xref: /freebsd/sys/kern/kern_sysctl.c (revision e40139ff33b48b56a24c808b166b04b8ee6f5b21)
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 
741 static int
742 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
743 {
744 	struct sysctl_oid *p, *tmp;
745 	int error;
746 
747 	SYSCTL_ASSERT_WLOCKED();
748 	if (oidp == NULL)
749 		return(EINVAL);
750 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
751 		printf("Warning: can't remove non-dynamic nodes (%s)!\n",
752 		    oidp->oid_name);
753 		return (EINVAL);
754 	}
755 	/*
756 	 * WARNING: normal method to do this should be through
757 	 * sysctl_ctx_free(). Use recursing as the last resort
758 	 * method to purge your sysctl tree of leftovers...
759 	 * However, if some other code still references these nodes,
760 	 * it will panic.
761 	 */
762 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
763 		if (oidp->oid_refcnt == 1) {
764 			SLIST_FOREACH_SAFE(p,
765 			    SYSCTL_CHILDREN(oidp), oid_link, tmp) {
766 				if (!recurse) {
767 					printf("Warning: failed attempt to "
768 					    "remove oid %s with child %s\n",
769 					    oidp->oid_name, p->oid_name);
770 					return (ENOTEMPTY);
771 				}
772 				error = sysctl_remove_oid_locked(p, del,
773 				    recurse);
774 				if (error)
775 					return (error);
776 			}
777 		}
778 	}
779 	if (oidp->oid_refcnt > 1 ) {
780 		oidp->oid_refcnt--;
781 	} else {
782 		if (oidp->oid_refcnt == 0) {
783 			printf("Warning: bad oid_refcnt=%u (%s)!\n",
784 				oidp->oid_refcnt, oidp->oid_name);
785 			return (EINVAL);
786 		}
787 		sysctl_unregister_oid(oidp);
788 		if (del) {
789 			/*
790 			 * Wait for all threads running the handler to drain.
791 			 * This preserves the previous behavior when the
792 			 * sysctl lock was held across a handler invocation,
793 			 * and is necessary for module unload correctness.
794 			 */
795 			while (oidp->oid_running > 0) {
796 				oidp->oid_kind |= CTLFLAG_DYING;
797 				SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
798 			}
799 			if (oidp->oid_descr)
800 				free(__DECONST(char *, oidp->oid_descr),
801 				    M_SYSCTLOID);
802 			if (oidp->oid_label)
803 				free(__DECONST(char *, oidp->oid_label),
804 				    M_SYSCTLOID);
805 			free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
806 			free(oidp, M_SYSCTLOID);
807 		}
808 	}
809 	return (0);
810 }
811 /*
812  * Create new sysctls at run time.
813  * clist may point to a valid context initialized with sysctl_ctx_init().
814  */
815 struct sysctl_oid *
816 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
817 	int number, const char *name, int kind, void *arg1, intmax_t arg2,
818 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr,
819 	const char *label)
820 {
821 	struct sysctl_oid *oidp;
822 
823 	/* You have to hook up somewhere.. */
824 	if (parent == NULL)
825 		return(NULL);
826 	/* Check if the node already exists, otherwise create it */
827 	SYSCTL_WLOCK();
828 	oidp = sysctl_find_oidname(name, parent);
829 	if (oidp != NULL) {
830 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
831 			oidp->oid_refcnt++;
832 			/* Update the context */
833 			if (clist != NULL)
834 				sysctl_ctx_entry_add(clist, oidp);
835 			SYSCTL_WUNLOCK();
836 			return (oidp);
837 		} else {
838 			sysctl_warn_reuse(__func__, oidp);
839 			SYSCTL_WUNLOCK();
840 			return (NULL);
841 		}
842 	}
843 	oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
844 	oidp->oid_parent = parent;
845 	SLIST_INIT(&oidp->oid_children);
846 	oidp->oid_number = number;
847 	oidp->oid_refcnt = 1;
848 	oidp->oid_name = strdup(name, M_SYSCTLOID);
849 	oidp->oid_handler = handler;
850 	oidp->oid_kind = CTLFLAG_DYN | kind;
851 	oidp->oid_arg1 = arg1;
852 	oidp->oid_arg2 = arg2;
853 	oidp->oid_fmt = fmt;
854 	if (descr != NULL)
855 		oidp->oid_descr = strdup(descr, M_SYSCTLOID);
856 	if (label != NULL)
857 		oidp->oid_label = strdup(label, M_SYSCTLOID);
858 	/* Update the context, if used */
859 	if (clist != NULL)
860 		sysctl_ctx_entry_add(clist, oidp);
861 	/* Register this oid */
862 	sysctl_register_oid(oidp);
863 	SYSCTL_WUNLOCK();
864 	return (oidp);
865 }
866 
867 /*
868  * Rename an existing oid.
869  */
870 void
871 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
872 {
873 	char *newname;
874 	char *oldname;
875 
876 	newname = strdup(name, M_SYSCTLOID);
877 	SYSCTL_WLOCK();
878 	oldname = __DECONST(char *, oidp->oid_name);
879 	oidp->oid_name = newname;
880 	SYSCTL_WUNLOCK();
881 	free(oldname, M_SYSCTLOID);
882 }
883 
884 /*
885  * Reparent an existing oid.
886  */
887 int
888 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
889 {
890 	struct sysctl_oid *oidp;
891 
892 	SYSCTL_WLOCK();
893 	if (oid->oid_parent == parent) {
894 		SYSCTL_WUNLOCK();
895 		return (0);
896 	}
897 	oidp = sysctl_find_oidname(oid->oid_name, parent);
898 	if (oidp != NULL) {
899 		SYSCTL_WUNLOCK();
900 		return (EEXIST);
901 	}
902 	sysctl_unregister_oid(oid);
903 	oid->oid_parent = parent;
904 	oid->oid_number = OID_AUTO;
905 	sysctl_register_oid(oid);
906 	SYSCTL_WUNLOCK();
907 	return (0);
908 }
909 
910 /*
911  * Register the kernel's oids on startup.
912  */
913 SET_DECLARE(sysctl_set, struct sysctl_oid);
914 
915 static void
916 sysctl_register_all(void *arg)
917 {
918 	struct sysctl_oid **oidp;
919 
920 	sx_init(&sysctlmemlock, "sysctl mem");
921 	SYSCTL_INIT();
922 	SYSCTL_WLOCK();
923 	SET_FOREACH(oidp, sysctl_set)
924 		sysctl_register_oid(*oidp);
925 	SYSCTL_WUNLOCK();
926 }
927 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_FIRST, sysctl_register_all, NULL);
928 
929 /*
930  * "Staff-functions"
931  *
932  * These functions implement a presently undocumented interface
933  * used by the sysctl program to walk the tree, and get the type
934  * so it can print the value.
935  * This interface is under work and consideration, and should probably
936  * be killed with a big axe by the first person who can find the time.
937  * (be aware though, that the proper interface isn't as obvious as it
938  * may seem, there are various conflicting requirements.
939  *
940  * {CTL_SYSCTL, CTL_SYSCTL_DEBUG}		printf the entire MIB-tree.
941  * {CTL_SYSCTL, CTL_SYSCTL_NAME, ...}		return the name of the "..."
942  *						OID.
943  * {CTL_SYSCTL, CTL_SYSCTL_NEXT, ...}		return the next OID.
944  * {CTL_SYSCTL, CTL_SYSCTL_NAME2OID}		return the OID of the name in
945  *						"new"
946  * {CTL_SYSCTL, CTL_SYSCTL_OIDFMT, ...}		return the kind & format info
947  *						for the "..." OID.
948  * {CTL_SYSCTL, CTL_SYSCTL_OIDDESCR, ...}	return the description of the
949  *						"..." OID.
950  * {CTL_SYSCTL, CTL_SYSCTL_OIDLABEL, ...}	return the aggregation label of
951  *						the "..." OID.
952  */
953 
954 #ifdef SYSCTL_DEBUG
955 static void
956 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
957 {
958 	int k;
959 	struct sysctl_oid *oidp;
960 
961 	SYSCTL_ASSERT_LOCKED();
962 	SLIST_FOREACH(oidp, l, oid_link) {
963 
964 		for (k=0; k<i; k++)
965 			printf(" ");
966 
967 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
968 
969 		printf("%c%c",
970 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
971 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
972 
973 		if (oidp->oid_handler)
974 			printf(" *Handler");
975 
976 		switch (oidp->oid_kind & CTLTYPE) {
977 			case CTLTYPE_NODE:
978 				printf(" Node\n");
979 				if (!oidp->oid_handler) {
980 					sysctl_sysctl_debug_dump_node(
981 					    SYSCTL_CHILDREN(oidp), i + 2);
982 				}
983 				break;
984 			case CTLTYPE_INT:    printf(" Int\n"); break;
985 			case CTLTYPE_UINT:   printf(" u_int\n"); break;
986 			case CTLTYPE_LONG:   printf(" Long\n"); break;
987 			case CTLTYPE_ULONG:  printf(" u_long\n"); break;
988 			case CTLTYPE_STRING: printf(" String\n"); break;
989 			case CTLTYPE_S8:     printf(" int8_t\n"); break;
990 			case CTLTYPE_S16:    printf(" int16_t\n"); break;
991 			case CTLTYPE_S32:    printf(" int32_t\n"); break;
992 			case CTLTYPE_S64:    printf(" int64_t\n"); break;
993 			case CTLTYPE_U8:     printf(" uint8_t\n"); break;
994 			case CTLTYPE_U16:    printf(" uint16_t\n"); break;
995 			case CTLTYPE_U32:    printf(" uint32_t\n"); break;
996 			case CTLTYPE_U64:    printf(" uint64_t\n"); break;
997 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
998 			default:	     printf("\n");
999 		}
1000 
1001 	}
1002 }
1003 
1004 static int
1005 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
1006 {
1007 	struct rm_priotracker tracker;
1008 	int error;
1009 
1010 	error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
1011 	if (error)
1012 		return (error);
1013 	SYSCTL_RLOCK(&tracker);
1014 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
1015 	SYSCTL_RUNLOCK(&tracker);
1016 	return (ENOENT);
1017 }
1018 
1019 SYSCTL_PROC(_sysctl, CTL_SYSCTL_DEBUG, debug, CTLTYPE_STRING | CTLFLAG_RD |
1020     CTLFLAG_MPSAFE, 0, 0, sysctl_sysctl_debug, "-", "");
1021 #endif
1022 
1023 static int
1024 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
1025 {
1026 	int *name = (int *) arg1;
1027 	u_int namelen = arg2;
1028 	int error;
1029 	struct sysctl_oid *oid;
1030 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
1031 	struct rm_priotracker tracker;
1032 	char buf[10];
1033 
1034 	error = sysctl_wire_old_buffer(req, 0);
1035 	if (error)
1036 		return (error);
1037 
1038 	SYSCTL_RLOCK(&tracker);
1039 	while (namelen) {
1040 		if (!lsp) {
1041 			snprintf(buf,sizeof(buf),"%d",*name);
1042 			if (req->oldidx)
1043 				error = SYSCTL_OUT(req, ".", 1);
1044 			if (!error)
1045 				error = SYSCTL_OUT(req, buf, strlen(buf));
1046 			if (error)
1047 				goto out;
1048 			namelen--;
1049 			name++;
1050 			continue;
1051 		}
1052 		lsp2 = NULL;
1053 		SLIST_FOREACH(oid, lsp, oid_link) {
1054 			if (oid->oid_number != *name)
1055 				continue;
1056 
1057 			if (req->oldidx)
1058 				error = SYSCTL_OUT(req, ".", 1);
1059 			if (!error)
1060 				error = SYSCTL_OUT(req, oid->oid_name,
1061 					strlen(oid->oid_name));
1062 			if (error)
1063 				goto out;
1064 
1065 			namelen--;
1066 			name++;
1067 
1068 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1069 				break;
1070 
1071 			if (oid->oid_handler)
1072 				break;
1073 
1074 			lsp2 = SYSCTL_CHILDREN(oid);
1075 			break;
1076 		}
1077 		lsp = lsp2;
1078 	}
1079 	error = SYSCTL_OUT(req, "", 1);
1080  out:
1081 	SYSCTL_RUNLOCK(&tracker);
1082 	return (error);
1083 }
1084 
1085 /*
1086  * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
1087  * capability mode.
1088  */
1089 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NAME, name, CTLFLAG_RD |
1090     CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_name, "");
1091 
1092 static int
1093 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
1094 	int *next, int *len, int level, struct sysctl_oid **oidpp)
1095 {
1096 	struct sysctl_oid *oidp;
1097 
1098 	SYSCTL_ASSERT_LOCKED();
1099 	*len = level;
1100 	SLIST_FOREACH(oidp, lsp, oid_link) {
1101 		*next = oidp->oid_number;
1102 		*oidpp = oidp;
1103 
1104 		if ((oidp->oid_kind & (CTLFLAG_SKIP | CTLFLAG_DORMANT)) != 0)
1105 			continue;
1106 
1107 		if (!namelen) {
1108 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1109 				return (0);
1110 			if (oidp->oid_handler)
1111 				/* We really should call the handler here...*/
1112 				return (0);
1113 			lsp = SYSCTL_CHILDREN(oidp);
1114 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
1115 				len, level+1, oidpp))
1116 				return (0);
1117 			goto emptynode;
1118 		}
1119 
1120 		if (oidp->oid_number < *name)
1121 			continue;
1122 
1123 		if (oidp->oid_number > *name) {
1124 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1125 				return (0);
1126 			if (oidp->oid_handler)
1127 				return (0);
1128 			lsp = SYSCTL_CHILDREN(oidp);
1129 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
1130 				next+1, len, level+1, oidpp))
1131 				return (0);
1132 			goto next;
1133 		}
1134 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1135 			continue;
1136 
1137 		if (oidp->oid_handler)
1138 			continue;
1139 
1140 		lsp = SYSCTL_CHILDREN(oidp);
1141 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
1142 			len, level+1, oidpp))
1143 			return (0);
1144 	next:
1145 		namelen = 1;
1146 	emptynode:
1147 		*len = level;
1148 	}
1149 	return (1);
1150 }
1151 
1152 static int
1153 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
1154 {
1155 	int *name = (int *) arg1;
1156 	u_int namelen = arg2;
1157 	int i, j, error;
1158 	struct sysctl_oid *oid;
1159 	struct sysctl_oid_list *lsp = &sysctl__children;
1160 	struct rm_priotracker tracker;
1161 	int newoid[CTL_MAXNAME];
1162 
1163 	SYSCTL_RLOCK(&tracker);
1164 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
1165 	SYSCTL_RUNLOCK(&tracker);
1166 	if (i)
1167 		return (ENOENT);
1168 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
1169 	return (error);
1170 }
1171 
1172 /*
1173  * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
1174  * capability mode.
1175  */
1176 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NEXT, next, CTLFLAG_RD |
1177     CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_next, "");
1178 
1179 static int
1180 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
1181 {
1182 	struct sysctl_oid *oidp;
1183 	struct sysctl_oid_list *lsp = &sysctl__children;
1184 	char *p;
1185 
1186 	SYSCTL_ASSERT_LOCKED();
1187 
1188 	for (*len = 0; *len < CTL_MAXNAME;) {
1189 		p = strsep(&name, ".");
1190 
1191 		oidp = SLIST_FIRST(lsp);
1192 		for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
1193 			if (oidp == NULL)
1194 				return (ENOENT);
1195 			if (strcmp(p, oidp->oid_name) == 0)
1196 				break;
1197 		}
1198 		*oid++ = oidp->oid_number;
1199 		(*len)++;
1200 
1201 		if (name == NULL || *name == '\0') {
1202 			if (oidpp)
1203 				*oidpp = oidp;
1204 			return (0);
1205 		}
1206 
1207 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1208 			break;
1209 
1210 		if (oidp->oid_handler)
1211 			break;
1212 
1213 		lsp = SYSCTL_CHILDREN(oidp);
1214 	}
1215 	return (ENOENT);
1216 }
1217 
1218 static int
1219 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
1220 {
1221 	char *p;
1222 	int error, oid[CTL_MAXNAME], len = 0;
1223 	struct sysctl_oid *op = NULL;
1224 	struct rm_priotracker tracker;
1225 	char buf[32];
1226 
1227 	if (!req->newlen)
1228 		return (ENOENT);
1229 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
1230 		return (ENAMETOOLONG);
1231 
1232 	p = buf;
1233 	if (req->newlen >= sizeof(buf))
1234 		p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
1235 
1236 	error = SYSCTL_IN(req, p, req->newlen);
1237 	if (error) {
1238 		if (p != buf)
1239 			free(p, M_SYSCTL);
1240 		return (error);
1241 	}
1242 
1243 	p [req->newlen] = '\0';
1244 
1245 	SYSCTL_RLOCK(&tracker);
1246 	error = name2oid(p, oid, &len, &op);
1247 	SYSCTL_RUNLOCK(&tracker);
1248 
1249 	if (p != buf)
1250 		free(p, M_SYSCTL);
1251 
1252 	if (error)
1253 		return (error);
1254 
1255 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
1256 	return (error);
1257 }
1258 
1259 /*
1260  * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
1261  * capability mode.
1262  */
1263 SYSCTL_PROC(_sysctl, CTL_SYSCTL_NAME2OID, name2oid, CTLTYPE_INT | CTLFLAG_RW |
1264     CTLFLAG_ANYBODY | CTLFLAG_MPSAFE | CTLFLAG_CAPRW, 0, 0,
1265     sysctl_sysctl_name2oid, "I", "");
1266 
1267 static int
1268 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
1269 {
1270 	struct sysctl_oid *oid;
1271 	struct rm_priotracker tracker;
1272 	int error;
1273 
1274 	error = sysctl_wire_old_buffer(req, 0);
1275 	if (error)
1276 		return (error);
1277 
1278 	SYSCTL_RLOCK(&tracker);
1279 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1280 	if (error)
1281 		goto out;
1282 
1283 	if (oid->oid_fmt == NULL) {
1284 		error = ENOENT;
1285 		goto out;
1286 	}
1287 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
1288 	if (error)
1289 		goto out;
1290 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
1291  out:
1292 	SYSCTL_RUNLOCK(&tracker);
1293 	return (error);
1294 }
1295 
1296 
1297 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDFMT, oidfmt, CTLFLAG_RD |
1298     CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidfmt, "");
1299 
1300 static int
1301 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
1302 {
1303 	struct sysctl_oid *oid;
1304 	struct rm_priotracker tracker;
1305 	int error;
1306 
1307 	error = sysctl_wire_old_buffer(req, 0);
1308 	if (error)
1309 		return (error);
1310 
1311 	SYSCTL_RLOCK(&tracker);
1312 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1313 	if (error)
1314 		goto out;
1315 
1316 	if (oid->oid_descr == NULL) {
1317 		error = ENOENT;
1318 		goto out;
1319 	}
1320 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
1321  out:
1322 	SYSCTL_RUNLOCK(&tracker);
1323 	return (error);
1324 }
1325 
1326 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDDESCR, oiddescr, CTLFLAG_RD |
1327     CTLFLAG_MPSAFE|CTLFLAG_CAPRD, sysctl_sysctl_oiddescr, "");
1328 
1329 static int
1330 sysctl_sysctl_oidlabel(SYSCTL_HANDLER_ARGS)
1331 {
1332 	struct sysctl_oid *oid;
1333 	struct rm_priotracker tracker;
1334 	int error;
1335 
1336 	error = sysctl_wire_old_buffer(req, 0);
1337 	if (error)
1338 		return (error);
1339 
1340 	SYSCTL_RLOCK(&tracker);
1341 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1342 	if (error)
1343 		goto out;
1344 
1345 	if (oid->oid_label == NULL) {
1346 		error = ENOENT;
1347 		goto out;
1348 	}
1349 	error = SYSCTL_OUT(req, oid->oid_label, strlen(oid->oid_label) + 1);
1350  out:
1351 	SYSCTL_RUNLOCK(&tracker);
1352 	return (error);
1353 }
1354 
1355 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDLABEL, oidlabel, CTLFLAG_RD |
1356     CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidlabel, "");
1357 
1358 /*
1359  * Default "handler" functions.
1360  */
1361 
1362 /*
1363  * Handle a bool.
1364  * Two cases:
1365  *     a variable:  point arg1 at it.
1366  *     a constant:  pass it in arg2.
1367  */
1368 
1369 int
1370 sysctl_handle_bool(SYSCTL_HANDLER_ARGS)
1371 {
1372 	uint8_t temp;
1373 	int error;
1374 
1375 	/*
1376 	 * Attempt to get a coherent snapshot by making a copy of the data.
1377 	 */
1378 	if (arg1)
1379 		temp = *(bool *)arg1 ? 1 : 0;
1380 	else
1381 		temp = arg2 ? 1 : 0;
1382 
1383 	error = SYSCTL_OUT(req, &temp, sizeof(temp));
1384 	if (error || !req->newptr)
1385 		return (error);
1386 
1387 	if (!arg1)
1388 		error = EPERM;
1389 	else {
1390 		error = SYSCTL_IN(req, &temp, sizeof(temp));
1391 		if (!error)
1392 			*(bool *)arg1 = temp ? 1 : 0;
1393 	}
1394 	return (error);
1395 }
1396 
1397 /*
1398  * Handle an int8_t, signed or unsigned.
1399  * Two cases:
1400  *     a variable:  point arg1 at it.
1401  *     a constant:  pass it in arg2.
1402  */
1403 
1404 int
1405 sysctl_handle_8(SYSCTL_HANDLER_ARGS)
1406 {
1407 	int8_t tmpout;
1408 	int error = 0;
1409 
1410 	/*
1411 	 * Attempt to get a coherent snapshot by making a copy of the data.
1412 	 */
1413 	if (arg1)
1414 		tmpout = *(int8_t *)arg1;
1415 	else
1416 		tmpout = arg2;
1417 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1418 
1419 	if (error || !req->newptr)
1420 		return (error);
1421 
1422 	if (!arg1)
1423 		error = EPERM;
1424 	else
1425 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1426 	return (error);
1427 }
1428 
1429 /*
1430  * Handle an int16_t, signed or unsigned.
1431  * Two cases:
1432  *     a variable:  point arg1 at it.
1433  *     a constant:  pass it in arg2.
1434  */
1435 
1436 int
1437 sysctl_handle_16(SYSCTL_HANDLER_ARGS)
1438 {
1439 	int16_t tmpout;
1440 	int error = 0;
1441 
1442 	/*
1443 	 * Attempt to get a coherent snapshot by making a copy of the data.
1444 	 */
1445 	if (arg1)
1446 		tmpout = *(int16_t *)arg1;
1447 	else
1448 		tmpout = arg2;
1449 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1450 
1451 	if (error || !req->newptr)
1452 		return (error);
1453 
1454 	if (!arg1)
1455 		error = EPERM;
1456 	else
1457 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1458 	return (error);
1459 }
1460 
1461 /*
1462  * Handle an int32_t, signed or unsigned.
1463  * Two cases:
1464  *     a variable:  point arg1 at it.
1465  *     a constant:  pass it in arg2.
1466  */
1467 
1468 int
1469 sysctl_handle_32(SYSCTL_HANDLER_ARGS)
1470 {
1471 	int32_t tmpout;
1472 	int error = 0;
1473 
1474 	/*
1475 	 * Attempt to get a coherent snapshot by making a copy of the data.
1476 	 */
1477 	if (arg1)
1478 		tmpout = *(int32_t *)arg1;
1479 	else
1480 		tmpout = arg2;
1481 	error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1482 
1483 	if (error || !req->newptr)
1484 		return (error);
1485 
1486 	if (!arg1)
1487 		error = EPERM;
1488 	else
1489 		error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1490 	return (error);
1491 }
1492 
1493 /*
1494  * Handle an int, signed or unsigned.
1495  * Two cases:
1496  *     a variable:  point arg1 at it.
1497  *     a constant:  pass it in arg2.
1498  */
1499 
1500 int
1501 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
1502 {
1503 	int tmpout, error = 0;
1504 
1505 	/*
1506 	 * Attempt to get a coherent snapshot by making a copy of the data.
1507 	 */
1508 	if (arg1)
1509 		tmpout = *(int *)arg1;
1510 	else
1511 		tmpout = arg2;
1512 	error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1513 
1514 	if (error || !req->newptr)
1515 		return (error);
1516 
1517 	if (!arg1)
1518 		error = EPERM;
1519 	else
1520 		error = SYSCTL_IN(req, arg1, sizeof(int));
1521 	return (error);
1522 }
1523 
1524 /*
1525  * Based on on sysctl_handle_int() convert milliseconds into ticks.
1526  * Note: this is used by TCP.
1527  */
1528 
1529 int
1530 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1531 {
1532 	int error, s, tt;
1533 
1534 	tt = *(int *)arg1;
1535 	s = (int)((int64_t)tt * 1000 / hz);
1536 
1537 	error = sysctl_handle_int(oidp, &s, 0, req);
1538 	if (error || !req->newptr)
1539 		return (error);
1540 
1541 	tt = (int)((int64_t)s * hz / 1000);
1542 	if (tt < 1)
1543 		return (EINVAL);
1544 
1545 	*(int *)arg1 = tt;
1546 	return (0);
1547 }
1548 
1549 
1550 /*
1551  * Handle a long, signed or unsigned.
1552  * Two cases:
1553  *     a variable:  point arg1 at it.
1554  *     a constant:  pass it in arg2.
1555  */
1556 
1557 int
1558 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1559 {
1560 	int error = 0;
1561 	long tmplong;
1562 #ifdef SCTL_MASK32
1563 	int tmpint;
1564 #endif
1565 
1566 	/*
1567 	 * Attempt to get a coherent snapshot by making a copy of the data.
1568 	 */
1569 	if (arg1)
1570 		tmplong = *(long *)arg1;
1571 	else
1572 		tmplong = arg2;
1573 #ifdef SCTL_MASK32
1574 	if (req->flags & SCTL_MASK32) {
1575 		tmpint = tmplong;
1576 		error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1577 	} else
1578 #endif
1579 		error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1580 
1581 	if (error || !req->newptr)
1582 		return (error);
1583 
1584 	if (!arg1)
1585 		error = EPERM;
1586 #ifdef SCTL_MASK32
1587 	else if (req->flags & SCTL_MASK32) {
1588 		error = SYSCTL_IN(req, &tmpint, sizeof(int));
1589 		*(long *)arg1 = (long)tmpint;
1590 	}
1591 #endif
1592 	else
1593 		error = SYSCTL_IN(req, arg1, sizeof(long));
1594 	return (error);
1595 }
1596 
1597 /*
1598  * Handle a 64 bit int, signed or unsigned.
1599  * Two cases:
1600  *     a variable:  point arg1 at it.
1601  *     a constant:  pass it in arg2.
1602  */
1603 int
1604 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1605 {
1606 	int error = 0;
1607 	uint64_t tmpout;
1608 
1609 	/*
1610 	 * Attempt to get a coherent snapshot by making a copy of the data.
1611 	 */
1612 	if (arg1)
1613 		tmpout = *(uint64_t *)arg1;
1614 	else
1615 		tmpout = arg2;
1616 	error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1617 
1618 	if (error || !req->newptr)
1619 		return (error);
1620 
1621 	if (!arg1)
1622 		error = EPERM;
1623 	else
1624 		error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1625 	return (error);
1626 }
1627 
1628 /*
1629  * Handle our generic '\0' terminated 'C' string.
1630  * Two cases:
1631  * 	a variable string:  point arg1 at it, arg2 is max length.
1632  * 	a constant string:  point arg1 at it, arg2 is zero.
1633  */
1634 
1635 int
1636 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1637 {
1638 	size_t outlen;
1639 	int error = 0, ro_string = 0;
1640 
1641 	/*
1642 	 * A zero-length buffer indicates a fixed size read-only
1643 	 * string.  In ddb, don't worry about trying to make a malloced
1644 	 * snapshot.
1645 	 */
1646 	if (arg2 == 0 || kdb_active) {
1647 		arg2 = strlen((char *)arg1) + 1;
1648 		ro_string = 1;
1649 	}
1650 
1651 	if (req->oldptr != NULL) {
1652 		char *tmparg;
1653 
1654 		if (ro_string) {
1655 			tmparg = arg1;
1656 		} else {
1657 			/* try to make a coherent snapshot of the string */
1658 			tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK);
1659 			memcpy(tmparg, arg1, arg2);
1660 		}
1661 
1662 		outlen = strnlen(tmparg, arg2 - 1) + 1;
1663 		error = SYSCTL_OUT(req, tmparg, outlen);
1664 
1665 		if (!ro_string)
1666 			free(tmparg, M_SYSCTLTMP);
1667 	} else {
1668 		outlen = strnlen((char *)arg1, arg2 - 1) + 1;
1669 		error = SYSCTL_OUT(req, NULL, outlen);
1670 	}
1671 	if (error || !req->newptr)
1672 		return (error);
1673 
1674 	if ((req->newlen - req->newidx) >= arg2) {
1675 		error = EINVAL;
1676 	} else {
1677 		arg2 = (req->newlen - req->newidx);
1678 		error = SYSCTL_IN(req, arg1, arg2);
1679 		((char *)arg1)[arg2] = '\0';
1680 	}
1681 	return (error);
1682 }
1683 
1684 /*
1685  * Handle any kind of opaque data.
1686  * arg1 points to it, arg2 is the size.
1687  */
1688 
1689 int
1690 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1691 {
1692 	int error, tries;
1693 	u_int generation;
1694 	struct sysctl_req req2;
1695 
1696 	/*
1697 	 * Attempt to get a coherent snapshot, by using the thread
1698 	 * pre-emption counter updated from within mi_switch() to
1699 	 * determine if we were pre-empted during a bcopy() or
1700 	 * copyout(). Make 3 attempts at doing this before giving up.
1701 	 * If we encounter an error, stop immediately.
1702 	 */
1703 	tries = 0;
1704 	req2 = *req;
1705 retry:
1706 	generation = curthread->td_generation;
1707 	error = SYSCTL_OUT(req, arg1, arg2);
1708 	if (error)
1709 		return (error);
1710 	tries++;
1711 	if (generation != curthread->td_generation && tries < 3) {
1712 		*req = req2;
1713 		goto retry;
1714 	}
1715 
1716 	error = SYSCTL_IN(req, arg1, arg2);
1717 
1718 	return (error);
1719 }
1720 
1721 /*
1722  * Based on on sysctl_handle_int() convert microseconds to a sbintime.
1723  */
1724 int
1725 sysctl_usec_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 = sbttous(tt);
1733 
1734 	error = sysctl_handle_64(oidp, &sb, 0, req);
1735 	if (error || !req->newptr)
1736 		return (error);
1737 
1738 	tt = ustosbt(sb);
1739 	*(int64_t *)arg1 = tt;
1740 
1741 	return (0);
1742 }
1743 
1744 /*
1745  * Based on on sysctl_handle_int() convert milliseconds to a sbintime.
1746  */
1747 int
1748 sysctl_msec_to_sbintime(SYSCTL_HANDLER_ARGS)
1749 {
1750 	int error;
1751 	int64_t tt;
1752 	sbintime_t sb;
1753 
1754 	tt = *(int64_t *)arg1;
1755 	sb = sbttoms(tt);
1756 
1757 	error = sysctl_handle_64(oidp, &sb, 0, req);
1758 	if (error || !req->newptr)
1759 		return (error);
1760 
1761 	tt = mstosbt(sb);
1762 	*(int64_t *)arg1 = tt;
1763 
1764 	return (0);
1765 }
1766 
1767 /*
1768  * Convert seconds to a struct timeval.  Intended for use with
1769  * intervals and thus does not permit negative seconds.
1770  */
1771 int
1772 sysctl_sec_to_timeval(SYSCTL_HANDLER_ARGS)
1773 {
1774 	struct timeval *tv;
1775 	int error, secs;
1776 
1777 	tv = arg1;
1778 	secs = tv->tv_sec;
1779 
1780 	error = sysctl_handle_int(oidp, &secs, 0, req);
1781 	if (error || req->newptr == NULL)
1782 		return (error);
1783 
1784 	if (secs < 0)
1785 		return (EINVAL);
1786 	tv->tv_sec = secs;
1787 
1788 	return (0);
1789 }
1790 
1791 /*
1792  * Transfer functions to/from kernel space.
1793  * XXX: rather untested at this point
1794  */
1795 static int
1796 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1797 {
1798 	size_t i = 0;
1799 
1800 	if (req->oldptr) {
1801 		i = l;
1802 		if (req->oldlen <= req->oldidx)
1803 			i = 0;
1804 		else
1805 			if (i > req->oldlen - req->oldidx)
1806 				i = req->oldlen - req->oldidx;
1807 		if (i > 0)
1808 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
1809 	}
1810 	req->oldidx += l;
1811 	if (req->oldptr && i != l)
1812 		return (ENOMEM);
1813 	return (0);
1814 }
1815 
1816 static int
1817 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1818 {
1819 	if (!req->newptr)
1820 		return (0);
1821 	if (req->newlen - req->newidx < l)
1822 		return (EINVAL);
1823 	bcopy((const char *)req->newptr + req->newidx, p, l);
1824 	req->newidx += l;
1825 	return (0);
1826 }
1827 
1828 int
1829 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1830     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1831 {
1832 	int error = 0;
1833 	struct sysctl_req req;
1834 
1835 	bzero(&req, sizeof req);
1836 
1837 	req.td = td;
1838 	req.flags = flags;
1839 
1840 	if (oldlenp) {
1841 		req.oldlen = *oldlenp;
1842 	}
1843 	req.validlen = req.oldlen;
1844 
1845 	if (old) {
1846 		req.oldptr= old;
1847 	}
1848 
1849 	if (new != NULL) {
1850 		req.newlen = newlen;
1851 		req.newptr = new;
1852 	}
1853 
1854 	req.oldfunc = sysctl_old_kernel;
1855 	req.newfunc = sysctl_new_kernel;
1856 	req.lock = REQ_UNWIRED;
1857 
1858 	error = sysctl_root(0, name, namelen, &req);
1859 
1860 	if (req.lock == REQ_WIRED && req.validlen > 0)
1861 		vsunlock(req.oldptr, req.validlen);
1862 
1863 	if (error && error != ENOMEM)
1864 		return (error);
1865 
1866 	if (retval) {
1867 		if (req.oldptr && req.oldidx > req.validlen)
1868 			*retval = req.validlen;
1869 		else
1870 			*retval = req.oldidx;
1871 	}
1872 	return (error);
1873 }
1874 
1875 int
1876 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1877     void *new, size_t newlen, size_t *retval, int flags)
1878 {
1879         int oid[CTL_MAXNAME];
1880         size_t oidlen, plen;
1881 	int error;
1882 
1883 	oid[0] = CTL_SYSCTL;
1884 	oid[1] = CTL_SYSCTL_NAME2OID;
1885 	oidlen = sizeof(oid);
1886 
1887 	error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1888 	    (void *)name, strlen(name), &plen, flags);
1889 	if (error)
1890 		return (error);
1891 
1892 	error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1893 	    new, newlen, retval, flags);
1894 	return (error);
1895 }
1896 
1897 /*
1898  * Transfer function to/from user space.
1899  */
1900 static int
1901 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1902 {
1903 	size_t i, len, origidx;
1904 	int error;
1905 
1906 	origidx = req->oldidx;
1907 	req->oldidx += l;
1908 	if (req->oldptr == NULL)
1909 		return (0);
1910 	/*
1911 	 * If we have not wired the user supplied buffer and we are currently
1912 	 * holding locks, drop a witness warning, as it's possible that
1913 	 * write operations to the user page can sleep.
1914 	 */
1915 	if (req->lock != REQ_WIRED)
1916 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1917 		    "sysctl_old_user()");
1918 	i = l;
1919 	len = req->validlen;
1920 	if (len <= origidx)
1921 		i = 0;
1922 	else {
1923 		if (i > len - origidx)
1924 			i = len - origidx;
1925 		if (req->lock == REQ_WIRED) {
1926 			error = copyout_nofault(p, (char *)req->oldptr +
1927 			    origidx, i);
1928 		} else
1929 			error = copyout(p, (char *)req->oldptr + origidx, i);
1930 		if (error != 0)
1931 			return (error);
1932 	}
1933 	if (i < l)
1934 		return (ENOMEM);
1935 	return (0);
1936 }
1937 
1938 static int
1939 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1940 {
1941 	int error;
1942 
1943 	if (!req->newptr)
1944 		return (0);
1945 	if (req->newlen - req->newidx < l)
1946 		return (EINVAL);
1947 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1948 	    "sysctl_new_user()");
1949 	error = copyin((const char *)req->newptr + req->newidx, p, l);
1950 	req->newidx += l;
1951 	return (error);
1952 }
1953 
1954 /*
1955  * Wire the user space destination buffer.  If set to a value greater than
1956  * zero, the len parameter limits the maximum amount of wired memory.
1957  */
1958 int
1959 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1960 {
1961 	int ret;
1962 	size_t wiredlen;
1963 
1964 	wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1965 	ret = 0;
1966 	if (req->lock != REQ_WIRED && req->oldptr &&
1967 	    req->oldfunc == sysctl_old_user) {
1968 		if (wiredlen != 0) {
1969 			ret = vslock(req->oldptr, wiredlen);
1970 			if (ret != 0) {
1971 				if (ret != ENOMEM)
1972 					return (ret);
1973 				wiredlen = 0;
1974 			}
1975 		}
1976 		req->lock = REQ_WIRED;
1977 		req->validlen = wiredlen;
1978 	}
1979 	return (0);
1980 }
1981 
1982 int
1983 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1984     int *nindx, struct sysctl_req *req)
1985 {
1986 	struct sysctl_oid_list *lsp;
1987 	struct sysctl_oid *oid;
1988 	int indx;
1989 
1990 	SYSCTL_ASSERT_LOCKED();
1991 	lsp = &sysctl__children;
1992 	indx = 0;
1993 	while (indx < CTL_MAXNAME) {
1994 		SLIST_FOREACH(oid, lsp, oid_link) {
1995 			if (oid->oid_number == name[indx])
1996 				break;
1997 		}
1998 		if (oid == NULL)
1999 			return (ENOENT);
2000 
2001 		indx++;
2002 		if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2003 			if (oid->oid_handler != NULL || indx == namelen) {
2004 				*noid = oid;
2005 				if (nindx != NULL)
2006 					*nindx = indx;
2007 				KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
2008 				    ("%s found DYING node %p", __func__, oid));
2009 				return (0);
2010 			}
2011 			lsp = SYSCTL_CHILDREN(oid);
2012 		} else if (indx == namelen) {
2013 			if ((oid->oid_kind & CTLFLAG_DORMANT) != 0)
2014 				return (ENOENT);
2015 			*noid = oid;
2016 			if (nindx != NULL)
2017 				*nindx = indx;
2018 			KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
2019 			    ("%s found DYING node %p", __func__, oid));
2020 			return (0);
2021 		} else {
2022 			return (ENOTDIR);
2023 		}
2024 	}
2025 	return (ENOENT);
2026 }
2027 
2028 /*
2029  * Traverse our tree, and find the right node, execute whatever it points
2030  * to, and return the resulting error code.
2031  */
2032 
2033 static int
2034 sysctl_root(SYSCTL_HANDLER_ARGS)
2035 {
2036 	struct sysctl_oid *oid;
2037 	struct rm_priotracker tracker;
2038 	int error, indx, lvl;
2039 
2040 	SYSCTL_RLOCK(&tracker);
2041 
2042 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
2043 	if (error)
2044 		goto out;
2045 
2046 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2047 		/*
2048 		 * You can't call a sysctl when it's a node, but has
2049 		 * no handler.  Inform the user that it's a node.
2050 		 * The indx may or may not be the same as namelen.
2051 		 */
2052 		if (oid->oid_handler == NULL) {
2053 			error = EISDIR;
2054 			goto out;
2055 		}
2056 	}
2057 
2058 	/* Is this sysctl writable? */
2059 	if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) {
2060 		error = EPERM;
2061 		goto out;
2062 	}
2063 
2064 	KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
2065 
2066 #ifdef CAPABILITY_MODE
2067 	/*
2068 	 * If the process is in capability mode, then don't permit reading or
2069 	 * writing unless specifically granted for the node.
2070 	 */
2071 	if (IN_CAPABILITY_MODE(req->td)) {
2072 		if ((req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) ||
2073 		    (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))) {
2074 			error = EPERM;
2075 			goto out;
2076 		}
2077 	}
2078 #endif
2079 
2080 	/* Is this sysctl sensitive to securelevels? */
2081 	if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
2082 		lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
2083 		error = securelevel_gt(req->td->td_ucred, lvl);
2084 		if (error)
2085 			goto out;
2086 	}
2087 
2088 	/* Is this sysctl writable by only privileged users? */
2089 	if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
2090 		int priv;
2091 
2092 		if (oid->oid_kind & CTLFLAG_PRISON)
2093 			priv = PRIV_SYSCTL_WRITEJAIL;
2094 #ifdef VIMAGE
2095 		else if ((oid->oid_kind & CTLFLAG_VNET) &&
2096 		     prison_owns_vnet(req->td->td_ucred))
2097 			priv = PRIV_SYSCTL_WRITEJAIL;
2098 #endif
2099 		else
2100 			priv = PRIV_SYSCTL_WRITE;
2101 		error = priv_check(req->td, priv);
2102 		if (error)
2103 			goto out;
2104 	}
2105 
2106 	if (!oid->oid_handler) {
2107 		error = EINVAL;
2108 		goto out;
2109 	}
2110 
2111 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2112 		arg1 = (int *)arg1 + indx;
2113 		arg2 -= indx;
2114 	} else {
2115 		arg1 = oid->oid_arg1;
2116 		arg2 = oid->oid_arg2;
2117 	}
2118 #ifdef MAC
2119 	error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
2120 	    req);
2121 	if (error != 0)
2122 		goto out;
2123 #endif
2124 #ifdef VIMAGE
2125 	if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
2126 		arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
2127 #endif
2128 	error = sysctl_root_handler_locked(oid, arg1, arg2, req, &tracker);
2129 
2130 out:
2131 	SYSCTL_RUNLOCK(&tracker);
2132 	return (error);
2133 }
2134 
2135 #ifndef _SYS_SYSPROTO_H_
2136 struct sysctl_args {
2137 	int	*name;
2138 	u_int	namelen;
2139 	void	*old;
2140 	size_t	*oldlenp;
2141 	void	*new;
2142 	size_t	newlen;
2143 };
2144 #endif
2145 int
2146 sys___sysctl(struct thread *td, struct sysctl_args *uap)
2147 {
2148 	int error, i, name[CTL_MAXNAME];
2149 	size_t j;
2150 
2151 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
2152 		return (EINVAL);
2153 
2154  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
2155  	if (error)
2156 		return (error);
2157 
2158 	error = userland_sysctl(td, name, uap->namelen,
2159 		uap->old, uap->oldlenp, 0,
2160 		uap->new, uap->newlen, &j, 0);
2161 	if (error && error != ENOMEM)
2162 		return (error);
2163 	if (uap->oldlenp) {
2164 		i = copyout(&j, uap->oldlenp, sizeof(j));
2165 		if (i)
2166 			return (i);
2167 	}
2168 	return (error);
2169 }
2170 
2171 int
2172 kern___sysctlbyname(struct thread *td, const char *oname, size_t namelen,
2173     void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval,
2174     int flags, bool inkernel)
2175 {
2176 	int oid[CTL_MAXNAME];
2177 	char namebuf[16];
2178 	char *name;
2179 	size_t oidlen;
2180 	int error;
2181 
2182 	if (namelen > MAXPATHLEN || namelen == 0)
2183 		return (EINVAL);
2184 	name = namebuf;
2185 	if (namelen > sizeof(namebuf))
2186 		name = malloc(namelen, M_SYSCTL, M_WAITOK);
2187 	error = copyin(oname, name, namelen);
2188 	if (error != 0)
2189 		goto out;
2190 
2191 	oid[0] = CTL_SYSCTL;
2192 	oid[1] = CTL_SYSCTL_NAME2OID;
2193 	oidlen = sizeof(oid);
2194 	error = kernel_sysctl(td, oid, 2, oid, &oidlen, (void *)name, namelen,
2195 	    retval, flags);
2196 	if (error != 0)
2197 		goto out;
2198 	error = userland_sysctl(td, oid, *retval / sizeof(int), old, oldlenp,
2199 	    inkernel, new, newlen, retval, flags);
2200 
2201 out:
2202 	if (namelen > sizeof(namebuf))
2203 		free(name, M_SYSCTL);
2204 	return (error);
2205 }
2206 
2207 #ifndef	_SYS_SYSPROTO_H_
2208 struct __sysctlbyname_args {
2209 	const char	*name;
2210 	size_t	namelen;
2211 	void	*old;
2212 	size_t	*oldlenp;
2213 	void	*new;
2214 	size_t	newlen;
2215 };
2216 #endif
2217 int
2218 sys___sysctlbyname(struct thread *td, struct __sysctlbyname_args *uap)
2219 {
2220 	size_t rv;
2221 	int error;
2222 
2223 	error = kern___sysctlbyname(td, uap->name, uap->namelen, uap->old,
2224 	    uap->oldlenp, uap->new, uap->newlen, &rv, 0, 0);
2225 	if (error != 0)
2226 		return (error);
2227 	if (uap->oldlenp != NULL)
2228 		error = copyout(&rv, uap->oldlenp, sizeof(rv));
2229 
2230 	return (error);
2231 }
2232 
2233 /*
2234  * This is used from various compatibility syscalls too.  That's why name
2235  * must be in kernel space.
2236  */
2237 int
2238 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
2239     size_t *oldlenp, int inkernel, const void *new, size_t newlen,
2240     size_t *retval, int flags)
2241 {
2242 	int error = 0, memlocked;
2243 	struct sysctl_req req;
2244 
2245 	bzero(&req, sizeof req);
2246 
2247 	req.td = td;
2248 	req.flags = flags;
2249 
2250 	if (oldlenp) {
2251 		if (inkernel) {
2252 			req.oldlen = *oldlenp;
2253 		} else {
2254 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
2255 			if (error)
2256 				return (error);
2257 		}
2258 	}
2259 	req.validlen = req.oldlen;
2260 	req.oldptr = old;
2261 
2262 	if (new != NULL) {
2263 		req.newlen = newlen;
2264 		req.newptr = new;
2265 	}
2266 
2267 	req.oldfunc = sysctl_old_user;
2268 	req.newfunc = sysctl_new_user;
2269 	req.lock = REQ_UNWIRED;
2270 
2271 #ifdef KTRACE
2272 	if (KTRPOINT(curthread, KTR_SYSCTL))
2273 		ktrsysctl(name, namelen);
2274 #endif
2275 	memlocked = 0;
2276 	if (req.oldptr && req.oldlen > 4 * PAGE_SIZE) {
2277 		memlocked = 1;
2278 		sx_xlock(&sysctlmemlock);
2279 	}
2280 	CURVNET_SET(TD_TO_VNET(td));
2281 
2282 	for (;;) {
2283 		req.oldidx = 0;
2284 		req.newidx = 0;
2285 		error = sysctl_root(0, name, namelen, &req);
2286 		if (error != EAGAIN)
2287 			break;
2288 		kern_yield(PRI_USER);
2289 	}
2290 
2291 	CURVNET_RESTORE();
2292 
2293 	if (req.lock == REQ_WIRED && req.validlen > 0)
2294 		vsunlock(req.oldptr, req.validlen);
2295 	if (memlocked)
2296 		sx_xunlock(&sysctlmemlock);
2297 
2298 	if (error && error != ENOMEM)
2299 		return (error);
2300 
2301 	if (retval) {
2302 		if (req.oldptr && req.oldidx > req.validlen)
2303 			*retval = req.validlen;
2304 		else
2305 			*retval = req.oldidx;
2306 	}
2307 	return (error);
2308 }
2309 
2310 /*
2311  * Drain into a sysctl struct.  The user buffer should be wired if a page
2312  * fault would cause issue.
2313  */
2314 static int
2315 sbuf_sysctl_drain(void *arg, const char *data, int len)
2316 {
2317 	struct sysctl_req *req = arg;
2318 	int error;
2319 
2320 	error = SYSCTL_OUT(req, data, len);
2321 	KASSERT(error >= 0, ("Got unexpected negative value %d", error));
2322 	return (error == 0 ? len : -error);
2323 }
2324 
2325 struct sbuf *
2326 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
2327     struct sysctl_req *req)
2328 {
2329 
2330 	/* Supply a default buffer size if none given. */
2331 	if (buf == NULL && length == 0)
2332 		length = 64;
2333 	s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
2334 	sbuf_set_drain(s, sbuf_sysctl_drain, req);
2335 	return (s);
2336 }
2337 
2338 #ifdef DDB
2339 
2340 /* The current OID the debugger is working with */
2341 static struct sysctl_oid *g_ddb_oid;
2342 
2343 /* The current flags specified by the user */
2344 static int g_ddb_sysctl_flags;
2345 
2346 /* Check to see if the last sysctl printed */
2347 static int g_ddb_sysctl_printed;
2348 
2349 static const int ctl_sign[CTLTYPE+1] = {
2350 	[CTLTYPE_INT] = 1,
2351 	[CTLTYPE_LONG] = 1,
2352 	[CTLTYPE_S8] = 1,
2353 	[CTLTYPE_S16] = 1,
2354 	[CTLTYPE_S32] = 1,
2355 	[CTLTYPE_S64] = 1,
2356 };
2357 
2358 static const int ctl_size[CTLTYPE+1] = {
2359 	[CTLTYPE_INT] = sizeof(int),
2360 	[CTLTYPE_UINT] = sizeof(u_int),
2361 	[CTLTYPE_LONG] = sizeof(long),
2362 	[CTLTYPE_ULONG] = sizeof(u_long),
2363 	[CTLTYPE_S8] = sizeof(int8_t),
2364 	[CTLTYPE_S16] = sizeof(int16_t),
2365 	[CTLTYPE_S32] = sizeof(int32_t),
2366 	[CTLTYPE_S64] = sizeof(int64_t),
2367 	[CTLTYPE_U8] = sizeof(uint8_t),
2368 	[CTLTYPE_U16] = sizeof(uint16_t),
2369 	[CTLTYPE_U32] = sizeof(uint32_t),
2370 	[CTLTYPE_U64] = sizeof(uint64_t),
2371 };
2372 
2373 #define DB_SYSCTL_NAME_ONLY	0x001	/* Compare with -N */
2374 #define DB_SYSCTL_VALUE_ONLY	0x002	/* Compare with -n */
2375 #define DB_SYSCTL_OPAQUE	0x004	/* Compare with -o */
2376 #define DB_SYSCTL_HEX		0x008	/* Compare with -x */
2377 
2378 #define DB_SYSCTL_SAFE_ONLY	0x100	/* Only simple types */
2379 
2380 static const char db_sysctl_modifs[] = {
2381 	'N', 'n', 'o', 'x',
2382 };
2383 
2384 static const int db_sysctl_modif_values[] = {
2385 	DB_SYSCTL_NAME_ONLY, DB_SYSCTL_VALUE_ONLY,
2386 	DB_SYSCTL_OPAQUE, DB_SYSCTL_HEX,
2387 };
2388 
2389 /* Handlers considered safe to print while recursing */
2390 static int (* const db_safe_handlers[])(SYSCTL_HANDLER_ARGS) = {
2391 	sysctl_handle_bool,
2392 	sysctl_handle_8,
2393 	sysctl_handle_16,
2394 	sysctl_handle_32,
2395 	sysctl_handle_64,
2396 	sysctl_handle_int,
2397 	sysctl_handle_long,
2398 	sysctl_handle_string,
2399 	sysctl_handle_opaque,
2400 };
2401 
2402 /*
2403  * Use in place of sysctl_old_kernel to print sysctl values.
2404  *
2405  * Compare to the output handling in show_var from sbin/sysctl/sysctl.c
2406  */
2407 static int
2408 sysctl_old_ddb(struct sysctl_req *req, const void *ptr, size_t len)
2409 {
2410 	const u_char *val, *p;
2411 	const char *sep1;
2412 	size_t intlen, slen;
2413 	uintmax_t umv;
2414 	intmax_t mv;
2415 	int sign, ctltype, hexlen, xflag, error;
2416 
2417 	/* Suppress false-positive GCC uninitialized variable warnings */
2418 	mv = 0;
2419 	umv = 0;
2420 
2421 	slen = len;
2422 	val = p = ptr;
2423 
2424 	if (ptr == NULL) {
2425 		error = 0;
2426 		goto out;
2427 	}
2428 
2429 	/* We are going to print */
2430 	g_ddb_sysctl_printed = 1;
2431 
2432 	xflag = g_ddb_sysctl_flags & DB_SYSCTL_HEX;
2433 
2434 	ctltype = (g_ddb_oid->oid_kind & CTLTYPE);
2435 	sign = ctl_sign[ctltype];
2436 	intlen = ctl_size[ctltype];
2437 
2438 	switch (ctltype) {
2439 	case CTLTYPE_NODE:
2440 	case CTLTYPE_STRING:
2441 		db_printf("%.*s", (int) len, (const char *) p);
2442 		error = 0;
2443 		goto out;
2444 
2445 	case CTLTYPE_INT:
2446 	case CTLTYPE_UINT:
2447 	case CTLTYPE_LONG:
2448 	case CTLTYPE_ULONG:
2449 	case CTLTYPE_S8:
2450 	case CTLTYPE_S16:
2451 	case CTLTYPE_S32:
2452 	case CTLTYPE_S64:
2453 	case CTLTYPE_U8:
2454 	case CTLTYPE_U16:
2455 	case CTLTYPE_U32:
2456 	case CTLTYPE_U64:
2457 		hexlen = 2 + (intlen * CHAR_BIT + 3) / 4;
2458 		sep1 = "";
2459 		while (len >= intlen) {
2460 			switch (ctltype) {
2461 			case CTLTYPE_INT:
2462 			case CTLTYPE_UINT:
2463 				umv = *(const u_int *)p;
2464 				mv = *(const int *)p;
2465 				break;
2466 			case CTLTYPE_LONG:
2467 			case CTLTYPE_ULONG:
2468 				umv = *(const u_long *)p;
2469 				mv = *(const long *)p;
2470 				break;
2471 			case CTLTYPE_S8:
2472 			case CTLTYPE_U8:
2473 				umv = *(const uint8_t *)p;
2474 				mv = *(const int8_t *)p;
2475 				break;
2476 			case CTLTYPE_S16:
2477 			case CTLTYPE_U16:
2478 				umv = *(const uint16_t *)p;
2479 				mv = *(const int16_t *)p;
2480 				break;
2481 			case CTLTYPE_S32:
2482 			case CTLTYPE_U32:
2483 				umv = *(const uint32_t *)p;
2484 				mv = *(const int32_t *)p;
2485 				break;
2486 			case CTLTYPE_S64:
2487 			case CTLTYPE_U64:
2488 				umv = *(const uint64_t *)p;
2489 				mv = *(const int64_t *)p;
2490 				break;
2491 			}
2492 
2493 			db_printf("%s", sep1);
2494 			if (xflag)
2495 				db_printf("%#0*jx", hexlen, umv);
2496 			else if (!sign)
2497 				db_printf("%ju", umv);
2498 			else if (g_ddb_oid->oid_fmt[1] == 'K') {
2499 				/* Kelvins are currently unsupported. */
2500 				error = EOPNOTSUPP;
2501 				goto out;
2502 			} else
2503 				db_printf("%jd", mv);
2504 
2505 			sep1 = " ";
2506 			len -= intlen;
2507 			p += intlen;
2508 		}
2509 		error = 0;
2510 		goto out;
2511 
2512 	case CTLTYPE_OPAQUE:
2513 		/* TODO: Support struct functions. */
2514 
2515 		/* FALLTHROUGH */
2516 	default:
2517 		db_printf("Format:%s Length:%zu Dump:0x",
2518 		    g_ddb_oid->oid_fmt, len);
2519 		while (len-- && (xflag || p < val + 16))
2520 			db_printf("%02x", *p++);
2521 		if (!xflag && len > 16)
2522 			db_printf("...");
2523 		error = 0;
2524 		goto out;
2525 	}
2526 
2527 out:
2528 	req->oldidx += slen;
2529 	return (error);
2530 }
2531 
2532 /*
2533  * Avoid setting new sysctl values from the debugger
2534  */
2535 static int
2536 sysctl_new_ddb(struct sysctl_req *req, void *p, size_t l)
2537 {
2538 
2539 	if (!req->newptr)
2540 		return (0);
2541 
2542 	/* Changing sysctls from the debugger is currently unsupported */
2543 	return (EPERM);
2544 }
2545 
2546 /*
2547  * Run a sysctl handler with the DDB oldfunc and newfunc attached.
2548  * Instead of copying any output to a buffer we'll dump it right to
2549  * the console.
2550  */
2551 static int
2552 db_sysctl(struct sysctl_oid *oidp, int *name, u_int namelen,
2553     void *old, size_t *oldlenp, size_t *retval, int flags)
2554 {
2555 	struct sysctl_req req;
2556 	int error;
2557 
2558 	/* Setup the request */
2559 	bzero(&req, sizeof req);
2560 	req.td = kdb_thread;
2561 	req.oldfunc = sysctl_old_ddb;
2562 	req.newfunc = sysctl_new_ddb;
2563 	req.lock = REQ_UNWIRED;
2564 	if (oldlenp) {
2565 		req.oldlen = *oldlenp;
2566 	}
2567 	req.validlen = req.oldlen;
2568 	if (old) {
2569 		req.oldptr = old;
2570 	}
2571 
2572 	/* Setup our globals for sysctl_old_ddb */
2573 	g_ddb_oid = oidp;
2574 	g_ddb_sysctl_flags = flags;
2575 	g_ddb_sysctl_printed = 0;
2576 
2577 	error = sysctl_root(0, name, namelen, &req);
2578 
2579 	/* Reset globals */
2580 	g_ddb_oid = NULL;
2581 	g_ddb_sysctl_flags = 0;
2582 
2583 	if (retval) {
2584 		if (req.oldptr && req.oldidx > req.validlen)
2585 			*retval = req.validlen;
2586 		else
2587 			*retval = req.oldidx;
2588 	}
2589 	return (error);
2590 }
2591 
2592 /*
2593  * Show a sysctl's name
2594  */
2595 static void
2596 db_show_oid_name(int *oid, size_t nlen)
2597 {
2598 	struct sysctl_oid *oidp;
2599 	int qoid[CTL_MAXNAME+2];
2600 	int error;
2601 
2602 	qoid[0] = 0;
2603 	memcpy(qoid + 2, oid, nlen * sizeof(int));
2604 	qoid[1] = 1;
2605 
2606 	error = sysctl_find_oid(qoid, nlen + 2, &oidp, NULL, NULL);
2607 	if (error)
2608 		db_error("sysctl name oid");
2609 
2610 	error = db_sysctl(oidp, qoid, nlen + 2, NULL, NULL, NULL, 0);
2611 	if (error)
2612 		db_error("sysctl name");
2613 }
2614 
2615 /*
2616  * Check to see if an OID is safe to print from ddb.
2617  */
2618 static bool
2619 db_oid_safe(const struct sysctl_oid *oidp)
2620 {
2621 	for (unsigned int i = 0; i < nitems(db_safe_handlers); ++i) {
2622 		if (oidp->oid_handler == db_safe_handlers[i])
2623 			return (true);
2624 	}
2625 
2626 	return (false);
2627 }
2628 
2629 /*
2630  * Show a sysctl at a specific OID
2631  * Compare to the input handling in show_var from sbin/sysctl/sysctl.c
2632  */
2633 static int
2634 db_show_oid(struct sysctl_oid *oidp, int *oid, size_t nlen, int flags)
2635 {
2636 	int error, xflag, oflag, Nflag, nflag;
2637 	size_t len;
2638 
2639 	xflag = flags & DB_SYSCTL_HEX;
2640 	oflag = flags & DB_SYSCTL_OPAQUE;
2641 	nflag = flags & DB_SYSCTL_VALUE_ONLY;
2642 	Nflag = flags & DB_SYSCTL_NAME_ONLY;
2643 
2644 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_OPAQUE &&
2645 	    (!xflag && !oflag))
2646 		return (0);
2647 
2648 	if (Nflag) {
2649 		db_show_oid_name(oid, nlen);
2650 		error = 0;
2651 		goto out;
2652 	}
2653 
2654 	if (!nflag) {
2655 		db_show_oid_name(oid, nlen);
2656 		db_printf(": ");
2657 	}
2658 
2659 	if ((flags & DB_SYSCTL_SAFE_ONLY) && !db_oid_safe(oidp)) {
2660 		db_printf("Skipping, unsafe to print while recursing.");
2661 		error = 0;
2662 		goto out;
2663 	}
2664 
2665 	/* Try once, and ask about the size */
2666 	len = 0;
2667 	error = db_sysctl(oidp, oid, nlen,
2668 	    NULL, NULL, &len, flags);
2669 	if (error)
2670 		goto out;
2671 
2672 	if (!g_ddb_sysctl_printed)
2673 		/* Lie about the size */
2674 		error = db_sysctl(oidp, oid, nlen,
2675 		    (void *) 1, &len, NULL, flags);
2676 
2677 out:
2678 	db_printf("\n");
2679 	return (error);
2680 }
2681 
2682 /*
2683  * Show all sysctls under a specific OID
2684  * Compare to sysctl_all from sbin/sysctl/sysctl.c
2685  */
2686 static int
2687 db_show_sysctl_all(int *oid, size_t len, int flags)
2688 {
2689 	struct sysctl_oid *oidp;
2690 	int name1[CTL_MAXNAME + 2], name2[CTL_MAXNAME + 2];
2691 	size_t l1, l2;
2692 
2693 	name1[0] = CTL_SYSCTL;
2694 	name1[1] = CTL_SYSCTL_NEXT;
2695 	l1 = 2;
2696 	if (len) {
2697 		memcpy(name1+2, oid, len * sizeof(int));
2698 		l1 +=len;
2699 	} else {
2700 		name1[2] = 1;
2701 		l1++;
2702 	}
2703 	for (;;) {
2704 		int i, error;
2705 
2706 		l2 = sizeof(name2);
2707 		error = kernel_sysctl(kdb_thread, name1, l1,
2708 		    name2, &l2, NULL, 0, &l2, 0);
2709 		if (error != 0) {
2710 			if (error == ENOENT)
2711 				return (0);
2712 			else
2713 				db_error("sysctl(getnext)");
2714 		}
2715 
2716 		l2 /= sizeof(int);
2717 
2718 		if (l2 < (unsigned int)len)
2719 			return (0);
2720 
2721 		for (i = 0; i < len; i++)
2722 			if (name2[i] != oid[i])
2723 				return (0);
2724 
2725 		/* Find the OID in question */
2726 		error = sysctl_find_oid(name2, l2, &oidp, NULL, NULL);
2727 		if (error)
2728 			return (error);
2729 
2730 		i = db_show_oid(oidp, name2, l2, flags | DB_SYSCTL_SAFE_ONLY);
2731 
2732 		if (db_pager_quit)
2733 			return (0);
2734 
2735 		memcpy(name1+2, name2, l2 * sizeof(int));
2736 		l1 = 2 + l2;
2737 	}
2738 }
2739 
2740 /*
2741  * Show a sysctl by its user facing string
2742  */
2743 static int
2744 db_sysctlbyname(char *name, int flags)
2745 {
2746 	struct sysctl_oid *oidp;
2747 	int oid[CTL_MAXNAME];
2748 	int error, nlen;
2749 
2750 	error = name2oid(name, oid, &nlen, &oidp);
2751 	if (error) {
2752 		return (error);
2753 	}
2754 
2755 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2756 		db_show_sysctl_all(oid, nlen, flags);
2757 	} else {
2758 		error = db_show_oid(oidp, oid, nlen, flags);
2759 	}
2760 
2761 	return (error);
2762 }
2763 
2764 static void
2765 db_sysctl_cmd_usage(void)
2766 {
2767 	db_printf(
2768 	    " sysctl [/Nnox] <sysctl>					    \n"
2769 	    "								    \n"
2770 	    " <sysctl> The name of the sysctl to show.			    \n"
2771 	    "								    \n"
2772 	    " Show a sysctl by hooking into SYSCTL_IN and SYSCTL_OUT.	    \n"
2773 	    " This will work for most sysctls, but should not be used	    \n"
2774 	    " with sysctls that are known to malloc.			    \n"
2775 	    "								    \n"
2776 	    " While recursing any \"unsafe\" sysctls will be skipped.	    \n"
2777 	    " Call sysctl directly on the sysctl to try printing the	    \n"
2778 	    " skipped sysctl. This is unsafe and may make the ddb	    \n"
2779 	    " session unusable.						    \n"
2780 	    "								    \n"
2781 	    " Arguments:						    \n"
2782 	    "	/N	Display only the name of the sysctl.		    \n"
2783 	    "	/n	Display only the value of the sysctl.		    \n"
2784 	    "	/o	Display opaque values.				    \n"
2785 	    "	/x	Display the sysctl in hex.			    \n"
2786 	    "								    \n"
2787 	    "For example:						    \n"
2788 	    "sysctl vm.v_free_min					    \n"
2789 	    "vn.v_free_min: 12669					    \n"
2790 	    );
2791 }
2792 
2793 /*
2794  * Show a specific sysctl similar to sysctl (8).
2795  */
2796 DB_FUNC(sysctl, db_sysctl_cmd, db_cmd_table, CS_OWN, NULL)
2797 {
2798 	char name[TOK_STRING_SIZE];
2799 	int error, i, t, flags;
2800 
2801 	/* Parse the modifiers */
2802 	t = db_read_token();
2803 	if (t == tSLASH || t == tMINUS) {
2804 		t = db_read_token();
2805 		if (t != tIDENT) {
2806 			db_printf("Bad modifier\n");
2807 			error = EINVAL;
2808 			goto out;
2809 		}
2810 		db_strcpy(modif, db_tok_string);
2811 	}
2812 	else {
2813 		db_unread_token(t);
2814 		modif[0] = '\0';
2815 	}
2816 
2817 	flags = 0;
2818 	for (i = 0; i < nitems(db_sysctl_modifs); i++) {
2819 		if (strchr(modif, db_sysctl_modifs[i])) {
2820 			flags |= db_sysctl_modif_values[i];
2821 		}
2822 	}
2823 
2824 	/* Parse the sysctl names */
2825 	t = db_read_token();
2826 	if (t != tIDENT) {
2827 		db_printf("Need sysctl name\n");
2828 		error = EINVAL;
2829 		goto out;
2830 	}
2831 
2832 	/* Copy the name into a temporary buffer */
2833 	db_strcpy(name, db_tok_string);
2834 
2835 	/* Ensure there is no trailing cruft */
2836 	t = db_read_token();
2837 	if (t != tEOL) {
2838 		db_printf("Unexpected sysctl argument\n");
2839 		error = EINVAL;
2840 		goto out;
2841 	}
2842 
2843 	error = db_sysctlbyname(name, flags);
2844 	if (error == ENOENT) {
2845 		db_printf("unknown oid: '%s'\n", db_tok_string);
2846 		goto out;
2847 	} else if (error) {
2848 		db_printf("%s: error: %d\n", db_tok_string, error);
2849 		goto out;
2850 	}
2851 
2852 out:
2853 	/* Ensure we eat all of our text */
2854 	db_flush_lex();
2855 
2856 	if (error == EINVAL) {
2857 		db_sysctl_cmd_usage();
2858 	}
2859 }
2860 
2861 #endif /* DDB */
2862