 
2184 #ifndef	_SYS_SYSPROTO_H_
2185 struct __sysctlbyname_args {
2186 	const char	*name;
2187 	size_t	namelen;
2188 	void	*old;
2189 	size_t	*oldlenp;
2190 	void	*new;
2191 	size_t	newlen;
2192 };
2193 #endif
2194 int
2195 sys___sysctlbyname(struct thread *td, struct __sysctlbyname_args *uap)
2196 {
2197 	size_t rv;
2198 	int error;
2199 
2200 	error = kern___sysctlbyname(td, uap->name, uap->namelen, uap->old,
2201 	    uap->oldlenp, uap->new, uap->newlen, &rv, 0, 0);
2202 	if (error != 0)
2203 		return (error);
2204 	if (uap->oldlenp != NULL)
2205 		error = copyout(&rv, uap->oldlenp, sizeof(rv));
2206 
2207 	return (error);
2208 }
2209 
2210 /*
2211  * This is used from various compatibility syscalls too.  That's why name
2212  * must be in kernel space.
2213  */
2214 int
2215 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
2216     size_t *oldlenp, int inkernel, const void *new, size_t newlen,
2217     size_t *retval, int flags)
2218 {
2219 	int error = 0, memlocked;
2220 	struct sysctl_req req;
2221 
2222 	bzero(&req, sizeof req);
2223 
2224 	req.td = td;
2225 	req.flags = flags;
2226 
2227 	if (oldlenp) {
2228 		if (inkernel) {
2229 			req.oldlen = *oldlenp;
2230 		} else {
2231 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
2232 			if (error)
2233 				return (error);
2234 		}
2235 	}
2236 	req.validlen = req.oldlen;
2237 	req.oldptr = old;
2238 
2239 	if (new != NULL) {
2240 		req.newlen = newlen;
2241 		req.newptr = new;
2242 	}
2243 
2244 	req.oldfunc = sysctl_old_user;
2245 	req.newfunc = sysctl_new_user;
2246 	req.lock = REQ_UNWIRED;
2247 
2248 #ifdef KTRACE
2249 	if (KTRPOINT(curthread, KTR_SYSCTL))
2250 		ktrsysctl(name, namelen);
2251 #endif
2252 	memlocked = 0;
2253 	if (req.oldptr && req.oldlen > 4 * PAGE_SIZE) {
2254 		memlocked = 1;
2255 		sx_xlock(&sysctlmemlock);
2256 	}
2257 	CURVNET_SET(TD_TO_VNET(td));
2258 
2259 	for (;;) {
2260 		req.oldidx = 0;
2261 		req.newidx = 0;
2262 		error = sysctl_root(0, name, namelen, &req);
2263 		if (error != EAGAIN)
2264 			break;
2265 		kern_yield(PRI_USER);
2266 	}
2267 
2268 	CURVNET_RESTORE();
2269 
2270 	if (req.lock == REQ_WIRED && req.validlen > 0)
2271 		vsunlock(req.oldptr, req.validlen);
2272 	if (memlocked)
2273 		sx_xunlock(&sysctlmemlock);
2274 
2275 	if (error && error != ENOMEM)
2276 		return (error);
2277 
2278 	if (retval) {
2279 		if (req.oldptr && req.oldidx > req.validlen)
2280 			*retval = req.validlen;
2281 		else
2282 			*retval = req.oldidx;
2283 	}
2284 	return (error);
2285 }
2286 
2287 /*
2288  * Drain into a sysctl struct.  The user buffer should be wired if a page
2289  * fault would cause issue.
2290  */
2291 static int
2292 sbuf_sysctl_drain(void *arg, const char *data, int len)
2293 {
2294 	struct sysctl_req *req = arg;
2295 	int error;
2296 
2297 	error = SYSCTL_OUT(req, data, len);
2298 	KASSERT(error >= 0, ("Got unexpected negative value %d", error));
2299 	return (error == 0 ? len : -error);
2300 }
2301 
2302 struct sbuf *
2303 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
2304     struct sysctl_req *req)
2305 {
2306 
2307 	/* Supply a default buffer size if none given. */
2308 	if (buf == NULL && length == 0)
2309 		length = 64;
2310 	s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
2311 	sbuf_set_drain(s, sbuf_sysctl_drain, req);
2312 	return (s);
2313 }
2314 
2315 #ifdef DDB
2316 
2317 /* The current OID the debugger is working with */
2318 static struct sysctl_oid *g_ddb_oid;
2319 
2320 /* The current flags specified by the user */
2321 static int g_ddb_sysctl_flags;
2322 
2323 /* Check to see if the last sysctl printed */
2324 static int g_ddb_sysctl_printed;
2325 
2326 static const int ctl_sign[CTLTYPE+1] = {
2327 	[CTLTYPE_INT] = 1,
2328 	[CTLTYPE_LONG] = 1,
2329 	[CTLTYPE_S8] = 1,
2330 	[CTLTYPE_S16] = 1,
2331 	[CTLTYPE_S32] = 1,
2332 	[CTLTYPE_S64] = 1,
2333 };
2334 
2335 static const int ctl_size[CTLTYPE+1] = {
2336 	[CTLTYPE_INT] = sizeof(int),
2337 	[CTLTYPE_UINT] = sizeof(u_int),
2338 	[CTLTYPE_LONG] = sizeof(long),
2339 	[CTLTYPE_ULONG] = sizeof(u_long),
2340 	[CTLTYPE_S8] = sizeof(int8_t),
2341 	[CTLTYPE_S16] = sizeof(int16_t),
2342 	[CTLTYPE_S32] = sizeof(int32_t),
2343 	[CTLTYPE_S64] = sizeof(int64_t),
2344 	[CTLTYPE_U8] = sizeof(uint8_t),
2345 	[CTLTYPE_U16] = sizeof(uint16_t),
2346 	[CTLTYPE_U32] = sizeof(uint32_t),
2347 	[CTLTYPE_U64] = sizeof(uint64_t),
2348 };
2349 
2350 #define DB_SYSCTL_NAME_ONLY	0x001	/* Compare with -N */
2351 #define DB_SYSCTL_VALUE_ONLY	0x002	/* Compare with -n */
2352 #define DB_SYSCTL_OPAQUE	0x004	/* Compare with -o */
2353 #define DB_SYSCTL_HEX		0x008	/* Compare with -x */
2354 
2355 #define DB_SYSCTL_SAFE_ONLY	0x100	/* Only simple types */
2356 
2357 static const char db_sysctl_modifs[] = {
2358 	'N', 'n', 'o', 'x',
2359 };
2360 
2361 static const int db_sysctl_modif_values[] = {
2362 	DB_SYSCTL_NAME_ONLY, DB_SYSCTL_VALUE_ONLY,
2363 	DB_SYSCTL_OPAQUE, DB_SYSCTL_HEX,
2364 };
2365 
2366 /* Handlers considered safe to print while recursing */
2367 static int (* const db_safe_handlers[])(SYSCTL_HANDLER_ARGS) = {
2368 	sysctl_handle_bool,
2369 	sysctl_handle_8,
2370 	sysctl_handle_16,
2371 	sysctl_handle_32,
2372 	sysctl_handle_64,
2373 	sysctl_handle_int,
2374 	sysctl_handle_long,
2375 	sysctl_handle_string,
2376 	sysctl_handle_opaque,
2377 };
2378 
2379 /*
2380  * Use in place of sysctl_old_kernel to print sysctl values.
2381  *
2382  * Compare to the output handling in show_var from sbin/sysctl/sysctl.c
2383  */
2384 static int
2385 sysctl_old_ddb(struct sysctl_req *req, const void *ptr, size_t len)
2386 {
2387 	const u_char *val, *p;
2388 	const char *sep1;
2389 	size_t intlen, slen;
2390 	uintmax_t umv;
2391 	intmax_t mv;
2392 	int sign, ctltype, hexlen, xflag, error;
2393 
2394 	/* Suppress false-positive GCC uninitialized variable warnings */
2395 	mv = 0;
2396 	umv = 0;
2397 
2398 	slen = len;
2399 	val = p = ptr;
2400 
2401 	if (ptr == NULL) {
2402 		error = 0;
2403 		goto out;
2404 	}
2405 
2406 	/* We are going to print */
2407 	g_ddb_sysctl_printed = 1;
2408 
2409 	xflag = g_ddb_sysctl_flags & DB_SYSCTL_HEX;
2410 
2411 	ctltype = (g_ddb_oid->oid_kind & CTLTYPE);
2412 	sign = ctl_sign[ctltype];
2413 	intlen = ctl_size[ctltype];
2414 
2415 	switch (ctltype) {
2416 	case CTLTYPE_NODE:
2417 	case CTLTYPE_STRING:
2418 		db_printf("%.*s", (int) len, (const char *) p);
2419 		error = 0;
2420 		goto out;
2421 
2422 	case CTLTYPE_INT:
2423 	case CTLTYPE_UINT:
2424 	case CTLTYPE_LONG:
2425 	case CTLTYPE_ULONG:
2426 	case CTLTYPE_S8:
2427 	case CTLTYPE_S16:
2428 	case CTLTYPE_S32:
2429 	case CTLTYPE_S64:
2430 	case CTLTYPE_U8:
2431 	case CTLTYPE_U16:
2432 	case CTLTYPE_U32:
2433 	case CTLTYPE_U64:
2434 		hexlen = 2 + (intlen * CHAR_BIT + 3) / 4;
2435 		sep1 = "";
2436 		while (len >= intlen) {
2437 			switch (ctltype) {
2438 			case CTLTYPE_INT:
2439 			case CTLTYPE_UINT:
2440 				umv = *(const u_int *)p;
2441 				mv = *(const int *)p;
2442 				break;
2443 			case CTLTYPE_LONG:
2444 			case CTLTYPE_ULONG:
2445 				umv = *(const u_long *)p;
2446 				mv = *(const long *)p;
2447 				break;
2448 			case CTLTYPE_S8:
2449 			case CTLTYPE_U8:
2450 				umv = *(const uint8_t *)p;
2451 				mv = *(const int8_t *)p;
2452 				break;
2453 			case CTLTYPE_S16:
2454 			case CTLTYPE_U16:
2455 				umv = *(const uint16_t *)p;
2456 				mv = *(const int16_t *)p;
2457 				break;
2458 			case CTLTYPE_S32:
2459 			case CTLTYPE_U32:
2460 				umv = *(const uint32_t *)p;
2461 				mv = *(const int32_t *)p;
2462 				break;
2463 			case CTLTYPE_S64:
2464 			case CTLTYPE_U64:
2465 				umv = *(const uint64_t *)p;
2466 				mv = *(const int64_t *)p;
2467 				break;
2468 			}
2469 
2470 			db_printf("%s", sep1);
2471 			if (xflag)
2472 				db_printf("%#0*jx", hexlen, umv);
2473 			else if (!sign)
2474 				db_printf("%ju", umv);
2475 			else if (g_ddb_oid->oid_fmt[1] == 'K') {
2476 				/* Kelvins are currently unsupported. */
2477 				error = EOPNOTSUPP;
2478 				goto out;
2479 			} else
2480 				db_printf("%jd", mv);
2481 
2482 			sep1 = " ";
2483 			len -= intlen;
2484 			p += intlen;
2485 		}
2486 		error = 0;
2487 		goto out;
2488 
2489 	case CTLTYPE_OPAQUE:
2490 		/* TODO: Support struct functions. */
2491 
2492 		/* FALLTHROUGH */
2493 	default:
2494 		db_printf("Format:%s Length:%zu Dump:0x",
2495 		    g_ddb_oid->oid_fmt, len);
2496 		while (len-- && (xflag || p < val + 16))
2497 			db_printf("%02x", *p++);
2498 		if (!xflag && len > 16)
2499 			db_printf("...");
2500 		error = 0;
2501 		goto out;
2502 	}
2503 
2504 out:
2505 	req->oldidx += slen;
2506 	return (error);
2507 }
2508 
2509 /*
2510  * Avoid setting new sysctl values from the debugger
2511  */
2512 static int
2513 sysctl_new_ddb(struct sysctl_req *req, void *p, size_t l)
2514 {
2515 
2516 	if (!req->newptr)
2517 		return (0);
2518 
2519 	/* Changing sysctls from the debugger is currently unsupported */
2520 	return (EPERM);
2521 }
2522 
2523 /*
2524  * Run a sysctl handler with the DDB oldfunc and newfunc attached.
2525  * Instead of copying any output to a buffer we'll dump it right to
2526  * the console.
2527  */
2528 static int
2529 db_sysctl(struct sysctl_oid *oidp, int *name, u_int namelen,
2530     void *old, size_t *oldlenp, size_t *retval, int flags)
2531 {
2532 	struct sysctl_req req;
2533 	int error;
2534 
2535 	/* Setup the request */
2536 	bzero(&req, sizeof req);
2537 	req.td = kdb_thread;
2538 	req.oldfunc = sysctl_old_ddb;
2539 	req.newfunc = sysctl_new_ddb;
2540 	req.lock = REQ_UNWIRED;
2541 	if (oldlenp) {
2542 		req.oldlen = *oldlenp;
2543 	}
2544 	req.validlen = req.oldlen;
2545 	if (old) {
2546 		req.oldptr = old;
2547 	}
2548 
2549 	/* Setup our globals for sysctl_old_ddb */
2550 	g_ddb_oid = oidp;
2551 	g_ddb_sysctl_flags = flags;
2552 	g_ddb_sysctl_printed = 0;
2553 
2554 	error = sysctl_root(0, name, namelen, &req);
2555 
2556 	/* Reset globals */
2557 	g_ddb_oid = NULL;
2558 	g_ddb_sysctl_flags = 0;
2559 
2560 	if (retval) {
2561 		if (req.oldptr && req.oldidx > req.validlen)
2562 			*retval = req.validlen;
2563 		else
2564 			*retval = req.oldidx;
2565 	}
2566 	return (error);
2567 }
2568 
2569 /*
2570  * Show a sysctl's name
2571  */
2572 static void
2573 db_show_oid_name(int *oid, size_t nlen)
2574 {
2575 	struct sysctl_oid *oidp;
2576 	int qoid[CTL_MAXNAME+2];
2577 	int error;
2578 
2579 	qoid[0] = 0;
2580 	memcpy(qoid + 2, oid, nlen * sizeof(int));
2581 	qoid[1] = 1;
2582 
2583 	error = sysctl_find_oid(qoid, nlen + 2, &oidp, NULL, NULL);
2584 	if (error)
2585 		db_error("sysctl name oid");
2586 
2587 	error = db_sysctl(oidp, qoid, nlen + 2, NULL, NULL, NULL, 0);
2588 	if (error)
2589 		db_error("sysctl name");
2590 }
2591 
2592 /*
2593  * Check to see if an OID is safe to print from ddb.
2594  */
2595 static bool
2596 db_oid_safe(const struct sysctl_oid *oidp)
2597 {
2598 	for (unsigned int i = 0; i < nitems(db_safe_handlers); ++i) {
2599 		if (oidp->oid_handler == db_safe_handlers[i])
2600 			return (true);
2601 	}
2602 
2603 	return (false);
2604 }
2605 
2606 /*
2607  * Show a sysctl at a specific OID
2608  * Compare to the input handling in show_var from sbin/sysctl/sysctl.c
2609  */
2610 static int
2611 db_show_oid(struct sysctl_oid *oidp, int *oid, size_t nlen, int flags)
2612 {
2613 	int error, xflag, oflag, Nflag, nflag;
2614 	size_t len;
2615 
2616 	xflag = flags & DB_SYSCTL_HEX;
2617 	oflag = flags & DB_SYSCTL_OPAQUE;
2618 	nflag = flags & DB_SYSCTL_VALUE_ONLY;
2619 	Nflag = flags & DB_SYSCTL_NAME_ONLY;
2620 
2621 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_OPAQUE &&
2622 	    (!xflag && !oflag))
2623 		return (0);
2624 
2625 	if (Nflag) {
2626 		db_show_oid_name(oid, nlen);
2627 		error = 0;
2628 		goto out;
2629 	}
2630 
2631 	if (!nflag) {
2632 		db_show_oid_name(oid, nlen);
2633 		db_printf(": ");
2634 	}
2635 
2636 	if ((flags & DB_SYSCTL_SAFE_ONLY) && !db_oid_safe(oidp)) {
2637 		db_printf("Skipping, unsafe to print while recursing.");
2638 		error = 0;
2639 		goto out;
2640 	}
2641 
2642 	/* Try once, and ask about the size */
2643 	len = 0;
2644 	error = db_sysctl(oidp, oid, nlen,
2645 	    NULL, NULL, &len, flags);
2646 	if (error)
2647 		goto out;
2648 
2649 	if (!g_ddb_sysctl_printed)
2650 		/* Lie about the size */
2651 		error = db_sysctl(oidp, oid, nlen,
2652 		    (void *) 1, &len, NULL, flags);
2653 
2654 out:
2655 	db_printf("\n");
2656 	return (error);
2657 }
2658 
2659 /*
2660  * Show all sysctls under a specific OID
2661  * Compare to sysctl_all from sbin/sysctl/sysctl.c
2662  */
2663 static int
2664 db_show_sysctl_all(int *oid, size_t len, int flags)
2665 {
2666 	struct sysctl_oid *oidp;
2667 	int name1[CTL_MAXNAME + 2], name2[CTL_MAXNAME + 2];
2668 	size_t l1, l2;
2669 
2670 	name1[0] = 0;
2671 	name1[1] = 2;
2672 	l1 = 2;
2673 	if (len) {
2674 		memcpy(name1+2, oid, len * sizeof(int));
2675 		l1 +=len;
2676 	} else {
2677 		name1[2] = 1;
2678 		l1++;
2679 	}
2680 	for (;;) {
2681 		int i, error;
2682 
2683 		l2 = sizeof(name2);
2684 		error = kernel_sysctl(kdb_thread, name1, l1,
2685 		    name2, &l2, NULL, 0, &l2, 0);
2686 		if (error != 0) {
2687 			if (error == ENOENT)
2688 				return (0);
2689 			else
2690 				db_error("sysctl(getnext)");
2691 		}
2692 
2693 		l2 /= sizeof(int);
2694 
2695 		if (l2 < (unsigned int)len)
2696 			return (0);
2697 
2698 		for (i = 0; i < len; i++)
2699 			if (name2[i] != oid[i])
2700 				return (0);
2701 
2702 		/* Find the OID in question */
2703 		error = sysctl_find_oid(name2, l2, &oidp, NULL, NULL);
2704 		if (error)
2705 			return (error);
2706 
2707 		i = db_show_oid(oidp, name2, l2, flags | DB_SYSCTL_SAFE_ONLY);
2708 
2709 		if (db_pager_quit)
2710 			return (0);
2711 
2712 		memcpy(name1+2, name2, l2 * sizeof(int));
2713 		l1 = 2 + l2;
2714 	}
2715 }
2716 
2717 /*
2718  * Show a sysctl by its user facing string
2719  */
2720 static int
2721 db_sysctlbyname(char *name, int flags)
2722 {
2723 	struct sysctl_oid *oidp;
2724 	int oid[CTL_MAXNAME];
2725 	int error, nlen;
2726 
2727 	error = name2oid(name, oid, &nlen, &oidp);
2728 	if (error) {
2729 		return (error);
2730 	}
2731 
2732 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2733 		db_show_sysctl_all(oid, nlen, flags);
2734 	} else {
2735 		error = db_show_oid(oidp, oid, nlen, flags);
2736 	}
2737 
2738 	return (error);
2739 }
2740 
2741 static void
2742 db_sysctl_cmd_usage(void)
2743 {
2744 	db_printf(
2745 	    " sysctl [/Nnox] <sysctl>					    \n"
2746 	    "								    \n"
2747 	    " <sysctl> The name of the sysctl to show.			    \n"
2748 	    "								    \n"
2749 	    " Show a sysctl by hooking into SYSCTL_IN and SYSCTL_OUT.	    \n"
2750 	    " This will work for most sysctls, but should not be used	    \n"
2751 	    " with sysctls that are known to malloc.			    \n"
2752 	    "								    \n"
2753 	    " While recursing any \"unsafe\" sysctls will be skipped.	    \n"
2754 	    " Call sysctl directly on the sysctl to try printing the	    \n"
2755 	    " skipped sysctl. This is unsafe and may make the ddb	    \n"
2756 	    " session unusable.						    \n"
2757 	    "								    \n"
2758 	    " Arguments:						    \n"
2759 	    "	/N	Display only the name of the sysctl.		    \n"
2760 	    "	/n	Display only the value of the sysctl.		    \n"
2761 	    "	/o	Display opaque values.				    \n"
2762 	    "	/x	Display the sysctl in hex.			    \n"
2763 	    "								    \n"
2764 	    "For example:						    \n"
2765 	    "sysctl vm.v_free_min					    \n"
2766 	    "vn.v_free_min: 12669					    \n"
2767 	    );
2768 }
2769 
2770 /*
2771  * Show a specific sysctl similar to sysctl (8).
2772  */
2773 DB_FUNC(sysctl, db_sysctl_cmd, db_cmd_table, CS_OWN, NULL)
2774 {
2775 	char name[TOK_STRING_SIZE];
2776 	int error, i, t, flags;
2777 
2778 	/* Parse the modifiers */
2779 	t = db_read_token();
2780 	if (t == tSLASH || t == tMINUS) {
2781 		t = db_read_token();
2782 		if (t != tIDENT) {
2783 			db_printf("Bad modifier\n");
2784 			error = EINVAL;
2785 			goto out;
2786 		}
2787 		db_strcpy(modif, db_tok_string);
2788 	}
2789 	else {
2790 		db_unread_token(t);
2791 		modif[0] = '\0';
2792 	}
2793 
2794 	flags = 0;
2795 	for (i = 0; i < nitems(db_sysctl_modifs); i++) {
2796 		if (strchr(modif, db_sysctl_modifs[i])) {
2797 			flags |= db_sysctl_modif_values[i];
2798 		}
2799 	}
2800 
2801 	/* Parse the sysctl names */
2802 	t = db_read_token();
2803 	if (t != tIDENT) {
2804 		db_printf("Need sysctl name\n");
2805 		error = EINVAL;
2806 		goto out;
2807 	}
2808 
2809 	/* Copy the name into a temporary buffer */
2810 	db_strcpy(name, db_tok_string);
2811 
2812 	/* Ensure there is no trailing cruft */
2813 	t = db_read_token();
2814 	if (t != tEOL) {
2815 		db_printf("Unexpected sysctl argument\n");
2816 		error = EINVAL;
2817 		goto out;
2818 	}
2819 
2820 	error = db_sysctlbyname(name, flags);
2821 	if (error == ENOENT) {
2822 		db_printf("unknown oid: '%s'\n", db_tok_string);
2823 		goto out;
2824 	} else if (error) {
2825 		db_printf("%s: error: %d\n", db_tok_string, error);
2826 		goto out;
2827 	}
2828 
2829 out:
2830 	/* Ensure we eat all of our text */
2831 	db_flush_lex();
2832 
2833 	if (error == EINVAL) {
2834 		db_sysctl_cmd_usage();
2835 	}
2836 }
2837 
2838 #endif /* DDB */
2839