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 * 3. 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
36 * $FreeBSD: src/sys/kern/kern_sysctl.c,v 1.92.2.9 2003/05/01 22:48:09 trhodes Exp $
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
43 #include <sys/sysctl.h>
44 #include <sys/malloc.h>
47 #include <sys/sysproto.h>
52 #include <vm/vm_extern.h>
54 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
55 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
58 * The sysctllock protects the MIB tree. It also protects sysctl
59 * contexts used with dynamic sysctls. The sysctl_register_oid() and
60 * sysctl_unregister_oid() routines require the sysctllock to already
61 * be held, so the sysctl_lock() and sysctl_unlock() routines are
62 * provided for the few places in the kernel which need to use that
63 * API rather than using the dynamic API. Use of the dynamic API is
64 * strongly encouraged for most code.
67 static int sysctl_root(SYSCTL_HANDLER_ARGS);
68 static void sysctl_register_oid_int(struct sysctl_oid *oipd);
69 static void sysctl_unregister_oid_int(struct sysctl_oid *oipd);
71 struct sysctl_oid_list sysctl__children; /* root list */
73 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
76 static struct sysctl_oid *
77 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list, int lock)
79 struct sysctl_oid *oidp;
81 SLIST_FOREACH(oidp, list, oid_link) {
82 if (strcmp(oidp->oid_name, name) == 0) {
90 * Initialization of the MIB tree.
92 * Order by number in each list.
96 sysctl_register_oid(struct sysctl_oid *oidp)
99 sysctl_register_oid_int(oidp);
104 sysctl_register_oid_int(struct sysctl_oid *oidp)
106 struct sysctl_oid_list *parent = oidp->oid_parent;
107 struct sysctl_oid *p;
108 struct sysctl_oid *q;
111 * Finish initialization from sysctl_set or add.
113 lockinit(&oidp->oid_lock, "oidlk", 0, LK_CANRECURSE);
116 * First check if another oid with the same name already
117 * exists in the parent's list.
119 p = sysctl_find_oidname(oidp->oid_name, parent, 0);
121 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE)
124 kprintf("can't re-use a leaf (%s)!\n", p->oid_name);
129 * If this oid has a number OID_AUTO, give it a number which
130 * is greater than any current oid. Make sure it is at least
131 * 256 to leave space for pre-assigned oid numbers.
133 if (oidp->oid_number == OID_AUTO) {
134 int newoid = 0x100; /* minimum AUTO oid */
137 * Adjust based on highest oid in parent list
139 SLIST_FOREACH(p, parent, oid_link) {
140 if (newoid <= p->oid_number)
141 newoid = p->oid_number + 1;
143 oidp->oid_number = newoid;
147 * Insert the oid into the parent's list in order.
150 SLIST_FOREACH(p, parent, oid_link) {
151 if (oidp->oid_number < p->oid_number)
156 SLIST_INSERT_AFTER(q, oidp, oid_link);
158 SLIST_INSERT_HEAD(parent, oidp, oid_link);
162 sysctl_unregister_oid(struct sysctl_oid *oidp)
165 sysctl_unregister_oid_int(oidp);
170 sysctl_unregister_oid_int(struct sysctl_oid *oidp)
172 struct sysctl_oid *p;
174 if (oidp->oid_number == OID_AUTO)
175 panic("Trying to unregister OID_AUTO entry: %p", oidp);
177 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
180 SLIST_REMOVE(oidp->oid_parent, oidp, sysctl_oid, oid_link);
185 * This can happen when a module fails to register and is
186 * being unloaded afterwards. It should not be a panic()
189 kprintf("%s: failed to unregister sysctl\n", __func__);
192 /* Initialize a new context to keep track of dynamically added sysctls. */
194 sysctl_ctx_init(struct sysctl_ctx_list *c)
202 /* Free the context, and destroy all dynamic oids registered in this context */
204 sysctl_ctx_free(struct sysctl_ctx_list *clist)
206 struct sysctl_ctx_entry *e, *e1;
211 * First perform a "dry run" to check if it's ok to remove oids.
213 * XXX This algorithm is a hack. But I don't know any
214 * XXX better solution for now...
217 TAILQ_FOREACH(e, clist, link) {
218 error = sysctl_remove_oid_locked(e->entry, 0, 0);
223 * Restore deregistered entries, either from the end,
224 * or from the place where error occured.
225 * e contains the entry that was not unregistered
228 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
230 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
232 sysctl_register_oid(e1->entry);
233 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
239 /* Now really delete the entries */
240 e = TAILQ_FIRST(clist);
242 e1 = TAILQ_NEXT(e, link);
243 error = sysctl_remove_oid_locked(e->entry, 1, 0);
245 panic("sysctl_remove_oid: corrupt tree, entry: %s",
247 kfree(e, M_SYSCTLOID);
254 /* Add an entry to the context */
255 struct sysctl_ctx_entry *
256 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
258 struct sysctl_ctx_entry *e;
260 SYSCTL_ASSERT_LOCKED();
261 if (clist == NULL || oidp == NULL)
263 e = kmalloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
265 TAILQ_INSERT_HEAD(clist, e, link);
269 /* Find an entry in the context */
270 struct sysctl_ctx_entry *
271 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
273 struct sysctl_ctx_entry *e;
275 SYSCTL_ASSERT_LOCKED();
276 if (clist == NULL || oidp == NULL)
278 TAILQ_FOREACH(e, clist, link) {
286 * Delete an entry from the context.
287 * NOTE: this function doesn't free oidp! You have to remove it
288 * with sysctl_remove_oid().
291 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
293 struct sysctl_ctx_entry *e;
295 if (clist == NULL || oidp == NULL)
298 e = sysctl_ctx_entry_find(clist, oidp);
300 TAILQ_REMOVE(clist, e, link);
302 kfree(e, M_SYSCTLOID);
311 * Remove dynamically created sysctl trees.
312 * oidp - top of the tree to be removed
313 * del - if 0 - just deregister, otherwise free up entries as well
314 * recurse - if != 0 traverse the subtree to be deleted
317 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
322 error = sysctl_remove_oid_locked(oidp, del, recurse);
328 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
330 struct sysctl_oid *p, *tmp;
333 SYSCTL_ASSERT_LOCKED();
336 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
337 kprintf("can't remove non-dynamic nodes!\n");
341 * WARNING: normal method to do this should be through
342 * sysctl_ctx_free(). Use recursing as the last resort
343 * method to purge your sysctl tree of leftovers...
344 * However, if some other code still references these nodes,
347 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
348 if (oidp->oid_refcnt == 1) {
349 SLIST_FOREACH_MUTABLE(p,
350 SYSCTL_CHILDREN(oidp), oid_link, tmp) {
352 kprintf("Warning: failed attempt to "
353 "remove oid %s with child %s\n",
354 oidp->oid_name, p->oid_name);
357 error = sysctl_remove_oid_locked(p, del,
363 kfree(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
366 if (oidp->oid_refcnt > 1 ) {
369 if (oidp->oid_refcnt == 0) {
370 kprintf("Warning: bad oid_refcnt=%u (%s)!\n",
371 oidp->oid_refcnt, oidp->oid_name);
374 sysctl_unregister_oid(oidp);
377 * Wait for all threads running the handler to drain.
378 * This preserves the previous behavior when the
379 * sysctl lock was held across a handler invocation,
380 * and is necessary for module unload correctness.
382 while (oidp->oid_running > 0) {
383 oidp->oid_kind |= CTLFLAG_DYING;
384 tsleep_interlock(&oidp->oid_running, 0);
386 tsleep(&oidp->oid_running, PINTERLOCKED,
391 kfree(__DECONST(char *, oidp->oid_descr),
393 kfree(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
394 lockuninit(&oidp->oid_lock);
395 kfree(oidp, M_SYSCTLOID);
402 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
403 int del, int recurse)
405 struct sysctl_oid *p, *tmp;
410 SLIST_FOREACH_MUTABLE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
411 if (strcmp(p->oid_name, name) == 0) {
412 error = sysctl_remove_oid_locked(p, del, recurse);
422 * Create new sysctls at run time.
423 * clist may point to a valid context initialized with sysctl_ctx_init().
426 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
427 int number, const char *name, int kind, void *arg1, int arg2,
428 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
430 struct sysctl_oid *oidp;
434 /* You have to hook up somewhere.. */
438 /* Check if the node already exists, otherwise create it */
439 oidp = sysctl_find_oidname(name, parent, 0);
441 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
443 /* Update the context */
445 sysctl_ctx_entry_add(clist, oidp);
449 kprintf("can't re-use a leaf (%s)!\n", name);
454 oidp = kmalloc(sizeof(struct sysctl_oid), M_SYSCTLOID,
456 oidp->oid_parent = parent;
457 SLIST_NEXT(oidp, oid_link) = NULL;
458 oidp->oid_number = number;
459 oidp->oid_refcnt = 1;
461 newname = kmalloc(len + 1, M_SYSCTLOID, M_WAITOK);
462 bcopy(name, newname, len + 1);
464 oidp->oid_name = newname;
465 oidp->oid_handler = handler;
466 oidp->oid_kind = CTLFLAG_DYN | kind;
467 if ((kind & CTLTYPE) == CTLTYPE_NODE) {
468 struct sysctl_oid_list *children;
470 /* Allocate space for children */
471 children = kmalloc(sizeof(*children), M_SYSCTLOID, M_WAITOK);
472 SYSCTL_SET_CHILDREN(oidp, children);
473 SLIST_INIT(children);
475 oidp->oid_arg1 = arg1;
476 oidp->oid_arg2 = arg2;
480 int len = strlen(descr) + 1;
481 oidp->oid_descr = kmalloc(len, M_SYSCTLOID, M_WAITOK);
482 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
484 /* Update the context, if used */
486 sysctl_ctx_entry_add(clist, oidp);
487 /* Register this oid */
488 sysctl_register_oid_int(oidp);
494 * Rename an existing oid.
497 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
502 newname = kstrdup(name, M_SYSCTLOID);
504 oldname = __DECONST(char *, oidp->oid_name);
505 oidp->oid_name = newname;
507 kfree(oldname, M_SYSCTLOID);
511 * Register the kernel's oids on startup.
513 SET_DECLARE(sysctl_set, struct sysctl_oid);
516 sysctl_register_all(void *arg)
518 struct sysctl_oid **oidp;
521 SET_FOREACH(oidp, sysctl_set)
522 sysctl_register_oid(*oidp);
525 SYSINIT(sysctl, SI_BOOT1_POST, SI_ORDER_ANY, sysctl_register_all, 0);
530 * These functions implement a presently undocumented interface
531 * used by the sysctl program to walk the tree, and get the type
532 * so it can print the value.
533 * This interface is under work and consideration, and should probably
534 * be killed with a big axe by the first person who can find the time.
535 * (be aware though, that the proper interface isn't as obvious as it
536 * may seem, there are various conflicting requirements.
538 * {0,0} kprintf the entire MIB-tree.
539 * {0,1,...} return the name of the "..." OID.
540 * {0,2,...} return the next OID.
541 * {0,3} return the OID of the name in "new"
542 * {0,4,...} return the kind & format info for the "..." OID.
546 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
549 struct sysctl_oid *oidp;
551 SLIST_FOREACH(oidp, l, oid_link) {
556 kprintf("%d %s ", oidp->oid_number, oidp->oid_name);
559 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
560 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
562 if (oidp->oid_handler)
563 kprintf(" *Handler");
565 switch (oidp->oid_kind & CTLTYPE) {
568 if (!oidp->oid_handler) {
569 sysctl_sysctl_debug_dump_node(
570 oidp->oid_arg1, i+2);
573 case CTLTYPE_INT: kprintf(" Int\n"); break;
574 case CTLTYPE_STRING: kprintf(" String\n"); break;
575 case CTLTYPE_QUAD: kprintf(" Quad\n"); break;
576 case CTLTYPE_OPAQUE: kprintf(" Opaque/struct\n"); break;
577 default: kprintf("\n");
584 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
588 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
591 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
596 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
597 0, 0, sysctl_sysctl_debug, "-", "");
600 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
602 int *name = (int *) arg1;
603 u_int namelen = arg2;
605 struct sysctl_oid *oid;
606 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
611 ksnprintf(buf, sizeof(buf), "%d", *name);
613 error = SYSCTL_OUT(req, ".", 1);
615 error = SYSCTL_OUT(req, buf, strlen(buf));
623 SLIST_FOREACH(oid, lsp, oid_link) {
624 if (oid->oid_number != *name)
628 error = SYSCTL_OUT(req, ".", 1);
630 error = SYSCTL_OUT(req, oid->oid_name,
631 strlen(oid->oid_name));
638 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
641 if (oid->oid_handler)
644 lsp2 = SYSCTL_CHILDREN(oid);
649 error = SYSCTL_OUT(req, "", 1);
654 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
657 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
658 int *next, int *len, int level, struct sysctl_oid **oidpp)
660 struct sysctl_oid *oidp;
663 SLIST_FOREACH(oidp, lsp, oid_link) {
664 *next = oidp->oid_number;
667 if (oidp->oid_kind & CTLFLAG_SKIP)
671 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
673 if (oidp->oid_handler)
674 /* We really should call the handler here...*/
676 lsp = SYSCTL_CHILDREN(oidp);
677 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
678 len, level+1, oidpp))
683 if (oidp->oid_number < *name)
686 if (oidp->oid_number > *name) {
687 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
689 if (oidp->oid_handler)
691 lsp = SYSCTL_CHILDREN(oidp);
692 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
693 next+1, len, level+1, oidpp))
697 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
700 if (oidp->oid_handler)
703 lsp = SYSCTL_CHILDREN(oidp);
704 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
705 len, level+1, oidpp))
716 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
718 int *name = (int *) arg1;
719 u_int namelen = arg2;
721 struct sysctl_oid *oid;
722 struct sysctl_oid_list *lsp = &sysctl__children;
723 int newoid[CTL_MAXNAME];
725 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
728 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
733 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
736 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
738 struct sysctl_oid *oidp;
739 struct sysctl_oid_list *lsp = &sysctl__children;
742 SYSCTL_ASSERT_LOCKED();
744 for (*len = 0; *len < CTL_MAXNAME;) {
745 p = strsep(&name, ".");
747 oidp = SLIST_FIRST(lsp);
748 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
751 if (strcmp(p, oidp->oid_name) == 0)
754 *oid++ = oidp->oid_number;
757 if (name == NULL || *name == '\0') {
763 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
766 if (oidp->oid_handler)
769 lsp = SYSCTL_CHILDREN(oidp);
775 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
778 int error, oid[CTL_MAXNAME], len;
779 struct sysctl_oid *op = NULL;
783 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
784 return (ENAMETOOLONG);
786 p = kmalloc(req->newlen+1, M_SYSCTL, M_WAITOK);
788 error = SYSCTL_IN(req, p, req->newlen);
794 p [req->newlen] = '\0';
796 error = name2oid(p, oid, &len, &op);
803 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
807 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_SHLOCK,
808 0, 0, sysctl_sysctl_name2oid, "I", "");
811 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
813 struct sysctl_oid *oid;
816 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
822 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
825 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
830 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
833 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
835 struct sysctl_oid *oid;
838 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
844 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
848 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
851 * Default "handler" functions.
855 * Handle an int, signed or unsigned.
857 * a variable: point arg1 at it.
858 * a constant: pass it in arg2.
862 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
867 error = SYSCTL_OUT(req, arg1, sizeof(int));
869 error = SYSCTL_OUT(req, &arg2, sizeof(int));
871 if (error || !req->newptr)
877 error = SYSCTL_IN(req, arg1, sizeof(int));
882 * Handle a long, signed or unsigned. arg1 points to it.
886 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
892 error = SYSCTL_OUT(req, arg1, sizeof(long));
894 if (error || !req->newptr)
897 error = SYSCTL_IN(req, arg1, sizeof(long));
902 * Handle a quad, signed or unsigned. arg1 points to it.
906 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
912 error = SYSCTL_OUT(req, arg1, sizeof(quad_t));
914 if (error || !req->newptr)
917 error = SYSCTL_IN(req, arg1, sizeof(quad_t));
922 * Handle our generic '\0' terminated 'C' string.
924 * a variable string: point arg1 at it, arg2 is max length.
925 * a constant string: point arg1 at it, arg2 is zero.
929 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
933 error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
935 if (error || !req->newptr)
938 if ((req->newlen - req->newidx) >= arg2) {
941 arg2 = (req->newlen - req->newidx);
942 error = SYSCTL_IN(req, arg1, arg2);
943 ((char *)arg1)[arg2] = '\0';
950 * Handle any kind of opaque data.
951 * arg1 points to it, arg2 is the size.
955 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
959 error = SYSCTL_OUT(req, arg1, arg2);
961 if (error || !req->newptr)
964 error = SYSCTL_IN(req, arg1, arg2);
970 * Transfer functions to/from kernel space.
971 * XXX: rather untested at this point
974 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
980 if (i > req->oldlen - req->oldidx)
981 i = req->oldlen - req->oldidx;
983 bcopy(p, (char *)req->oldptr + req->oldidx, i);
986 if (req->oldptr && i != l)
992 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
997 if (req->newlen - req->newidx < l)
999 bcopy((char *)req->newptr + req->newidx, p, l);
1005 kernel_sysctl(int *name, u_int namelen,
1006 void *old, size_t *oldlenp,
1007 void *new, size_t newlen, size_t *retval)
1010 struct sysctl_req req;
1012 bzero(&req, sizeof req);
1017 req.oldlen = *oldlenp;
1019 req.validlen = req.oldlen;
1026 req.newlen = newlen;
1030 req.oldfunc = sysctl_old_kernel;
1031 req.newfunc = sysctl_new_kernel;
1033 req.lock = REQ_UNWIRED;
1037 error = sysctl_root(0, name, namelen, &req);
1041 if (req.lock == REQ_WIRED && req.validlen > 0)
1042 vsunlock(req.oldptr, req.validlen);
1045 if (error && error != ENOMEM)
1049 if (req.oldptr && req.oldidx > req.validlen)
1050 *retval = req.validlen;
1052 *retval = req.oldidx;
1058 kernel_sysctlbyname(char *name,
1059 void *old, size_t *oldlenp,
1060 void *new, size_t newlen, size_t *retval)
1062 int oid[CTL_MAXNAME];
1063 size_t oidlen, plen;
1066 oid[0] = 0; /* sysctl internal magic */
1067 oid[1] = 3; /* name2oid */
1068 oidlen = sizeof(oid);
1070 error = kernel_sysctl(oid, 2, oid, &oidlen, name, strlen(name), &plen);
1074 error = kernel_sysctl(oid, plen / sizeof(int), old, oldlenp,
1075 new, newlen, retval);
1080 * Transfer function to/from user space.
1083 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1089 if (req->lock == 1 && req->oldptr) {
1090 vslock(req->oldptr, req->oldlen);
1096 if (i > req->oldlen - req->oldidx)
1097 i = req->oldlen - req->oldidx;
1099 error = copyout(p, (char *)req->oldptr + req->oldidx,
1105 if (req->oldptr && i < l)
1111 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1117 if (req->newlen - req->newidx < l)
1119 error = copyin((char *)req->newptr + req->newidx, p, l);
1125 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1126 int *nindx, struct sysctl_req *req)
1128 struct sysctl_oid_list *lsp;
1129 struct sysctl_oid *oid;
1132 lsp = &sysctl__children;
1134 while (indx < CTL_MAXNAME) {
1135 SLIST_FOREACH(oid, lsp, oid_link) {
1136 if (oid->oid_number == name[indx])
1143 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1144 if (oid->oid_handler != NULL || indx == namelen) {
1148 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1149 ("%s found DYING node %p", __func__, oid));
1152 lsp = SYSCTL_CHILDREN(oid);
1153 } else if (indx == namelen) {
1157 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1158 ("%s found DYING node %p", __func__, oid));
1168 * Traverse our tree, and find the right node, execute whatever it points
1169 * to, and return the resulting error code.
1172 sysctl_root(SYSCTL_HANDLER_ARGS)
1174 struct thread *td = req->td;
1175 struct proc *p = td ? td->td_proc : NULL;
1176 struct sysctl_oid *oid;
1180 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1184 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1186 * You can't call a sysctl when it's a node, but has
1187 * no handler. Inform the user that it's a node.
1188 * The indx may or may not be the same as namelen.
1190 if (oid->oid_handler == NULL)
1194 /* If writing isn't allowed */
1195 if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) ||
1196 ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0)))
1199 /* Most likely only root can write */
1200 if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p &&
1201 (error = priv_check_cred(td->td_ucred,
1202 (oid->oid_kind & CTLFLAG_PRISON) ? PRIV_SYSCTL_WRITEJAIL :
1203 PRIV_SYSCTL_WRITE, 0)))
1206 if (oid->oid_handler == NULL)
1210 * Default oid locking is exclusive when modifying (newptr),
1211 * shared otherwise, unless overridden with a control flag.
1213 if ((oid->oid_kind & CTLFLAG_NOLOCK) == 0) {
1214 lktype = (req->newptr != NULL) ? LK_EXCLUSIVE : LK_SHARED;
1215 if (oid->oid_kind & CTLFLAG_SHLOCK)
1217 if (oid->oid_kind & CTLFLAG_EXLOCK)
1218 lktype = LK_EXCLUSIVE;
1220 lockmgr(&oid->oid_lock, lktype);
1223 if (lockmgr(&oid->oid_lock, lktype | LK_SLEEPFAIL)) {
1224 kprintf("%s\n", oid->oid_name);
1225 lockmgr(&oid->oid_lock, lktype);
1230 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE)
1231 error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx,
1234 error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2,
1237 if ((oid->oid_kind & CTLFLAG_NOLOCK) == 0)
1238 lockmgr(&oid->oid_lock, LK_RELEASE);
1243 sys___sysctl(struct sysctl_args *uap)
1245 int error, i, name[CTL_MAXNAME];
1248 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1251 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1255 error = userland_sysctl(name, uap->namelen,
1256 uap->old, uap->oldlenp, 0,
1257 uap->new, uap->newlen, &j);
1258 if (error && error != ENOMEM)
1261 i = copyout(&j, uap->oldlenp, sizeof(j));
1269 * This is used from various compatibility syscalls too. That's why name
1270 * must be in kernel space.
1273 userland_sysctl(int *name, u_int namelen,
1274 void *old, size_t *oldlenp, int inkernel,
1275 void *new, size_t newlen, size_t *retval)
1278 struct sysctl_req req;
1280 bzero(&req, sizeof req);
1287 req.oldlen = *oldlenp;
1289 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1294 req.validlen = req.oldlen;
1297 * NOTE: User supplied buffers are not guaranteed to be good,
1298 * the sysctl copyins and copyouts can fail.
1304 req.newlen = newlen;
1308 req.oldfunc = sysctl_old_user;
1309 req.newfunc = sysctl_new_user;
1311 req.lock = REQ_UNWIRED;
1315 if (KTRPOINT(curthread, KTR_SYSCTL))
1316 ktrsysctl(name, namelen);
1323 error = sysctl_root(0, name, namelen, &req);
1325 if (error != EAGAIN)
1331 if (req.lock == REQ_WIRED && req.validlen > 0)
1332 vsunlock(req.oldptr, req.validlen);
1334 if (error && error != ENOMEM)
1338 if (req.oldptr && req.oldidx > req.validlen)
1339 *retval = req.validlen;
1341 *retval = req.oldidx;
1347 sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high)
1351 value = *(int *)arg1;
1352 error = sysctl_handle_int(oidp, &value, 0, req);
1353 if (error || !req->newptr)
1355 if (value < low || value > high)
1357 *(int *)arg1 = value;
1362 * Drain into a sysctl struct. The user buffer should be wired if a page
1363 * fault would cause issue.
1366 sbuf_sysctl_drain(void *arg, const char *data, int len)
1368 struct sysctl_req *req = arg;
1371 error = SYSCTL_OUT(req, data, len);
1372 KASSERT(error >= 0, ("Got unexpected negative value %d", error));
1373 return (error == 0 ? len : -error);
1377 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
1378 struct sysctl_req *req)
1381 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN);
1382 sbuf_set_drain(s, sbuf_sysctl_drain, req);
1387 * The exclusive sysctl lock only protects its topology, and is
1388 * very expensive, but allows us to use a pcpu shared lock for
1389 * critical path accesses.
1397 for (i = 0; i < ncpus; ++i) {
1398 gd = globaldata_find(i);
1399 lockmgr(&gd->gd_sysctllock, LK_EXCLUSIVE);
1404 _sysctl_xunlock(void)
1409 for (i = 0; i < ncpus; ++i) {
1410 gd = globaldata_find(i);
1411 lockmgr(&gd->gd_sysctllock, LK_RELEASE);