2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1989, 1993
5 * The Regents of the University of California. All rights reserved.
7 * This code is derived from software contributed to Berkeley by
8 * Mike Karels at Berkeley Software Design, Inc.
10 * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
11 * project, to make these variables more userfriendly.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
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.
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
37 * @(#)kern_sysctl.c 8.4 (Berkeley) 4/14/94
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
43 #include "opt_capsicum.h"
45 #include "opt_ktrace.h"
46 #include "opt_sysctl.h"
48 #include <sys/param.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>
61 #include <sys/mutex.h>
62 #include <sys/rmlock.h>
65 #include <sys/sysproto.h>
68 #include <sys/ktrace.h>
73 #include <ddb/db_lex.h>
78 #include <security/mac/mac_framework.h>
81 #include <vm/vm_extern.h>
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");
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.
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.
100 static struct rmlock sysctllock;
101 static struct sx __exclusive_cache_line sysctlmemlock;
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", \
113 #define SYSCTL_SLEEP(ch, wmesg, timo) \
114 rm_sleep(ch, &sysctllock, 0, wmesg, timo)
116 static int sysctl_root(SYSCTL_HANDLER_ARGS);
119 struct sysctl_oid_list sysctl__children = SLIST_HEAD_INITIALIZER(&sysctl__children);
121 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
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);
126 static struct sysctl_oid *
127 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
129 struct sysctl_oid *oidp;
131 SYSCTL_ASSERT_LOCKED();
132 SLIST_FOREACH(oidp, list, oid_link) {
133 if (strcmp(oidp->oid_name, name) == 0) {
141 * Initialization of the MIB tree.
143 * Order by number in each list.
160 sysctl_root_handler_locked(struct sysctl_oid *oid, void *arg1, intmax_t arg2,
161 struct sysctl_req *req, struct rm_priotracker *tracker)
165 if (oid->oid_kind & CTLFLAG_DYN)
166 atomic_add_int(&oid->oid_running, 1);
169 SYSCTL_RUNLOCK(tracker);
173 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
175 error = oid->oid_handler(oid, arg1, arg2, req);
176 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
179 KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
182 SYSCTL_RLOCK(tracker);
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);
196 sysctl_load_tunable_by_oid_locked(struct sysctl_oid *oidp)
198 struct sysctl_req req;
199 struct sysctl_oid *curr;
202 ssize_t rem = sizeof(path);
204 uint8_t data[512] __aligned(sizeof(uint64_t));
210 for (curr = oidp; curr != NULL; curr = SYSCTL_PARENT(curr)) {
211 len = strlen(curr->oid_name);
216 printf("OID path exceeds %d bytes\n", (int)sizeof(path));
219 memcpy(path + rem, curr->oid_name, len);
221 path[rem + len] = '.';
224 memset(&req, 0, sizeof(req));
227 req.oldfunc = sysctl_old_kernel;
228 req.newfunc = sysctl_new_kernel;
229 req.lock = REQ_UNWIRED;
231 switch (oidp->oid_kind & CTLTYPE) {
233 if (getenv_array(path + rem, data, sizeof(data), &size,
234 sizeof(int), GETENV_SIGNED) == 0)
240 if (getenv_array(path + rem, data, sizeof(data), &size,
241 sizeof(int), GETENV_UNSIGNED) == 0)
247 if (getenv_array(path + rem, data, sizeof(data), &size,
248 sizeof(long), GETENV_SIGNED) == 0)
254 if (getenv_array(path + rem, data, sizeof(data), &size,
255 sizeof(long), GETENV_UNSIGNED) == 0)
261 if (getenv_array(path + rem, data, sizeof(data), &size,
262 sizeof(int8_t), GETENV_SIGNED) == 0)
268 if (getenv_array(path + rem, data, sizeof(data), &size,
269 sizeof(int16_t), GETENV_SIGNED) == 0)
275 if (getenv_array(path + rem, data, sizeof(data), &size,
276 sizeof(int32_t), GETENV_SIGNED) == 0)
282 if (getenv_array(path + rem, data, sizeof(data), &size,
283 sizeof(int64_t), GETENV_SIGNED) == 0)
289 if (getenv_array(path + rem, data, sizeof(data), &size,
290 sizeof(uint8_t), GETENV_UNSIGNED) == 0)
296 if (getenv_array(path + rem, data, sizeof(data), &size,
297 sizeof(uint16_t), GETENV_UNSIGNED) == 0)
303 if (getenv_array(path + rem, data, sizeof(data), &size,
304 sizeof(uint32_t), GETENV_UNSIGNED) == 0)
310 if (getenv_array(path + rem, data, sizeof(data), &size,
311 sizeof(uint64_t), GETENV_UNSIGNED) == 0)
317 penv = kern_getenv(path + rem);
320 req.newlen = strlen(penv);
326 error = sysctl_root_handler_locked(oidp, oidp->oid_arg1,
327 oidp->oid_arg2, &req, NULL);
329 printf("Setting sysctl %s failed: %d\n", path + rem, error);
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.
340 sysctl_search_oid(struct sysctl_oid **nodes, struct sysctl_oid *needle)
344 SYSCTL_ASSERT_LOCKED();
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);
352 nodes[indx] = SLIST_NEXT(nodes[indx], oid_link);
354 if (nodes[indx] == needle)
357 if (nodes[indx] == NULL) {
362 if ((nodes[indx]->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
371 sysctl_warn_reuse(const char *func, struct sysctl_oid *leaf)
373 struct sysctl_oid *nodes[CTL_MAXNAME];
378 (void)sbuf_new(&sb, buf, sizeof(buf), SBUF_FIXEDLEN | SBUF_INCLUDENUL);
379 sbuf_set_drain(&sb, sbuf_printf_drain, NULL);
381 sbuf_printf(&sb, "%s: can't re-use a leaf (", __func__);
383 memset(nodes, 0, sizeof(nodes));
384 rc = sysctl_search_oid(nodes, leaf);
386 for (i = 0; i < rc; i++)
387 sbuf_printf(&sb, "%s%.*s", nodes[i]->oid_name,
390 sbuf_printf(&sb, "%s", leaf->oid_name);
392 sbuf_printf(&sb, ")!\n");
394 (void)sbuf_finish(&sb);
399 sysctl_reuse_test(SYSCTL_HANDLER_ARGS)
401 struct rm_priotracker tracker;
403 SYSCTL_RLOCK(&tracker);
404 sysctl_warn_reuse(__func__, oidp);
405 SYSCTL_RUNLOCK(&tracker);
408 SYSCTL_PROC(_sysctl, OID_AUTO, reuse_test,
409 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0, sysctl_reuse_test, "-",
414 sysctl_register_oid(struct sysctl_oid *oidp)
416 struct sysctl_oid_list *parent = oidp->oid_parent;
417 struct sysctl_oid *p;
418 struct sysctl_oid *q;
423 * First check if another oid with the same name already
424 * exists in the parent's list.
426 SYSCTL_ASSERT_WLOCKED();
427 p = sysctl_find_oidname(oidp->oid_name, parent);
429 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
433 sysctl_warn_reuse(__func__, p);
437 /* get current OID number */
438 oid_number = oidp->oid_number;
441 #error "OID_AUTO is expected to be a negative value"
444 * Any negative OID number qualifies as OID_AUTO. Valid OID
445 * numbers should always be positive.
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.
451 if (oid_number < 0) {
455 * By decrementing the next OID number we spend less
456 * time inserting the OIDs into a sorted list.
458 if (--newoid < CTL_AUTO_START)
465 * Insert the OID into the parent's list sorted by OID number.
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 */
479 panic("sysctl: Out of OID numbers\n");
484 } else if (oid_number < p->oid_number)
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);
494 /* update the OID number, if any */
495 oidp->oid_number = oid_number;
497 SLIST_INSERT_AFTER(q, oidp, oid_link);
499 SLIST_INSERT_HEAD(parent, oidp, oid_link);
501 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE &&
503 (oidp->oid_kind & CTLFLAG_VNET) == 0 &&
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);
515 sysctl_register_disabled_oid(struct sysctl_oid *oidp)
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.
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);
531 sysctl_enable_oid(struct sysctl_oid *oidp)
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"));
540 KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) != 0,
541 ("enabling already enabled sysctl oid"));
542 oidp->oid_kind &= ~CTLFLAG_DORMANT;
546 sysctl_unregister_oid(struct sysctl_oid *oidp)
548 struct sysctl_oid *p;
551 SYSCTL_ASSERT_WLOCKED();
552 if (oidp->oid_number == OID_AUTO) {
556 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
558 SLIST_REMOVE(oidp->oid_parent, oidp,
559 sysctl_oid, oid_link);
567 * This can happen when a module fails to register and is
568 * being unloaded afterwards. It should not be a panic()
572 printf("%s: failed(%d) to unregister sysctl(%s)\n",
573 __func__, error, oidp->oid_name);
577 /* Initialize a new context to keep track of dynamically added sysctls. */
579 sysctl_ctx_init(struct sysctl_ctx_list *c)
587 * No locking here, the caller is responsible for not adding
588 * new nodes to a context until after this function has
595 /* Free the context, and destroy all dynamic oids registered in this context */
597 sysctl_ctx_free(struct sysctl_ctx_list *clist)
599 struct sysctl_ctx_entry *e, *e1;
604 * First perform a "dry run" to check if it's ok to remove oids.
606 * XXX This algorithm is a hack. But I don't know any
607 * XXX better solution for now...
610 TAILQ_FOREACH(e, clist, link) {
611 error = sysctl_remove_oid_locked(e->entry, 0, 0);
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
621 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
623 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
625 sysctl_register_oid(e1->entry);
626 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
632 /* Now really delete the entries */
633 e = TAILQ_FIRST(clist);
635 e1 = TAILQ_NEXT(e, link);
636 error = sysctl_remove_oid_locked(e->entry, 1, 0);
638 panic("sysctl_remove_oid: corrupt tree, entry: %s",
640 free(e, M_SYSCTLOID);
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)
651 struct sysctl_ctx_entry *e;
653 SYSCTL_ASSERT_WLOCKED();
654 if (clist == NULL || oidp == NULL)
656 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
658 TAILQ_INSERT_HEAD(clist, e, link);
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)
666 struct sysctl_ctx_entry *e;
668 SYSCTL_ASSERT_WLOCKED();
669 if (clist == NULL || oidp == NULL)
671 TAILQ_FOREACH(e, clist, link) {
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().
684 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
686 struct sysctl_ctx_entry *e;
688 if (clist == NULL || oidp == NULL)
691 e = sysctl_ctx_entry_find(clist, oidp);
693 TAILQ_REMOVE(clist, e, link);
695 free(e, M_SYSCTLOID);
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
710 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
715 error = sysctl_remove_oid_locked(oidp, del, recurse);
721 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
722 int del, int recurse)
724 struct sysctl_oid *p, *tmp;
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);
742 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
744 struct sysctl_oid *p, *tmp;
747 SYSCTL_ASSERT_WLOCKED();
750 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
751 printf("Warning: can't remove non-dynamic nodes (%s)!\n",
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,
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) {
767 printf("Warning: failed attempt to "
768 "remove oid %s with child %s\n",
769 oidp->oid_name, p->oid_name);
772 error = sysctl_remove_oid_locked(p, del,
779 if (oidp->oid_refcnt > 1 ) {
782 if (oidp->oid_refcnt == 0) {
783 printf("Warning: bad oid_refcnt=%u (%s)!\n",
784 oidp->oid_refcnt, oidp->oid_name);
787 sysctl_unregister_oid(oidp);
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.
795 while (oidp->oid_running > 0) {
796 oidp->oid_kind |= CTLFLAG_DYING;
797 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
800 free(__DECONST(char *, oidp->oid_descr),
803 free(__DECONST(char *, oidp->oid_label),
805 free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
806 free(oidp, M_SYSCTLOID);
812 * Create new sysctls at run time.
813 * clist may point to a valid context initialized with sysctl_ctx_init().
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,
821 struct sysctl_oid *oidp;
823 /* You have to hook up somewhere.. */
826 /* Check if the node already exists, otherwise create it */
828 oidp = sysctl_find_oidname(name, parent);
830 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
832 /* Update the context */
834 sysctl_ctx_entry_add(clist, oidp);
838 sysctl_warn_reuse(__func__, oidp);
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;
855 oidp->oid_descr = strdup(descr, M_SYSCTLOID);
857 oidp->oid_label = strdup(label, M_SYSCTLOID);
858 /* Update the context, if used */
860 sysctl_ctx_entry_add(clist, oidp);
861 /* Register this oid */
862 sysctl_register_oid(oidp);
868 * Rename an existing oid.
871 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
876 newname = strdup(name, M_SYSCTLOID);
878 oldname = __DECONST(char *, oidp->oid_name);
879 oidp->oid_name = newname;
881 free(oldname, M_SYSCTLOID);
885 * Reparent an existing oid.
888 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
890 struct sysctl_oid *oidp;
893 if (oid->oid_parent == parent) {
897 oidp = sysctl_find_oidname(oid->oid_name, parent);
902 sysctl_unregister_oid(oid);
903 oid->oid_parent = parent;
904 oid->oid_number = OID_AUTO;
905 sysctl_register_oid(oid);
911 * Register the kernel's oids on startup.
913 SET_DECLARE(sysctl_set, struct sysctl_oid);
916 sysctl_register_all(void *arg)
918 struct sysctl_oid **oidp;
920 sx_init(&sysctlmemlock, "sysctl mem");
923 SET_FOREACH(oidp, sysctl_set)
924 sysctl_register_oid(*oidp);
927 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_FIRST, sysctl_register_all, NULL);
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.
940 * {CTL_SYSCTL, CTL_SYSCTL_DEBUG} printf the entire MIB-tree.
941 * {CTL_SYSCTL, CTL_SYSCTL_NAME, ...} return the name of the "..."
943 * {CTL_SYSCTL, CTL_SYSCTL_NEXT, ...} return the next OID.
944 * {CTL_SYSCTL, CTL_SYSCTL_NAME2OID} return the OID of the name in
946 * {CTL_SYSCTL, CTL_SYSCTL_OIDFMT, ...} return the kind & format info
948 * {CTL_SYSCTL, CTL_SYSCTL_OIDDESCR, ...} return the description of the
950 * {CTL_SYSCTL, CTL_SYSCTL_OIDLABEL, ...} return the aggregation label of
956 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
959 struct sysctl_oid *oidp;
961 SYSCTL_ASSERT_LOCKED();
962 SLIST_FOREACH(oidp, l, oid_link) {
967 printf("%d %s ", oidp->oid_number, oidp->oid_name);
970 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
971 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
973 if (oidp->oid_handler)
976 switch (oidp->oid_kind & CTLTYPE) {
979 if (!oidp->oid_handler) {
980 sysctl_sysctl_debug_dump_node(
981 SYSCTL_CHILDREN(oidp), i + 2);
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");
1005 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
1007 struct rm_priotracker tracker;
1010 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
1013 SYSCTL_RLOCK(&tracker);
1014 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
1015 SYSCTL_RUNLOCK(&tracker);
1019 SYSCTL_PROC(_sysctl, CTL_SYSCTL_DEBUG, debug, CTLTYPE_STRING | CTLFLAG_RD |
1020 CTLFLAG_MPSAFE, 0, 0, sysctl_sysctl_debug, "-", "");
1024 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
1026 int *name = (int *) arg1;
1027 u_int namelen = arg2;
1029 struct sysctl_oid *oid;
1030 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
1031 struct rm_priotracker tracker;
1034 error = sysctl_wire_old_buffer(req, 0);
1038 SYSCTL_RLOCK(&tracker);
1041 snprintf(buf,sizeof(buf),"%d",*name);
1043 error = SYSCTL_OUT(req, ".", 1);
1045 error = SYSCTL_OUT(req, buf, strlen(buf));
1053 SLIST_FOREACH(oid, lsp, oid_link) {
1054 if (oid->oid_number != *name)
1058 error = SYSCTL_OUT(req, ".", 1);
1060 error = SYSCTL_OUT(req, oid->oid_name,
1061 strlen(oid->oid_name));
1068 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1071 if (oid->oid_handler)
1074 lsp2 = SYSCTL_CHILDREN(oid);
1079 error = SYSCTL_OUT(req, "", 1);
1081 SYSCTL_RUNLOCK(&tracker);
1086 * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
1089 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NAME, name, CTLFLAG_RD |
1090 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_name, "");
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)
1096 struct sysctl_oid *oidp;
1098 SYSCTL_ASSERT_LOCKED();
1100 SLIST_FOREACH(oidp, lsp, oid_link) {
1101 *next = oidp->oid_number;
1104 if ((oidp->oid_kind & (CTLFLAG_SKIP | CTLFLAG_DORMANT)) != 0)
1108 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1110 if (oidp->oid_handler)
1111 /* We really should call the handler here...*/
1113 lsp = SYSCTL_CHILDREN(oidp);
1114 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
1115 len, level+1, oidpp))
1120 if (oidp->oid_number < *name)
1123 if (oidp->oid_number > *name) {
1124 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1126 if (oidp->oid_handler)
1128 lsp = SYSCTL_CHILDREN(oidp);
1129 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
1130 next+1, len, level+1, oidpp))
1134 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1137 if (oidp->oid_handler)
1140 lsp = SYSCTL_CHILDREN(oidp);
1141 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
1142 len, level+1, oidpp))
1153 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
1155 int *name = (int *) arg1;
1156 u_int namelen = arg2;
1158 struct sysctl_oid *oid;
1159 struct sysctl_oid_list *lsp = &sysctl__children;
1160 struct rm_priotracker tracker;
1161 int newoid[CTL_MAXNAME];
1163 SYSCTL_RLOCK(&tracker);
1164 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
1165 SYSCTL_RUNLOCK(&tracker);
1168 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
1173 * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
1176 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NEXT, next, CTLFLAG_RD |
1177 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_next, "");
1180 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
1182 struct sysctl_oid *oidp;
1183 struct sysctl_oid_list *lsp = &sysctl__children;
1186 SYSCTL_ASSERT_LOCKED();
1188 for (*len = 0; *len < CTL_MAXNAME;) {
1189 p = strsep(&name, ".");
1191 oidp = SLIST_FIRST(lsp);
1192 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
1195 if (strcmp(p, oidp->oid_name) == 0)
1198 *oid++ = oidp->oid_number;
1201 if (name == NULL || *name == '\0') {
1207 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1210 if (oidp->oid_handler)
1213 lsp = SYSCTL_CHILDREN(oidp);
1219 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
1222 int error, oid[CTL_MAXNAME], len = 0;
1223 struct sysctl_oid *op = NULL;
1224 struct rm_priotracker tracker;
1229 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
1230 return (ENAMETOOLONG);
1233 if (req->newlen >= sizeof(buf))
1234 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
1236 error = SYSCTL_IN(req, p, req->newlen);
1243 p [req->newlen] = '\0';
1245 SYSCTL_RLOCK(&tracker);
1246 error = name2oid(p, oid, &len, &op);
1247 SYSCTL_RUNLOCK(&tracker);
1255 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
1260 * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
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", "");
1268 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
1270 struct sysctl_oid *oid;
1271 struct rm_priotracker tracker;
1274 error = sysctl_wire_old_buffer(req, 0);
1278 SYSCTL_RLOCK(&tracker);
1279 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1283 if (oid->oid_fmt == NULL) {
1287 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
1290 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
1292 SYSCTL_RUNLOCK(&tracker);
1297 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDFMT, oidfmt, CTLFLAG_RD |
1298 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidfmt, "");
1301 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
1303 struct sysctl_oid *oid;
1304 struct rm_priotracker tracker;
1307 error = sysctl_wire_old_buffer(req, 0);
1311 SYSCTL_RLOCK(&tracker);
1312 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1316 if (oid->oid_descr == NULL) {
1320 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
1322 SYSCTL_RUNLOCK(&tracker);
1326 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDDESCR, oiddescr, CTLFLAG_RD |
1327 CTLFLAG_MPSAFE|CTLFLAG_CAPRD, sysctl_sysctl_oiddescr, "");
1330 sysctl_sysctl_oidlabel(SYSCTL_HANDLER_ARGS)
1332 struct sysctl_oid *oid;
1333 struct rm_priotracker tracker;
1336 error = sysctl_wire_old_buffer(req, 0);
1340 SYSCTL_RLOCK(&tracker);
1341 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1345 if (oid->oid_label == NULL) {
1349 error = SYSCTL_OUT(req, oid->oid_label, strlen(oid->oid_label) + 1);
1351 SYSCTL_RUNLOCK(&tracker);
1355 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDLABEL, oidlabel, CTLFLAG_RD |
1356 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidlabel, "");
1359 * Default "handler" functions.
1365 * a variable: point arg1 at it.
1366 * a constant: pass it in arg2.
1370 sysctl_handle_bool(SYSCTL_HANDLER_ARGS)
1376 * Attempt to get a coherent snapshot by making a copy of the data.
1379 temp = *(bool *)arg1 ? 1 : 0;
1381 temp = arg2 ? 1 : 0;
1383 error = SYSCTL_OUT(req, &temp, sizeof(temp));
1384 if (error || !req->newptr)
1390 error = SYSCTL_IN(req, &temp, sizeof(temp));
1392 *(bool *)arg1 = temp ? 1 : 0;
1398 * Handle an int8_t, signed or unsigned.
1400 * a variable: point arg1 at it.
1401 * a constant: pass it in arg2.
1405 sysctl_handle_8(SYSCTL_HANDLER_ARGS)
1411 * Attempt to get a coherent snapshot by making a copy of the data.
1414 tmpout = *(int8_t *)arg1;
1417 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1419 if (error || !req->newptr)
1425 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1430 * Handle an int16_t, signed or unsigned.
1432 * a variable: point arg1 at it.
1433 * a constant: pass it in arg2.
1437 sysctl_handle_16(SYSCTL_HANDLER_ARGS)
1443 * Attempt to get a coherent snapshot by making a copy of the data.
1446 tmpout = *(int16_t *)arg1;
1449 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1451 if (error || !req->newptr)
1457 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1462 * Handle an int32_t, signed or unsigned.
1464 * a variable: point arg1 at it.
1465 * a constant: pass it in arg2.
1469 sysctl_handle_32(SYSCTL_HANDLER_ARGS)
1475 * Attempt to get a coherent snapshot by making a copy of the data.
1478 tmpout = *(int32_t *)arg1;
1481 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1483 if (error || !req->newptr)
1489 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1494 * Handle an int, signed or unsigned.
1496 * a variable: point arg1 at it.
1497 * a constant: pass it in arg2.
1501 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
1503 int tmpout, error = 0;
1506 * Attempt to get a coherent snapshot by making a copy of the data.
1509 tmpout = *(int *)arg1;
1512 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1514 if (error || !req->newptr)
1520 error = SYSCTL_IN(req, arg1, sizeof(int));
1525 * Based on on sysctl_handle_int() convert milliseconds into ticks.
1526 * Note: this is used by TCP.
1530 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1535 s = (int)((int64_t)tt * 1000 / hz);
1537 error = sysctl_handle_int(oidp, &s, 0, req);
1538 if (error || !req->newptr)
1541 tt = (int)((int64_t)s * hz / 1000);
1551 * Handle a long, signed or unsigned.
1553 * a variable: point arg1 at it.
1554 * a constant: pass it in arg2.
1558 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1567 * Attempt to get a coherent snapshot by making a copy of the data.
1570 tmplong = *(long *)arg1;
1574 if (req->flags & SCTL_MASK32) {
1576 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1579 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1581 if (error || !req->newptr)
1587 else if (req->flags & SCTL_MASK32) {
1588 error = SYSCTL_IN(req, &tmpint, sizeof(int));
1589 *(long *)arg1 = (long)tmpint;
1593 error = SYSCTL_IN(req, arg1, sizeof(long));
1598 * Handle a 64 bit int, signed or unsigned.
1600 * a variable: point arg1 at it.
1601 * a constant: pass it in arg2.
1604 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1610 * Attempt to get a coherent snapshot by making a copy of the data.
1613 tmpout = *(uint64_t *)arg1;
1616 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1618 if (error || !req->newptr)
1624 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1629 * Handle our generic '\0' terminated 'C' string.
1631 * a variable string: point arg1 at it, arg2 is max length.
1632 * a constant string: point arg1 at it, arg2 is zero.
1636 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1639 int error = 0, ro_string = 0;
1642 * A zero-length buffer indicates a fixed size read-only
1643 * string. In ddb, don't worry about trying to make a malloced
1646 if (arg2 == 0 || kdb_active) {
1647 arg2 = strlen((char *)arg1) + 1;
1651 if (req->oldptr != NULL) {
1657 /* try to make a coherent snapshot of the string */
1658 tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK);
1659 memcpy(tmparg, arg1, arg2);
1662 outlen = strnlen(tmparg, arg2 - 1) + 1;
1663 error = SYSCTL_OUT(req, tmparg, outlen);
1666 free(tmparg, M_SYSCTLTMP);
1668 outlen = strnlen((char *)arg1, arg2 - 1) + 1;
1669 error = SYSCTL_OUT(req, NULL, outlen);
1671 if (error || !req->newptr)
1674 if ((req->newlen - req->newidx) >= arg2) {
1677 arg2 = (req->newlen - req->newidx);
1678 error = SYSCTL_IN(req, arg1, arg2);
1679 ((char *)arg1)[arg2] = '\0';
1685 * Handle any kind of opaque data.
1686 * arg1 points to it, arg2 is the size.
1690 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1694 struct sysctl_req req2;
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.
1706 generation = curthread->td_generation;
1707 error = SYSCTL_OUT(req, arg1, arg2);
1711 if (generation != curthread->td_generation && tries < 3) {
1716 error = SYSCTL_IN(req, arg1, arg2);
1722 * Based on on sysctl_handle_int() convert microseconds to a sbintime.
1725 sysctl_usec_to_sbintime(SYSCTL_HANDLER_ARGS)
1731 tt = *(int64_t *)arg1;
1734 error = sysctl_handle_64(oidp, &sb, 0, req);
1735 if (error || !req->newptr)
1739 *(int64_t *)arg1 = tt;
1745 * Based on on sysctl_handle_int() convert milliseconds to a sbintime.
1748 sysctl_msec_to_sbintime(SYSCTL_HANDLER_ARGS)
1754 tt = *(int64_t *)arg1;
1757 error = sysctl_handle_64(oidp, &sb, 0, req);
1758 if (error || !req->newptr)
1762 *(int64_t *)arg1 = tt;
1768 * Convert seconds to a struct timeval. Intended for use with
1769 * intervals and thus does not permit negative seconds.
1772 sysctl_sec_to_timeval(SYSCTL_HANDLER_ARGS)
1780 error = sysctl_handle_int(oidp, &secs, 0, req);
1781 if (error || req->newptr == NULL)
1792 * Transfer functions to/from kernel space.
1793 * XXX: rather untested at this point
1796 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1802 if (req->oldlen <= req->oldidx)
1805 if (i > req->oldlen - req->oldidx)
1806 i = req->oldlen - req->oldidx;
1808 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1811 if (req->oldptr && i != l)
1817 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1821 if (req->newlen - req->newidx < l)
1823 bcopy((const char *)req->newptr + req->newidx, p, l);
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)
1833 struct sysctl_req req;
1835 bzero(&req, sizeof req);
1841 req.oldlen = *oldlenp;
1843 req.validlen = req.oldlen;
1850 req.newlen = newlen;
1854 req.oldfunc = sysctl_old_kernel;
1855 req.newfunc = sysctl_new_kernel;
1856 req.lock = REQ_UNWIRED;
1858 error = sysctl_root(0, name, namelen, &req);
1860 if (req.lock == REQ_WIRED && req.validlen > 0)
1861 vsunlock(req.oldptr, req.validlen);
1863 if (error && error != ENOMEM)
1867 if (req.oldptr && req.oldidx > req.validlen)
1868 *retval = req.validlen;
1870 *retval = req.oldidx;
1876 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1877 void *new, size_t newlen, size_t *retval, int flags)
1879 int oid[CTL_MAXNAME];
1880 size_t oidlen, plen;
1883 oid[0] = CTL_SYSCTL;
1884 oid[1] = CTL_SYSCTL_NAME2OID;
1885 oidlen = sizeof(oid);
1887 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1888 (void *)name, strlen(name), &plen, flags);
1892 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1893 new, newlen, retval, flags);
1898 * Transfer function to/from user space.
1901 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1903 size_t i, len, origidx;
1906 origidx = req->oldidx;
1908 if (req->oldptr == NULL)
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.
1915 if (req->lock != REQ_WIRED)
1916 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1917 "sysctl_old_user()");
1919 len = req->validlen;
1923 if (i > len - origidx)
1925 if (req->lock == REQ_WIRED) {
1926 error = copyout_nofault(p, (char *)req->oldptr +
1929 error = copyout(p, (char *)req->oldptr + origidx, i);
1939 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1945 if (req->newlen - req->newidx < l)
1947 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1948 "sysctl_new_user()");
1949 error = copyin((const char *)req->newptr + req->newidx, p, l);
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.
1959 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1964 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1966 if (req->lock != REQ_WIRED && req->oldptr &&
1967 req->oldfunc == sysctl_old_user) {
1968 if (wiredlen != 0) {
1969 ret = vslock(req->oldptr, wiredlen);
1976 req->lock = REQ_WIRED;
1977 req->validlen = wiredlen;
1983 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1984 int *nindx, struct sysctl_req *req)
1986 struct sysctl_oid_list *lsp;
1987 struct sysctl_oid *oid;
1990 SYSCTL_ASSERT_LOCKED();
1991 lsp = &sysctl__children;
1993 while (indx < CTL_MAXNAME) {
1994 SLIST_FOREACH(oid, lsp, oid_link) {
1995 if (oid->oid_number == name[indx])
2002 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2003 if (oid->oid_handler != NULL || indx == namelen) {
2007 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
2008 ("%s found DYING node %p", __func__, oid));
2011 lsp = SYSCTL_CHILDREN(oid);
2012 } else if (indx == namelen) {
2013 if ((oid->oid_kind & CTLFLAG_DORMANT) != 0)
2018 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
2019 ("%s found DYING node %p", __func__, oid));
2029 * Traverse our tree, and find the right node, execute whatever it points
2030 * to, and return the resulting error code.
2034 sysctl_root(SYSCTL_HANDLER_ARGS)
2036 struct sysctl_oid *oid;
2037 struct rm_priotracker tracker;
2038 int error, indx, lvl;
2040 SYSCTL_RLOCK(&tracker);
2042 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
2046 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
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.
2052 if (oid->oid_handler == NULL) {
2058 /* Is this sysctl writable? */
2059 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) {
2064 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
2066 #ifdef CAPABILITY_MODE
2068 * If the process is in capability mode, then don't permit reading or
2069 * writing unless specifically granted for the node.
2071 if (IN_CAPABILITY_MODE(req->td)) {
2072 if ((req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) ||
2073 (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))) {
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);
2088 /* Is this sysctl writable by only privileged users? */
2089 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
2092 if (oid->oid_kind & CTLFLAG_PRISON)
2093 priv = PRIV_SYSCTL_WRITEJAIL;
2095 else if ((oid->oid_kind & CTLFLAG_VNET) &&
2096 prison_owns_vnet(req->td->td_ucred))
2097 priv = PRIV_SYSCTL_WRITEJAIL;
2100 priv = PRIV_SYSCTL_WRITE;
2101 error = priv_check(req->td, priv);
2106 if (!oid->oid_handler) {
2111 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2112 arg1 = (int *)arg1 + indx;
2115 arg1 = oid->oid_arg1;
2116 arg2 = oid->oid_arg2;
2119 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
2125 if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
2126 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
2128 error = sysctl_root_handler_locked(oid, arg1, arg2, req, &tracker);
2131 SYSCTL_RUNLOCK(&tracker);
2135 #ifndef _SYS_SYSPROTO_H_
2136 struct sysctl_args {
2146 sys___sysctl(struct thread *td, struct sysctl_args *uap)
2148 int error, i, name[CTL_MAXNAME];
2151 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
2154 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
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)
2164 i = copyout(&j, uap->oldlenp, sizeof(j));
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)
2176 int oid[CTL_MAXNAME];
2182 if (namelen > MAXPATHLEN || namelen == 0)
2185 if (namelen > sizeof(namebuf))
2186 name = malloc(namelen, M_SYSCTL, M_WAITOK);
2187 error = copyin(oname, name, namelen);
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,
2198 error = userland_sysctl(td, oid, *retval / sizeof(int), old, oldlenp,
2199 inkernel, new, newlen, retval, flags);
2202 if (namelen > sizeof(namebuf))
2203 free(name, M_SYSCTL);
2207 #ifndef _SYS_SYSPROTO_H_
2208 struct __sysctlbyname_args {
2218 sys___sysctlbyname(struct thread *td, struct __sysctlbyname_args *uap)
2223 error = kern___sysctlbyname(td, uap->name, uap->namelen, uap->old,
2224 uap->oldlenp, uap->new, uap->newlen, &rv, 0, 0);
2227 if (uap->oldlenp != NULL)
2228 error = copyout(&rv, uap->oldlenp, sizeof(rv));
2234 * This is used from various compatibility syscalls too. That's why name
2235 * must be in kernel space.
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)
2242 int error = 0, memlocked;
2243 struct sysctl_req req;
2245 bzero(&req, sizeof req);
2252 req.oldlen = *oldlenp;
2254 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
2259 req.validlen = req.oldlen;
2263 req.newlen = newlen;
2267 req.oldfunc = sysctl_old_user;
2268 req.newfunc = sysctl_new_user;
2269 req.lock = REQ_UNWIRED;
2272 if (KTRPOINT(curthread, KTR_SYSCTL))
2273 ktrsysctl(name, namelen);
2276 if (req.oldptr && req.oldlen > 4 * PAGE_SIZE) {
2278 sx_xlock(&sysctlmemlock);
2280 CURVNET_SET(TD_TO_VNET(td));
2285 error = sysctl_root(0, name, namelen, &req);
2286 if (error != EAGAIN)
2288 kern_yield(PRI_USER);
2293 if (req.lock == REQ_WIRED && req.validlen > 0)
2294 vsunlock(req.oldptr, req.validlen);
2296 sx_xunlock(&sysctlmemlock);
2298 if (error && error != ENOMEM)
2302 if (req.oldptr && req.oldidx > req.validlen)
2303 *retval = req.validlen;
2305 *retval = req.oldidx;
2311 * Drain into a sysctl struct. The user buffer should be wired if a page
2312 * fault would cause issue.
2315 sbuf_sysctl_drain(void *arg, const char *data, int len)
2317 struct sysctl_req *req = arg;
2320 error = SYSCTL_OUT(req, data, len);
2321 KASSERT(error >= 0, ("Got unexpected negative value %d", error));
2322 return (error == 0 ? len : -error);
2326 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
2327 struct sysctl_req *req)
2330 /* Supply a default buffer size if none given. */
2331 if (buf == NULL && length == 0)
2333 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
2334 sbuf_set_drain(s, sbuf_sysctl_drain, req);
2340 /* The current OID the debugger is working with */
2341 static struct sysctl_oid *g_ddb_oid;
2343 /* The current flags specified by the user */
2344 static int g_ddb_sysctl_flags;
2346 /* Check to see if the last sysctl printed */
2347 static int g_ddb_sysctl_printed;
2349 static const int ctl_sign[CTLTYPE+1] = {
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),
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 */
2378 #define DB_SYSCTL_SAFE_ONLY 0x100 /* Only simple types */
2380 static const char db_sysctl_modifs[] = {
2384 static const int db_sysctl_modif_values[] = {
2385 DB_SYSCTL_NAME_ONLY, DB_SYSCTL_VALUE_ONLY,
2386 DB_SYSCTL_OPAQUE, DB_SYSCTL_HEX,
2389 /* Handlers considered safe to print while recursing */
2390 static int (* const db_safe_handlers[])(SYSCTL_HANDLER_ARGS) = {
2398 sysctl_handle_string,
2399 sysctl_handle_opaque,
2403 * Use in place of sysctl_old_kernel to print sysctl values.
2405 * Compare to the output handling in show_var from sbin/sysctl/sysctl.c
2408 sysctl_old_ddb(struct sysctl_req *req, const void *ptr, size_t len)
2410 const u_char *val, *p;
2412 size_t intlen, slen;
2415 int sign, ctltype, hexlen, xflag, error;
2417 /* Suppress false-positive GCC uninitialized variable warnings */
2429 /* We are going to print */
2430 g_ddb_sysctl_printed = 1;
2432 xflag = g_ddb_sysctl_flags & DB_SYSCTL_HEX;
2434 ctltype = (g_ddb_oid->oid_kind & CTLTYPE);
2435 sign = ctl_sign[ctltype];
2436 intlen = ctl_size[ctltype];
2440 case CTLTYPE_STRING:
2441 db_printf("%.*s", (int) len, (const char *) p);
2457 hexlen = 2 + (intlen * CHAR_BIT + 3) / 4;
2459 while (len >= intlen) {
2463 umv = *(const u_int *)p;
2464 mv = *(const int *)p;
2468 umv = *(const u_long *)p;
2469 mv = *(const long *)p;
2473 umv = *(const uint8_t *)p;
2474 mv = *(const int8_t *)p;
2478 umv = *(const uint16_t *)p;
2479 mv = *(const int16_t *)p;
2483 umv = *(const uint32_t *)p;
2484 mv = *(const int32_t *)p;
2488 umv = *(const uint64_t *)p;
2489 mv = *(const int64_t *)p;
2493 db_printf("%s", sep1);
2495 db_printf("%#0*jx", hexlen, umv);
2497 db_printf("%ju", umv);
2498 else if (g_ddb_oid->oid_fmt[1] == 'K') {
2499 /* Kelvins are currently unsupported. */
2503 db_printf("%jd", mv);
2512 case CTLTYPE_OPAQUE:
2513 /* TODO: Support struct functions. */
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)
2528 req->oldidx += slen;
2533 * Avoid setting new sysctl values from the debugger
2536 sysctl_new_ddb(struct sysctl_req *req, void *p, size_t l)
2542 /* Changing sysctls from the debugger is currently unsupported */
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
2552 db_sysctl(struct sysctl_oid *oidp, int *name, u_int namelen,
2553 void *old, size_t *oldlenp, size_t *retval, int flags)
2555 struct sysctl_req req;
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;
2565 req.oldlen = *oldlenp;
2567 req.validlen = req.oldlen;
2572 /* Setup our globals for sysctl_old_ddb */
2574 g_ddb_sysctl_flags = flags;
2575 g_ddb_sysctl_printed = 0;
2577 error = sysctl_root(0, name, namelen, &req);
2581 g_ddb_sysctl_flags = 0;
2584 if (req.oldptr && req.oldidx > req.validlen)
2585 *retval = req.validlen;
2587 *retval = req.oldidx;
2593 * Show a sysctl's name
2596 db_show_oid_name(int *oid, size_t nlen)
2598 struct sysctl_oid *oidp;
2599 int qoid[CTL_MAXNAME+2];
2603 memcpy(qoid + 2, oid, nlen * sizeof(int));
2606 error = sysctl_find_oid(qoid, nlen + 2, &oidp, NULL, NULL);
2608 db_error("sysctl name oid");
2610 error = db_sysctl(oidp, qoid, nlen + 2, NULL, NULL, NULL, 0);
2612 db_error("sysctl name");
2616 * Check to see if an OID is safe to print from ddb.
2619 db_oid_safe(const struct sysctl_oid *oidp)
2621 for (unsigned int i = 0; i < nitems(db_safe_handlers); ++i) {
2622 if (oidp->oid_handler == db_safe_handlers[i])
2630 * Show a sysctl at a specific OID
2631 * Compare to the input handling in show_var from sbin/sysctl/sysctl.c
2634 db_show_oid(struct sysctl_oid *oidp, int *oid, size_t nlen, int flags)
2636 int error, xflag, oflag, Nflag, nflag;
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;
2644 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_OPAQUE &&
2649 db_show_oid_name(oid, nlen);
2655 db_show_oid_name(oid, nlen);
2659 if ((flags & DB_SYSCTL_SAFE_ONLY) && !db_oid_safe(oidp)) {
2660 db_printf("Skipping, unsafe to print while recursing.");
2665 /* Try once, and ask about the size */
2667 error = db_sysctl(oidp, oid, nlen,
2668 NULL, NULL, &len, flags);
2672 if (!g_ddb_sysctl_printed)
2673 /* Lie about the size */
2674 error = db_sysctl(oidp, oid, nlen,
2675 (void *) 1, &len, NULL, flags);
2683 * Show all sysctls under a specific OID
2684 * Compare to sysctl_all from sbin/sysctl/sysctl.c
2687 db_show_sysctl_all(int *oid, size_t len, int flags)
2689 struct sysctl_oid *oidp;
2690 int name1[CTL_MAXNAME + 2], name2[CTL_MAXNAME + 2];
2693 name1[0] = CTL_SYSCTL;
2694 name1[1] = CTL_SYSCTL_NEXT;
2697 memcpy(name1+2, oid, len * sizeof(int));
2707 error = kernel_sysctl(kdb_thread, name1, l1,
2708 name2, &l2, NULL, 0, &l2, 0);
2710 if (error == ENOENT)
2713 db_error("sysctl(getnext)");
2718 if (l2 < (unsigned int)len)
2721 for (i = 0; i < len; i++)
2722 if (name2[i] != oid[i])
2725 /* Find the OID in question */
2726 error = sysctl_find_oid(name2, l2, &oidp, NULL, NULL);
2730 i = db_show_oid(oidp, name2, l2, flags | DB_SYSCTL_SAFE_ONLY);
2735 memcpy(name1+2, name2, l2 * sizeof(int));
2741 * Show a sysctl by its user facing string
2744 db_sysctlbyname(char *name, int flags)
2746 struct sysctl_oid *oidp;
2747 int oid[CTL_MAXNAME];
2750 error = name2oid(name, oid, &nlen, &oidp);
2755 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2756 db_show_sysctl_all(oid, nlen, flags);
2758 error = db_show_oid(oidp, oid, nlen, flags);
2765 db_sysctl_cmd_usage(void)
2768 " sysctl [/Nnox] <sysctl> \n"
2770 " <sysctl> The name of the sysctl to show. \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"
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"
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"
2788 "sysctl vm.v_free_min \n"
2789 "vn.v_free_min: 12669 \n"
2794 * Show a specific sysctl similar to sysctl (8).
2796 DB_FUNC(sysctl, db_sysctl_cmd, db_cmd_table, CS_OWN, NULL)
2798 char name[TOK_STRING_SIZE];
2799 int error, i, t, flags;
2801 /* Parse the modifiers */
2802 t = db_read_token();
2803 if (t == tSLASH || t == tMINUS) {
2804 t = db_read_token();
2806 db_printf("Bad modifier\n");
2810 db_strcpy(modif, db_tok_string);
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];
2824 /* Parse the sysctl names */
2825 t = db_read_token();
2827 db_printf("Need sysctl name\n");
2832 /* Copy the name into a temporary buffer */
2833 db_strcpy(name, db_tok_string);
2835 /* Ensure there is no trailing cruft */
2836 t = db_read_token();
2838 db_printf("Unexpected sysctl argument\n");
2843 error = db_sysctlbyname(name, flags);
2844 if (error == ENOENT) {
2845 db_printf("unknown oid: '%s'\n", db_tok_string);
2848 db_printf("%s: error: %d\n", db_tok_string, error);
2853 /* Ensure we eat all of our text */
2856 if (error == EINVAL) {
2857 db_sysctl_cmd_usage();