xref: /freebsd/sys/kern/kern_sysctl.c (revision eb24e1491f9900e922c78e53af588f22a3e9535f)
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 #include "opt_sysctl.h"
47 
48 #include <sys/param.h>
49 #include <sys/fail.h>
50 #include <sys/systm.h>
51 #include <sys/capsicum.h>
52 #include <sys/kernel.h>
53 #include <sys/limits.h>
54 #include <sys/sysctl.h>
55 #include <sys/malloc.h>
56 #include <sys/priv.h>
57 #include <sys/proc.h>
58 #include <sys/jail.h>
59 #include <sys/kdb.h>
60 #include <sys/lock.h>
61 #include <sys/mutex.h>
62 #include <sys/rmlock.h>
63 #include <sys/sbuf.h>
64 #include <sys/sx.h>
65 #include <sys/sysproto.h>
66 #include <sys/uio.h>
67 #ifdef KTRACE
68 #include <sys/ktrace.h>
69 #endif
70 
71 #ifdef DDB
72 #include <ddb/ddb.h>
73 #include <ddb/db_lex.h>
74 #endif
75 
76 #include <net/vnet.h>
77 
78 #include <security/mac/mac_framework.h>
79 
80 #include <vm/vm.h>
81 #include <vm/vm_extern.h>
82 
83 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
84 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
85 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
86 
87 /*
88  * The sysctllock protects the MIB tree.  It also protects sysctl
89  * contexts used with dynamic sysctls.  The sysctl_register_oid() and
90  * sysctl_unregister_oid() routines require the sysctllock to already
91  * be held, so the sysctl_wlock() and sysctl_wunlock() routines are
92  * provided for the few places in the kernel which need to use that
93  * API rather than using the dynamic API.  Use of the dynamic API is
94  * strongly encouraged for most code.
95  *
96  * The sysctlmemlock is used to limit the amount of user memory wired for
97  * sysctl requests.  This is implemented by serializing any userland
98  * sysctl requests larger than a single page via an exclusive lock.
99  */
100 static struct rmlock sysctllock;
101 static struct sx __exclusive_cache_line sysctlmemlock;
102 
103 #define	SYSCTL_WLOCK()		rm_wlock(&sysctllock)
104 #define	SYSCTL_WUNLOCK()	rm_wunlock(&sysctllock)
105 #define	SYSCTL_RLOCK(tracker)	rm_rlock(&sysctllock, (tracker))
106 #define	SYSCTL_RUNLOCK(tracker)	rm_runlock(&sysctllock, (tracker))
107 #define	SYSCTL_WLOCKED()	rm_wowned(&sysctllock)
108 #define	SYSCTL_ASSERT_LOCKED()	rm_assert(&sysctllock, RA_LOCKED)
109 #define	SYSCTL_ASSERT_WLOCKED()	rm_assert(&sysctllock, RA_WLOCKED)
110 #define	SYSCTL_ASSERT_RLOCKED()	rm_assert(&sysctllock, RA_RLOCKED)
111 #define	SYSCTL_INIT()		rm_init_flags(&sysctllock, "sysctl lock", \
112 				    RM_SLEEPABLE)
113 #define	SYSCTL_SLEEP(ch, wmesg, timo)					\
114 				rm_sleep(ch, &sysctllock, 0, wmesg, timo)
115 
116 static int sysctl_root(SYSCTL_HANDLER_ARGS);
117 
118 /* Root list */
119 struct sysctl_oid_list sysctl__children = SLIST_HEAD_INITIALIZER(&sysctl__children);
120 
121 static int	sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
122 		    int recurse);
123 static int	sysctl_old_kernel(struct sysctl_req *, const void *, size_t);
124 static int	sysctl_new_kernel(struct sysctl_req *, void *, size_t);
125 
126 static struct sysctl_oid *
127 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
128 {
129 	struct sysctl_oid *oidp;
130 
131 	SYSCTL_ASSERT_LOCKED();
132 	SLIST_FOREACH(oidp, list, oid_link) {
133 		if (strcmp(oidp->oid_name, name) == 0) {
134 			return (oidp);
135 		}
136 	}
137 	return (NULL);
138 }
139 
140 /*
141  * Initialization of the MIB tree.
142  *
143  * Order by number in each list.
144  */
145 void
146 sysctl_wlock(void)
147 {
148 
149 	SYSCTL_WLOCK();
150 }
151 
152 void
153 sysctl_wunlock(void)
154 {
155 
156 	SYSCTL_WUNLOCK();
157 }
158 
159 static int
160 sysctl_root_handler_locked(struct sysctl_oid *oid, void *arg1, intmax_t arg2,
161     struct sysctl_req *req, struct rm_priotracker *tracker)
162 {
163 	int error;
164 
165 	if (oid->oid_kind & CTLFLAG_DYN)
166 		atomic_add_int(&oid->oid_running, 1);
167 
168 	if (tracker != NULL)
169 		SYSCTL_RUNLOCK(tracker);
170 	else
171 		SYSCTL_WUNLOCK();
172 
173 	if (!(oid->oid_kind & CTLFLAG_MPSAFE))
174 		mtx_lock(&Giant);
175 	error = oid->oid_handler(oid, arg1, arg2, req);
176 	if (!(oid->oid_kind & CTLFLAG_MPSAFE))
177 		mtx_unlock(&Giant);
178 
179 	KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
180 
181 	if (tracker != NULL)
182 		SYSCTL_RLOCK(tracker);
183 	else
184 		SYSCTL_WLOCK();
185 
186 	if (oid->oid_kind & CTLFLAG_DYN) {
187 		if (atomic_fetchadd_int(&oid->oid_running, -1) == 1 &&
188 		    (oid->oid_kind & CTLFLAG_DYING) != 0)
189 			wakeup(&oid->oid_running);
190 	}
191 
192 	return (error);
193 }
194 
195 static void
196 sysctl_load_tunable_by_oid_locked(struct sysctl_oid *oidp)
197 {
198 	struct sysctl_req req;
199 	struct sysctl_oid *curr;
200 	char *penv = NULL;
201 	char path[96];
202 	ssize_t rem = sizeof(path);
203 	ssize_t len;
204 	uint8_t data[512] __aligned(sizeof(uint64_t));
205 	int size;
206 	int error;
207 
208 	path[--rem] = 0;
209 
210 	for (curr = oidp; curr != NULL; curr = SYSCTL_PARENT(curr)) {
211 		len = strlen(curr->oid_name);
212 		rem -= len;
213 		if (curr != oidp)
214 			rem -= 1;
215 		if (rem < 0) {
216 			printf("OID path exceeds %d bytes\n", (int)sizeof(path));
217 			return;
218 		}
219 		memcpy(path + rem, curr->oid_name, len);
220 		if (curr != oidp)
221 			path[rem + len] = '.';
222 	}
223 
224 	memset(&req, 0, sizeof(req));
225 
226 	req.td = curthread;
227 	req.oldfunc = sysctl_old_kernel;
228 	req.newfunc = sysctl_new_kernel;
229 	req.lock = REQ_UNWIRED;
230 
231 	switch (oidp->oid_kind & CTLTYPE) {
232 	case CTLTYPE_INT:
233 		if (getenv_array(path + rem, data, sizeof(data), &size,
234 		    sizeof(int), GETENV_SIGNED) == 0)
235 			return;
236 		req.newlen = size;
237 		req.newptr = data;
238 		break;
239 	case CTLTYPE_UINT:
240 		if (getenv_array(path + rem, data, sizeof(data), &size,
241 		    sizeof(int), GETENV_UNSIGNED) == 0)
242 			return;
243 		req.newlen = size;
244 		req.newptr = data;
245 		break;
246 	case CTLTYPE_LONG:
247 		if (getenv_array(path + rem, data, sizeof(data), &size,
248 		    sizeof(long), GETENV_SIGNED) == 0)
249 			return;
250 		req.newlen = size;
251 		req.newptr = data;
252 		break;
253 	case CTLTYPE_ULONG:
254 		if (getenv_array(path + rem, data, sizeof(data), &size,
255 		    sizeof(long), GETENV_UNSIGNED) == 0)
256 			return;
257 		req.newlen = size;
258 		req.newptr = data;
259 		break;
260 	case CTLTYPE_S8:
261 		if (getenv_array(path + rem, data, sizeof(data), &size,
262 		    sizeof(int8_t), GETENV_SIGNED) == 0)
263 			return;
264 		req.newlen = size;
265 		req.newptr = data;
266 		break;
267 	case CTLTYPE_S16:
268 		if (getenv_array(path + rem, data, sizeof(data), &size,
269 		    sizeof(int16_t), GETENV_SIGNED) == 0)
270 			return;
271 		req.newlen = size;
272 		req.newptr = data;
273 		break;
274 	case CTLTYPE_S32:
275 		if (getenv_array(path + rem, data, sizeof(data), &size,
276 		    sizeof(int32_t), GETENV_SIGNED) == 0)
277 			return;
278 		req.newlen = size;
279 		req.newptr = data;
280 		break;
281 	case CTLTYPE_S64:
282 		if (getenv_array(path + rem, data, sizeof(data), &size,
283 		    sizeof(int64_t), GETENV_SIGNED) == 0)
284 			return;
285 		req.newlen = size;
286 		req.newptr = data;
287 		break;
288 	case CTLTYPE_U8:
289 		if (getenv_array(path + rem, data, sizeof(data), &size,
290 		    sizeof(uint8_t), GETENV_UNSIGNED) == 0)
291 			return;
292 		req.newlen = size;
293 		req.newptr = data;
294 		break;
295 	case CTLTYPE_U16:
296 		if (getenv_array(path + rem, data, sizeof(data), &size,
297 		    sizeof(uint16_t), GETENV_UNSIGNED) == 0)
298 			return;
299 		req.newlen = size;
300 		req.newptr = data;
301 		break;
302 	case CTLTYPE_U32:
303 		if (getenv_array(path + rem, data, sizeof(data), &size,
304 		    sizeof(uint32_t), GETENV_UNSIGNED) == 0)
305 			return;
306 		req.newlen = size;
307 		req.newptr = data;
308 		break;
309 	case CTLTYPE_U64:
310 		if (getenv_array(path + rem, data, sizeof(data), &size,
311 		    sizeof(uint64_t), GETENV_UNSIGNED) == 0)
312 			return;
313 		req.newlen = size;
314 		req.newptr = data;
315 		break;
316 	case CTLTYPE_STRING:
317 		penv = kern_getenv(path + rem);
318 		if (penv == NULL)
319 			return;
320 		req.newlen = strlen(penv);
321 		req.newptr = penv;
322 		break;
323 	default:
324 		return;
325 	}
326 	error = sysctl_root_handler_locked(oidp, oidp->oid_arg1,
327 	    oidp->oid_arg2, &req, NULL);
328 	if (error != 0)
329 		printf("Setting sysctl %s failed: %d\n", path + rem, error);
330 	if (penv != NULL)
331 		freeenv(penv);
332 }
333 
334 /*
335  * Locate the path to a given oid.  Returns the length of the resulting path,
336  * or -1 if the oid was not found.  nodes must have room for CTL_MAXNAME
337  * elements and be NULL initialized.
338  */
339 static int
340 sysctl_search_oid(struct sysctl_oid **nodes, struct sysctl_oid *needle)
341 {
342 	int indx;
343 
344 	SYSCTL_ASSERT_LOCKED();
345 	indx = 0;
346 	while (indx < CTL_MAXNAME && indx >= 0) {
347 		if (nodes[indx] == NULL && indx == 0)
348 			nodes[indx] = SLIST_FIRST(&sysctl__children);
349 		else if (nodes[indx] == NULL)
350 			nodes[indx] = SLIST_FIRST(&nodes[indx - 1]->oid_children);
351 		else
352 			nodes[indx] = SLIST_NEXT(nodes[indx], oid_link);
353 
354 		if (nodes[indx] == needle)
355 			return (indx + 1);
356 
357 		if (nodes[indx] == NULL) {
358 			indx--;
359 			continue;
360 		}
361 
362 		if ((nodes[indx]->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
363 			indx++;
364 			continue;
365 		}
366 	}
367 	return (-1);
368 }
369 
370 static void
371 sysctl_warn_reuse(const char *func, struct sysctl_oid *leaf)
372 {
373 	struct sysctl_oid *nodes[CTL_MAXNAME];
374 	char buf[128];
375 	struct sbuf sb;
376 	int rc, i;
377 
378 	(void)sbuf_new(&sb, buf, sizeof(buf), SBUF_FIXEDLEN | SBUF_INCLUDENUL);
379 	sbuf_set_drain(&sb, sbuf_printf_drain, NULL);
380 
381 	sbuf_printf(&sb, "%s: can't re-use a leaf (", __func__);
382 
383 	memset(nodes, 0, sizeof(nodes));
384 	rc = sysctl_search_oid(nodes, leaf);
385 	if (rc > 0) {
386 		for (i = 0; i < rc; i++)
387 			sbuf_printf(&sb, "%s%.*s", nodes[i]->oid_name,
388 			    i != (rc - 1), ".");
389 	} else {
390 		sbuf_printf(&sb, "%s", leaf->oid_name);
391 	}
392 	sbuf_printf(&sb, ")!\n");
393 
394 	(void)sbuf_finish(&sb);
395 }
396 
397 #ifdef SYSCTL_DEBUG
398 static int
399 sysctl_reuse_test(SYSCTL_HANDLER_ARGS)
400 {
401 	struct rm_priotracker tracker;
402 
403 	SYSCTL_RLOCK(&tracker);
404 	sysctl_warn_reuse(__func__, oidp);
405 	SYSCTL_RUNLOCK(&tracker);
406 	return (0);
407 }
408 SYSCTL_PROC(_sysctl, OID_AUTO, reuse_test,
409     CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0, sysctl_reuse_test, "-",
410     "");
411 #endif
412 
413 void
414 sysctl_register_oid(struct sysctl_oid *oidp)
415 {
416 	struct sysctl_oid_list *parent = oidp->oid_parent;
417 	struct sysctl_oid *p;
418 	struct sysctl_oid *q;
419 	int oid_number;
420 	int timeout = 2;
421 
422 	/*
423 	 * First check if another oid with the same name already
424 	 * exists in the parent's list.
425 	 */
426 	SYSCTL_ASSERT_WLOCKED();
427 	p = sysctl_find_oidname(oidp->oid_name, parent);
428 	if (p != NULL) {
429 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
430 			p->oid_refcnt++;
431 			return;
432 		} else {
433 			sysctl_warn_reuse(__func__, p);
434 			return;
435 		}
436 	}
437 	/* get current OID number */
438 	oid_number = oidp->oid_number;
439 
440 #if (OID_AUTO >= 0)
441 #error "OID_AUTO is expected to be a negative value"
442 #endif
443 	/*
444 	 * Any negative OID number qualifies as OID_AUTO. Valid OID
445 	 * numbers should always be positive.
446 	 *
447 	 * NOTE: DO NOT change the starting value here, change it in
448 	 * <sys/sysctl.h>, and make sure it is at least 256 to
449 	 * accommodate e.g. net.inet.raw as a static sysctl node.
450 	 */
451 	if (oid_number < 0) {
452 		static int newoid;
453 
454 		/*
455 		 * By decrementing the next OID number we spend less
456 		 * time inserting the OIDs into a sorted list.
457 		 */
458 		if (--newoid < CTL_AUTO_START)
459 			newoid = 0x7fffffff;
460 
461 		oid_number = newoid;
462 	}
463 
464 	/*
465 	 * Insert the OID into the parent's list sorted by OID number.
466 	 */
467 retry:
468 	q = NULL;
469 	SLIST_FOREACH(p, parent, oid_link) {
470 		/* check if the current OID number is in use */
471 		if (oid_number == p->oid_number) {
472 			/* get the next valid OID number */
473 			if (oid_number < CTL_AUTO_START ||
474 			    oid_number == 0x7fffffff) {
475 				/* wraparound - restart */
476 				oid_number = CTL_AUTO_START;
477 				/* don't loop forever */
478 				if (!timeout--)
479 					panic("sysctl: Out of OID numbers\n");
480 				goto retry;
481 			} else {
482 				oid_number++;
483 			}
484 		} else if (oid_number < p->oid_number)
485 			break;
486 		q = p;
487 	}
488 	/* check for non-auto OID number collision */
489 	if (oidp->oid_number >= 0 && oidp->oid_number < CTL_AUTO_START &&
490 	    oid_number >= CTL_AUTO_START) {
491 		printf("sysctl: OID number(%d) is already in use for '%s'\n",
492 		    oidp->oid_number, oidp->oid_name);
493 	}
494 	/* update the OID number, if any */
495 	oidp->oid_number = oid_number;
496 	if (q != NULL)
497 		SLIST_INSERT_AFTER(q, oidp, oid_link);
498 	else
499 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
500 
501 	if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE &&
502 #ifdef VIMAGE
503 	    (oidp->oid_kind & CTLFLAG_VNET) == 0 &&
504 #endif
505 	    (oidp->oid_kind & CTLFLAG_TUN) != 0 &&
506 	    (oidp->oid_kind & CTLFLAG_NOFETCH) == 0) {
507 		/* only fetch value once */
508 		oidp->oid_kind |= CTLFLAG_NOFETCH;
509 		/* try to fetch value from kernel environment */
510 		sysctl_load_tunable_by_oid_locked(oidp);
511 	}
512 }
513 
514 void
515 sysctl_register_disabled_oid(struct sysctl_oid *oidp)
516 {
517 
518 	/*
519 	 * Mark the leaf as dormant if it's not to be immediately enabled.
520 	 * We do not disable nodes as they can be shared between modules
521 	 * and it is always safe to access a node.
522 	 */
523 	KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
524 	    ("internal flag is set in oid_kind"));
525 	if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
526 		oidp->oid_kind |= CTLFLAG_DORMANT;
527 	sysctl_register_oid(oidp);
528 }
529 
530 void
531 sysctl_enable_oid(struct sysctl_oid *oidp)
532 {
533 
534 	SYSCTL_ASSERT_WLOCKED();
535 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
536 		KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
537 		    ("sysctl node is marked as dormant"));
538 		return;
539 	}
540 	KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) != 0,
541 	    ("enabling already enabled sysctl oid"));
542 	oidp->oid_kind &= ~CTLFLAG_DORMANT;
543 }
544 
545 void
546 sysctl_unregister_oid(struct sysctl_oid *oidp)
547 {
548 	struct sysctl_oid *p;
549 	int error;
550 
551 	SYSCTL_ASSERT_WLOCKED();
552 	if (oidp->oid_number == OID_AUTO) {
553 		error = EINVAL;
554 	} else {
555 		error = ENOENT;
556 		SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
557 			if (p == oidp) {
558 				SLIST_REMOVE(oidp->oid_parent, oidp,
559 				    sysctl_oid, oid_link);
560 				error = 0;
561 				break;
562 			}
563 		}
564 	}
565 
566 	/*
567 	 * This can happen when a module fails to register and is
568 	 * being unloaded afterwards.  It should not be a panic()
569 	 * for normal use.
570 	 */
571 	if (error) {
572 		printf("%s: failed(%d) to unregister sysctl(%s)\n",
573 		    __func__, error, oidp->oid_name);
574 	}
575 }
576 
577 /* Initialize a new context to keep track of dynamically added sysctls. */
578 int
579 sysctl_ctx_init(struct sysctl_ctx_list *c)
580 {
581 
582 	if (c == NULL) {
583 		return (EINVAL);
584 	}
585 
586 	/*
587 	 * No locking here, the caller is responsible for not adding
588 	 * new nodes to a context until after this function has
589 	 * returned.
590 	 */
591 	TAILQ_INIT(c);
592 	return (0);
593 }
594 
595 /* Free the context, and destroy all dynamic oids registered in this context */
596 int
597 sysctl_ctx_free(struct sysctl_ctx_list *clist)
598 {
599 	struct sysctl_ctx_entry *e, *e1;
600 	int error;
601 
602 	error = 0;
603 	/*
604 	 * First perform a "dry run" to check if it's ok to remove oids.
605 	 * XXX FIXME
606 	 * XXX This algorithm is a hack. But I don't know any
607 	 * XXX better solution for now...
608 	 */
609 	SYSCTL_WLOCK();
610 	TAILQ_FOREACH(e, clist, link) {
611 		error = sysctl_remove_oid_locked(e->entry, 0, 0);
612 		if (error)
613 			break;
614 	}
615 	/*
616 	 * Restore deregistered entries, either from the end,
617 	 * or from the place where error occurred.
618 	 * e contains the entry that was not unregistered
619 	 */
620 	if (error)
621 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
622 	else
623 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
624 	while (e1 != NULL) {
625 		sysctl_register_oid(e1->entry);
626 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
627 	}
628 	if (error) {
629 		SYSCTL_WUNLOCK();
630 		return(EBUSY);
631 	}
632 	/* Now really delete the entries */
633 	e = TAILQ_FIRST(clist);
634 	while (e != NULL) {
635 		e1 = TAILQ_NEXT(e, link);
636 		error = sysctl_remove_oid_locked(e->entry, 1, 0);
637 		if (error)
638 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
639 			    e->entry->oid_name);
640 		free(e, M_SYSCTLOID);
641 		e = e1;
642 	}
643 	SYSCTL_WUNLOCK();
644 	return (error);
645 }
646 
647 /* Add an entry to the context */
648 struct sysctl_ctx_entry *
649 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
650 {
651 	struct sysctl_ctx_entry *e;
652 
653 	SYSCTL_ASSERT_WLOCKED();
654 	if (clist == NULL || oidp == NULL)
655 		return(NULL);
656 	e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
657 	e->entry = oidp;
658 	TAILQ_INSERT_HEAD(clist, e, link);
659 	return (e);
660 }
661 
662 /* Find an entry in the context */
663 struct sysctl_ctx_entry *
664 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
665 {
666 	struct sysctl_ctx_entry *e;
667 
668 	SYSCTL_ASSERT_WLOCKED();
669 	if (clist == NULL || oidp == NULL)
670 		return(NULL);
671 	TAILQ_FOREACH(e, clist, link) {
672 		if(e->entry == oidp)
673 			return(e);
674 	}
675 	return (e);
676 }
677 
678 /*
679  * Delete an entry from the context.
680  * NOTE: this function doesn't free oidp! You have to remove it
681  * with sysctl_remove_oid().
682  */
683 int
684 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
685 {
686 	struct sysctl_ctx_entry *e;
687 
688 	if (clist == NULL || oidp == NULL)
689 		return (EINVAL);
690 	SYSCTL_WLOCK();
691 	e = sysctl_ctx_entry_find(clist, oidp);
692 	if (e != NULL) {
693 		TAILQ_REMOVE(clist, e, link);
694 		SYSCTL_WUNLOCK();
695 		free(e, M_SYSCTLOID);
696 		return (0);
697 	} else {
698 		SYSCTL_WUNLOCK();
699 		return (ENOENT);
700 	}
701 }
702 
703 /*
704  * Remove dynamically created sysctl trees.
705  * oidp - top of the tree to be removed
706  * del - if 0 - just deregister, otherwise free up entries as well
707  * recurse - if != 0 traverse the subtree to be deleted
708  */
709 int
710 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
711 {
712 	int error;
713 
714 	SYSCTL_WLOCK();
715 	error = sysctl_remove_oid_locked(oidp, del, recurse);
716 	SYSCTL_WUNLOCK();
717 	return (error);
718 }
719 
720 int
721 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
722     int del, int recurse)
723 {
724 	struct sysctl_oid *p, *tmp;
725 	int error;
726 
727 	error = ENOENT;
728 	SYSCTL_WLOCK();
729 	SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
730 		if (strcmp(p->oid_name, name) == 0) {
731 			error = sysctl_remove_oid_locked(p, del, recurse);
732 			break;
733 		}
734 	}
735 	SYSCTL_WUNLOCK();
736 
737 	return (error);
738 }
739 
740 static int
741 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
742 {
743 	struct sysctl_oid *p, *tmp;
744 	int error;
745 
746 	SYSCTL_ASSERT_WLOCKED();
747 	if (oidp == NULL)
748 		return(EINVAL);
749 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
750 		printf("Warning: can't remove non-dynamic nodes (%s)!\n",
751 		    oidp->oid_name);
752 		return (EINVAL);
753 	}
754 	/*
755 	 * WARNING: normal method to do this should be through
756 	 * sysctl_ctx_free(). Use recursing as the last resort
757 	 * method to purge your sysctl tree of leftovers...
758 	 * However, if some other code still references these nodes,
759 	 * it will panic.
760 	 */
761 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
762 		if (oidp->oid_refcnt == 1) {
763 			SLIST_FOREACH_SAFE(p,
764 			    SYSCTL_CHILDREN(oidp), oid_link, tmp) {
765 				if (!recurse) {
766 					printf("Warning: failed attempt to "
767 					    "remove oid %s with child %s\n",
768 					    oidp->oid_name, p->oid_name);
769 					return (ENOTEMPTY);
770 				}
771 				error = sysctl_remove_oid_locked(p, del,
772 				    recurse);
773 				if (error)
774 					return (error);
775 			}
776 		}
777 	}
778 	if (oidp->oid_refcnt > 1 ) {
779 		oidp->oid_refcnt--;
780 	} else {
781 		if (oidp->oid_refcnt == 0) {
782 			printf("Warning: bad oid_refcnt=%u (%s)!\n",
783 				oidp->oid_refcnt, oidp->oid_name);
784 			return (EINVAL);
785 		}
786 		sysctl_unregister_oid(oidp);
787 		if (del) {
788 			/*
789 			 * Wait for all threads running the handler to drain.
790 			 * This preserves the previous behavior when the
791 			 * sysctl lock was held across a handler invocation,
792 			 * and is necessary for module unload correctness.
793 			 */
794 			while (oidp->oid_running > 0) {
795 				oidp->oid_kind |= CTLFLAG_DYING;
796 				SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
797 			}
798 			if (oidp->oid_descr)
799 				free(__DECONST(char *, oidp->oid_descr),
800 				    M_SYSCTLOID);
801 			if (oidp->oid_label)
802 				free(__DECONST(char *, oidp->oid_label),
803 				    M_SYSCTLOID);
804 			free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
805 			free(oidp, M_SYSCTLOID);
806 		}
807 	}
808 	return (0);
809 }
810 /*
811  * Create new sysctls at run time.
812  * clist may point to a valid context initialized with sysctl_ctx_init().
813  */
814 struct sysctl_oid *
815 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
816 	int number, const char *name, int kind, void *arg1, intmax_t arg2,
817 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr,
818 	const char *label)
819 {
820 	struct sysctl_oid *oidp;
821 
822 	/* You have to hook up somewhere.. */
823 	if (parent == NULL)
824 		return(NULL);
825 	/* Check if the node already exists, otherwise create it */
826 	SYSCTL_WLOCK();
827 	oidp = sysctl_find_oidname(name, parent);
828 	if (oidp != NULL) {
829 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
830 			oidp->oid_refcnt++;
831 			/* Update the context */
832 			if (clist != NULL)
833 				sysctl_ctx_entry_add(clist, oidp);
834 			SYSCTL_WUNLOCK();
835 			return (oidp);
836 		} else {
837 			sysctl_warn_reuse(__func__, oidp);
838 			SYSCTL_WUNLOCK();
839 			return (NULL);
840 		}
841 	}
842 	oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
843 	oidp->oid_parent = parent;
844 	SLIST_INIT(&oidp->oid_children);
845 	oidp->oid_number = number;
846 	oidp->oid_refcnt = 1;
847 	oidp->oid_name = strdup(name, M_SYSCTLOID);
848 	oidp->oid_handler = handler;
849 	oidp->oid_kind = CTLFLAG_DYN | kind;
850 	oidp->oid_arg1 = arg1;
851 	oidp->oid_arg2 = arg2;
852 	oidp->oid_fmt = fmt;
853 	if (descr != NULL)
854 		oidp->oid_descr = strdup(descr, M_SYSCTLOID);
855 	if (label != NULL)
856 		oidp->oid_label = strdup(label, M_SYSCTLOID);
857 	/* Update the context, if used */
858 	if (clist != NULL)
859 		sysctl_ctx_entry_add(clist, oidp);
860 	/* Register this oid */
861 	sysctl_register_oid(oidp);
862 	SYSCTL_WUNLOCK();
863 	return (oidp);
864 }
865 
866 /*
867  * Rename an existing oid.
868  */
869 void
870 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
871 {
872 	char *newname;
873 	char *oldname;
874 
875 	newname = strdup(name, M_SYSCTLOID);
876 	SYSCTL_WLOCK();
877 	oldname = __DECONST(char *, oidp->oid_name);
878 	oidp->oid_name = newname;
879 	SYSCTL_WUNLOCK();
880 	free(oldname, M_SYSCTLOID);
881 }
882 
883 /*
884  * Reparent an existing oid.
885  */
886 int
887 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
888 {
889 	struct sysctl_oid *oidp;
890 
891 	SYSCTL_WLOCK();
892 	if (oid->oid_parent == parent) {
893 		SYSCTL_WUNLOCK();
894 		return (0);
895 	}
896 	oidp = sysctl_find_oidname(oid->oid_name, parent);
897 	if (oidp != NULL) {
898 		SYSCTL_WUNLOCK();
899 		return (EEXIST);
900 	}
901 	sysctl_unregister_oid(oid);
902 	oid->oid_parent = parent;
903 	oid->oid_number = OID_AUTO;
904 	sysctl_register_oid(oid);
905 	SYSCTL_WUNLOCK();
906 	return (0);
907 }
908 
909 /*
910  * Register the kernel's oids on startup.
911  */
912 SET_DECLARE(sysctl_set, struct sysctl_oid);
913 
914 static void
915 sysctl_register_all(void *arg)
916 {
917 	struct sysctl_oid **oidp;
918 
919 	sx_init(&sysctlmemlock, "sysctl mem");
920 	SYSCTL_INIT();
921 	SYSCTL_WLOCK();
922 	SET_FOREACH(oidp, sysctl_set)
923 		sysctl_register_oid(*oidp);
924 	SYSCTL_WUNLOCK();
925 }
926 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_FIRST, sysctl_register_all, NULL);
927 
928 /*
929  * "Staff-functions"
930  *
931  * These functions implement a presently undocumented interface
932  * used by the sysctl program to walk the tree, and get the type
933  * so it can print the value.
934  * This interface is under work and consideration, and should probably
935  * be killed with a big axe by the first person who can find the time.
936  * (be aware though, that the proper interface isn't as obvious as it
937  * may seem, there are various conflicting requirements.
938  *
939  * {CTL_SYSCTL, CTL_SYSCTL_DEBUG}		printf the entire MIB-tree.
940  * {CTL_SYSCTL, CTL_SYSCTL_NAME, ...}		return the name of the "..."
941  *						OID.
942  * {CTL_SYSCTL, CTL_SYSCTL_NEXT, ...}		return the next OID.
943  * {CTL_SYSCTL, CTL_SYSCTL_NAME2OID}		return the OID of the name in
944  *						"new"
945  * {CTL_SYSCTL, CTL_SYSCTL_OIDFMT, ...}		return the kind & format info
946  *						for the "..." OID.
947  * {CTL_SYSCTL, CTL_SYSCTL_OIDDESCR, ...}	return the description of the
948  *						"..." OID.
949  * {CTL_SYSCTL, CTL_SYSCTL_OIDLABEL, ...}	return the aggregation label of
950  *						the "..." OID.
951  */
952 
953 #ifdef SYSCTL_DEBUG
954 static void
955 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
956 {
957 	int k;
958 	struct sysctl_oid *oidp;
959 
960 	SYSCTL_ASSERT_LOCKED();
961 	SLIST_FOREACH(oidp, l, oid_link) {
962 
963 		for (k=0; k<i; k++)
964 			printf(" ");
965 
966 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
967 
968 		printf("%c%c",
969 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
970 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
971 
972 		if (oidp->oid_handler)
973 			printf(" *Handler");
974 
975 		switch (oidp->oid_kind & CTLTYPE) {
976 			case CTLTYPE_NODE:
977 				printf(" Node\n");
978 				if (!oidp->oid_handler) {
979 					sysctl_sysctl_debug_dump_node(
980 					    SYSCTL_CHILDREN(oidp), i + 2);
981 				}
982 				break;
983 			case CTLTYPE_INT:    printf(" Int\n"); break;
984 			case CTLTYPE_UINT:   printf(" u_int\n"); break;
985 			case CTLTYPE_LONG:   printf(" Long\n"); break;
986 			case CTLTYPE_ULONG:  printf(" u_long\n"); break;
987 			case CTLTYPE_STRING: printf(" String\n"); break;
988 			case CTLTYPE_S8:     printf(" int8_t\n"); break;
989 			case CTLTYPE_S16:    printf(" int16_t\n"); break;
990 			case CTLTYPE_S32:    printf(" int32_t\n"); break;
991 			case CTLTYPE_S64:    printf(" int64_t\n"); break;
992 			case CTLTYPE_U8:     printf(" uint8_t\n"); break;
993 			case CTLTYPE_U16:    printf(" uint16_t\n"); break;
994 			case CTLTYPE_U32:    printf(" uint32_t\n"); break;
995 			case CTLTYPE_U64:    printf(" uint64_t\n"); break;
996 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
997 			default:	     printf("\n");
998 		}
999 
1000 	}
1001 }
1002 
1003 static int
1004 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
1005 {
1006 	struct rm_priotracker tracker;
1007 	int error;
1008 
1009 	error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
1010 	if (error)
1011 		return (error);
1012 	SYSCTL_RLOCK(&tracker);
1013 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
1014 	SYSCTL_RUNLOCK(&tracker);
1015 	return (ENOENT);
1016 }
1017 
1018 SYSCTL_PROC(_sysctl, CTL_SYSCTL_DEBUG, debug, CTLTYPE_STRING | CTLFLAG_RD |
1019     CTLFLAG_MPSAFE, 0, 0, sysctl_sysctl_debug, "-", "");
1020 #endif
1021 
1022 static int
1023 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
1024 {
1025 	int *name = (int *) arg1;
1026 	u_int namelen = arg2;
1027 	int error;
1028 	struct sysctl_oid *oid;
1029 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
1030 	struct rm_priotracker tracker;
1031 	char buf[10];
1032 
1033 	error = sysctl_wire_old_buffer(req, 0);
1034 	if (error)
1035 		return (error);
1036 
1037 	SYSCTL_RLOCK(&tracker);
1038 	while (namelen) {
1039 		if (!lsp) {
1040 			snprintf(buf,sizeof(buf),"%d",*name);
1041 			if (req->oldidx)
1042 				error = SYSCTL_OUT(req, ".", 1);
1043 			if (!error)
1044 				error = SYSCTL_OUT(req, buf, strlen(buf));
1045 			if (error)
1046 				goto out;
1047 			namelen--;
1048 			name++;
1049 			continue;
1050 		}
1051 		lsp2 = NULL;
1052 		SLIST_FOREACH(oid, lsp, oid_link) {
1053 			if (oid->oid_number != *name)
1054 				continue;
1055 
1056 			if (req->oldidx)
1057 				error = SYSCTL_OUT(req, ".", 1);
1058 			if (!error)
1059 				error = SYSCTL_OUT(req, oid->oid_name,
1060 					strlen(oid->oid_name));
1061 			if (error)
1062 				goto out;
1063 
1064 			namelen--;
1065 			name++;
1066 
1067 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1068 				break;
1069 
1070 			if (oid->oid_handler)
1071 				break;
1072 
1073 			lsp2 = SYSCTL_CHILDREN(oid);
1074 			break;
1075 		}
1076 		lsp = lsp2;
1077 	}
1078 	error = SYSCTL_OUT(req, "", 1);
1079  out:
1080 	SYSCTL_RUNLOCK(&tracker);
1081 	return (error);
1082 }
1083 
1084 /*
1085  * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
1086  * capability mode.
1087  */
1088 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NAME, name, CTLFLAG_RD |
1089     CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_name, "");
1090 
1091 static int
1092 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
1093 	int *next, int *len, int level, struct sysctl_oid **oidpp)
1094 {
1095 	struct sysctl_oid *oidp;
1096 
1097 	SYSCTL_ASSERT_LOCKED();
1098 	*len = level;
1099 	SLIST_FOREACH(oidp, lsp, oid_link) {
1100 		*next = oidp->oid_number;
1101 		*oidpp = oidp;
1102 
1103 		if ((oidp->oid_kind & (CTLFLAG_SKIP | CTLFLAG_DORMANT)) != 0)
1104 			continue;
1105 
1106 		if (!namelen) {
1107 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1108 				return (0);
1109 			if (oidp->oid_handler)
1110 				/* We really should call the handler here...*/
1111 				return (0);
1112 			lsp = SYSCTL_CHILDREN(oidp);
1113 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
1114 				len, level+1, oidpp))
1115 				return (0);
1116 			goto emptynode;
1117 		}
1118 
1119 		if (oidp->oid_number < *name)
1120 			continue;
1121 
1122 		if (oidp->oid_number > *name) {
1123 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1124 				return (0);
1125 			if (oidp->oid_handler)
1126 				return (0);
1127 			lsp = SYSCTL_CHILDREN(oidp);
1128 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
1129 				next+1, len, level+1, oidpp))
1130 				return (0);
1131 			goto next;
1132 		}
1133 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1134 			continue;
1135 
1136 		if (oidp->oid_handler)
1137 			continue;
1138 
1139 		lsp = SYSCTL_CHILDREN(oidp);
1140 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
1141 			len, level+1, oidpp))
1142 			return (0);
1143 	next:
1144 		namelen = 1;
1145 	emptynode:
1146 		*len = level;
1147 	}
1148 	return (1);
1149 }
1150 
1151 static int
1152 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
1153 {
1154 	int *name = (int *) arg1;
1155 	u_int namelen = arg2;
1156 	int i, j, error;
1157 	struct sysctl_oid *oid;
1158 	struct sysctl_oid_list *lsp = &sysctl__children;
1159 	struct rm_priotracker tracker;
1160 	int newoid[CTL_MAXNAME];
1161 
1162 	SYSCTL_RLOCK(&tracker);
1163 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
1164 	SYSCTL_RUNLOCK(&tracker);
1165 	if (i)
1166 		return (ENOENT);
1167 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
1168 	return (error);
1169 }
1170 
1171 /*
1172  * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
1173  * capability mode.
1174  */
1175 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NEXT, next, CTLFLAG_RD |
1176     CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_next, "");
1177 
1178 static int
1179 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
1180 {
1181 	struct sysctl_oid *oidp;
1182 	struct sysctl_oid_list *lsp = &sysctl__children;
1183 	char *p;
1184 
1185 	SYSCTL_ASSERT_LOCKED();
1186 
1187 	for (*len = 0; *len < CTL_MAXNAME;) {
1188 		p = strsep(&name, ".");
1189 
1190 		oidp = SLIST_FIRST(lsp);
1191 		for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
1192 			if (oidp == NULL)
1193 				return (ENOENT);
1194 			if (strcmp(p, oidp->oid_name) == 0)
1195 				break;
1196 		}
1197 		*oid++ = oidp->oid_number;
1198 		(*len)++;
1199 
1200 		if (name == NULL || *name == '\0') {
1201 			if (oidpp)
1202 				*oidpp = oidp;
1203 			return (0);
1204 		}
1205 
1206 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1207 			break;
1208 
1209 		if (oidp->oid_handler)
1210 			break;
1211 
1212 		lsp = SYSCTL_CHILDREN(oidp);
1213 	}
1214 	return (ENOENT);
1215 }
1216 
1217 static int
1218 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
1219 {
1220 	char *p;
1221 	int error, oid[CTL_MAXNAME], len = 0;
1222 	struct sysctl_oid *op = NULL;
1223 	struct rm_priotracker tracker;
1224 	char buf[32];
1225 
1226 	if (!req->newlen)
1227 		return (ENOENT);
1228 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
1229 		return (ENAMETOOLONG);
1230 
1231 	p = buf;
1232 	if (req->newlen >= sizeof(buf))
1233 		p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
1234 
1235 	error = SYSCTL_IN(req, p, req->newlen);
1236 	if (error) {
1237 		if (p != buf)
1238 			free(p, M_SYSCTL);
1239 		return (error);
1240 	}
1241 
1242 	p [req->newlen] = '\0';
1243 
1244 	SYSCTL_RLOCK(&tracker);
1245 	error = name2oid(p, oid, &len, &op);
1246 	SYSCTL_RUNLOCK(&tracker);
1247 
1248 	if (p != buf)
1249 		free(p, M_SYSCTL);
1250 
1251 	if (error)
1252 		return (error);
1253 
1254 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
1255 	return (error);
1256 }
1257 
1258 /*
1259  * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
1260  * capability mode.
1261  */
1262 SYSCTL_PROC(_sysctl, CTL_SYSCTL_NAME2OID, name2oid, CTLTYPE_INT | CTLFLAG_RW |
1263     CTLFLAG_ANYBODY | CTLFLAG_MPSAFE | CTLFLAG_CAPRW, 0, 0,
1264     sysctl_sysctl_name2oid, "I", "");
1265 
1266 static int
1267 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
1268 {
1269 	struct sysctl_oid *oid;
1270 	struct rm_priotracker tracker;
1271 	int error;
1272 
1273 	error = sysctl_wire_old_buffer(req, 0);
1274 	if (error)
1275 		return (error);
1276 
1277 	SYSCTL_RLOCK(&tracker);
1278 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1279 	if (error)
1280 		goto out;
1281 
1282 	if (oid->oid_fmt == NULL) {
1283 		error = ENOENT;
1284 		goto out;
1285 	}
1286 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
1287 	if (error)
1288 		goto out;
1289 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
1290  out:
1291 	SYSCTL_RUNLOCK(&tracker);
1292 	return (error);
1293 }
1294 
1295 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDFMT, oidfmt, CTLFLAG_RD |
1296     CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidfmt, "");
1297 
1298 static int
1299 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
1300 {
1301 	struct sysctl_oid *oid;
1302 	struct rm_priotracker tracker;
1303 	int error;
1304 
1305 	error = sysctl_wire_old_buffer(req, 0);
1306 	if (error)
1307 		return (error);
1308 
1309 	SYSCTL_RLOCK(&tracker);
1310 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1311 	if (error)
1312 		goto out;
1313 
1314 	if (oid->oid_descr == NULL) {
1315 		error = ENOENT;
1316 		goto out;
1317 	}
1318 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
1319  out:
1320 	SYSCTL_RUNLOCK(&tracker);
1321 	return (error);
1322 }
1323 
1324 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDDESCR, oiddescr, CTLFLAG_RD |
1325     CTLFLAG_MPSAFE|CTLFLAG_CAPRD, sysctl_sysctl_oiddescr, "");
1326 
1327 static int
1328 sysctl_sysctl_oidlabel(SYSCTL_HANDLER_ARGS)
1329 {
1330 	struct sysctl_oid *oid;
1331 	struct rm_priotracker tracker;
1332 	int error;
1333 
1334 	error = sysctl_wire_old_buffer(req, 0);
1335 	if (error)
1336 		return (error);
1337 
1338 	SYSCTL_RLOCK(&tracker);
1339 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1340 	if (error)
1341 		goto out;
1342 
1343 	if (oid->oid_label == NULL) {
1344 		error = ENOENT;
1345 		goto out;
1346 	}
1347 	error = SYSCTL_OUT(req, oid->oid_label, strlen(oid->oid_label) + 1);
1348  out:
1349 	SYSCTL_RUNLOCK(&tracker);
1350 	return (error);
1351 }
1352 
1353 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDLABEL, oidlabel, CTLFLAG_RD |
1354     CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidlabel, "");
1355 
1356 /*
1357  * Default "handler" functions.
1358  */
1359 
1360 /*
1361  * Handle a bool.
1362  * Two cases:
1363  *     a variable:  point arg1 at it.
1364  *     a constant:  pass it in arg2.
1365  */
1366 
1367 int
1368 sysctl_handle_bool(SYSCTL_HANDLER_ARGS)
1369 {
1370 	uint8_t temp;
1371 	int error;
1372 
1373 	/*
1374 	 * Attempt to get a coherent snapshot by making a copy of the data.
1375 	 */
1376 	if (arg1)
1377 		temp = *(bool *)arg1 ? 1 : 0;
1378 	else
1379 		temp = arg2 ? 1 : 0;
1380 
1381 	error = SYSCTL_OUT(req, &temp, sizeof(temp));
1382 	if (error || !req->newptr)
1383 		return (error);
1384 
1385 	if (!arg1)
1386 		error = EPERM;
1387 	else {
1388 		error = SYSCTL_IN(req, &temp, sizeof(temp));
1389 		if (!error)
1390 			*(bool *)arg1 = temp ? 1 : 0;
1391 	}
1392 	return (error);
1393 }
1394 
1395 /*
1396  * Handle an int8_t, signed or unsigned.
1397  * Two cases:
1398  *     a variable:  point arg1 at it.
1399  *     a constant:  pass it in arg2.
1400  */
1401 
1402 int
1403 sysctl_handle_8(SYSCTL_HANDLER_ARGS)
1404 {
1405 	int8_t tmpout;
1406 	int error = 0;
1407 
1408 	/*
1409 	 * Attempt to get a coherent snapshot by making a copy of the data.
1410 	 */
1411 	if (arg1)
1412 		tmpout = *(int8_t *)arg1;
1413 	else
1414 		tmpout = arg2;
1415 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1416 
1417 	if (error || !req->newptr)
1418 		return (error);
1419 
1420 	if (!arg1)
1421 		error = EPERM;
1422 	else
1423 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1424 	return (error);
1425 }
1426 
1427 /*
1428  * Handle an int16_t, signed or unsigned.
1429  * Two cases:
1430  *     a variable:  point arg1 at it.
1431  *     a constant:  pass it in arg2.
1432  */
1433 
1434 int
1435 sysctl_handle_16(SYSCTL_HANDLER_ARGS)
1436 {
1437 	int16_t tmpout;
1438 	int error = 0;
1439 
1440 	/*
1441 	 * Attempt to get a coherent snapshot by making a copy of the data.
1442 	 */
1443 	if (arg1)
1444 		tmpout = *(int16_t *)arg1;
1445 	else
1446 		tmpout = arg2;
1447 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1448 
1449 	if (error || !req->newptr)
1450 		return (error);
1451 
1452 	if (!arg1)
1453 		error = EPERM;
1454 	else
1455 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1456 	return (error);
1457 }
1458 
1459 /*
1460  * Handle an int32_t, signed or unsigned.
1461  * Two cases:
1462  *     a variable:  point arg1 at it.
1463  *     a constant:  pass it in arg2.
1464  */
1465 
1466 int
1467 sysctl_handle_32(SYSCTL_HANDLER_ARGS)
1468 {
1469 	int32_t tmpout;
1470 	int error = 0;
1471 
1472 	/*
1473 	 * Attempt to get a coherent snapshot by making a copy of the data.
1474 	 */
1475 	if (arg1)
1476 		tmpout = *(int32_t *)arg1;
1477 	else
1478 		tmpout = arg2;
1479 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1480 
1481 	if (error || !req->newptr)
1482 		return (error);
1483 
1484 	if (!arg1)
1485 		error = EPERM;
1486 	else
1487 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1488 	return (error);
1489 }
1490 
1491 /*
1492  * Handle an int, signed or unsigned.
1493  * Two cases:
1494  *     a variable:  point arg1 at it.
1495  *     a constant:  pass it in arg2.
1496  */
1497 
1498 int
1499 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
1500 {
1501 	int tmpout, error = 0;
1502 
1503 	/*
1504 	 * Attempt to get a coherent snapshot by making a copy of the data.
1505 	 */
1506 	if (arg1)
1507 		tmpout = *(int *)arg1;
1508 	else
1509 		tmpout = arg2;
1510 	error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1511 
1512 	if (error || !req->newptr)
1513 		return (error);
1514 
1515 	if (!arg1)
1516 		error = EPERM;
1517 	else
1518 		error = SYSCTL_IN(req, arg1, sizeof(int));
1519 	return (error);
1520 }
1521 
1522 /*
1523  * Based on on sysctl_handle_int() convert milliseconds into ticks.
1524  * Note: this is used by TCP.
1525  */
1526 
1527 int
1528 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1529 {
1530 	int error, s, tt;
1531 
1532 	tt = *(int *)arg1;
1533 	s = (int)((int64_t)tt * 1000 / hz);
1534 
1535 	error = sysctl_handle_int(oidp, &s, 0, req);
1536 	if (error || !req->newptr)
1537 		return (error);
1538 
1539 	tt = (int)((int64_t)s * hz / 1000);
1540 	if (tt < 1)
1541 		return (EINVAL);
1542 
1543 	*(int *)arg1 = tt;
1544 	return (0);
1545 }
1546 
1547 /*
1548  * Handle a long, signed or unsigned.
1549  * Two cases:
1550  *     a variable:  point arg1 at it.
1551  *     a constant:  pass it in arg2.
1552  */
1553 
1554 int
1555 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1556 {
1557 	int error = 0;
1558 	long tmplong;
1559 #ifdef SCTL_MASK32
1560 	int tmpint;
1561 #endif
1562 
1563 	/*
1564 	 * Attempt to get a coherent snapshot by making a copy of the data.
1565 	 */
1566 	if (arg1)
1567 		tmplong = *(long *)arg1;
1568 	else
1569 		tmplong = arg2;
1570 #ifdef SCTL_MASK32
1571 	if (req->flags & SCTL_MASK32) {
1572 		tmpint = tmplong;
1573 		error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1574 	} else
1575 #endif
1576 		error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1577 
1578 	if (error || !req->newptr)
1579 		return (error);
1580 
1581 	if (!arg1)
1582 		error = EPERM;
1583 #ifdef SCTL_MASK32
1584 	else if (req->flags & SCTL_MASK32) {
1585 		error = SYSCTL_IN(req, &tmpint, sizeof(int));
1586 		*(long *)arg1 = (long)tmpint;
1587 	}
1588 #endif
1589 	else
1590 		error = SYSCTL_IN(req, arg1, sizeof(long));
1591 	return (error);
1592 }
1593 
1594 /*
1595  * Handle a 64 bit int, signed or unsigned.
1596  * Two cases:
1597  *     a variable:  point arg1 at it.
1598  *     a constant:  pass it in arg2.
1599  */
1600 int
1601 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1602 {
1603 	int error = 0;
1604 	uint64_t tmpout;
1605 
1606 	/*
1607 	 * Attempt to get a coherent snapshot by making a copy of the data.
1608 	 */
1609 	if (arg1)
1610 		tmpout = *(uint64_t *)arg1;
1611 	else
1612 		tmpout = arg2;
1613 	error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1614 
1615 	if (error || !req->newptr)
1616 		return (error);
1617 
1618 	if (!arg1)
1619 		error = EPERM;
1620 	else
1621 		error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1622 	return (error);
1623 }
1624 
1625 /*
1626  * Handle our generic '\0' terminated 'C' string.
1627  * Two cases:
1628  * 	a variable string:  point arg1 at it, arg2 is max length.
1629  * 	a constant string:  point arg1 at it, arg2 is zero.
1630  */
1631 
1632 int
1633 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1634 {
1635 	size_t outlen;
1636 	int error = 0, ro_string = 0;
1637 
1638 	/*
1639 	 * A zero-length buffer indicates a fixed size read-only
1640 	 * string.  In ddb, don't worry about trying to make a malloced
1641 	 * snapshot.
1642 	 */
1643 	if (arg2 == 0 || kdb_active) {
1644 		arg2 = strlen((char *)arg1) + 1;
1645 		ro_string = 1;
1646 	}
1647 
1648 	if (req->oldptr != NULL) {
1649 		char *tmparg;
1650 
1651 		if (ro_string) {
1652 			tmparg = arg1;
1653 		} else {
1654 			/* try to make a coherent snapshot of the string */
1655 			tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK);
1656 			memcpy(tmparg, arg1, arg2);
1657 		}
1658 
1659 		outlen = strnlen(tmparg, arg2 - 1) + 1;
1660 		error = SYSCTL_OUT(req, tmparg, outlen);
1661 
1662 		if (!ro_string)
1663 			free(tmparg, M_SYSCTLTMP);
1664 	} else {
1665 		outlen = strnlen((char *)arg1, arg2 - 1) + 1;
1666 		error = SYSCTL_OUT(req, NULL, outlen);
1667 	}
1668 	if (error || !req->newptr)
1669 		return (error);
1670 
1671 	if ((req->newlen - req->newidx) >= arg2) {
1672 		error = EINVAL;
1673 	} else {
1674 		arg2 = (req->newlen - req->newidx);
1675 		error = SYSCTL_IN(req, arg1, arg2);
1676 		((char *)arg1)[arg2] = '\0';
1677 	}
1678 	return (error);
1679 }
1680 
1681 /*
1682  * Handle any kind of opaque data.
1683  * arg1 points to it, arg2 is the size.
1684  */
1685 
1686 int
1687 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1688 {
1689 	int error, tries;
1690 	u_int generation;
1691 	struct sysctl_req req2;
1692 
1693 	/*
1694 	 * Attempt to get a coherent snapshot, by using the thread
1695 	 * pre-emption counter updated from within mi_switch() to
1696 	 * determine if we were pre-empted during a bcopy() or
1697 	 * copyout(). Make 3 attempts at doing this before giving up.
1698 	 * If we encounter an error, stop immediately.
1699 	 */
1700 	tries = 0;
1701 	req2 = *req;
1702 retry:
1703 	generation = curthread->td_generation;
1704 	error = SYSCTL_OUT(req, arg1, arg2);
1705 	if (error)
1706 		return (error);
1707 	tries++;
1708 	if (generation != curthread->td_generation && tries < 3) {
1709 		*req = req2;
1710 		goto retry;
1711 	}
1712 
1713 	error = SYSCTL_IN(req, arg1, arg2);
1714 
1715 	return (error);
1716 }
1717 
1718 /*
1719  * Based on on sysctl_handle_int() convert microseconds to a sbintime.
1720  */
1721 int
1722 sysctl_usec_to_sbintime(SYSCTL_HANDLER_ARGS)
1723 {
1724 	int error;
1725 	int64_t tt;
1726 	sbintime_t sb;
1727 
1728 	tt = *(int64_t *)arg1;
1729 	sb = sbttous(tt);
1730 
1731 	error = sysctl_handle_64(oidp, &sb, 0, req);
1732 	if (error || !req->newptr)
1733 		return (error);
1734 
1735 	tt = ustosbt(sb);
1736 	*(int64_t *)arg1 = tt;
1737 
1738 	return (0);
1739 }
1740 
1741 /*
1742  * Based on on sysctl_handle_int() convert milliseconds to a sbintime.
1743  */
1744 int
1745 sysctl_msec_to_sbintime(SYSCTL_HANDLER_ARGS)
1746 {
1747 	int error;
1748 	int64_t tt;
1749 	sbintime_t sb;
1750 
1751 	tt = *(int64_t *)arg1;
1752 	sb = sbttoms(tt);
1753 
1754 	error = sysctl_handle_64(oidp, &sb, 0, req);
1755 	if (error || !req->newptr)
1756 		return (error);
1757 
1758 	tt = mstosbt(sb);
1759 	*(int64_t *)arg1 = tt;
1760 
1761 	return (0);
1762 }
1763 
1764 /*
1765  * Convert seconds to a struct timeval.  Intended for use with
1766  * intervals and thus does not permit negative seconds.
1767  */
1768 int
1769 sysctl_sec_to_timeval(SYSCTL_HANDLER_ARGS)
1770 {
1771 	struct timeval *tv;
1772 	int error, secs;
1773 
1774 	tv = arg1;
1775 	secs = tv->tv_sec;
1776 
1777 	error = sysctl_handle_int(oidp, &secs, 0, req);
1778 	if (error || req->newptr == NULL)
1779 		return (error);
1780 
1781 	if (secs < 0)
1782 		return (EINVAL);
1783 	tv->tv_sec = secs;
1784 
1785 	return (0);
1786 }
1787 
1788 /*
1789  * Transfer functions to/from kernel space.
1790  * XXX: rather untested at this point
1791  */
1792 static int
1793 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1794 {
1795 	size_t i = 0;
1796 
1797 	if (req->oldptr) {
1798 		i = l;
1799 		if (req->oldlen <= req->oldidx)
1800 			i = 0;
1801 		else
1802 			if (i > req->oldlen - req->oldidx)
1803 				i = req->oldlen - req->oldidx;
1804 		if (i > 0)
1805 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
1806 	}
1807 	req->oldidx += l;
1808 	if (req->oldptr && i != l)
1809 		return (ENOMEM);
1810 	return (0);
1811 }
1812 
1813 static int
1814 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1815 {
1816 	if (!req->newptr)
1817 		return (0);
1818 	if (req->newlen - req->newidx < l)
1819 		return (EINVAL);
1820 	bcopy((const char *)req->newptr + req->newidx, p, l);
1821 	req->newidx += l;
1822 	return (0);
1823 }
1824 
1825 int
1826 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1827     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1828 {
1829 	int error = 0;
1830 	struct sysctl_req req;
1831 
1832 	bzero(&req, sizeof req);
1833 
1834 	req.td = td;
1835 	req.flags = flags;
1836 
1837 	if (oldlenp) {
1838 		req.oldlen = *oldlenp;
1839 	}
1840 	req.validlen = req.oldlen;
1841 
1842 	if (old) {
1843 		req.oldptr= old;
1844 	}
1845 
1846 	if (new != NULL) {
1847 		req.newlen = newlen;
1848 		req.newptr = new;
1849 	}
1850 
1851 	req.oldfunc = sysctl_old_kernel;
1852 	req.newfunc = sysctl_new_kernel;
1853 	req.lock = REQ_UNWIRED;
1854 
1855 	error = sysctl_root(0, name, namelen, &req);
1856 
1857 	if (req.lock == REQ_WIRED && req.validlen > 0)
1858 		vsunlock(req.oldptr, req.validlen);
1859 
1860 	if (error && error != ENOMEM)
1861 		return (error);
1862 
1863 	if (retval) {
1864 		if (req.oldptr && req.oldidx > req.validlen)
1865 			*retval = req.validlen;
1866 		else
1867 			*retval = req.oldidx;
1868 	}
1869 	return (error);
1870 }
1871 
1872 int
1873 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1874     void *new, size_t newlen, size_t *retval, int flags)
1875 {
1876         int oid[CTL_MAXNAME];
1877         size_t oidlen, plen;
1878 	int error;
1879 
1880 	oid[0] = CTL_SYSCTL;
1881 	oid[1] = CTL_SYSCTL_NAME2OID;
1882 	oidlen = sizeof(oid);
1883 
1884 	error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1885 	    (void *)name, strlen(name), &plen, flags);
1886 	if (error)
1887 		return (error);
1888 
1889 	error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1890 	    new, newlen, retval, flags);
1891 	return (error);
1892 }
1893 
1894 /*
1895  * Transfer function to/from user space.
1896  */
1897 static int
1898 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1899 {
1900 	size_t i, len, origidx;
1901 	int error;
1902 
1903 	origidx = req->oldidx;
1904 	req->oldidx += l;
1905 	if (req->oldptr == NULL)
1906 		return (0);
1907 	/*
1908 	 * If we have not wired the user supplied buffer and we are currently
1909 	 * holding locks, drop a witness warning, as it's possible that
1910 	 * write operations to the user page can sleep.
1911 	 */
1912 	if (req->lock != REQ_WIRED)
1913 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1914 		    "sysctl_old_user()");
1915 	i = l;
1916 	len = req->validlen;
1917 	if (len <= origidx)
1918 		i = 0;
1919 	else {
1920 		if (i > len - origidx)
1921 			i = len - origidx;
1922 		if (req->lock == REQ_WIRED) {
1923 			error = copyout_nofault(p, (char *)req->oldptr +
1924 			    origidx, i);
1925 		} else
1926 			error = copyout(p, (char *)req->oldptr + origidx, i);
1927 		if (error != 0)
1928 			return (error);
1929 	}
1930 	if (i < l)
1931 		return (ENOMEM);
1932 	return (0);
1933 }
1934 
1935 static int
1936 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1937 {
1938 	int error;
1939 
1940 	if (!req->newptr)
1941 		return (0);
1942 	if (req->newlen - req->newidx < l)
1943 		return (EINVAL);
1944 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1945 	    "sysctl_new_user()");
1946 	error = copyin((const char *)req->newptr + req->newidx, p, l);
1947 	req->newidx += l;
1948 	return (error);
1949 }
1950 
1951 /*
1952  * Wire the user space destination buffer.  If set to a value greater than
1953  * zero, the len parameter limits the maximum amount of wired memory.
1954  */
1955 int
1956 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1957 {
1958 	int ret;
1959 	size_t wiredlen;
1960 
1961 	wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1962 	ret = 0;
1963 	if (req->lock != REQ_WIRED && req->oldptr &&
1964 	    req->oldfunc == sysctl_old_user) {
1965 		if (wiredlen != 0) {
1966 			ret = vslock(req->oldptr, wiredlen);
1967 			if (ret != 0) {
1968 				if (ret != ENOMEM)
1969 					return (ret);
1970 				wiredlen = 0;
1971 			}
1972 		}
1973 		req->lock = REQ_WIRED;
1974 		req->validlen = wiredlen;
1975 	}
1976 	return (0);
1977 }
1978 
1979 int
1980 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1981     int *nindx, struct sysctl_req *req)
1982 {
1983 	struct sysctl_oid_list *lsp;
1984 	struct sysctl_oid *oid;
1985 	int indx;
1986 
1987 	SYSCTL_ASSERT_LOCKED();
1988 	lsp = &sysctl__children;
1989 	indx = 0;
1990 	while (indx < CTL_MAXNAME) {
1991 		SLIST_FOREACH(oid, lsp, oid_link) {
1992 			if (oid->oid_number == name[indx])
1993 				break;
1994 		}
1995 		if (oid == NULL)
1996 			return (ENOENT);
1997 
1998 		indx++;
1999 		if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2000 			if (oid->oid_handler != NULL || indx == namelen) {
2001 				*noid = oid;
2002 				if (nindx != NULL)
2003 					*nindx = indx;
2004 				KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
2005 				    ("%s found DYING node %p", __func__, oid));
2006 				return (0);
2007 			}
2008 			lsp = SYSCTL_CHILDREN(oid);
2009 		} else if (indx == namelen) {
2010 			if ((oid->oid_kind & CTLFLAG_DORMANT) != 0)
2011 				return (ENOENT);
2012 			*noid = oid;
2013 			if (nindx != NULL)
2014 				*nindx = indx;
2015 			KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
2016 			    ("%s found DYING node %p", __func__, oid));
2017 			return (0);
2018 		} else {
2019 			return (ENOTDIR);
2020 		}
2021 	}
2022 	return (ENOENT);
2023 }
2024 
2025 /*
2026  * Traverse our tree, and find the right node, execute whatever it points
2027  * to, and return the resulting error code.
2028  */
2029 
2030 static int
2031 sysctl_root(SYSCTL_HANDLER_ARGS)
2032 {
2033 	struct sysctl_oid *oid;
2034 	struct rm_priotracker tracker;
2035 	int error, indx, lvl;
2036 
2037 	SYSCTL_RLOCK(&tracker);
2038 
2039 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
2040 	if (error)
2041 		goto out;
2042 
2043 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2044 		/*
2045 		 * You can't call a sysctl when it's a node, but has
2046 		 * no handler.  Inform the user that it's a node.
2047 		 * The indx may or may not be the same as namelen.
2048 		 */
2049 		if (oid->oid_handler == NULL) {
2050 			error = EISDIR;
2051 			goto out;
2052 		}
2053 	}
2054 
2055 	/* Is this sysctl writable? */
2056 	if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) {
2057 		error = EPERM;
2058 		goto out;
2059 	}
2060 
2061 	KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
2062 
2063 #ifdef CAPABILITY_MODE
2064 	/*
2065 	 * If the process is in capability mode, then don't permit reading or
2066 	 * writing unless specifically granted for the node.
2067 	 */
2068 	if (IN_CAPABILITY_MODE(req->td)) {
2069 		if ((req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) ||
2070 		    (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))) {
2071 			error = EPERM;
2072 			goto out;
2073 		}
2074 	}
2075 #endif
2076 
2077 	/* Is this sysctl sensitive to securelevels? */
2078 	if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
2079 		lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
2080 		error = securelevel_gt(req->td->td_ucred, lvl);
2081 		if (error)
2082 			goto out;
2083 	}
2084 
2085 	/* Is this sysctl writable by only privileged users? */
2086 	if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
2087 		int priv;
2088 
2089 		if (oid->oid_kind & CTLFLAG_PRISON)
2090 			priv = PRIV_SYSCTL_WRITEJAIL;
2091 #ifdef VIMAGE
2092 		else if ((oid->oid_kind & CTLFLAG_VNET) &&
2093 		     prison_owns_vnet(req->td->td_ucred))
2094 			priv = PRIV_SYSCTL_WRITEJAIL;
2095 #endif
2096 		else
2097 			priv = PRIV_SYSCTL_WRITE;
2098 		error = priv_check(req->td, priv);
2099 		if (error)
2100 			goto out;
2101 	}
2102 
2103 	if (!oid->oid_handler) {
2104 		error = EINVAL;
2105 		goto out;
2106 	}
2107 
2108 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2109 		arg1 = (int *)arg1 + indx;
2110 		arg2 -= indx;
2111 	} else {
2112 		arg1 = oid->oid_arg1;
2113 		arg2 = oid->oid_arg2;
2114 	}
2115 #ifdef MAC
2116 	error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
2117 	    req);
2118 	if (error != 0)
2119 		goto out;
2120 #endif
2121 #ifdef VIMAGE
2122 	if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
2123 		arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
2124 #endif
2125 	error = sysctl_root_handler_locked(oid, arg1, arg2, req, &tracker);
2126 
2127 out:
2128 	SYSCTL_RUNLOCK(&tracker);
2129 	return (error);
2130 }
2131 
2132 #ifndef _SYS_SYSPROTO_H_
2133 struct sysctl_args {
2134 	int	*name;
2135 	u_int	namelen;
2136 	void	*old;
2137 	size_t	*oldlenp;
2138 	void	*new;
2139 	size_t	newlen;
2140 };
2141 #endif
2142 int
2143 sys___sysctl(struct thread *td, struct sysctl_args *uap)
2144 {
2145 	int error, i, name[CTL_MAXNAME];
2146 	size_t j;
2147 
2148 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
2149 		return (EINVAL);
2150 
2151  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
2152  	if (error)
2153 		return (error);
2154 
2155 	error = userland_sysctl(td, name, uap->namelen,
2156 		uap->old, uap->oldlenp, 0,
2157 		uap->new, uap->newlen, &j, 0);
2158 	if (error && error != ENOMEM)
2159 		return (error);
2160 	if (uap->oldlenp) {
2161 		i = copyout(&j, uap->oldlenp, sizeof(j));
2162 		if (i)
2163 			return (i);
2164 	}
2165 	return (error);
2166 }
2167 
2168 int
2169 kern___sysctlbyname(struct thread *td, const char *oname, size_t namelen,
2170     void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval,
2171     int flags, bool inkernel)
2172 {
2173 	int oid[CTL_MAXNAME];
2174 	char namebuf[16];
2175 	char *name;
2176 	size_t oidlen;
2177 	int error;
2178 
2179 	if (namelen > MAXPATHLEN || namelen == 0)
2180 		return (EINVAL);
2181 	name = namebuf;
2182 	if (namelen > sizeof(namebuf))
2183 		name = malloc(namelen, M_SYSCTL, M_WAITOK);
2184 	error = copyin(oname, name, namelen);
2185 	if (error != 0)
2186 		goto out;
2187 
2188 	oid[0] = CTL_SYSCTL;
2189 	oid[1] = CTL_SYSCTL_NAME2OID;
2190 	oidlen = sizeof(oid);
2191 	error = kernel_sysctl(td, oid, 2, oid, &oidlen, (void *)name, namelen,
2192 	    retval, flags);
2193 	if (error != 0)
2194 		goto out;
2195 	error = userland_sysctl(td, oid, *retval / sizeof(int), old, oldlenp,
2196 	    inkernel, new, newlen, retval, flags);
2197 
2198 out:
2199 	if (namelen > sizeof(namebuf))
2200 		free(name, M_SYSCTL);
2201 	return (error);
2202 }
2203 
2204 #ifndef	_SYS_SYSPROTO_H_
2205 struct __sysctlbyname_args {
2206 	const char	*name;
2207 	size_t	namelen;
2208 	void	*old;
2209 	size_t	*oldlenp;
2210 	void	*new;
2211 	size_t	newlen;
2212 };
2213 #endif
2214 int
2215 sys___sysctlbyname(struct thread *td, struct __sysctlbyname_args *uap)
2216 {
2217 	size_t rv;
2218 	int error;
2219 
2220 	error = kern___sysctlbyname(td, uap->name, uap->namelen, uap->old,
2221 	    uap->oldlenp, uap->new, uap->newlen, &rv, 0, 0);
2222 	if (error != 0)
2223 		return (error);
2224 	if (uap->oldlenp != NULL)
2225 		error = copyout(&rv, uap->oldlenp, sizeof(rv));
2226 
2227 	return (error);
2228 }
2229 
2230 /*
2231  * This is used from various compatibility syscalls too.  That's why name
2232  * must be in kernel space.
2233  */
2234 int
2235 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
2236     size_t *oldlenp, int inkernel, const void *new, size_t newlen,
2237     size_t *retval, int flags)
2238 {
2239 	int error = 0, memlocked;
2240 	struct sysctl_req req;
2241 
2242 	bzero(&req, sizeof req);
2243 
2244 	req.td = td;
2245 	req.flags = flags;
2246 
2247 	if (oldlenp) {
2248 		if (inkernel) {
2249 			req.oldlen = *oldlenp;
2250 		} else {
2251 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
2252 			if (error)
2253 				return (error);
2254 		}
2255 	}
2256 	req.validlen = req.oldlen;
2257 	req.oldptr = old;
2258 
2259 	if (new != NULL) {
2260 		req.newlen = newlen;
2261 		req.newptr = new;
2262 	}
2263 
2264 	req.oldfunc = sysctl_old_user;
2265 	req.newfunc = sysctl_new_user;
2266 	req.lock = REQ_UNWIRED;
2267 
2268 #ifdef KTRACE
2269 	if (KTRPOINT(curthread, KTR_SYSCTL))
2270 		ktrsysctl(name, namelen);
2271 #endif
2272 	memlocked = 0;
2273 	if (req.oldptr && req.oldlen > 4 * PAGE_SIZE) {
2274 		memlocked = 1;
2275 		sx_xlock(&sysctlmemlock);
2276 	}
2277 	CURVNET_SET(TD_TO_VNET(td));
2278 
2279 	for (;;) {
2280 		req.oldidx = 0;
2281 		req.newidx = 0;
2282 		error = sysctl_root(0, name, namelen, &req);
2283 		if (error != EAGAIN)
2284 			break;
2285 		kern_yield(PRI_USER);
2286 	}
2287 
2288 	CURVNET_RESTORE();
2289 
2290 	if (req.lock == REQ_WIRED && req.validlen > 0)
2291 		vsunlock(req.oldptr, req.validlen);
2292 	if (memlocked)
2293 		sx_xunlock(&sysctlmemlock);
2294 
2295 	if (error && error != ENOMEM)
2296 		return (error);
2297 
2298 	if (retval) {
2299 		if (req.oldptr && req.oldidx > req.validlen)
2300 			*retval = req.validlen;
2301 		else
2302 			*retval = req.oldidx;
2303 	}
2304 	return (error);
2305 }
2306 
2307 /*
2308  * Drain into a sysctl struct.  The user buffer should be wired if a page
2309  * fault would cause issue.
2310  */
2311 static int
2312 sbuf_sysctl_drain(void *arg, const char *data, int len)
2313 {
2314 	struct sysctl_req *req = arg;
2315 	int error;
2316 
2317 	error = SYSCTL_OUT(req, data, len);
2318 	KASSERT(error >= 0, ("Got unexpected negative value %d", error));
2319 	return (error == 0 ? len : -error);
2320 }
2321 
2322 struct sbuf *
2323 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
2324     struct sysctl_req *req)
2325 {
2326 
2327 	/* Supply a default buffer size if none given. */
2328 	if (buf == NULL && length == 0)
2329 		length = 64;
2330 	s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
2331 	sbuf_set_drain(s, sbuf_sysctl_drain, req);
2332 	return (s);
2333 }
2334 
2335 #ifdef DDB
2336 
2337 /* The current OID the debugger is working with */
2338 static struct sysctl_oid *g_ddb_oid;
2339 
2340 /* The current flags specified by the user */
2341 static int g_ddb_sysctl_flags;
2342 
2343 /* Check to see if the last sysctl printed */
2344 static int g_ddb_sysctl_printed;
2345 
2346 static const int ctl_sign[CTLTYPE+1] = {
2347 	[CTLTYPE_INT] = 1,
2348 	[CTLTYPE_LONG] = 1,
2349 	[CTLTYPE_S8] = 1,
2350 	[CTLTYPE_S16] = 1,
2351 	[CTLTYPE_S32] = 1,
2352 	[CTLTYPE_S64] = 1,
2353 };
2354 
2355 static const int ctl_size[CTLTYPE+1] = {
2356 	[CTLTYPE_INT] = sizeof(int),
2357 	[CTLTYPE_UINT] = sizeof(u_int),
2358 	[CTLTYPE_LONG] = sizeof(long),
2359 	[CTLTYPE_ULONG] = sizeof(u_long),
2360 	[CTLTYPE_S8] = sizeof(int8_t),
2361 	[CTLTYPE_S16] = sizeof(int16_t),
2362 	[CTLTYPE_S32] = sizeof(int32_t),
2363 	[CTLTYPE_S64] = sizeof(int64_t),
2364 	[CTLTYPE_U8] = sizeof(uint8_t),
2365 	[CTLTYPE_U16] = sizeof(uint16_t),
2366 	[CTLTYPE_U32] = sizeof(uint32_t),
2367 	[CTLTYPE_U64] = sizeof(uint64_t),
2368 };
2369 
2370 #define DB_SYSCTL_NAME_ONLY	0x001	/* Compare with -N */
2371 #define DB_SYSCTL_VALUE_ONLY	0x002	/* Compare with -n */
2372 #define DB_SYSCTL_OPAQUE	0x004	/* Compare with -o */
2373 #define DB_SYSCTL_HEX		0x008	/* Compare with -x */
2374 
2375 #define DB_SYSCTL_SAFE_ONLY	0x100	/* Only simple types */
2376 
2377 static const char db_sysctl_modifs[] = {
2378 	'N', 'n', 'o', 'x',
2379 };
2380 
2381 static const int db_sysctl_modif_values[] = {
2382 	DB_SYSCTL_NAME_ONLY, DB_SYSCTL_VALUE_ONLY,
2383 	DB_SYSCTL_OPAQUE, DB_SYSCTL_HEX,
2384 };
2385 
2386 /* Handlers considered safe to print while recursing */
2387 static int (* const db_safe_handlers[])(SYSCTL_HANDLER_ARGS) = {
2388 	sysctl_handle_bool,
2389 	sysctl_handle_8,
2390 	sysctl_handle_16,
2391 	sysctl_handle_32,
2392 	sysctl_handle_64,
2393 	sysctl_handle_int,
2394 	sysctl_handle_long,
2395 	sysctl_handle_string,
2396 	sysctl_handle_opaque,
2397 };
2398 
2399 /*
2400  * Use in place of sysctl_old_kernel to print sysctl values.
2401  *
2402  * Compare to the output handling in show_var from sbin/sysctl/sysctl.c
2403  */
2404 static int
2405 sysctl_old_ddb(struct sysctl_req *req, const void *ptr, size_t len)
2406 {
2407 	const u_char *val, *p;
2408 	const char *sep1;
2409 	size_t intlen, slen;
2410 	uintmax_t umv;
2411 	intmax_t mv;
2412 	int sign, ctltype, hexlen, xflag, error;
2413 
2414 	/* Suppress false-positive GCC uninitialized variable warnings */
2415 	mv = 0;
2416 	umv = 0;
2417 
2418 	slen = len;
2419 	val = p = ptr;
2420 
2421 	if (ptr == NULL) {
2422 		error = 0;
2423 		goto out;
2424 	}
2425 
2426 	/* We are going to print */
2427 	g_ddb_sysctl_printed = 1;
2428 
2429 	xflag = g_ddb_sysctl_flags & DB_SYSCTL_HEX;
2430 
2431 	ctltype = (g_ddb_oid->oid_kind & CTLTYPE);
2432 	sign = ctl_sign[ctltype];
2433 	intlen = ctl_size[ctltype];
2434 
2435 	switch (ctltype) {
2436 	case CTLTYPE_NODE:
2437 	case CTLTYPE_STRING:
2438 		db_printf("%.*s", (int) len, (const char *) p);
2439 		error = 0;
2440 		goto out;
2441 
2442 	case CTLTYPE_INT:
2443 	case CTLTYPE_UINT:
2444 	case CTLTYPE_LONG:
2445 	case CTLTYPE_ULONG:
2446 	case CTLTYPE_S8:
2447 	case CTLTYPE_S16:
2448 	case CTLTYPE_S32:
2449 	case CTLTYPE_S64:
2450 	case CTLTYPE_U8:
2451 	case CTLTYPE_U16:
2452 	case CTLTYPE_U32:
2453 	case CTLTYPE_U64:
2454 		hexlen = 2 + (intlen * CHAR_BIT + 3) / 4;
2455 		sep1 = "";
2456 		while (len >= intlen) {
2457 			switch (ctltype) {
2458 			case CTLTYPE_INT:
2459 			case CTLTYPE_UINT:
2460 				umv = *(const u_int *)p;
2461 				mv = *(const int *)p;
2462 				break;
2463 			case CTLTYPE_LONG:
2464 			case CTLTYPE_ULONG:
2465 				umv = *(const u_long *)p;
2466 				mv = *(const long *)p;
2467 				break;
2468 			case CTLTYPE_S8:
2469 			case CTLTYPE_U8:
2470 				umv = *(const uint8_t *)p;
2471 				mv = *(const int8_t *)p;
2472 				break;
2473 			case CTLTYPE_S16:
2474 			case CTLTYPE_U16:
2475 				umv = *(const uint16_t *)p;
2476 				mv = *(const int16_t *)p;
2477 				break;
2478 			case CTLTYPE_S32:
2479 			case CTLTYPE_U32:
2480 				umv = *(const uint32_t *)p;
2481 				mv = *(const int32_t *)p;
2482 				break;
2483 			case CTLTYPE_S64:
2484 			case CTLTYPE_U64:
2485 				umv = *(const uint64_t *)p;
2486 				mv = *(const int64_t *)p;
2487 				break;
2488 			}
2489 
2490 			db_printf("%s", sep1);
2491 			if (xflag)
2492 				db_printf("%#0*jx", hexlen, umv);
2493 			else if (!sign)
2494 				db_printf("%ju", umv);
2495 			else if (g_ddb_oid->oid_fmt[1] == 'K') {
2496 				/* Kelvins are currently unsupported. */
2497 				error = EOPNOTSUPP;
2498 				goto out;
2499 			} else
2500 				db_printf("%jd", mv);
2501 
2502 			sep1 = " ";
2503 			len -= intlen;
2504 			p += intlen;
2505 		}
2506 		error = 0;
2507 		goto out;
2508 
2509 	case CTLTYPE_OPAQUE:
2510 		/* TODO: Support struct functions. */
2511 
2512 		/* FALLTHROUGH */
2513 	default:
2514 		db_printf("Format:%s Length:%zu Dump:0x",
2515 		    g_ddb_oid->oid_fmt, len);
2516 		while (len-- && (xflag || p < val + 16))
2517 			db_printf("%02x", *p++);
2518 		if (!xflag && len > 16)
2519 			db_printf("...");
2520 		error = 0;
2521 		goto out;
2522 	}
2523 
2524 out:
2525 	req->oldidx += slen;
2526 	return (error);
2527 }
2528 
2529 /*
2530  * Avoid setting new sysctl values from the debugger
2531  */
2532 static int
2533 sysctl_new_ddb(struct sysctl_req *req, void *p, size_t l)
2534 {
2535 
2536 	if (!req->newptr)
2537 		return (0);
2538 
2539 	/* Changing sysctls from the debugger is currently unsupported */
2540 	return (EPERM);
2541 }
2542 
2543 /*
2544  * Run a sysctl handler with the DDB oldfunc and newfunc attached.
2545  * Instead of copying any output to a buffer we'll dump it right to
2546  * the console.
2547  */
2548 static int
2549 db_sysctl(struct sysctl_oid *oidp, int *name, u_int namelen,
2550     void *old, size_t *oldlenp, size_t *retval, int flags)
2551 {
2552 	struct sysctl_req req;
2553 	int error;
2554 
2555 	/* Setup the request */
2556 	bzero(&req, sizeof req);
2557 	req.td = kdb_thread;
2558 	req.oldfunc = sysctl_old_ddb;
2559 	req.newfunc = sysctl_new_ddb;
2560 	req.lock = REQ_UNWIRED;
2561 	if (oldlenp) {
2562 		req.oldlen = *oldlenp;
2563 	}
2564 	req.validlen = req.oldlen;
2565 	if (old) {
2566 		req.oldptr = old;
2567 	}
2568 
2569 	/* Setup our globals for sysctl_old_ddb */
2570 	g_ddb_oid = oidp;
2571 	g_ddb_sysctl_flags = flags;
2572 	g_ddb_sysctl_printed = 0;
2573 
2574 	error = sysctl_root(0, name, namelen, &req);
2575 
2576 	/* Reset globals */
2577 	g_ddb_oid = NULL;
2578 	g_ddb_sysctl_flags = 0;
2579 
2580 	if (retval) {
2581 		if (req.oldptr && req.oldidx > req.validlen)
2582 			*retval = req.validlen;
2583 		else
2584 			*retval = req.oldidx;
2585 	}
2586 	return (error);
2587 }
2588 
2589 /*
2590  * Show a sysctl's name
2591  */
2592 static void
2593 db_show_oid_name(int *oid, size_t nlen)
2594 {
2595 	struct sysctl_oid *oidp;
2596 	int qoid[CTL_MAXNAME+2];
2597 	int error;
2598 
2599 	qoid[0] = 0;
2600 	memcpy(qoid + 2, oid, nlen * sizeof(int));
2601 	qoid[1] = 1;
2602 
2603 	error = sysctl_find_oid(qoid, nlen + 2, &oidp, NULL, NULL);
2604 	if (error)
2605 		db_error("sysctl name oid");
2606 
2607 	error = db_sysctl(oidp, qoid, nlen + 2, NULL, NULL, NULL, 0);
2608 	if (error)
2609 		db_error("sysctl name");
2610 }
2611 
2612 /*
2613  * Check to see if an OID is safe to print from ddb.
2614  */
2615 static bool
2616 db_oid_safe(const struct sysctl_oid *oidp)
2617 {
2618 	for (unsigned int i = 0; i < nitems(db_safe_handlers); ++i) {
2619 		if (oidp->oid_handler == db_safe_handlers[i])
2620 			return (true);
2621 	}
2622 
2623 	return (false);
2624 }
2625 
2626 /*
2627  * Show a sysctl at a specific OID
2628  * Compare to the input handling in show_var from sbin/sysctl/sysctl.c
2629  */
2630 static int
2631 db_show_oid(struct sysctl_oid *oidp, int *oid, size_t nlen, int flags)
2632 {
2633 	int error, xflag, oflag, Nflag, nflag;
2634 	size_t len;
2635 
2636 	xflag = flags & DB_SYSCTL_HEX;
2637 	oflag = flags & DB_SYSCTL_OPAQUE;
2638 	nflag = flags & DB_SYSCTL_VALUE_ONLY;
2639 	Nflag = flags & DB_SYSCTL_NAME_ONLY;
2640 
2641 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_OPAQUE &&
2642 	    (!xflag && !oflag))
2643 		return (0);
2644 
2645 	if (Nflag) {
2646 		db_show_oid_name(oid, nlen);
2647 		error = 0;
2648 		goto out;
2649 	}
2650 
2651 	if (!nflag) {
2652 		db_show_oid_name(oid, nlen);
2653 		db_printf(": ");
2654 	}
2655 
2656 	if ((flags & DB_SYSCTL_SAFE_ONLY) && !db_oid_safe(oidp)) {
2657 		db_printf("Skipping, unsafe to print while recursing.");
2658 		error = 0;
2659 		goto out;
2660 	}
2661 
2662 	/* Try once, and ask about the size */
2663 	len = 0;
2664 	error = db_sysctl(oidp, oid, nlen,
2665 	    NULL, NULL, &len, flags);
2666 	if (error)
2667 		goto out;
2668 
2669 	if (!g_ddb_sysctl_printed)
2670 		/* Lie about the size */
2671 		error = db_sysctl(oidp, oid, nlen,
2672 		    (void *) 1, &len, NULL, flags);
2673 
2674 out:
2675 	db_printf("\n");
2676 	return (error);
2677 }
2678 
2679 /*
2680  * Show all sysctls under a specific OID
2681  * Compare to sysctl_all from sbin/sysctl/sysctl.c
2682  */
2683 static int
2684 db_show_sysctl_all(int *oid, size_t len, int flags)
2685 {
2686 	struct sysctl_oid *oidp;
2687 	int name1[CTL_MAXNAME + 2], name2[CTL_MAXNAME + 2];
2688 	size_t l1, l2;
2689 
2690 	name1[0] = CTL_SYSCTL;
2691 	name1[1] = CTL_SYSCTL_NEXT;
2692 	l1 = 2;
2693 	if (len) {
2694 		memcpy(name1+2, oid, len * sizeof(int));
2695 		l1 +=len;
2696 	} else {
2697 		name1[2] = 1;
2698 		l1++;
2699 	}
2700 	for (;;) {
2701 		int i, error;
2702 
2703 		l2 = sizeof(name2);
2704 		error = kernel_sysctl(kdb_thread, name1, l1,
2705 		    name2, &l2, NULL, 0, &l2, 0);
2706 		if (error != 0) {
2707 			if (error == ENOENT)
2708 				return (0);
2709 			else
2710 				db_error("sysctl(getnext)");
2711 		}
2712 
2713 		l2 /= sizeof(int);
2714 
2715 		if (l2 < (unsigned int)len)
2716 			return (0);
2717 
2718 		for (i = 0; i < len; i++)
2719 			if (name2[i] != oid[i])
2720 				return (0);
2721 
2722 		/* Find the OID in question */
2723 		error = sysctl_find_oid(name2, l2, &oidp, NULL, NULL);
2724 		if (error)
2725 			return (error);
2726 
2727 		i = db_show_oid(oidp, name2, l2, flags | DB_SYSCTL_SAFE_ONLY);
2728 
2729 		if (db_pager_quit)
2730 			return (0);
2731 
2732 		memcpy(name1+2, name2, l2 * sizeof(int));
2733 		l1 = 2 + l2;
2734 	}
2735 }
2736 
2737 /*
2738  * Show a sysctl by its user facing string
2739  */
2740 static int
2741 db_sysctlbyname(char *name, int flags)
2742 {
2743 	struct sysctl_oid *oidp;
2744 	int oid[CTL_MAXNAME];
2745 	int error, nlen;
2746 
2747 	error = name2oid(name, oid, &nlen, &oidp);
2748 	if (error) {
2749 		return (error);
2750 	}
2751 
2752 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2753 		db_show_sysctl_all(oid, nlen, flags);
2754 	} else {
2755 		error = db_show_oid(oidp, oid, nlen, flags);
2756 	}
2757 
2758 	return (error);
2759 }
2760 
2761 static void
2762 db_sysctl_cmd_usage(void)
2763 {
2764 	db_printf(
2765 	    " sysctl [/Nnox] <sysctl>					    \n"
2766 	    "								    \n"
2767 	    " <sysctl> The name of the sysctl to show.			    \n"
2768 	    "								    \n"
2769 	    " Show a sysctl by hooking into SYSCTL_IN and SYSCTL_OUT.	    \n"
2770 	    " This will work for most sysctls, but should not be used	    \n"
2771 	    " with sysctls that are known to malloc.			    \n"
2772 	    "								    \n"
2773 	    " While recursing any \"unsafe\" sysctls will be skipped.	    \n"
2774 	    " Call sysctl directly on the sysctl to try printing the	    \n"
2775 	    " skipped sysctl. This is unsafe and may make the ddb	    \n"
2776 	    " session unusable.						    \n"
2777 	    "								    \n"
2778 	    " Arguments:						    \n"
2779 	    "	/N	Display only the name of the sysctl.		    \n"
2780 	    "	/n	Display only the value of the sysctl.		    \n"
2781 	    "	/o	Display opaque values.				    \n"
2782 	    "	/x	Display the sysctl in hex.			    \n"
2783 	    "								    \n"
2784 	    "For example:						    \n"
2785 	    "sysctl vm.v_free_min					    \n"
2786 	    "vn.v_free_min: 12669					    \n"
2787 	    );
2788 }
2789 
2790 /*
2791  * Show a specific sysctl similar to sysctl (8).
2792  */
2793 DB_FUNC(sysctl, db_sysctl_cmd, db_cmd_table, CS_OWN, NULL)
2794 {
2795 	char name[TOK_STRING_SIZE];
2796 	int error, i, t, flags;
2797 
2798 	/* Parse the modifiers */
2799 	t = db_read_token();
2800 	if (t == tSLASH || t == tMINUS) {
2801 		t = db_read_token();
2802 		if (t != tIDENT) {
2803 			db_printf("Bad modifier\n");
2804 			error = EINVAL;
2805 			goto out;
2806 		}
2807 		db_strcpy(modif, db_tok_string);
2808 	}
2809 	else {
2810 		db_unread_token(t);
2811 		modif[0] = '\0';
2812 	}
2813 
2814 	flags = 0;
2815 	for (i = 0; i < nitems(db_sysctl_modifs); i++) {
2816 		if (strchr(modif, db_sysctl_modifs[i])) {
2817 			flags |= db_sysctl_modif_values[i];
2818 		}
2819 	}
2820 
2821 	/* Parse the sysctl names */
2822 	t = db_read_token();
2823 	if (t != tIDENT) {
2824 		db_printf("Need sysctl name\n");
2825 		error = EINVAL;
2826 		goto out;
2827 	}
2828 
2829 	/* Copy the name into a temporary buffer */
2830 	db_strcpy(name, db_tok_string);
2831 
2832 	/* Ensure there is no trailing cruft */
2833 	t = db_read_token();
2834 	if (t != tEOL) {
2835 		db_printf("Unexpected sysctl argument\n");
2836 		error = EINVAL;
2837 		goto out;
2838 	}
2839 
2840 	error = db_sysctlbyname(name, flags);
2841 	if (error == ENOENT) {
2842 		db_printf("unknown oid: '%s'\n", db_tok_string);
2843 		goto out;
2844 	} else if (error) {
2845 		db_printf("%s: error: %d\n", db_tok_string, error);
2846 		goto out;
2847 	}
2848 
2849 out:
2850 	/* Ensure we eat all of our text */
2851 	db_flush_lex();
2852 
2853 	if (error == EINVAL) {
2854 		db_sysctl_cmd_usage();
2855 	}
2856 }
2857 
2858 #endif /* DDB */
2859