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"
44 #include "opt_ktrace.h"
46 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/capsicum.h>
50 #include <sys/kernel.h>
51 #include <sys/sysctl.h>
52 #include <sys/malloc.h>
57 #include <sys/mutex.h>
58 #include <sys/rmlock.h>
61 #include <sys/sysproto.h>
64 #include <sys/ktrace.h>
69 #include <security/mac/mac_framework.h>
72 #include <vm/vm_extern.h>
74 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
75 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
76 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
79 * The sysctllock protects the MIB tree. It also protects sysctl
80 * contexts used with dynamic sysctls. The sysctl_register_oid() and
81 * sysctl_unregister_oid() routines require the sysctllock to already
82 * be held, so the sysctl_wlock() and sysctl_wunlock() routines are
83 * provided for the few places in the kernel which need to use that
84 * API rather than using the dynamic API. Use of the dynamic API is
85 * strongly encouraged for most code.
87 * The sysctlmemlock is used to limit the amount of user memory wired for
88 * sysctl requests. This is implemented by serializing any userland
89 * sysctl requests larger than a single page via an exclusive lock.
91 static struct rmlock sysctllock;
92 static struct sx __exclusive_cache_line sysctlmemlock;
94 #define SYSCTL_WLOCK() rm_wlock(&sysctllock)
95 #define SYSCTL_WUNLOCK() rm_wunlock(&sysctllock)
96 #define SYSCTL_RLOCK(tracker) rm_rlock(&sysctllock, (tracker))
97 #define SYSCTL_RUNLOCK(tracker) rm_runlock(&sysctllock, (tracker))
98 #define SYSCTL_WLOCKED() rm_wowned(&sysctllock)
99 #define SYSCTL_ASSERT_LOCKED() rm_assert(&sysctllock, RA_LOCKED)
100 #define SYSCTL_ASSERT_WLOCKED() rm_assert(&sysctllock, RA_WLOCKED)
101 #define SYSCTL_ASSERT_RLOCKED() rm_assert(&sysctllock, RA_RLOCKED)
102 #define SYSCTL_INIT() rm_init_flags(&sysctllock, "sysctl lock", \
104 #define SYSCTL_SLEEP(ch, wmesg, timo) \
105 rm_sleep(ch, &sysctllock, 0, wmesg, timo)
107 static int sysctl_root(SYSCTL_HANDLER_ARGS);
110 struct sysctl_oid_list sysctl__children = SLIST_HEAD_INITIALIZER(&sysctl__children);
112 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
114 static int sysctl_old_kernel(struct sysctl_req *, const void *, size_t);
115 static int sysctl_new_kernel(struct sysctl_req *, void *, size_t);
117 static struct sysctl_oid *
118 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
120 struct sysctl_oid *oidp;
122 SYSCTL_ASSERT_LOCKED();
123 SLIST_FOREACH(oidp, list, oid_link) {
124 if (strcmp(oidp->oid_name, name) == 0) {
132 * Initialization of the MIB tree.
134 * Order by number in each list.
151 sysctl_root_handler_locked(struct sysctl_oid *oid, void *arg1, intmax_t arg2,
152 struct sysctl_req *req, struct rm_priotracker *tracker)
156 if (oid->oid_kind & CTLFLAG_DYN)
157 atomic_add_int(&oid->oid_running, 1);
160 SYSCTL_RUNLOCK(tracker);
164 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
166 error = oid->oid_handler(oid, arg1, arg2, req);
167 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
170 KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
173 SYSCTL_RLOCK(tracker);
177 if (oid->oid_kind & CTLFLAG_DYN) {
178 if (atomic_fetchadd_int(&oid->oid_running, -1) == 1 &&
179 (oid->oid_kind & CTLFLAG_DYING) != 0)
180 wakeup(&oid->oid_running);
187 sysctl_load_tunable_by_oid_locked(struct sysctl_oid *oidp)
189 struct sysctl_req req;
190 struct sysctl_oid *curr;
193 ssize_t rem = sizeof(path);
195 uint8_t data[512] __aligned(sizeof(uint64_t));
201 for (curr = oidp; curr != NULL; curr = SYSCTL_PARENT(curr)) {
202 len = strlen(curr->oid_name);
207 printf("OID path exceeds %d bytes\n", (int)sizeof(path));
210 memcpy(path + rem, curr->oid_name, len);
212 path[rem + len] = '.';
215 memset(&req, 0, sizeof(req));
218 req.oldfunc = sysctl_old_kernel;
219 req.newfunc = sysctl_new_kernel;
220 req.lock = REQ_UNWIRED;
222 switch (oidp->oid_kind & CTLTYPE) {
224 if (getenv_array(path + rem, data, sizeof(data), &size,
225 sizeof(int), GETENV_SIGNED) == 0)
231 if (getenv_array(path + rem, data, sizeof(data), &size,
232 sizeof(int), GETENV_UNSIGNED) == 0)
238 if (getenv_array(path + rem, data, sizeof(data), &size,
239 sizeof(long), GETENV_SIGNED) == 0)
245 if (getenv_array(path + rem, data, sizeof(data), &size,
246 sizeof(long), GETENV_UNSIGNED) == 0)
252 if (getenv_array(path + rem, data, sizeof(data), &size,
253 sizeof(int8_t), GETENV_SIGNED) == 0)
259 if (getenv_array(path + rem, data, sizeof(data), &size,
260 sizeof(int16_t), GETENV_SIGNED) == 0)
266 if (getenv_array(path + rem, data, sizeof(data), &size,
267 sizeof(int32_t), GETENV_SIGNED) == 0)
273 if (getenv_array(path + rem, data, sizeof(data), &size,
274 sizeof(int64_t), GETENV_SIGNED) == 0)
280 if (getenv_array(path + rem, data, sizeof(data), &size,
281 sizeof(uint8_t), GETENV_UNSIGNED) == 0)
287 if (getenv_array(path + rem, data, sizeof(data), &size,
288 sizeof(uint16_t), GETENV_UNSIGNED) == 0)
294 if (getenv_array(path + rem, data, sizeof(data), &size,
295 sizeof(uint32_t), GETENV_UNSIGNED) == 0)
301 if (getenv_array(path + rem, data, sizeof(data), &size,
302 sizeof(uint64_t), GETENV_UNSIGNED) == 0)
308 penv = kern_getenv(path + rem);
311 req.newlen = strlen(penv);
317 error = sysctl_root_handler_locked(oidp, oidp->oid_arg1,
318 oidp->oid_arg2, &req, NULL);
320 printf("Setting sysctl %s failed: %d\n", path + rem, error);
326 sbuf_printf_drain(void *arg __unused, const char *data, int len)
329 return (printf("%.*s", len, data));
333 * Locate the path to a given oid. Returns the length of the resulting path,
334 * or -1 if the oid was not found. nodes must have room for CTL_MAXNAME
335 * elements and be NULL initialized.
338 sysctl_search_oid(struct sysctl_oid **nodes, struct sysctl_oid *needle)
342 SYSCTL_ASSERT_LOCKED();
344 while (indx < CTL_MAXNAME && indx >= 0) {
345 if (nodes[indx] == NULL && indx == 0)
346 nodes[indx] = SLIST_FIRST(&sysctl__children);
347 else if (nodes[indx] == NULL)
348 nodes[indx] = SLIST_FIRST(&nodes[indx - 1]->oid_children);
350 nodes[indx] = SLIST_NEXT(nodes[indx], oid_link);
352 if (nodes[indx] == needle)
355 if (nodes[indx] == NULL) {
360 if ((nodes[indx]->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
369 sysctl_warn_reuse(const char *func, struct sysctl_oid *leaf)
371 struct sysctl_oid *nodes[CTL_MAXNAME];
376 (void)sbuf_new(&sb, buf, sizeof(buf), SBUF_FIXEDLEN | SBUF_INCLUDENUL);
377 sbuf_set_drain(&sb, sbuf_printf_drain, NULL);
379 sbuf_printf(&sb, "%s: can't re-use a leaf (", __func__);
381 memset(nodes, 0, sizeof(nodes));
382 rc = sysctl_search_oid(nodes, leaf);
384 for (i = 0; i < rc; i++)
385 sbuf_printf(&sb, "%s%.*s", nodes[i]->oid_name,
388 sbuf_printf(&sb, "%s", leaf->oid_name);
390 sbuf_printf(&sb, ")!\n");
392 (void)sbuf_finish(&sb);
397 sysctl_reuse_test(SYSCTL_HANDLER_ARGS)
399 struct rm_priotracker tracker;
401 SYSCTL_RLOCK(&tracker);
402 sysctl_warn_reuse(__func__, oidp);
403 SYSCTL_RUNLOCK(&tracker);
406 SYSCTL_PROC(_sysctl, 0, reuse_test, CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE,
407 0, 0, sysctl_reuse_test, "-", "");
411 sysctl_register_oid(struct sysctl_oid *oidp)
413 struct sysctl_oid_list *parent = oidp->oid_parent;
414 struct sysctl_oid *p;
415 struct sysctl_oid *q;
420 * First check if another oid with the same name already
421 * exists in the parent's list.
423 SYSCTL_ASSERT_WLOCKED();
424 p = sysctl_find_oidname(oidp->oid_name, parent);
426 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
430 sysctl_warn_reuse(__func__, p);
434 /* get current OID number */
435 oid_number = oidp->oid_number;
438 #error "OID_AUTO is expected to be a negative value"
441 * Any negative OID number qualifies as OID_AUTO. Valid OID
442 * numbers should always be positive.
444 * NOTE: DO NOT change the starting value here, change it in
445 * <sys/sysctl.h>, and make sure it is at least 256 to
446 * accommodate e.g. net.inet.raw as a static sysctl node.
448 if (oid_number < 0) {
452 * By decrementing the next OID number we spend less
453 * time inserting the OIDs into a sorted list.
455 if (--newoid < CTL_AUTO_START)
462 * Insert the OID into the parent's list sorted by OID number.
466 SLIST_FOREACH(p, parent, oid_link) {
467 /* check if the current OID number is in use */
468 if (oid_number == p->oid_number) {
469 /* get the next valid OID number */
470 if (oid_number < CTL_AUTO_START ||
471 oid_number == 0x7fffffff) {
472 /* wraparound - restart */
473 oid_number = CTL_AUTO_START;
474 /* don't loop forever */
476 panic("sysctl: Out of OID numbers\n");
481 } else if (oid_number < p->oid_number)
485 /* check for non-auto OID number collision */
486 if (oidp->oid_number >= 0 && oidp->oid_number < CTL_AUTO_START &&
487 oid_number >= CTL_AUTO_START) {
488 printf("sysctl: OID number(%d) is already in use for '%s'\n",
489 oidp->oid_number, oidp->oid_name);
491 /* update the OID number, if any */
492 oidp->oid_number = oid_number;
494 SLIST_INSERT_AFTER(q, oidp, oid_link);
496 SLIST_INSERT_HEAD(parent, oidp, oid_link);
498 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE &&
500 (oidp->oid_kind & CTLFLAG_VNET) == 0 &&
502 (oidp->oid_kind & CTLFLAG_TUN) != 0 &&
503 (oidp->oid_kind & CTLFLAG_NOFETCH) == 0) {
504 /* only fetch value once */
505 oidp->oid_kind |= CTLFLAG_NOFETCH;
506 /* try to fetch value from kernel environment */
507 sysctl_load_tunable_by_oid_locked(oidp);
512 sysctl_register_disabled_oid(struct sysctl_oid *oidp)
516 * Mark the leaf as dormant if it's not to be immediately enabled.
517 * We do not disable nodes as they can be shared between modules
518 * and it is always safe to access a node.
520 KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
521 ("internal flag is set in oid_kind"));
522 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
523 oidp->oid_kind |= CTLFLAG_DORMANT;
524 sysctl_register_oid(oidp);
528 sysctl_enable_oid(struct sysctl_oid *oidp)
531 SYSCTL_ASSERT_WLOCKED();
532 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
533 KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
534 ("sysctl node is marked as dormant"));
537 KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) != 0,
538 ("enabling already enabled sysctl oid"));
539 oidp->oid_kind &= ~CTLFLAG_DORMANT;
543 sysctl_unregister_oid(struct sysctl_oid *oidp)
545 struct sysctl_oid *p;
548 SYSCTL_ASSERT_WLOCKED();
550 if (oidp->oid_number == OID_AUTO) {
553 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
555 SLIST_REMOVE(oidp->oid_parent, oidp,
556 sysctl_oid, oid_link);
564 * This can happen when a module fails to register and is
565 * being unloaded afterwards. It should not be a panic()
569 printf("%s: failed to unregister sysctl\n", __func__);
572 /* Initialize a new context to keep track of dynamically added sysctls. */
574 sysctl_ctx_init(struct sysctl_ctx_list *c)
582 * No locking here, the caller is responsible for not adding
583 * new nodes to a context until after this function has
590 /* Free the context, and destroy all dynamic oids registered in this context */
592 sysctl_ctx_free(struct sysctl_ctx_list *clist)
594 struct sysctl_ctx_entry *e, *e1;
599 * First perform a "dry run" to check if it's ok to remove oids.
601 * XXX This algorithm is a hack. But I don't know any
602 * XXX better solution for now...
605 TAILQ_FOREACH(e, clist, link) {
606 error = sysctl_remove_oid_locked(e->entry, 0, 0);
611 * Restore deregistered entries, either from the end,
612 * or from the place where error occurred.
613 * e contains the entry that was not unregistered
616 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
618 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
620 sysctl_register_oid(e1->entry);
621 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
627 /* Now really delete the entries */
628 e = TAILQ_FIRST(clist);
630 e1 = TAILQ_NEXT(e, link);
631 error = sysctl_remove_oid_locked(e->entry, 1, 0);
633 panic("sysctl_remove_oid: corrupt tree, entry: %s",
635 free(e, M_SYSCTLOID);
642 /* Add an entry to the context */
643 struct sysctl_ctx_entry *
644 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
646 struct sysctl_ctx_entry *e;
648 SYSCTL_ASSERT_WLOCKED();
649 if (clist == NULL || oidp == NULL)
651 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
653 TAILQ_INSERT_HEAD(clist, e, link);
657 /* Find an entry in the context */
658 struct sysctl_ctx_entry *
659 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
661 struct sysctl_ctx_entry *e;
663 SYSCTL_ASSERT_WLOCKED();
664 if (clist == NULL || oidp == NULL)
666 TAILQ_FOREACH(e, clist, link) {
674 * Delete an entry from the context.
675 * NOTE: this function doesn't free oidp! You have to remove it
676 * with sysctl_remove_oid().
679 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
681 struct sysctl_ctx_entry *e;
683 if (clist == NULL || oidp == NULL)
686 e = sysctl_ctx_entry_find(clist, oidp);
688 TAILQ_REMOVE(clist, e, link);
690 free(e, M_SYSCTLOID);
699 * Remove dynamically created sysctl trees.
700 * oidp - top of the tree to be removed
701 * del - if 0 - just deregister, otherwise free up entries as well
702 * recurse - if != 0 traverse the subtree to be deleted
705 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
710 error = sysctl_remove_oid_locked(oidp, del, recurse);
716 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
717 int del, int recurse)
719 struct sysctl_oid *p, *tmp;
724 SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
725 if (strcmp(p->oid_name, name) == 0) {
726 error = sysctl_remove_oid_locked(p, del, recurse);
737 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
739 struct sysctl_oid *p, *tmp;
742 SYSCTL_ASSERT_WLOCKED();
745 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
746 printf("Warning: can't remove non-dynamic nodes (%s)!\n",
751 * WARNING: normal method to do this should be through
752 * sysctl_ctx_free(). Use recursing as the last resort
753 * method to purge your sysctl tree of leftovers...
754 * However, if some other code still references these nodes,
757 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
758 if (oidp->oid_refcnt == 1) {
759 SLIST_FOREACH_SAFE(p,
760 SYSCTL_CHILDREN(oidp), oid_link, tmp) {
762 printf("Warning: failed attempt to "
763 "remove oid %s with child %s\n",
764 oidp->oid_name, p->oid_name);
767 error = sysctl_remove_oid_locked(p, del,
774 if (oidp->oid_refcnt > 1 ) {
777 if (oidp->oid_refcnt == 0) {
778 printf("Warning: bad oid_refcnt=%u (%s)!\n",
779 oidp->oid_refcnt, oidp->oid_name);
782 sysctl_unregister_oid(oidp);
785 * Wait for all threads running the handler to drain.
786 * This preserves the previous behavior when the
787 * sysctl lock was held across a handler invocation,
788 * and is necessary for module unload correctness.
790 while (oidp->oid_running > 0) {
791 oidp->oid_kind |= CTLFLAG_DYING;
792 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
795 free(__DECONST(char *, oidp->oid_descr),
798 free(__DECONST(char *, oidp->oid_label),
800 free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
801 free(oidp, M_SYSCTLOID);
807 * Create new sysctls at run time.
808 * clist may point to a valid context initialized with sysctl_ctx_init().
811 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
812 int number, const char *name, int kind, void *arg1, intmax_t arg2,
813 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr,
816 struct sysctl_oid *oidp;
818 /* You have to hook up somewhere.. */
821 /* Check if the node already exists, otherwise create it */
823 oidp = sysctl_find_oidname(name, parent);
825 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
827 /* Update the context */
829 sysctl_ctx_entry_add(clist, oidp);
833 sysctl_warn_reuse(__func__, oidp);
838 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
839 oidp->oid_parent = parent;
840 SLIST_INIT(&oidp->oid_children);
841 oidp->oid_number = number;
842 oidp->oid_refcnt = 1;
843 oidp->oid_name = strdup(name, M_SYSCTLOID);
844 oidp->oid_handler = handler;
845 oidp->oid_kind = CTLFLAG_DYN | kind;
846 oidp->oid_arg1 = arg1;
847 oidp->oid_arg2 = arg2;
850 oidp->oid_descr = strdup(descr, M_SYSCTLOID);
852 oidp->oid_label = strdup(label, M_SYSCTLOID);
853 /* Update the context, if used */
855 sysctl_ctx_entry_add(clist, oidp);
856 /* Register this oid */
857 sysctl_register_oid(oidp);
863 * Rename an existing oid.
866 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
871 newname = strdup(name, M_SYSCTLOID);
873 oldname = __DECONST(char *, oidp->oid_name);
874 oidp->oid_name = newname;
876 free(oldname, M_SYSCTLOID);
880 * Reparent an existing oid.
883 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
885 struct sysctl_oid *oidp;
888 if (oid->oid_parent == parent) {
892 oidp = sysctl_find_oidname(oid->oid_name, parent);
897 sysctl_unregister_oid(oid);
898 oid->oid_parent = parent;
899 oid->oid_number = OID_AUTO;
900 sysctl_register_oid(oid);
906 * Register the kernel's oids on startup.
908 SET_DECLARE(sysctl_set, struct sysctl_oid);
911 sysctl_register_all(void *arg)
913 struct sysctl_oid **oidp;
915 sx_init(&sysctlmemlock, "sysctl mem");
918 SET_FOREACH(oidp, sysctl_set)
919 sysctl_register_oid(*oidp);
922 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_FIRST, sysctl_register_all, NULL);
927 * These functions implement a presently undocumented interface
928 * used by the sysctl program to walk the tree, and get the type
929 * so it can print the value.
930 * This interface is under work and consideration, and should probably
931 * be killed with a big axe by the first person who can find the time.
932 * (be aware though, that the proper interface isn't as obvious as it
933 * may seem, there are various conflicting requirements.
935 * {0,0} printf the entire MIB-tree.
936 * {0,1,...} return the name of the "..." OID.
937 * {0,2,...} return the next OID.
938 * {0,3} return the OID of the name in "new"
939 * {0,4,...} return the kind & format info for the "..." OID.
940 * {0,5,...} return the description of the "..." OID.
941 * {0,6,...} return the aggregation label of the "..." OID.
946 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
949 struct sysctl_oid *oidp;
951 SYSCTL_ASSERT_LOCKED();
952 SLIST_FOREACH(oidp, l, oid_link) {
957 printf("%d %s ", oidp->oid_number, oidp->oid_name);
960 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
961 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
963 if (oidp->oid_handler)
966 switch (oidp->oid_kind & CTLTYPE) {
969 if (!oidp->oid_handler) {
970 sysctl_sysctl_debug_dump_node(
971 SYSCTL_CHILDREN(oidp), i + 2);
974 case CTLTYPE_INT: printf(" Int\n"); break;
975 case CTLTYPE_UINT: printf(" u_int\n"); break;
976 case CTLTYPE_LONG: printf(" Long\n"); break;
977 case CTLTYPE_ULONG: printf(" u_long\n"); break;
978 case CTLTYPE_STRING: printf(" String\n"); break;
979 case CTLTYPE_S8: printf(" int8_t\n"); break;
980 case CTLTYPE_S16: printf(" int16_t\n"); break;
981 case CTLTYPE_S32: printf(" int32_t\n"); break;
982 case CTLTYPE_S64: printf(" int64_t\n"); break;
983 case CTLTYPE_U8: printf(" uint8_t\n"); break;
984 case CTLTYPE_U16: printf(" uint16_t\n"); break;
985 case CTLTYPE_U32: printf(" uint32_t\n"); break;
986 case CTLTYPE_U64: printf(" uint64_t\n"); break;
987 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
988 default: printf("\n");
995 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
997 struct rm_priotracker tracker;
1000 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
1003 SYSCTL_RLOCK(&tracker);
1004 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
1005 SYSCTL_RUNLOCK(&tracker);
1009 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE,
1010 0, 0, sysctl_sysctl_debug, "-", "");
1014 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
1016 int *name = (int *) arg1;
1017 u_int namelen = arg2;
1019 struct sysctl_oid *oid;
1020 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
1021 struct rm_priotracker tracker;
1024 SYSCTL_RLOCK(&tracker);
1027 snprintf(buf,sizeof(buf),"%d",*name);
1029 error = SYSCTL_OUT(req, ".", 1);
1031 error = SYSCTL_OUT(req, buf, strlen(buf));
1039 SLIST_FOREACH(oid, lsp, oid_link) {
1040 if (oid->oid_number != *name)
1044 error = SYSCTL_OUT(req, ".", 1);
1046 error = SYSCTL_OUT(req, oid->oid_name,
1047 strlen(oid->oid_name));
1054 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1057 if (oid->oid_handler)
1060 lsp2 = SYSCTL_CHILDREN(oid);
1065 error = SYSCTL_OUT(req, "", 1);
1067 SYSCTL_RUNLOCK(&tracker);
1072 * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
1075 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD,
1076 sysctl_sysctl_name, "");
1079 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
1080 int *next, int *len, int level, struct sysctl_oid **oidpp)
1082 struct sysctl_oid *oidp;
1084 SYSCTL_ASSERT_LOCKED();
1086 SLIST_FOREACH(oidp, lsp, oid_link) {
1087 *next = oidp->oid_number;
1090 if ((oidp->oid_kind & (CTLFLAG_SKIP | CTLFLAG_DORMANT)) != 0)
1094 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1096 if (oidp->oid_handler)
1097 /* We really should call the handler here...*/
1099 lsp = SYSCTL_CHILDREN(oidp);
1100 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
1101 len, level+1, oidpp))
1106 if (oidp->oid_number < *name)
1109 if (oidp->oid_number > *name) {
1110 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1112 if (oidp->oid_handler)
1114 lsp = SYSCTL_CHILDREN(oidp);
1115 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
1116 next+1, len, level+1, oidpp))
1120 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1123 if (oidp->oid_handler)
1126 lsp = SYSCTL_CHILDREN(oidp);
1127 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
1128 len, level+1, oidpp))
1139 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
1141 int *name = (int *) arg1;
1142 u_int namelen = arg2;
1144 struct sysctl_oid *oid;
1145 struct sysctl_oid_list *lsp = &sysctl__children;
1146 struct rm_priotracker tracker;
1147 int newoid[CTL_MAXNAME];
1149 SYSCTL_RLOCK(&tracker);
1150 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
1151 SYSCTL_RUNLOCK(&tracker);
1154 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
1159 * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
1162 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD,
1163 sysctl_sysctl_next, "");
1166 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
1168 struct sysctl_oid *oidp;
1169 struct sysctl_oid_list *lsp = &sysctl__children;
1172 SYSCTL_ASSERT_LOCKED();
1174 for (*len = 0; *len < CTL_MAXNAME;) {
1175 p = strsep(&name, ".");
1177 oidp = SLIST_FIRST(lsp);
1178 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
1181 if (strcmp(p, oidp->oid_name) == 0)
1184 *oid++ = oidp->oid_number;
1187 if (name == NULL || *name == '\0') {
1193 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1196 if (oidp->oid_handler)
1199 lsp = SYSCTL_CHILDREN(oidp);
1205 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
1208 int error, oid[CTL_MAXNAME], len = 0;
1209 struct sysctl_oid *op = NULL;
1210 struct rm_priotracker tracker;
1215 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
1216 return (ENAMETOOLONG);
1219 if (req->newlen >= sizeof(buf))
1220 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
1222 error = SYSCTL_IN(req, p, req->newlen);
1229 p [req->newlen] = '\0';
1231 SYSCTL_RLOCK(&tracker);
1232 error = name2oid(p, oid, &len, &op);
1233 SYSCTL_RUNLOCK(&tracker);
1241 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
1246 * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
1249 SYSCTL_PROC(_sysctl, 3, name2oid,
1250 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE
1251 | CTLFLAG_CAPRW, 0, 0, sysctl_sysctl_name2oid, "I", "");
1254 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
1256 struct sysctl_oid *oid;
1257 struct rm_priotracker tracker;
1260 SYSCTL_RLOCK(&tracker);
1261 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1265 if (oid->oid_fmt == NULL) {
1269 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
1272 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
1274 SYSCTL_RUNLOCK(&tracker);
1279 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
1280 sysctl_sysctl_oidfmt, "");
1283 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
1285 struct sysctl_oid *oid;
1286 struct rm_priotracker tracker;
1289 SYSCTL_RLOCK(&tracker);
1290 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1294 if (oid->oid_descr == NULL) {
1298 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
1300 SYSCTL_RUNLOCK(&tracker);
1304 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
1305 sysctl_sysctl_oiddescr, "");
1308 sysctl_sysctl_oidlabel(SYSCTL_HANDLER_ARGS)
1310 struct sysctl_oid *oid;
1311 struct rm_priotracker tracker;
1314 SYSCTL_RLOCK(&tracker);
1315 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1319 if (oid->oid_label == NULL) {
1323 error = SYSCTL_OUT(req, oid->oid_label, strlen(oid->oid_label) + 1);
1325 SYSCTL_RUNLOCK(&tracker);
1329 static SYSCTL_NODE(_sysctl, 6, oidlabel,
1330 CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidlabel, "");
1333 * Default "handler" functions.
1339 * a variable: point arg1 at it.
1340 * a constant: pass it in arg2.
1344 sysctl_handle_bool(SYSCTL_HANDLER_ARGS)
1350 * Attempt to get a coherent snapshot by making a copy of the data.
1353 temp = *(bool *)arg1 ? 1 : 0;
1355 temp = arg2 ? 1 : 0;
1357 error = SYSCTL_OUT(req, &temp, sizeof(temp));
1358 if (error || !req->newptr)
1364 error = SYSCTL_IN(req, &temp, sizeof(temp));
1366 *(bool *)arg1 = temp ? 1 : 0;
1372 * Handle an int8_t, signed or unsigned.
1374 * a variable: point arg1 at it.
1375 * a constant: pass it in arg2.
1379 sysctl_handle_8(SYSCTL_HANDLER_ARGS)
1385 * Attempt to get a coherent snapshot by making a copy of the data.
1388 tmpout = *(int8_t *)arg1;
1391 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1393 if (error || !req->newptr)
1399 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1404 * Handle an int16_t, signed or unsigned.
1406 * a variable: point arg1 at it.
1407 * a constant: pass it in arg2.
1411 sysctl_handle_16(SYSCTL_HANDLER_ARGS)
1417 * Attempt to get a coherent snapshot by making a copy of the data.
1420 tmpout = *(int16_t *)arg1;
1423 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1425 if (error || !req->newptr)
1431 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1436 * Handle an int32_t, signed or unsigned.
1438 * a variable: point arg1 at it.
1439 * a constant: pass it in arg2.
1443 sysctl_handle_32(SYSCTL_HANDLER_ARGS)
1449 * Attempt to get a coherent snapshot by making a copy of the data.
1452 tmpout = *(int32_t *)arg1;
1455 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1457 if (error || !req->newptr)
1463 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1468 * Handle an int, signed or unsigned.
1470 * a variable: point arg1 at it.
1471 * a constant: pass it in arg2.
1475 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
1477 int tmpout, error = 0;
1480 * Attempt to get a coherent snapshot by making a copy of the data.
1483 tmpout = *(int *)arg1;
1486 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1488 if (error || !req->newptr)
1494 error = SYSCTL_IN(req, arg1, sizeof(int));
1499 * Based on on sysctl_handle_int() convert milliseconds into ticks.
1500 * Note: this is used by TCP.
1504 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1509 s = (int)((int64_t)tt * 1000 / hz);
1511 error = sysctl_handle_int(oidp, &s, 0, req);
1512 if (error || !req->newptr)
1515 tt = (int)((int64_t)s * hz / 1000);
1525 * Handle a long, signed or unsigned.
1527 * a variable: point arg1 at it.
1528 * a constant: pass it in arg2.
1532 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1541 * Attempt to get a coherent snapshot by making a copy of the data.
1544 tmplong = *(long *)arg1;
1548 if (req->flags & SCTL_MASK32) {
1550 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1553 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1555 if (error || !req->newptr)
1561 else if (req->flags & SCTL_MASK32) {
1562 error = SYSCTL_IN(req, &tmpint, sizeof(int));
1563 *(long *)arg1 = (long)tmpint;
1567 error = SYSCTL_IN(req, arg1, sizeof(long));
1572 * Handle a 64 bit int, signed or unsigned.
1574 * a variable: point arg1 at it.
1575 * a constant: pass it in arg2.
1578 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1584 * Attempt to get a coherent snapshot by making a copy of the data.
1587 tmpout = *(uint64_t *)arg1;
1590 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1592 if (error || !req->newptr)
1598 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1603 * Handle our generic '\0' terminated 'C' string.
1605 * a variable string: point arg1 at it, arg2 is max length.
1606 * a constant string: point arg1 at it, arg2 is zero.
1610 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1613 int error = 0, ro_string = 0;
1616 * A zero-length buffer indicates a fixed size read-only
1620 arg2 = strlen((char *)arg1) + 1;
1624 if (req->oldptr != NULL) {
1630 /* try to make a coherent snapshot of the string */
1631 tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK);
1632 memcpy(tmparg, arg1, arg2);
1635 outlen = strnlen(tmparg, arg2 - 1) + 1;
1636 error = SYSCTL_OUT(req, tmparg, outlen);
1639 free(tmparg, M_SYSCTLTMP);
1641 outlen = strnlen((char *)arg1, arg2 - 1) + 1;
1642 error = SYSCTL_OUT(req, NULL, outlen);
1644 if (error || !req->newptr)
1647 if ((req->newlen - req->newidx) >= arg2) {
1650 arg2 = (req->newlen - req->newidx);
1651 error = SYSCTL_IN(req, arg1, arg2);
1652 ((char *)arg1)[arg2] = '\0';
1658 * Handle any kind of opaque data.
1659 * arg1 points to it, arg2 is the size.
1663 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1667 struct sysctl_req req2;
1670 * Attempt to get a coherent snapshot, by using the thread
1671 * pre-emption counter updated from within mi_switch() to
1672 * determine if we were pre-empted during a bcopy() or
1673 * copyout(). Make 3 attempts at doing this before giving up.
1674 * If we encounter an error, stop immediately.
1679 generation = curthread->td_generation;
1680 error = SYSCTL_OUT(req, arg1, arg2);
1684 if (generation != curthread->td_generation && tries < 3) {
1689 error = SYSCTL_IN(req, arg1, arg2);
1695 * Based on on sysctl_handle_int() convert microseconds to a sbintime.
1698 sysctl_usec_to_sbintime(SYSCTL_HANDLER_ARGS)
1704 tt = *(int64_t *)arg1;
1707 error = sysctl_handle_64(oidp, &sb, 0, req);
1708 if (error || !req->newptr)
1712 *(int64_t *)arg1 = tt;
1718 * Based on on sysctl_handle_int() convert milliseconds to a sbintime.
1721 sysctl_msec_to_sbintime(SYSCTL_HANDLER_ARGS)
1727 tt = *(int64_t *)arg1;
1730 error = sysctl_handle_64(oidp, &sb, 0, req);
1731 if (error || !req->newptr)
1735 *(int64_t *)arg1 = tt;
1742 * Transfer functions to/from kernel space.
1743 * XXX: rather untested at this point
1746 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1752 if (req->oldlen <= req->oldidx)
1755 if (i > req->oldlen - req->oldidx)
1756 i = req->oldlen - req->oldidx;
1758 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1761 if (req->oldptr && i != l)
1767 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1771 if (req->newlen - req->newidx < l)
1773 bcopy((char *)req->newptr + req->newidx, p, l);
1779 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1780 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1783 struct sysctl_req req;
1785 bzero(&req, sizeof req);
1791 req.oldlen = *oldlenp;
1793 req.validlen = req.oldlen;
1800 req.newlen = newlen;
1804 req.oldfunc = sysctl_old_kernel;
1805 req.newfunc = sysctl_new_kernel;
1806 req.lock = REQ_UNWIRED;
1808 error = sysctl_root(0, name, namelen, &req);
1810 if (req.lock == REQ_WIRED && req.validlen > 0)
1811 vsunlock(req.oldptr, req.validlen);
1813 if (error && error != ENOMEM)
1817 if (req.oldptr && req.oldidx > req.validlen)
1818 *retval = req.validlen;
1820 *retval = req.oldidx;
1826 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1827 void *new, size_t newlen, size_t *retval, int flags)
1829 int oid[CTL_MAXNAME];
1830 size_t oidlen, plen;
1833 oid[0] = 0; /* sysctl internal magic */
1834 oid[1] = 3; /* name2oid */
1835 oidlen = sizeof(oid);
1837 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1838 (void *)name, strlen(name), &plen, flags);
1842 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1843 new, newlen, retval, flags);
1848 * Transfer function to/from user space.
1851 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1853 size_t i, len, origidx;
1856 origidx = req->oldidx;
1858 if (req->oldptr == NULL)
1861 * If we have not wired the user supplied buffer and we are currently
1862 * holding locks, drop a witness warning, as it's possible that
1863 * write operations to the user page can sleep.
1865 if (req->lock != REQ_WIRED)
1866 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1867 "sysctl_old_user()");
1869 len = req->validlen;
1873 if (i > len - origidx)
1875 if (req->lock == REQ_WIRED) {
1876 error = copyout_nofault(p, (char *)req->oldptr +
1879 error = copyout(p, (char *)req->oldptr + origidx, i);
1889 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1895 if (req->newlen - req->newidx < l)
1897 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1898 "sysctl_new_user()");
1899 error = copyin((char *)req->newptr + req->newidx, p, l);
1905 * Wire the user space destination buffer. If set to a value greater than
1906 * zero, the len parameter limits the maximum amount of wired memory.
1909 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1914 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1916 if (req->lock != REQ_WIRED && req->oldptr &&
1917 req->oldfunc == sysctl_old_user) {
1918 if (wiredlen != 0) {
1919 ret = vslock(req->oldptr, wiredlen);
1926 req->lock = REQ_WIRED;
1927 req->validlen = wiredlen;
1933 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1934 int *nindx, struct sysctl_req *req)
1936 struct sysctl_oid_list *lsp;
1937 struct sysctl_oid *oid;
1940 SYSCTL_ASSERT_LOCKED();
1941 lsp = &sysctl__children;
1943 while (indx < CTL_MAXNAME) {
1944 SLIST_FOREACH(oid, lsp, oid_link) {
1945 if (oid->oid_number == name[indx])
1952 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1953 if (oid->oid_handler != NULL || indx == namelen) {
1957 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1958 ("%s found DYING node %p", __func__, oid));
1961 lsp = SYSCTL_CHILDREN(oid);
1962 } else if (indx == namelen) {
1963 if ((oid->oid_kind & CTLFLAG_DORMANT) != 0)
1968 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1969 ("%s found DYING node %p", __func__, oid));
1979 * Traverse our tree, and find the right node, execute whatever it points
1980 * to, and return the resulting error code.
1984 sysctl_root(SYSCTL_HANDLER_ARGS)
1986 struct sysctl_oid *oid;
1987 struct rm_priotracker tracker;
1988 int error, indx, lvl;
1990 SYSCTL_RLOCK(&tracker);
1992 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1996 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1998 * You can't call a sysctl when it's a node, but has
1999 * no handler. Inform the user that it's a node.
2000 * The indx may or may not be the same as namelen.
2002 if (oid->oid_handler == NULL) {
2008 /* Is this sysctl writable? */
2009 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) {
2014 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
2016 #ifdef CAPABILITY_MODE
2018 * If the process is in capability mode, then don't permit reading or
2019 * writing unless specifically granted for the node.
2021 if (IN_CAPABILITY_MODE(req->td)) {
2022 if ((req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) ||
2023 (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))) {
2030 /* Is this sysctl sensitive to securelevels? */
2031 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
2032 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
2033 error = securelevel_gt(req->td->td_ucred, lvl);
2038 /* Is this sysctl writable by only privileged users? */
2039 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
2042 if (oid->oid_kind & CTLFLAG_PRISON)
2043 priv = PRIV_SYSCTL_WRITEJAIL;
2045 else if ((oid->oid_kind & CTLFLAG_VNET) &&
2046 prison_owns_vnet(req->td->td_ucred))
2047 priv = PRIV_SYSCTL_WRITEJAIL;
2050 priv = PRIV_SYSCTL_WRITE;
2051 error = priv_check(req->td, priv);
2056 if (!oid->oid_handler) {
2061 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2062 arg1 = (int *)arg1 + indx;
2065 arg1 = oid->oid_arg1;
2066 arg2 = oid->oid_arg2;
2069 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
2075 if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
2076 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
2078 error = sysctl_root_handler_locked(oid, arg1, arg2, req, &tracker);
2081 SYSCTL_RUNLOCK(&tracker);
2085 #ifndef _SYS_SYSPROTO_H_
2086 struct sysctl_args {
2096 sys___sysctl(struct thread *td, struct sysctl_args *uap)
2098 int error, i, name[CTL_MAXNAME];
2101 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
2104 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
2108 error = userland_sysctl(td, name, uap->namelen,
2109 uap->old, uap->oldlenp, 0,
2110 uap->new, uap->newlen, &j, 0);
2111 if (error && error != ENOMEM)
2114 i = copyout(&j, uap->oldlenp, sizeof(j));
2122 * This is used from various compatibility syscalls too. That's why name
2123 * must be in kernel space.
2126 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
2127 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
2130 int error = 0, memlocked;
2131 struct sysctl_req req;
2133 bzero(&req, sizeof req);
2140 req.oldlen = *oldlenp;
2142 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
2147 req.validlen = req.oldlen;
2151 req.newlen = newlen;
2155 req.oldfunc = sysctl_old_user;
2156 req.newfunc = sysctl_new_user;
2157 req.lock = REQ_UNWIRED;
2160 if (KTRPOINT(curthread, KTR_SYSCTL))
2161 ktrsysctl(name, namelen);
2164 if (req.oldptr && req.oldlen > 4 * PAGE_SIZE) {
2166 sx_xlock(&sysctlmemlock);
2168 CURVNET_SET(TD_TO_VNET(td));
2173 error = sysctl_root(0, name, namelen, &req);
2174 if (error != EAGAIN)
2176 kern_yield(PRI_USER);
2181 if (req.lock == REQ_WIRED && req.validlen > 0)
2182 vsunlock(req.oldptr, req.validlen);
2184 sx_xunlock(&sysctlmemlock);
2186 if (error && error != ENOMEM)
2190 if (req.oldptr && req.oldidx > req.validlen)
2191 *retval = req.validlen;
2193 *retval = req.oldidx;
2199 * Drain into a sysctl struct. The user buffer should be wired if a page
2200 * fault would cause issue.
2203 sbuf_sysctl_drain(void *arg, const char *data, int len)
2205 struct sysctl_req *req = arg;
2208 error = SYSCTL_OUT(req, data, len);
2209 KASSERT(error >= 0, ("Got unexpected negative value %d", error));
2210 return (error == 0 ? len : -error);
2214 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
2215 struct sysctl_req *req)
2218 /* Supply a default buffer size if none given. */
2219 if (buf == NULL && length == 0)
2221 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
2222 sbuf_set_drain(s, sbuf_sysctl_drain, req);