2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Mike Karels at Berkeley Software Design, Inc.
8 * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
9 * project, to make these variables more userfriendly.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * @(#)kern_sysctl.c 8.4 (Berkeley) 4/14/94
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
41 #include "opt_capsicum.h"
42 #include "opt_compat.h"
43 #include "opt_ktrace.h"
45 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/capsicum.h>
49 #include <sys/kernel.h>
50 #include <sys/sysctl.h>
51 #include <sys/malloc.h>
56 #include <sys/mutex.h>
57 #include <sys/rmlock.h>
60 #include <sys/sysproto.h>
63 #include <sys/ktrace.h>
68 #include <security/mac/mac_framework.h>
71 #include <vm/vm_extern.h>
73 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
74 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
75 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
78 * The sysctllock protects the MIB tree. It also protects sysctl
79 * contexts used with dynamic sysctls. The sysctl_register_oid() and
80 * sysctl_unregister_oid() routines require the sysctllock to already
81 * be held, so the sysctl_wlock() and sysctl_wunlock() routines are
82 * provided for the few places in the kernel which need to use that
83 * API rather than using the dynamic API. Use of the dynamic API is
84 * strongly encouraged for most code.
86 * The sysctlmemlock is used to limit the amount of user memory wired for
87 * sysctl requests. This is implemented by serializing any userland
88 * sysctl requests larger than a single page via an exclusive lock.
90 static struct rmlock sysctllock;
91 static struct sx sysctlmemlock;
93 #define SYSCTL_WLOCK() rm_wlock(&sysctllock)
94 #define SYSCTL_WUNLOCK() rm_wunlock(&sysctllock)
95 #define SYSCTL_RLOCK(tracker) rm_rlock(&sysctllock, (tracker))
96 #define SYSCTL_RUNLOCK(tracker) rm_runlock(&sysctllock, (tracker))
97 #define SYSCTL_WLOCKED() rm_wowned(&sysctllock)
98 #define SYSCTL_ASSERT_LOCKED() rm_assert(&sysctllock, RA_LOCKED)
99 #define SYSCTL_ASSERT_WLOCKED() rm_assert(&sysctllock, RA_WLOCKED)
100 #define SYSCTL_ASSERT_RLOCKED() rm_assert(&sysctllock, RA_RLOCKED)
101 #define SYSCTL_INIT() rm_init_flags(&sysctllock, "sysctl lock", \
103 #define SYSCTL_SLEEP(ch, wmesg, timo) \
104 rm_sleep(ch, &sysctllock, 0, wmesg, timo)
106 static int sysctl_root(SYSCTL_HANDLER_ARGS);
109 struct sysctl_oid_list sysctl__children = SLIST_HEAD_INITIALIZER(&sysctl__children);
111 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
113 static int sysctl_old_kernel(struct sysctl_req *, const void *, size_t);
114 static int sysctl_new_kernel(struct sysctl_req *, void *, size_t);
116 static struct sysctl_oid *
117 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
119 struct sysctl_oid *oidp;
121 SYSCTL_ASSERT_LOCKED();
122 SLIST_FOREACH(oidp, list, oid_link) {
123 if (strcmp(oidp->oid_name, name) == 0) {
131 * Initialization of the MIB tree.
133 * Order by number in each list.
150 sysctl_root_handler_locked(struct sysctl_oid *oid, void *arg1, intmax_t arg2,
151 struct sysctl_req *req, struct rm_priotracker *tracker)
155 if (oid->oid_kind & CTLFLAG_DYN)
156 atomic_add_int(&oid->oid_running, 1);
159 SYSCTL_RUNLOCK(tracker);
163 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
165 error = oid->oid_handler(oid, arg1, arg2, req);
166 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
170 SYSCTL_RLOCK(tracker);
174 if (oid->oid_kind & CTLFLAG_DYN) {
175 if (atomic_fetchadd_int(&oid->oid_running, -1) == 1 &&
176 (oid->oid_kind & CTLFLAG_DYING) != 0)
177 wakeup(&oid->oid_running);
184 sysctl_load_tunable_by_oid_locked(struct sysctl_oid *oidp)
186 struct sysctl_req req;
187 struct sysctl_oid *curr;
190 ssize_t rem = sizeof(path);
203 for (curr = oidp; curr != NULL; curr = SYSCTL_PARENT(curr)) {
204 len = strlen(curr->oid_name);
209 printf("OID path exceeds %d bytes\n", (int)sizeof(path));
212 memcpy(path + rem, curr->oid_name, len);
214 path[rem + len] = '.';
217 memset(&req, 0, sizeof(req));
220 req.oldfunc = sysctl_old_kernel;
221 req.newfunc = sysctl_new_kernel;
222 req.lock = REQ_UNWIRED;
224 switch (oidp->oid_kind & CTLTYPE) {
226 if (getenv_int(path + rem, &val_int) == 0)
228 req.newlen = sizeof(val_int);
229 req.newptr = &val_int;
232 if (getenv_uint(path + rem, (unsigned int *)&val_int) == 0)
234 req.newlen = sizeof(val_int);
235 req.newptr = &val_int;
238 if (getenv_long(path + rem, &val_long) == 0)
240 req.newlen = sizeof(val_long);
241 req.newptr = &val_long;
244 if (getenv_ulong(path + rem, (unsigned long *)&val_long) == 0)
246 req.newlen = sizeof(val_long);
247 req.newptr = &val_long;
250 if (getenv_int(path + rem, &val_int) == 0)
253 req.newlen = sizeof(val_8);
257 if (getenv_int(path + rem, &val_int) == 0)
260 req.newlen = sizeof(val_16);
261 req.newptr = &val_16;
264 if (getenv_long(path + rem, &val_long) == 0)
267 req.newlen = sizeof(val_32);
268 req.newptr = &val_32;
271 if (getenv_quad(path + rem, &val_quad) == 0)
274 req.newlen = sizeof(val_64);
275 req.newptr = &val_64;
278 if (getenv_uint(path + rem, (unsigned int *)&val_int) == 0)
281 req.newlen = sizeof(val_8);
285 if (getenv_uint(path + rem, (unsigned int *)&val_int) == 0)
288 req.newlen = sizeof(val_16);
289 req.newptr = &val_16;
292 if (getenv_ulong(path + rem, (unsigned long *)&val_long) == 0)
295 req.newlen = sizeof(val_32);
296 req.newptr = &val_32;
299 /* XXX there is no getenv_uquad() */
300 if (getenv_quad(path + rem, &val_quad) == 0)
303 req.newlen = sizeof(val_64);
304 req.newptr = &val_64;
307 penv = kern_getenv(path + rem);
310 req.newlen = strlen(penv);
316 error = sysctl_root_handler_locked(oidp, oidp->oid_arg1,
317 oidp->oid_arg2, &req, NULL);
319 printf("Setting sysctl %s failed: %d\n", path + rem, error);
325 sysctl_register_oid(struct sysctl_oid *oidp)
327 struct sysctl_oid_list *parent = oidp->oid_parent;
328 struct sysctl_oid *p;
329 struct sysctl_oid *q;
334 * First check if another oid with the same name already
335 * exists in the parent's list.
337 SYSCTL_ASSERT_WLOCKED();
338 p = sysctl_find_oidname(oidp->oid_name, parent);
340 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
344 printf("can't re-use a leaf (%s)!\n", p->oid_name);
348 /* get current OID number */
349 oid_number = oidp->oid_number;
352 #error "OID_AUTO is expected to be a negative value"
355 * Any negative OID number qualifies as OID_AUTO. Valid OID
356 * numbers should always be positive.
358 * NOTE: DO NOT change the starting value here, change it in
359 * <sys/sysctl.h>, and make sure it is at least 256 to
360 * accomodate e.g. net.inet.raw as a static sysctl node.
362 if (oid_number < 0) {
366 * By decrementing the next OID number we spend less
367 * time inserting the OIDs into a sorted list.
369 if (--newoid < CTL_AUTO_START)
376 * Insert the OID into the parent's list sorted by OID number.
380 SLIST_FOREACH(p, parent, oid_link) {
381 /* check if the current OID number is in use */
382 if (oid_number == p->oid_number) {
383 /* get the next valid OID number */
384 if (oid_number < CTL_AUTO_START ||
385 oid_number == 0x7fffffff) {
386 /* wraparound - restart */
387 oid_number = CTL_AUTO_START;
388 /* don't loop forever */
390 panic("sysctl: Out of OID numbers\n");
395 } else if (oid_number < p->oid_number)
399 /* check for non-auto OID number collision */
400 if (oidp->oid_number >= 0 && oidp->oid_number < CTL_AUTO_START &&
401 oid_number >= CTL_AUTO_START) {
402 printf("sysctl: OID number(%d) is already in use for '%s'\n",
403 oidp->oid_number, oidp->oid_name);
405 /* update the OID number, if any */
406 oidp->oid_number = oid_number;
408 SLIST_INSERT_AFTER(q, oidp, oid_link);
410 SLIST_INSERT_HEAD(parent, oidp, oid_link);
412 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE &&
414 (oidp->oid_kind & CTLFLAG_VNET) == 0 &&
416 (oidp->oid_kind & CTLFLAG_TUN) != 0 &&
417 (oidp->oid_kind & CTLFLAG_NOFETCH) == 0) {
418 /* only fetch value once */
419 oidp->oid_kind |= CTLFLAG_NOFETCH;
420 /* try to fetch value from kernel environment */
421 sysctl_load_tunable_by_oid_locked(oidp);
426 sysctl_unregister_oid(struct sysctl_oid *oidp)
428 struct sysctl_oid *p;
431 SYSCTL_ASSERT_WLOCKED();
433 if (oidp->oid_number == OID_AUTO) {
436 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
438 SLIST_REMOVE(oidp->oid_parent, oidp,
439 sysctl_oid, oid_link);
447 * This can happen when a module fails to register and is
448 * being unloaded afterwards. It should not be a panic()
452 printf("%s: failed to unregister sysctl\n", __func__);
455 /* Initialize a new context to keep track of dynamically added sysctls. */
457 sysctl_ctx_init(struct sysctl_ctx_list *c)
465 * No locking here, the caller is responsible for not adding
466 * new nodes to a context until after this function has
473 /* Free the context, and destroy all dynamic oids registered in this context */
475 sysctl_ctx_free(struct sysctl_ctx_list *clist)
477 struct sysctl_ctx_entry *e, *e1;
482 * First perform a "dry run" to check if it's ok to remove oids.
484 * XXX This algorithm is a hack. But I don't know any
485 * XXX better solution for now...
488 TAILQ_FOREACH(e, clist, link) {
489 error = sysctl_remove_oid_locked(e->entry, 0, 0);
494 * Restore deregistered entries, either from the end,
495 * or from the place where error occured.
496 * e contains the entry that was not unregistered
499 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
501 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
503 sysctl_register_oid(e1->entry);
504 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
510 /* Now really delete the entries */
511 e = TAILQ_FIRST(clist);
513 e1 = TAILQ_NEXT(e, link);
514 error = sysctl_remove_oid_locked(e->entry, 1, 0);
516 panic("sysctl_remove_oid: corrupt tree, entry: %s",
518 free(e, M_SYSCTLOID);
525 /* Add an entry to the context */
526 struct sysctl_ctx_entry *
527 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
529 struct sysctl_ctx_entry *e;
531 SYSCTL_ASSERT_WLOCKED();
532 if (clist == NULL || oidp == NULL)
534 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
536 TAILQ_INSERT_HEAD(clist, e, link);
540 /* Find an entry in the context */
541 struct sysctl_ctx_entry *
542 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
544 struct sysctl_ctx_entry *e;
546 SYSCTL_ASSERT_WLOCKED();
547 if (clist == NULL || oidp == NULL)
549 TAILQ_FOREACH(e, clist, link) {
557 * Delete an entry from the context.
558 * NOTE: this function doesn't free oidp! You have to remove it
559 * with sysctl_remove_oid().
562 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
564 struct sysctl_ctx_entry *e;
566 if (clist == NULL || oidp == NULL)
569 e = sysctl_ctx_entry_find(clist, oidp);
571 TAILQ_REMOVE(clist, e, link);
573 free(e, M_SYSCTLOID);
582 * Remove dynamically created sysctl trees.
583 * oidp - top of the tree to be removed
584 * del - if 0 - just deregister, otherwise free up entries as well
585 * recurse - if != 0 traverse the subtree to be deleted
588 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
593 error = sysctl_remove_oid_locked(oidp, del, recurse);
599 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
600 int del, int recurse)
602 struct sysctl_oid *p, *tmp;
607 SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
608 if (strcmp(p->oid_name, name) == 0) {
609 error = sysctl_remove_oid_locked(p, del, recurse);
620 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
622 struct sysctl_oid *p, *tmp;
625 SYSCTL_ASSERT_WLOCKED();
628 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
629 printf("can't remove non-dynamic nodes!\n");
633 * WARNING: normal method to do this should be through
634 * sysctl_ctx_free(). Use recursing as the last resort
635 * method to purge your sysctl tree of leftovers...
636 * However, if some other code still references these nodes,
639 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
640 if (oidp->oid_refcnt == 1) {
641 SLIST_FOREACH_SAFE(p,
642 SYSCTL_CHILDREN(oidp), oid_link, tmp) {
644 printf("Warning: failed attempt to "
645 "remove oid %s with child %s\n",
646 oidp->oid_name, p->oid_name);
649 error = sysctl_remove_oid_locked(p, del,
656 if (oidp->oid_refcnt > 1 ) {
659 if (oidp->oid_refcnt == 0) {
660 printf("Warning: bad oid_refcnt=%u (%s)!\n",
661 oidp->oid_refcnt, oidp->oid_name);
664 sysctl_unregister_oid(oidp);
667 * Wait for all threads running the handler to drain.
668 * This preserves the previous behavior when the
669 * sysctl lock was held across a handler invocation,
670 * and is necessary for module unload correctness.
672 while (oidp->oid_running > 0) {
673 oidp->oid_kind |= CTLFLAG_DYING;
674 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
677 free(__DECONST(char *, oidp->oid_descr),
679 free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
680 free(oidp, M_SYSCTLOID);
686 * Create new sysctls at run time.
687 * clist may point to a valid context initialized with sysctl_ctx_init().
690 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
691 int number, const char *name, int kind, void *arg1, intmax_t arg2,
692 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
694 struct sysctl_oid *oidp;
696 /* You have to hook up somewhere.. */
699 /* Check if the node already exists, otherwise create it */
701 oidp = sysctl_find_oidname(name, parent);
703 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
705 /* Update the context */
707 sysctl_ctx_entry_add(clist, oidp);
712 printf("can't re-use a leaf (%s)!\n", name);
716 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
717 oidp->oid_parent = parent;
718 SLIST_INIT(&oidp->oid_children);
719 oidp->oid_number = number;
720 oidp->oid_refcnt = 1;
721 oidp->oid_name = strdup(name, M_SYSCTLOID);
722 oidp->oid_handler = handler;
723 oidp->oid_kind = CTLFLAG_DYN | kind;
724 oidp->oid_arg1 = arg1;
725 oidp->oid_arg2 = arg2;
728 oidp->oid_descr = strdup(descr, M_SYSCTLOID);
729 /* Update the context, if used */
731 sysctl_ctx_entry_add(clist, oidp);
732 /* Register this oid */
733 sysctl_register_oid(oidp);
739 * Rename an existing oid.
742 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
747 newname = strdup(name, M_SYSCTLOID);
749 oldname = __DECONST(char *, oidp->oid_name);
750 oidp->oid_name = newname;
752 free(oldname, M_SYSCTLOID);
756 * Reparent an existing oid.
759 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
761 struct sysctl_oid *oidp;
764 if (oid->oid_parent == parent) {
768 oidp = sysctl_find_oidname(oid->oid_name, parent);
773 sysctl_unregister_oid(oid);
774 oid->oid_parent = parent;
775 oid->oid_number = OID_AUTO;
776 sysctl_register_oid(oid);
782 * Register the kernel's oids on startup.
784 SET_DECLARE(sysctl_set, struct sysctl_oid);
787 sysctl_register_all(void *arg)
789 struct sysctl_oid **oidp;
791 sx_init(&sysctlmemlock, "sysctl mem");
794 SET_FOREACH(oidp, sysctl_set)
795 sysctl_register_oid(*oidp);
798 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_FIRST, sysctl_register_all, 0);
803 * These functions implement a presently undocumented interface
804 * used by the sysctl program to walk the tree, and get the type
805 * so it can print the value.
806 * This interface is under work and consideration, and should probably
807 * be killed with a big axe by the first person who can find the time.
808 * (be aware though, that the proper interface isn't as obvious as it
809 * may seem, there are various conflicting requirements.
811 * {0,0} printf the entire MIB-tree.
812 * {0,1,...} return the name of the "..." OID.
813 * {0,2,...} return the next OID.
814 * {0,3} return the OID of the name in "new"
815 * {0,4,...} return the kind & format info for the "..." OID.
816 * {0,5,...} return the description the "..." OID.
821 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
824 struct sysctl_oid *oidp;
826 SYSCTL_ASSERT_LOCKED();
827 SLIST_FOREACH(oidp, l, oid_link) {
832 printf("%d %s ", oidp->oid_number, oidp->oid_name);
835 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
836 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
838 if (oidp->oid_handler)
841 switch (oidp->oid_kind & CTLTYPE) {
844 if (!oidp->oid_handler) {
845 sysctl_sysctl_debug_dump_node(
846 SYSCTL_CHILDREN(oidp), i + 2);
849 case CTLTYPE_INT: printf(" Int\n"); break;
850 case CTLTYPE_UINT: printf(" u_int\n"); break;
851 case CTLTYPE_LONG: printf(" Long\n"); break;
852 case CTLTYPE_ULONG: printf(" u_long\n"); break;
853 case CTLTYPE_STRING: printf(" String\n"); break;
854 case CTLTYPE_S8: printf(" int8_t\n"); break;
855 case CTLTYPE_S16: printf(" int16_t\n"); break;
856 case CTLTYPE_S32: printf(" int32_t\n"); break;
857 case CTLTYPE_S64: printf(" int64_t\n"); break;
858 case CTLTYPE_U8: printf(" uint8_t\n"); break;
859 case CTLTYPE_U16: printf(" uint16_t\n"); break;
860 case CTLTYPE_U32: printf(" uint32_t\n"); break;
861 case CTLTYPE_U64: printf(" uint64_t\n"); break;
862 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
863 default: printf("\n");
870 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
872 struct rm_priotracker tracker;
875 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
878 SYSCTL_RLOCK(&tracker);
879 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
880 SYSCTL_RUNLOCK(&tracker);
884 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE,
885 0, 0, sysctl_sysctl_debug, "-", "");
889 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
891 int *name = (int *) arg1;
892 u_int namelen = arg2;
894 struct sysctl_oid *oid;
895 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
896 struct rm_priotracker tracker;
899 SYSCTL_RLOCK(&tracker);
902 snprintf(buf,sizeof(buf),"%d",*name);
904 error = SYSCTL_OUT(req, ".", 1);
906 error = SYSCTL_OUT(req, buf, strlen(buf));
914 SLIST_FOREACH(oid, lsp, oid_link) {
915 if (oid->oid_number != *name)
919 error = SYSCTL_OUT(req, ".", 1);
921 error = SYSCTL_OUT(req, oid->oid_name,
922 strlen(oid->oid_name));
929 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
932 if (oid->oid_handler)
935 lsp2 = SYSCTL_CHILDREN(oid);
940 error = SYSCTL_OUT(req, "", 1);
942 SYSCTL_RUNLOCK(&tracker);
947 * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
950 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD,
951 sysctl_sysctl_name, "");
954 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
955 int *next, int *len, int level, struct sysctl_oid **oidpp)
957 struct sysctl_oid *oidp;
959 SYSCTL_ASSERT_LOCKED();
961 SLIST_FOREACH(oidp, lsp, oid_link) {
962 *next = oidp->oid_number;
965 if (oidp->oid_kind & CTLFLAG_SKIP)
969 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
971 if (oidp->oid_handler)
972 /* We really should call the handler here...*/
974 lsp = SYSCTL_CHILDREN(oidp);
975 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
976 len, level+1, oidpp))
981 if (oidp->oid_number < *name)
984 if (oidp->oid_number > *name) {
985 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
987 if (oidp->oid_handler)
989 lsp = SYSCTL_CHILDREN(oidp);
990 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
991 next+1, len, level+1, oidpp))
995 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
998 if (oidp->oid_handler)
1001 lsp = SYSCTL_CHILDREN(oidp);
1002 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
1003 len, level+1, oidpp))
1014 sysctl_sysctl_next(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;
1021 struct rm_priotracker tracker;
1022 int newoid[CTL_MAXNAME];
1024 SYSCTL_RLOCK(&tracker);
1025 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
1026 SYSCTL_RUNLOCK(&tracker);
1029 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
1034 * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
1037 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD,
1038 sysctl_sysctl_next, "");
1041 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
1043 struct sysctl_oid *oidp;
1044 struct sysctl_oid_list *lsp = &sysctl__children;
1047 SYSCTL_ASSERT_LOCKED();
1049 for (*len = 0; *len < CTL_MAXNAME;) {
1050 p = strsep(&name, ".");
1052 oidp = SLIST_FIRST(lsp);
1053 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
1056 if (strcmp(p, oidp->oid_name) == 0)
1059 *oid++ = oidp->oid_number;
1062 if (name == NULL || *name == '\0') {
1068 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1071 if (oidp->oid_handler)
1074 lsp = SYSCTL_CHILDREN(oidp);
1080 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
1083 int error, oid[CTL_MAXNAME], len = 0;
1084 struct sysctl_oid *op = 0;
1085 struct rm_priotracker tracker;
1089 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
1090 return (ENAMETOOLONG);
1092 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
1094 error = SYSCTL_IN(req, p, req->newlen);
1100 p [req->newlen] = '\0';
1102 SYSCTL_RLOCK(&tracker);
1103 error = name2oid(p, oid, &len, &op);
1104 SYSCTL_RUNLOCK(&tracker);
1111 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
1116 * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
1119 SYSCTL_PROC(_sysctl, 3, name2oid,
1120 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE
1121 | CTLFLAG_CAPRW, 0, 0, sysctl_sysctl_name2oid, "I", "");
1124 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
1126 struct sysctl_oid *oid;
1127 struct rm_priotracker tracker;
1130 SYSCTL_RLOCK(&tracker);
1131 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1135 if (oid->oid_fmt == NULL) {
1139 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
1142 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
1144 SYSCTL_RUNLOCK(&tracker);
1149 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
1150 sysctl_sysctl_oidfmt, "");
1153 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
1155 struct sysctl_oid *oid;
1156 struct rm_priotracker tracker;
1159 SYSCTL_RLOCK(&tracker);
1160 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1164 if (oid->oid_descr == NULL) {
1168 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
1170 SYSCTL_RUNLOCK(&tracker);
1174 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
1175 sysctl_sysctl_oiddescr, "");
1178 * Default "handler" functions.
1182 * Handle an int8_t, signed or unsigned.
1184 * a variable: point arg1 at it.
1185 * a constant: pass it in arg2.
1189 sysctl_handle_8(SYSCTL_HANDLER_ARGS)
1195 * Attempt to get a coherent snapshot by making a copy of the data.
1198 tmpout = *(int8_t *)arg1;
1201 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1203 if (error || !req->newptr)
1209 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1214 * Handle an int16_t, signed or unsigned.
1216 * a variable: point arg1 at it.
1217 * a constant: pass it in arg2.
1221 sysctl_handle_16(SYSCTL_HANDLER_ARGS)
1227 * Attempt to get a coherent snapshot by making a copy of the data.
1230 tmpout = *(int16_t *)arg1;
1233 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1235 if (error || !req->newptr)
1241 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1246 * Handle an int32_t, signed or unsigned.
1248 * a variable: point arg1 at it.
1249 * a constant: pass it in arg2.
1253 sysctl_handle_32(SYSCTL_HANDLER_ARGS)
1259 * Attempt to get a coherent snapshot by making a copy of the data.
1262 tmpout = *(int32_t *)arg1;
1265 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1267 if (error || !req->newptr)
1273 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1278 * Handle an int, signed or unsigned.
1280 * a variable: point arg1 at it.
1281 * a constant: pass it in arg2.
1285 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
1287 int tmpout, error = 0;
1290 * Attempt to get a coherent snapshot by making a copy of the data.
1293 tmpout = *(int *)arg1;
1296 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1298 if (error || !req->newptr)
1304 error = SYSCTL_IN(req, arg1, sizeof(int));
1309 * Based on on sysctl_handle_int() convert milliseconds into ticks.
1310 * Note: this is used by TCP.
1314 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1319 s = (int)((int64_t)tt * 1000 / hz);
1321 error = sysctl_handle_int(oidp, &s, 0, req);
1322 if (error || !req->newptr)
1325 tt = (int)((int64_t)s * hz / 1000);
1335 * Handle a long, signed or unsigned.
1337 * a variable: point arg1 at it.
1338 * a constant: pass it in arg2.
1342 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1351 * Attempt to get a coherent snapshot by making a copy of the data.
1354 tmplong = *(long *)arg1;
1358 if (req->flags & SCTL_MASK32) {
1360 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1363 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1365 if (error || !req->newptr)
1371 else if (req->flags & SCTL_MASK32) {
1372 error = SYSCTL_IN(req, &tmpint, sizeof(int));
1373 *(long *)arg1 = (long)tmpint;
1377 error = SYSCTL_IN(req, arg1, sizeof(long));
1382 * Handle a 64 bit int, signed or unsigned.
1384 * a variable: point arg1 at it.
1385 * a constant: pass it in arg2.
1388 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1394 * Attempt to get a coherent snapshot by making a copy of the data.
1397 tmpout = *(uint64_t *)arg1;
1400 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1402 if (error || !req->newptr)
1408 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1413 * Handle our generic '\0' terminated 'C' string.
1415 * a variable string: point arg1 at it, arg2 is max length.
1416 * a constant string: point arg1 at it, arg2 is zero.
1420 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1423 int error = 0, ro_string = 0;
1426 * A zero-length buffer indicates a fixed size read-only
1430 arg2 = strlen((char *)arg1) + 1;
1434 if (req->oldptr != NULL) {
1440 /* try to make a coherent snapshot of the string */
1441 tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK);
1442 memcpy(tmparg, arg1, arg2);
1445 outlen = strnlen(tmparg, arg2 - 1) + 1;
1446 error = SYSCTL_OUT(req, tmparg, outlen);
1449 free(tmparg, M_SYSCTLTMP);
1451 outlen = strnlen((char *)arg1, arg2 - 1) + 1;
1452 error = SYSCTL_OUT(req, NULL, outlen);
1454 if (error || !req->newptr)
1457 if ((req->newlen - req->newidx) >= arg2) {
1460 arg2 = (req->newlen - req->newidx);
1461 error = SYSCTL_IN(req, arg1, arg2);
1462 ((char *)arg1)[arg2] = '\0';
1468 * Handle any kind of opaque data.
1469 * arg1 points to it, arg2 is the size.
1473 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1477 struct sysctl_req req2;
1480 * Attempt to get a coherent snapshot, by using the thread
1481 * pre-emption counter updated from within mi_switch() to
1482 * determine if we were pre-empted during a bcopy() or
1483 * copyout(). Make 3 attempts at doing this before giving up.
1484 * If we encounter an error, stop immediately.
1489 generation = curthread->td_generation;
1490 error = SYSCTL_OUT(req, arg1, arg2);
1494 if (generation != curthread->td_generation && tries < 3) {
1499 error = SYSCTL_IN(req, arg1, arg2);
1505 * Transfer functions to/from kernel space.
1506 * XXX: rather untested at this point
1509 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1515 if (req->oldlen <= req->oldidx)
1518 if (i > req->oldlen - req->oldidx)
1519 i = req->oldlen - req->oldidx;
1521 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1524 if (req->oldptr && i != l)
1530 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1534 if (req->newlen - req->newidx < l)
1536 bcopy((char *)req->newptr + req->newidx, p, l);
1542 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1543 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1546 struct sysctl_req req;
1548 bzero(&req, sizeof req);
1554 req.oldlen = *oldlenp;
1556 req.validlen = req.oldlen;
1563 req.newlen = newlen;
1567 req.oldfunc = sysctl_old_kernel;
1568 req.newfunc = sysctl_new_kernel;
1569 req.lock = REQ_UNWIRED;
1571 error = sysctl_root(0, name, namelen, &req);
1573 if (req.lock == REQ_WIRED && req.validlen > 0)
1574 vsunlock(req.oldptr, req.validlen);
1576 if (error && error != ENOMEM)
1580 if (req.oldptr && req.oldidx > req.validlen)
1581 *retval = req.validlen;
1583 *retval = req.oldidx;
1589 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1590 void *new, size_t newlen, size_t *retval, int flags)
1592 int oid[CTL_MAXNAME];
1593 size_t oidlen, plen;
1596 oid[0] = 0; /* sysctl internal magic */
1597 oid[1] = 3; /* name2oid */
1598 oidlen = sizeof(oid);
1600 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1601 (void *)name, strlen(name), &plen, flags);
1605 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1606 new, newlen, retval, flags);
1611 * Transfer function to/from user space.
1614 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1616 size_t i, len, origidx;
1619 origidx = req->oldidx;
1621 if (req->oldptr == NULL)
1624 * If we have not wired the user supplied buffer and we are currently
1625 * holding locks, drop a witness warning, as it's possible that
1626 * write operations to the user page can sleep.
1628 if (req->lock != REQ_WIRED)
1629 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1630 "sysctl_old_user()");
1632 len = req->validlen;
1636 if (i > len - origidx)
1638 if (req->lock == REQ_WIRED) {
1639 error = copyout_nofault(p, (char *)req->oldptr +
1642 error = copyout(p, (char *)req->oldptr + origidx, i);
1652 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1658 if (req->newlen - req->newidx < l)
1660 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1661 "sysctl_new_user()");
1662 error = copyin((char *)req->newptr + req->newidx, p, l);
1668 * Wire the user space destination buffer. If set to a value greater than
1669 * zero, the len parameter limits the maximum amount of wired memory.
1672 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1677 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1679 if (req->lock != REQ_WIRED && req->oldptr &&
1680 req->oldfunc == sysctl_old_user) {
1681 if (wiredlen != 0) {
1682 ret = vslock(req->oldptr, wiredlen);
1689 req->lock = REQ_WIRED;
1690 req->validlen = wiredlen;
1696 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1697 int *nindx, struct sysctl_req *req)
1699 struct sysctl_oid_list *lsp;
1700 struct sysctl_oid *oid;
1703 SYSCTL_ASSERT_LOCKED();
1704 lsp = &sysctl__children;
1706 while (indx < CTL_MAXNAME) {
1707 SLIST_FOREACH(oid, lsp, oid_link) {
1708 if (oid->oid_number == name[indx])
1715 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1716 if (oid->oid_handler != NULL || indx == namelen) {
1720 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1721 ("%s found DYING node %p", __func__, oid));
1724 lsp = SYSCTL_CHILDREN(oid);
1725 } else if (indx == namelen) {
1729 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1730 ("%s found DYING node %p", __func__, oid));
1740 * Traverse our tree, and find the right node, execute whatever it points
1741 * to, and return the resulting error code.
1745 sysctl_root(SYSCTL_HANDLER_ARGS)
1747 struct sysctl_oid *oid;
1748 struct rm_priotracker tracker;
1749 int error, indx, lvl;
1751 SYSCTL_RLOCK(&tracker);
1753 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1757 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1759 * You can't call a sysctl when it's a node, but has
1760 * no handler. Inform the user that it's a node.
1761 * The indx may or may not be the same as namelen.
1763 if (oid->oid_handler == NULL) {
1769 /* Is this sysctl writable? */
1770 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) {
1775 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1777 #ifdef CAPABILITY_MODE
1779 * If the process is in capability mode, then don't permit reading or
1780 * writing unless specifically granted for the node.
1782 if (IN_CAPABILITY_MODE(req->td)) {
1783 if ((req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) ||
1784 (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))) {
1791 /* Is this sysctl sensitive to securelevels? */
1792 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1793 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1794 error = securelevel_gt(req->td->td_ucred, lvl);
1799 /* Is this sysctl writable by only privileged users? */
1800 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1803 if (oid->oid_kind & CTLFLAG_PRISON)
1804 priv = PRIV_SYSCTL_WRITEJAIL;
1806 else if ((oid->oid_kind & CTLFLAG_VNET) &&
1807 prison_owns_vnet(req->td->td_ucred))
1808 priv = PRIV_SYSCTL_WRITEJAIL;
1811 priv = PRIV_SYSCTL_WRITE;
1812 error = priv_check(req->td, priv);
1817 if (!oid->oid_handler) {
1822 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1823 arg1 = (int *)arg1 + indx;
1826 arg1 = oid->oid_arg1;
1827 arg2 = oid->oid_arg2;
1830 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1836 if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
1837 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
1839 error = sysctl_root_handler_locked(oid, arg1, arg2, req, &tracker);
1841 KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
1844 SYSCTL_RUNLOCK(&tracker);
1848 #ifndef _SYS_SYSPROTO_H_
1849 struct sysctl_args {
1859 sys___sysctl(struct thread *td, struct sysctl_args *uap)
1861 int error, i, name[CTL_MAXNAME];
1864 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1867 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1871 error = userland_sysctl(td, name, uap->namelen,
1872 uap->old, uap->oldlenp, 0,
1873 uap->new, uap->newlen, &j, 0);
1874 if (error && error != ENOMEM)
1877 i = copyout(&j, uap->oldlenp, sizeof(j));
1885 * This is used from various compatibility syscalls too. That's why name
1886 * must be in kernel space.
1889 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1890 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1893 int error = 0, memlocked;
1894 struct sysctl_req req;
1896 bzero(&req, sizeof req);
1903 req.oldlen = *oldlenp;
1905 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1910 req.validlen = req.oldlen;
1913 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1919 if (!useracc(new, newlen, VM_PROT_READ))
1921 req.newlen = newlen;
1925 req.oldfunc = sysctl_old_user;
1926 req.newfunc = sysctl_new_user;
1927 req.lock = REQ_UNWIRED;
1930 if (KTRPOINT(curthread, KTR_SYSCTL))
1931 ktrsysctl(name, namelen);
1934 if (req.oldptr && req.oldlen > PAGE_SIZE) {
1936 sx_xlock(&sysctlmemlock);
1939 CURVNET_SET(TD_TO_VNET(td));
1944 error = sysctl_root(0, name, namelen, &req);
1945 if (error != EAGAIN)
1947 kern_yield(PRI_USER);
1952 if (req.lock == REQ_WIRED && req.validlen > 0)
1953 vsunlock(req.oldptr, req.validlen);
1955 sx_xunlock(&sysctlmemlock);
1957 if (error && error != ENOMEM)
1961 if (req.oldptr && req.oldidx > req.validlen)
1962 *retval = req.validlen;
1964 *retval = req.oldidx;
1970 * Drain into a sysctl struct. The user buffer should be wired if a page
1971 * fault would cause issue.
1974 sbuf_sysctl_drain(void *arg, const char *data, int len)
1976 struct sysctl_req *req = arg;
1979 error = SYSCTL_OUT(req, data, len);
1980 KASSERT(error >= 0, ("Got unexpected negative value %d", error));
1981 return (error == 0 ? len : -error);
1985 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
1986 struct sysctl_req *req)
1989 /* Supply a default buffer size if none given. */
1990 if (buf == NULL && length == 0)
1992 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
1993 sbuf_set_drain(s, sbuf_sysctl_drain, req);