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/capability.h>
49 #include <sys/kernel.h>
50 #include <sys/sysctl.h>
51 #include <sys/malloc.h>
56 #include <sys/mutex.h>
59 #include <sys/sysproto.h>
62 #include <sys/ktrace.h>
67 #include <security/mac/mac_framework.h>
70 #include <vm/vm_extern.h>
72 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
73 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
74 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
77 * The sysctllock protects the MIB tree. It also protects sysctl
78 * contexts used with dynamic sysctls. The sysctl_register_oid() and
79 * sysctl_unregister_oid() routines require the sysctllock to already
80 * be held, so the sysctl_lock() and sysctl_unlock() routines are
81 * provided for the few places in the kernel which need to use that
82 * API rather than using the dynamic API. Use of the dynamic API is
83 * strongly encouraged for most code.
85 * The sysctlmemlock is used to limit the amount of user memory wired for
86 * sysctl requests. This is implemented by serializing any userland
87 * sysctl requests larger than a single page via an exclusive lock.
89 static struct sx sysctllock;
90 static struct sx sysctlmemlock;
92 #define SYSCTL_XLOCK() sx_xlock(&sysctllock)
93 #define SYSCTL_XUNLOCK() sx_xunlock(&sysctllock)
94 #define SYSCTL_ASSERT_XLOCKED() sx_assert(&sysctllock, SA_XLOCKED)
95 #define SYSCTL_INIT() sx_init(&sysctllock, "sysctl lock")
96 #define SYSCTL_SLEEP(ch, wmesg, timo) \
97 sx_sleep(ch, &sysctllock, 0, wmesg, timo)
99 static int sysctl_root(SYSCTL_HANDLER_ARGS);
101 struct sysctl_oid_list sysctl__children; /* root list */
103 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
106 static struct sysctl_oid *
107 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
109 struct sysctl_oid *oidp;
111 SYSCTL_ASSERT_XLOCKED();
112 SLIST_FOREACH(oidp, list, oid_link) {
113 if (strcmp(oidp->oid_name, name) == 0) {
121 * Initialization of the MIB tree.
123 * Order by number in each list.
140 sysctl_register_oid(struct sysctl_oid *oidp)
142 struct sysctl_oid_list *parent = oidp->oid_parent;
143 struct sysctl_oid *p;
144 struct sysctl_oid *q;
147 * First check if another oid with the same name already
148 * exists in the parent's list.
150 SYSCTL_ASSERT_XLOCKED();
151 p = sysctl_find_oidname(oidp->oid_name, parent);
153 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
157 printf("can't re-use a leaf (%s)!\n", p->oid_name);
162 * If this oid has a number OID_AUTO, give it a number which
163 * is greater than any current oid.
164 * NOTE: DO NOT change the starting value here, change it in
165 * <sys/sysctl.h>, and make sure it is at least 256 to
166 * accomodate e.g. net.inet.raw as a static sysctl node.
168 if (oidp->oid_number == OID_AUTO) {
169 static int newoid = CTL_AUTO_START;
171 oidp->oid_number = newoid++;
172 if (newoid == 0x7fffffff)
173 panic("out of oids");
176 else if (oidp->oid_number >= CTL_AUTO_START) {
177 /* do not panic; this happens when unregistering sysctl sets */
178 printf("static sysctl oid too high: %d", oidp->oid_number);
183 * Insert the oid into the parent's list in order.
186 SLIST_FOREACH(p, parent, oid_link) {
187 if (oidp->oid_number < p->oid_number)
192 SLIST_INSERT_AFTER(q, oidp, oid_link);
194 SLIST_INSERT_HEAD(parent, oidp, oid_link);
198 sysctl_unregister_oid(struct sysctl_oid *oidp)
200 struct sysctl_oid *p;
203 SYSCTL_ASSERT_XLOCKED();
205 if (oidp->oid_number == OID_AUTO) {
208 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
210 SLIST_REMOVE(oidp->oid_parent, oidp,
211 sysctl_oid, oid_link);
219 * This can happen when a module fails to register and is
220 * being unloaded afterwards. It should not be a panic()
224 printf("%s: failed to unregister sysctl\n", __func__);
227 /* Initialize a new context to keep track of dynamically added sysctls. */
229 sysctl_ctx_init(struct sysctl_ctx_list *c)
237 * No locking here, the caller is responsible for not adding
238 * new nodes to a context until after this function has
245 /* Free the context, and destroy all dynamic oids registered in this context */
247 sysctl_ctx_free(struct sysctl_ctx_list *clist)
249 struct sysctl_ctx_entry *e, *e1;
254 * First perform a "dry run" to check if it's ok to remove oids.
256 * XXX This algorithm is a hack. But I don't know any
257 * XXX better solution for now...
260 TAILQ_FOREACH(e, clist, link) {
261 error = sysctl_remove_oid_locked(e->entry, 0, 0);
266 * Restore deregistered entries, either from the end,
267 * or from the place where error occured.
268 * e contains the entry that was not unregistered
271 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
273 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
275 sysctl_register_oid(e1->entry);
276 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
282 /* Now really delete the entries */
283 e = TAILQ_FIRST(clist);
285 e1 = TAILQ_NEXT(e, link);
286 error = sysctl_remove_oid_locked(e->entry, 1, 0);
288 panic("sysctl_remove_oid: corrupt tree, entry: %s",
290 free(e, M_SYSCTLOID);
297 /* Add an entry to the context */
298 struct sysctl_ctx_entry *
299 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
301 struct sysctl_ctx_entry *e;
303 SYSCTL_ASSERT_XLOCKED();
304 if (clist == NULL || oidp == NULL)
306 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
308 TAILQ_INSERT_HEAD(clist, e, link);
312 /* Find an entry in the context */
313 struct sysctl_ctx_entry *
314 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
316 struct sysctl_ctx_entry *e;
318 SYSCTL_ASSERT_XLOCKED();
319 if (clist == NULL || oidp == NULL)
321 TAILQ_FOREACH(e, clist, link) {
329 * Delete an entry from the context.
330 * NOTE: this function doesn't free oidp! You have to remove it
331 * with sysctl_remove_oid().
334 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
336 struct sysctl_ctx_entry *e;
338 if (clist == NULL || oidp == NULL)
341 e = sysctl_ctx_entry_find(clist, oidp);
343 TAILQ_REMOVE(clist, e, link);
345 free(e, M_SYSCTLOID);
354 * Remove dynamically created sysctl trees.
355 * oidp - top of the tree to be removed
356 * del - if 0 - just deregister, otherwise free up entries as well
357 * recurse - if != 0 traverse the subtree to be deleted
360 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
365 error = sysctl_remove_oid_locked(oidp, del, recurse);
371 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
372 int del, int recurse)
374 struct sysctl_oid *p, *tmp;
379 SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
380 if (strcmp(p->oid_name, name) == 0) {
381 error = sysctl_remove_oid_locked(p, del, recurse);
392 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
394 struct sysctl_oid *p, *tmp;
397 SYSCTL_ASSERT_XLOCKED();
400 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
401 printf("can't remove non-dynamic nodes!\n");
405 * WARNING: normal method to do this should be through
406 * sysctl_ctx_free(). Use recursing as the last resort
407 * method to purge your sysctl tree of leftovers...
408 * However, if some other code still references these nodes,
411 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
412 if (oidp->oid_refcnt == 1) {
413 SLIST_FOREACH_SAFE(p,
414 SYSCTL_CHILDREN(oidp), oid_link, tmp) {
417 error = sysctl_remove_oid_locked(p, del,
423 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
426 if (oidp->oid_refcnt > 1 ) {
429 if (oidp->oid_refcnt == 0) {
430 printf("Warning: bad oid_refcnt=%u (%s)!\n",
431 oidp->oid_refcnt, oidp->oid_name);
434 sysctl_unregister_oid(oidp);
437 * Wait for all threads running the handler to drain.
438 * This preserves the previous behavior when the
439 * sysctl lock was held across a handler invocation,
440 * and is necessary for module unload correctness.
442 while (oidp->oid_running > 0) {
443 oidp->oid_kind |= CTLFLAG_DYING;
444 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
447 free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
448 free((void *)(uintptr_t)(const void *)oidp->oid_name,
450 free(oidp, M_SYSCTLOID);
456 * Create new sysctls at run time.
457 * clist may point to a valid context initialized with sysctl_ctx_init().
460 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
461 int number, const char *name, int kind, void *arg1, intptr_t arg2,
462 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
464 struct sysctl_oid *oidp;
468 /* You have to hook up somewhere.. */
471 /* Check if the node already exists, otherwise create it */
473 oidp = sysctl_find_oidname(name, parent);
475 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
477 /* Update the context */
479 sysctl_ctx_entry_add(clist, oidp);
484 printf("can't re-use a leaf (%s)!\n", name);
488 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
489 oidp->oid_parent = parent;
490 SLIST_NEXT(oidp, oid_link) = NULL;
491 oidp->oid_number = number;
492 oidp->oid_refcnt = 1;
494 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
495 bcopy(name, newname, len + 1);
497 oidp->oid_name = newname;
498 oidp->oid_handler = handler;
499 oidp->oid_kind = CTLFLAG_DYN | kind;
500 if ((kind & CTLTYPE) == CTLTYPE_NODE) {
501 /* Allocate space for children */
502 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
503 M_SYSCTLOID, M_WAITOK));
504 SLIST_INIT(SYSCTL_CHILDREN(oidp));
505 oidp->oid_arg2 = arg2;
507 oidp->oid_arg1 = arg1;
508 oidp->oid_arg2 = arg2;
512 int len = strlen(descr) + 1;
513 oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
515 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
517 /* Update the context, if used */
519 sysctl_ctx_entry_add(clist, oidp);
520 /* Register this oid */
521 sysctl_register_oid(oidp);
527 * Rename an existing oid.
530 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
537 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
538 bcopy(name, newname, len + 1);
541 oldname = (void *)(uintptr_t)(const void *)oidp->oid_name;
542 oidp->oid_name = newname;
544 free(oldname, M_SYSCTLOID);
548 * Reparent an existing oid.
551 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
553 struct sysctl_oid *oidp;
556 if (oid->oid_parent == parent) {
560 oidp = sysctl_find_oidname(oid->oid_name, parent);
565 sysctl_unregister_oid(oid);
566 oid->oid_parent = parent;
567 oid->oid_number = OID_AUTO;
568 sysctl_register_oid(oid);
574 * Register the kernel's oids on startup.
576 SET_DECLARE(sysctl_set, struct sysctl_oid);
579 sysctl_register_all(void *arg)
581 struct sysctl_oid **oidp;
583 sx_init(&sysctlmemlock, "sysctl mem");
586 SET_FOREACH(oidp, sysctl_set)
587 sysctl_register_oid(*oidp);
590 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
595 * These functions implement a presently undocumented interface
596 * used by the sysctl program to walk the tree, and get the type
597 * so it can print the value.
598 * This interface is under work and consideration, and should probably
599 * be killed with a big axe by the first person who can find the time.
600 * (be aware though, that the proper interface isn't as obvious as it
601 * may seem, there are various conflicting requirements.
603 * {0,0} printf the entire MIB-tree.
604 * {0,1,...} return the name of the "..." OID.
605 * {0,2,...} return the next OID.
606 * {0,3} return the OID of the name in "new"
607 * {0,4,...} return the kind & format info for the "..." OID.
608 * {0,5,...} return the description the "..." OID.
613 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
616 struct sysctl_oid *oidp;
618 SYSCTL_ASSERT_XLOCKED();
619 SLIST_FOREACH(oidp, l, oid_link) {
624 printf("%d %s ", oidp->oid_number, oidp->oid_name);
627 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
628 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
630 if (oidp->oid_handler)
633 switch (oidp->oid_kind & CTLTYPE) {
636 if (!oidp->oid_handler) {
637 sysctl_sysctl_debug_dump_node(
638 oidp->oid_arg1, i+2);
641 case CTLTYPE_INT: printf(" Int\n"); break;
642 case CTLTYPE_UINT: printf(" u_int\n"); break;
643 case CTLTYPE_LONG: printf(" Long\n"); break;
644 case CTLTYPE_ULONG: printf(" u_long\n"); break;
645 case CTLTYPE_STRING: printf(" String\n"); break;
646 case CTLTYPE_U64: printf(" uint64_t\n"); break;
647 case CTLTYPE_S64: printf(" int64_t\n"); break;
648 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
649 default: printf("\n");
656 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
660 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
664 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
669 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
670 0, 0, sysctl_sysctl_debug, "-", "");
674 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
676 int *name = (int *) arg1;
677 u_int namelen = arg2;
679 struct sysctl_oid *oid;
680 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
686 snprintf(buf,sizeof(buf),"%d",*name);
688 error = SYSCTL_OUT(req, ".", 1);
690 error = SYSCTL_OUT(req, buf, strlen(buf));
698 SLIST_FOREACH(oid, lsp, oid_link) {
699 if (oid->oid_number != *name)
703 error = SYSCTL_OUT(req, ".", 1);
705 error = SYSCTL_OUT(req, oid->oid_name,
706 strlen(oid->oid_name));
713 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
716 if (oid->oid_handler)
719 lsp2 = SYSCTL_CHILDREN(oid);
724 error = SYSCTL_OUT(req, "", 1);
731 * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
734 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD | CTLFLAG_CAPRD,
735 sysctl_sysctl_name, "");
738 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
739 int *next, int *len, int level, struct sysctl_oid **oidpp)
741 struct sysctl_oid *oidp;
743 SYSCTL_ASSERT_XLOCKED();
745 SLIST_FOREACH(oidp, lsp, oid_link) {
746 *next = oidp->oid_number;
749 if (oidp->oid_kind & CTLFLAG_SKIP)
753 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
755 if (oidp->oid_handler)
756 /* We really should call the handler here...*/
758 lsp = SYSCTL_CHILDREN(oidp);
759 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
760 len, level+1, oidpp))
765 if (oidp->oid_number < *name)
768 if (oidp->oid_number > *name) {
769 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
771 if (oidp->oid_handler)
773 lsp = SYSCTL_CHILDREN(oidp);
774 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
775 next+1, len, level+1, oidpp))
779 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
782 if (oidp->oid_handler)
785 lsp = SYSCTL_CHILDREN(oidp);
786 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
787 len, level+1, oidpp))
798 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
800 int *name = (int *) arg1;
801 u_int namelen = arg2;
803 struct sysctl_oid *oid;
804 struct sysctl_oid_list *lsp = &sysctl__children;
805 int newoid[CTL_MAXNAME];
808 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
812 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
817 * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
820 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD | CTLFLAG_CAPRD,
821 sysctl_sysctl_next, "");
824 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
827 struct sysctl_oid *oidp;
828 struct sysctl_oid_list *lsp = &sysctl__children;
831 SYSCTL_ASSERT_XLOCKED();
836 p = name + strlen(name) - 1 ;
842 for (p = name; *p && *p != '.'; p++)
848 oidp = SLIST_FIRST(lsp);
850 while (oidp && *len < CTL_MAXNAME) {
851 if (strcmp(name, oidp->oid_name)) {
852 oidp = SLIST_NEXT(oidp, oid_link);
855 *oid++ = oidp->oid_number;
864 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
867 if (oidp->oid_handler)
870 lsp = SYSCTL_CHILDREN(oidp);
871 oidp = SLIST_FIRST(lsp);
873 for (p = name; *p && *p != '.'; p++)
883 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
886 int error, oid[CTL_MAXNAME], len = 0;
887 struct sysctl_oid *op = 0;
891 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
892 return (ENAMETOOLONG);
894 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
896 error = SYSCTL_IN(req, p, req->newlen);
902 p [req->newlen] = '\0';
905 error = name2oid(p, oid, &len, &op);
913 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
918 * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
921 SYSCTL_PROC(_sysctl, 3, name2oid,
922 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE
923 | CTLFLAG_CAPRW, 0, 0, sysctl_sysctl_name2oid, "I", "");
926 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
928 struct sysctl_oid *oid;
932 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
936 if (oid->oid_fmt == NULL) {
940 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
943 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
950 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
951 sysctl_sysctl_oidfmt, "");
954 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
956 struct sysctl_oid *oid;
960 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
964 if (oid->oid_descr == NULL) {
968 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
974 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD|CTLFLAG_CAPRD,
975 sysctl_sysctl_oiddescr, "");
978 * Default "handler" functions.
982 * Handle an int, signed or unsigned.
984 * a variable: point arg1 at it.
985 * a constant: pass it in arg2.
989 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
991 int tmpout, error = 0;
994 * Attempt to get a coherent snapshot by making a copy of the data.
997 tmpout = *(int *)arg1;
1000 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1002 if (error || !req->newptr)
1008 error = SYSCTL_IN(req, arg1, sizeof(int));
1013 * Based on on sysctl_handle_int() convert milliseconds into ticks.
1014 * Note: this is used by TCP.
1018 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1023 s = (int)((int64_t)tt * 1000 / hz);
1025 error = sysctl_handle_int(oidp, &s, 0, req);
1026 if (error || !req->newptr)
1029 tt = (int)((int64_t)s * hz / 1000);
1039 * Handle a long, signed or unsigned. arg1 points to it.
1043 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1052 * Attempt to get a coherent snapshot by making a copy of the data.
1056 tmplong = *(long *)arg1;
1058 if (req->flags & SCTL_MASK32) {
1060 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1063 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1065 if (error || !req->newptr)
1069 if (req->flags & SCTL_MASK32) {
1070 error = SYSCTL_IN(req, &tmpint, sizeof(int));
1071 *(long *)arg1 = (long)tmpint;
1074 error = SYSCTL_IN(req, arg1, sizeof(long));
1079 * Handle a 64 bit int, signed or unsigned. arg1 points to it.
1082 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1088 * Attempt to get a coherent snapshot by making a copy of the data.
1092 tmpout = *(uint64_t *)arg1;
1093 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1095 if (error || !req->newptr)
1098 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1103 * Handle our generic '\0' terminated 'C' string.
1105 * a variable string: point arg1 at it, arg2 is max length.
1106 * a constant string: point arg1 at it, arg2 is zero.
1110 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1117 * Attempt to get a coherent snapshot by copying to a
1118 * temporary kernel buffer.
1121 outlen = strlen((char *)arg1)+1;
1122 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
1124 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
1125 free(tmparg, M_SYSCTLTMP);
1129 error = SYSCTL_OUT(req, tmparg, outlen);
1130 free(tmparg, M_SYSCTLTMP);
1132 if (error || !req->newptr)
1135 if ((req->newlen - req->newidx) >= arg2) {
1138 arg2 = (req->newlen - req->newidx);
1139 error = SYSCTL_IN(req, arg1, arg2);
1140 ((char *)arg1)[arg2] = '\0';
1147 * Handle any kind of opaque data.
1148 * arg1 points to it, arg2 is the size.
1152 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1156 struct sysctl_req req2;
1159 * Attempt to get a coherent snapshot, by using the thread
1160 * pre-emption counter updated from within mi_switch() to
1161 * determine if we were pre-empted during a bcopy() or
1162 * copyout(). Make 3 attempts at doing this before giving up.
1163 * If we encounter an error, stop immediately.
1168 generation = curthread->td_generation;
1169 error = SYSCTL_OUT(req, arg1, arg2);
1173 if (generation != curthread->td_generation && tries < 3) {
1178 error = SYSCTL_IN(req, arg1, arg2);
1184 * Transfer functions to/from kernel space.
1185 * XXX: rather untested at this point
1188 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1194 if (req->oldlen <= req->oldidx)
1197 if (i > req->oldlen - req->oldidx)
1198 i = req->oldlen - req->oldidx;
1200 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1203 if (req->oldptr && i != l)
1209 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1213 if (req->newlen - req->newidx < l)
1215 bcopy((char *)req->newptr + req->newidx, p, l);
1221 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1222 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1225 struct sysctl_req req;
1227 bzero(&req, sizeof req);
1233 req.oldlen = *oldlenp;
1235 req.validlen = req.oldlen;
1242 req.newlen = newlen;
1246 req.oldfunc = sysctl_old_kernel;
1247 req.newfunc = sysctl_new_kernel;
1248 req.lock = REQ_UNWIRED;
1251 error = sysctl_root(0, name, namelen, &req);
1254 if (req.lock == REQ_WIRED && req.validlen > 0)
1255 vsunlock(req.oldptr, req.validlen);
1257 if (error && error != ENOMEM)
1261 if (req.oldptr && req.oldidx > req.validlen)
1262 *retval = req.validlen;
1264 *retval = req.oldidx;
1270 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1271 void *new, size_t newlen, size_t *retval, int flags)
1273 int oid[CTL_MAXNAME];
1274 size_t oidlen, plen;
1277 oid[0] = 0; /* sysctl internal magic */
1278 oid[1] = 3; /* name2oid */
1279 oidlen = sizeof(oid);
1281 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1282 (void *)name, strlen(name), &plen, flags);
1286 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1287 new, newlen, retval, flags);
1292 * Transfer function to/from user space.
1295 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1298 size_t i, len, origidx;
1300 origidx = req->oldidx;
1302 if (req->oldptr == NULL)
1305 * If we have not wired the user supplied buffer and we are currently
1306 * holding locks, drop a witness warning, as it's possible that
1307 * write operations to the user page can sleep.
1309 if (req->lock != REQ_WIRED)
1310 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1311 "sysctl_old_user()");
1313 len = req->validlen;
1317 if (i > len - origidx)
1319 error = copyout(p, (char *)req->oldptr + origidx, i);
1329 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1335 if (req->newlen - req->newidx < l)
1337 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1338 "sysctl_new_user()");
1339 error = copyin((char *)req->newptr + req->newidx, p, l);
1345 * Wire the user space destination buffer. If set to a value greater than
1346 * zero, the len parameter limits the maximum amount of wired memory.
1349 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1354 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1356 if (req->lock != REQ_WIRED && req->oldptr &&
1357 req->oldfunc == sysctl_old_user) {
1358 if (wiredlen != 0) {
1359 ret = vslock(req->oldptr, wiredlen);
1366 req->lock = REQ_WIRED;
1367 req->validlen = wiredlen;
1373 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1374 int *nindx, struct sysctl_req *req)
1376 struct sysctl_oid_list *lsp;
1377 struct sysctl_oid *oid;
1380 SYSCTL_ASSERT_XLOCKED();
1381 lsp = &sysctl__children;
1383 while (indx < CTL_MAXNAME) {
1384 SLIST_FOREACH(oid, lsp, oid_link) {
1385 if (oid->oid_number == name[indx])
1392 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1393 if (oid->oid_handler != NULL || indx == namelen) {
1397 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1398 ("%s found DYING node %p", __func__, oid));
1401 lsp = SYSCTL_CHILDREN(oid);
1402 } else if (indx == namelen) {
1406 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1407 ("%s found DYING node %p", __func__, oid));
1417 * Traverse our tree, and find the right node, execute whatever it points
1418 * to, and return the resulting error code.
1422 sysctl_root(SYSCTL_HANDLER_ARGS)
1424 struct sysctl_oid *oid;
1425 int error, indx, lvl;
1427 SYSCTL_ASSERT_XLOCKED();
1429 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1433 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1435 * You can't call a sysctl when it's a node, but has
1436 * no handler. Inform the user that it's a node.
1437 * The indx may or may not be the same as namelen.
1439 if (oid->oid_handler == NULL)
1443 /* Is this sysctl writable? */
1444 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1447 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1449 #ifdef CAPABILITY_MODE
1451 * If the process is in capability mode, then don't permit reading or
1452 * writing unless specifically granted for the node.
1454 if (IN_CAPABILITY_MODE(req->td)) {
1455 if (req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD))
1457 if (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))
1462 /* Is this sysctl sensitive to securelevels? */
1463 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1464 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1465 error = securelevel_gt(req->td->td_ucred, lvl);
1470 /* Is this sysctl writable by only privileged users? */
1471 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1474 if (oid->oid_kind & CTLFLAG_PRISON)
1475 priv = PRIV_SYSCTL_WRITEJAIL;
1477 else if ((oid->oid_kind & CTLFLAG_VNET) &&
1478 prison_owns_vnet(req->td->td_ucred))
1479 priv = PRIV_SYSCTL_WRITEJAIL;
1482 priv = PRIV_SYSCTL_WRITE;
1483 error = priv_check(req->td, priv);
1488 if (!oid->oid_handler)
1491 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1492 arg1 = (int *)arg1 + indx;
1495 arg1 = oid->oid_arg1;
1496 arg2 = oid->oid_arg2;
1499 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1507 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1509 error = oid->oid_handler(oid, arg1, arg2, req);
1510 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1513 KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
1517 if (oid->oid_running == 0 && (oid->oid_kind & CTLFLAG_DYING) != 0)
1518 wakeup(&oid->oid_running);
1522 #ifndef _SYS_SYSPROTO_H_
1523 struct sysctl_args {
1533 sys___sysctl(struct thread *td, struct sysctl_args *uap)
1535 int error, i, name[CTL_MAXNAME];
1538 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1541 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1545 error = userland_sysctl(td, name, uap->namelen,
1546 uap->old, uap->oldlenp, 0,
1547 uap->new, uap->newlen, &j, 0);
1548 if (error && error != ENOMEM)
1551 i = copyout(&j, uap->oldlenp, sizeof(j));
1559 * This is used from various compatibility syscalls too. That's why name
1560 * must be in kernel space.
1563 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1564 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1567 int error = 0, memlocked;
1568 struct sysctl_req req;
1570 bzero(&req, sizeof req);
1577 req.oldlen = *oldlenp;
1579 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1584 req.validlen = req.oldlen;
1587 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1593 if (!useracc(new, newlen, VM_PROT_READ))
1595 req.newlen = newlen;
1599 req.oldfunc = sysctl_old_user;
1600 req.newfunc = sysctl_new_user;
1601 req.lock = REQ_UNWIRED;
1604 if (KTRPOINT(curthread, KTR_SYSCTL))
1605 ktrsysctl(name, namelen);
1608 if (req.oldlen > PAGE_SIZE) {
1610 sx_xlock(&sysctlmemlock);
1613 CURVNET_SET(TD_TO_VNET(td));
1619 error = sysctl_root(0, name, namelen, &req);
1621 if (error != EAGAIN)
1623 kern_yield(PRI_USER);
1628 if (req.lock == REQ_WIRED && req.validlen > 0)
1629 vsunlock(req.oldptr, req.validlen);
1631 sx_xunlock(&sysctlmemlock);
1633 if (error && error != ENOMEM)
1637 if (req.oldptr && req.oldidx > req.validlen)
1638 *retval = req.validlen;
1640 *retval = req.oldidx;
1646 * Drain into a sysctl struct. The user buffer should be wired if a page
1647 * fault would cause issue.
1650 sbuf_sysctl_drain(void *arg, const char *data, int len)
1652 struct sysctl_req *req = arg;
1655 error = SYSCTL_OUT(req, data, len);
1656 KASSERT(error >= 0, ("Got unexpected negative value %d", error));
1657 return (error == 0 ? len : -error);
1661 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
1662 struct sysctl_req *req)
1665 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN);
1666 sbuf_set_drain(s, sbuf_sysctl_drain, req);