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_compat.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/sysctl.h>
48 #include <sys/malloc.h>
52 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
56 #include <security/mac/mac_framework.h>
59 #include <vm/vm_extern.h>
61 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
62 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
63 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
66 * Locking - this locks the sysctl tree in memory.
68 static struct sx sysctllock;
70 #define SYSCTL_LOCK() sx_xlock(&sysctllock)
71 #define SYSCTL_UNLOCK() sx_xunlock(&sysctllock)
72 #define SYSCTL_INIT() sx_init(&sysctllock, "sysctl lock")
74 static int sysctl_root(SYSCTL_HANDLER_ARGS);
76 struct sysctl_oid_list sysctl__children; /* root list */
78 static struct sysctl_oid *
79 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
81 struct sysctl_oid *oidp;
83 SLIST_FOREACH(oidp, list, oid_link) {
84 if (strcmp(oidp->oid_name, name) == 0) {
92 * Initialization of the MIB tree.
94 * Order by number in each list.
98 sysctl_register_oid(struct sysctl_oid *oidp)
100 struct sysctl_oid_list *parent = oidp->oid_parent;
101 struct sysctl_oid *p;
102 struct sysctl_oid *q;
105 * First check if another oid with the same name already
106 * exists in the parent's list.
108 p = sysctl_find_oidname(oidp->oid_name, parent);
110 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
114 printf("can't re-use a leaf (%s)!\n", p->oid_name);
119 * If this oid has a number OID_AUTO, give it a number which
120 * is greater than any current oid.
121 * NOTE: DO NOT change the starting value here, change it in
122 * <sys/sysctl.h>, and make sure it is at least 256 to
123 * accomodate e.g. net.inet.raw as a static sysctl node.
125 if (oidp->oid_number == OID_AUTO) {
126 static int newoid = CTL_AUTO_START;
128 oidp->oid_number = newoid++;
129 if (newoid == 0x7fffffff)
130 panic("out of oids");
133 else if (oidp->oid_number >= CTL_AUTO_START) {
134 /* do not panic; this happens when unregistering sysctl sets */
135 printf("static sysctl oid too high: %d", oidp->oid_number);
140 * Insert the oid into the parent's list in order.
143 SLIST_FOREACH(p, parent, oid_link) {
144 if (oidp->oid_number < p->oid_number)
149 SLIST_INSERT_AFTER(q, oidp, oid_link);
151 SLIST_INSERT_HEAD(parent, oidp, oid_link);
155 sysctl_unregister_oid(struct sysctl_oid *oidp)
157 struct sysctl_oid *p;
161 if (oidp->oid_number == OID_AUTO) {
164 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
166 SLIST_REMOVE(oidp->oid_parent, oidp,
167 sysctl_oid, oid_link);
175 * This can happen when a module fails to register and is
176 * being unloaded afterwards. It should not be a panic()
180 printf("%s: failed to unregister sysctl\n", __func__);
183 /* Initialize a new context to keep track of dynamically added sysctls. */
185 sysctl_ctx_init(struct sysctl_ctx_list *c)
195 /* Free the context, and destroy all dynamic oids registered in this context */
197 sysctl_ctx_free(struct sysctl_ctx_list *clist)
199 struct sysctl_ctx_entry *e, *e1;
204 * First perform a "dry run" to check if it's ok to remove oids.
206 * XXX This algorithm is a hack. But I don't know any
207 * XXX better solution for now...
209 TAILQ_FOREACH(e, clist, link) {
210 error = sysctl_remove_oid(e->entry, 0, 0);
215 * Restore deregistered entries, either from the end,
216 * or from the place where error occured.
217 * e contains the entry that was not unregistered
220 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
222 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
224 sysctl_register_oid(e1->entry);
225 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
229 /* Now really delete the entries */
230 e = TAILQ_FIRST(clist);
232 e1 = TAILQ_NEXT(e, link);
233 error = sysctl_remove_oid(e->entry, 1, 0);
235 panic("sysctl_remove_oid: corrupt tree, entry: %s",
237 free(e, M_SYSCTLOID);
243 /* Add an entry to the context */
244 struct sysctl_ctx_entry *
245 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
247 struct sysctl_ctx_entry *e;
249 if (clist == NULL || oidp == NULL)
251 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
253 TAILQ_INSERT_HEAD(clist, e, link);
257 /* Find an entry in the context */
258 struct sysctl_ctx_entry *
259 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
261 struct sysctl_ctx_entry *e;
263 if (clist == NULL || oidp == NULL)
265 TAILQ_FOREACH(e, clist, link) {
273 * Delete an entry from the context.
274 * NOTE: this function doesn't free oidp! You have to remove it
275 * with sysctl_remove_oid().
278 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
280 struct sysctl_ctx_entry *e;
282 if (clist == NULL || oidp == NULL)
284 e = sysctl_ctx_entry_find(clist, oidp);
286 TAILQ_REMOVE(clist, e, link);
287 free(e, M_SYSCTLOID);
294 * Remove dynamically created sysctl trees.
295 * oidp - top of the tree to be removed
296 * del - if 0 - just deregister, otherwise free up entries as well
297 * recurse - if != 0 traverse the subtree to be deleted
300 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
302 struct sysctl_oid *p;
307 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
308 printf("can't remove non-dynamic nodes!\n");
312 * WARNING: normal method to do this should be through
313 * sysctl_ctx_free(). Use recursing as the last resort
314 * method to purge your sysctl tree of leftovers...
315 * However, if some other code still references these nodes,
318 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
319 if (oidp->oid_refcnt == 1) {
320 SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
323 error = sysctl_remove_oid(p, del, recurse);
328 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
331 if (oidp->oid_refcnt > 1 ) {
334 if (oidp->oid_refcnt == 0) {
335 printf("Warning: bad oid_refcnt=%u (%s)!\n",
336 oidp->oid_refcnt, oidp->oid_name);
339 sysctl_unregister_oid(oidp);
342 free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
343 free((void *)(uintptr_t)(const void *)oidp->oid_name,
345 free(oidp, M_SYSCTLOID);
352 * Create new sysctls at run time.
353 * clist may point to a valid context initialized with sysctl_ctx_init().
356 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
357 int number, const char *name, int kind, void *arg1, int arg2,
358 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
360 struct sysctl_oid *oidp;
364 /* You have to hook up somewhere.. */
367 /* Check if the node already exists, otherwise create it */
368 oidp = sysctl_find_oidname(name, parent);
370 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
372 /* Update the context */
374 sysctl_ctx_entry_add(clist, oidp);
377 printf("can't re-use a leaf (%s)!\n", name);
381 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
382 oidp->oid_parent = parent;
383 SLIST_NEXT(oidp, oid_link) = NULL;
384 oidp->oid_number = number;
385 oidp->oid_refcnt = 1;
387 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
388 bcopy(name, newname, len + 1);
390 oidp->oid_name = newname;
391 oidp->oid_handler = handler;
392 oidp->oid_kind = CTLFLAG_DYN | kind;
393 if ((kind & CTLTYPE) == CTLTYPE_NODE) {
394 /* Allocate space for children */
395 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
396 M_SYSCTLOID, M_WAITOK));
397 SLIST_INIT(SYSCTL_CHILDREN(oidp));
399 oidp->oid_arg1 = arg1;
400 oidp->oid_arg2 = arg2;
404 int len = strlen(descr) + 1;
405 oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
407 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
409 /* Update the context, if used */
411 sysctl_ctx_entry_add(clist, oidp);
412 /* Register this oid */
413 sysctl_register_oid(oidp);
418 * Rename an existing oid.
421 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
427 oldname = (void *)(uintptr_t)(const void *)oidp->oid_name;
429 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
430 bcopy(name, newname, len + 1);
432 oidp->oid_name = newname;
433 free(oldname, M_SYSCTLOID);
437 * Reparent an existing oid.
440 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
442 struct sysctl_oid *oidp;
444 if (oid->oid_parent == parent)
446 oidp = sysctl_find_oidname(oid->oid_name, parent);
449 sysctl_unregister_oid(oid);
450 oid->oid_parent = parent;
451 oid->oid_number = OID_AUTO;
452 sysctl_register_oid(oid);
457 * Register the kernel's oids on startup.
459 SET_DECLARE(sysctl_set, struct sysctl_oid);
462 sysctl_register_all(void *arg)
464 struct sysctl_oid **oidp;
467 SET_FOREACH(oidp, sysctl_set)
468 sysctl_register_oid(*oidp);
470 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
475 * These functions implement a presently undocumented interface
476 * used by the sysctl program to walk the tree, and get the type
477 * so it can print the value.
478 * This interface is under work and consideration, and should probably
479 * be killed with a big axe by the first person who can find the time.
480 * (be aware though, that the proper interface isn't as obvious as it
481 * may seem, there are various conflicting requirements.
483 * {0,0} printf the entire MIB-tree.
484 * {0,1,...} return the name of the "..." OID.
485 * {0,2,...} return the next OID.
486 * {0,3} return the OID of the name in "new"
487 * {0,4,...} return the kind & format info for the "..." OID.
488 * {0,5,...} return the description the "..." OID.
493 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
496 struct sysctl_oid *oidp;
498 SLIST_FOREACH(oidp, l, oid_link) {
503 printf("%d %s ", oidp->oid_number, oidp->oid_name);
506 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
507 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
509 if (oidp->oid_handler)
512 switch (oidp->oid_kind & CTLTYPE) {
515 if (!oidp->oid_handler) {
516 sysctl_sysctl_debug_dump_node(
517 oidp->oid_arg1, i+2);
520 case CTLTYPE_INT: printf(" Int\n"); break;
521 case CTLTYPE_STRING: printf(" String\n"); break;
522 case CTLTYPE_QUAD: printf(" Quad\n"); break;
523 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
524 default: printf("\n");
531 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
535 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
538 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
542 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
543 0, 0, sysctl_sysctl_debug, "-", "");
547 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
549 int *name = (int *) arg1;
550 u_int namelen = arg2;
552 struct sysctl_oid *oid;
553 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
558 snprintf(buf,sizeof(buf),"%d",*name);
560 error = SYSCTL_OUT(req, ".", 1);
562 error = SYSCTL_OUT(req, buf, strlen(buf));
570 SLIST_FOREACH(oid, lsp, oid_link) {
571 if (oid->oid_number != *name)
575 error = SYSCTL_OUT(req, ".", 1);
577 error = SYSCTL_OUT(req, oid->oid_name,
578 strlen(oid->oid_name));
585 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
588 if (oid->oid_handler)
591 lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
596 return (SYSCTL_OUT(req, "", 1));
599 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
602 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
603 int *next, int *len, int level, struct sysctl_oid **oidpp)
605 struct sysctl_oid *oidp;
608 SLIST_FOREACH(oidp, lsp, oid_link) {
609 *next = oidp->oid_number;
612 if (oidp->oid_kind & CTLFLAG_SKIP)
616 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
618 if (oidp->oid_handler)
619 /* We really should call the handler here...*/
621 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
622 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
623 len, level+1, oidpp))
628 if (oidp->oid_number < *name)
631 if (oidp->oid_number > *name) {
632 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
634 if (oidp->oid_handler)
636 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
637 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
638 next+1, len, level+1, oidpp))
642 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
645 if (oidp->oid_handler)
648 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
649 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
650 len, level+1, oidpp))
661 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
663 int *name = (int *) arg1;
664 u_int namelen = arg2;
666 struct sysctl_oid *oid;
667 struct sysctl_oid_list *lsp = &sysctl__children;
668 int newoid[CTL_MAXNAME];
670 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
673 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
677 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
680 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
683 struct sysctl_oid *oidp;
684 struct sysctl_oid_list *lsp = &sysctl__children;
690 p = name + strlen(name) - 1 ;
696 for (p = name; *p && *p != '.'; p++)
702 oidp = SLIST_FIRST(lsp);
704 while (oidp && *len < CTL_MAXNAME) {
705 if (strcmp(name, oidp->oid_name)) {
706 oidp = SLIST_NEXT(oidp, oid_link);
709 *oid++ = oidp->oid_number;
718 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
721 if (oidp->oid_handler)
724 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
725 oidp = SLIST_FIRST(lsp);
727 for (p = name; *p && *p != '.'; p++)
737 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
740 int error, oid[CTL_MAXNAME], len;
741 struct sysctl_oid *op = 0;
745 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
746 return (ENAMETOOLONG);
748 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
750 error = SYSCTL_IN(req, p, req->newlen);
756 p [req->newlen] = '\0';
758 error = name2oid(p, oid, &len, &op);
765 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
769 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
770 sysctl_sysctl_name2oid, "I", "");
773 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
775 struct sysctl_oid *oid;
778 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
784 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
787 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
792 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
795 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
797 struct sysctl_oid *oid;
800 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
806 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
810 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
813 * Default "handler" functions.
817 * Handle an int, signed or unsigned.
819 * a variable: point arg1 at it.
820 * a constant: pass it in arg2.
824 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
826 int tmpout, error = 0;
829 * Attempt to get a coherent snapshot by making a copy of the data.
832 tmpout = *(int *)arg1;
835 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
837 if (error || !req->newptr)
843 error = SYSCTL_IN(req, arg1, sizeof(int));
849 * Based on on sysctl_handle_int() convert milliseconds into ticks.
853 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
857 tt = *(int *)oidp->oid_arg1;
858 s = (int)((int64_t)tt * 1000 / hz);
860 error = sysctl_handle_int(oidp, &s, 0, req);
861 if (error || !req->newptr)
864 tt = (int)((int64_t)s * hz / 1000);
868 *(int *)oidp->oid_arg1 = tt;
874 * Handle a long, signed or unsigned. arg1 points to it.
878 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
887 * Attempt to get a coherent snapshot by making a copy of the data.
891 tmplong = *(long *)arg1;
893 if (req->flags & SCTL_MASK32) {
895 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
898 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
900 if (error || !req->newptr)
904 if (req->flags & SCTL_MASK32) {
905 error = SYSCTL_IN(req, &tmpint, sizeof(int));
906 *(long *)arg1 = (long)tmpint;
909 error = SYSCTL_IN(req, arg1, sizeof(long));
914 * Handle a 64 bit int, signed or unsigned. arg1 points to it.
918 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
924 * Attempt to get a coherent snapshot by making a copy of the data.
928 tmpout = *(uint64_t *)arg1;
929 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
931 if (error || !req->newptr)
934 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
939 * Handle our generic '\0' terminated 'C' string.
941 * a variable string: point arg1 at it, arg2 is max length.
942 * a constant string: point arg1 at it, arg2 is zero.
946 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
953 * Attempt to get a coherent snapshot by copying to a
954 * temporary kernel buffer.
957 outlen = strlen((char *)arg1)+1;
958 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
960 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
961 free(tmparg, M_SYSCTLTMP);
965 error = SYSCTL_OUT(req, tmparg, outlen);
966 free(tmparg, M_SYSCTLTMP);
968 if (error || !req->newptr)
971 if ((req->newlen - req->newidx) >= arg2) {
974 arg2 = (req->newlen - req->newidx);
975 error = SYSCTL_IN(req, arg1, arg2);
976 ((char *)arg1)[arg2] = '\0';
983 * Handle any kind of opaque data.
984 * arg1 points to it, arg2 is the size.
988 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
992 struct sysctl_req req2;
995 * Attempt to get a coherent snapshot, by using the thread
996 * pre-emption counter updated from within mi_switch() to
997 * determine if we were pre-empted during a bcopy() or
998 * copyout(). Make 3 attempts at doing this before giving up.
999 * If we encounter an error, stop immediately.
1004 generation = curthread->td_generation;
1005 error = SYSCTL_OUT(req, arg1, arg2);
1009 if (generation != curthread->td_generation && tries < 3) {
1014 error = SYSCTL_IN(req, arg1, arg2);
1020 * Transfer functions to/from kernel space.
1021 * XXX: rather untested at this point
1024 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1030 if (req->oldlen <= req->oldidx)
1033 if (i > req->oldlen - req->oldidx)
1034 i = req->oldlen - req->oldidx;
1036 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1039 if (req->oldptr && i != l)
1045 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1049 if (req->newlen - req->newidx < l)
1051 bcopy((char *)req->newptr + req->newidx, p, l);
1057 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1058 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1061 struct sysctl_req req;
1063 bzero(&req, sizeof req);
1069 req.oldlen = *oldlenp;
1071 req.validlen = req.oldlen;
1078 req.newlen = newlen;
1082 req.oldfunc = sysctl_old_kernel;
1083 req.newfunc = sysctl_new_kernel;
1084 req.lock = REQ_LOCKED;
1088 error = sysctl_root(0, name, namelen, &req);
1090 if (req.lock == REQ_WIRED && req.validlen > 0)
1091 vsunlock(req.oldptr, req.validlen);
1095 if (error && error != ENOMEM)
1099 if (req.oldptr && req.oldidx > req.validlen)
1100 *retval = req.validlen;
1102 *retval = req.oldidx;
1108 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1109 void *new, size_t newlen, size_t *retval, int flags)
1111 int oid[CTL_MAXNAME];
1112 size_t oidlen, plen;
1115 oid[0] = 0; /* sysctl internal magic */
1116 oid[1] = 3; /* name2oid */
1117 oidlen = sizeof(oid);
1119 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1120 (void *)name, strlen(name), &plen, flags);
1124 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1125 new, newlen, retval, flags);
1130 * Transfer function to/from user space.
1133 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1136 size_t i, len, origidx;
1138 origidx = req->oldidx;
1140 if (req->oldptr == NULL)
1143 * If we have not wired the user supplied buffer and we are currently
1144 * holding locks, drop a witness warning, as it's possible that
1145 * write operations to the user page can sleep.
1147 if (req->lock != REQ_WIRED)
1148 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1149 "sysctl_old_user()");
1151 len = req->validlen;
1155 if (i > len - origidx)
1157 error = copyout(p, (char *)req->oldptr + origidx, i);
1167 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1173 if (req->newlen - req->newidx < l)
1175 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1176 "sysctl_new_user()");
1177 error = copyin((char *)req->newptr + req->newidx, p, l);
1183 * Wire the user space destination buffer. If set to a value greater than
1184 * zero, the len parameter limits the maximum amount of wired memory.
1187 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1193 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1195 if (req->lock == REQ_LOCKED && req->oldptr &&
1196 req->oldfunc == sysctl_old_user) {
1197 if (wiredlen != 0) {
1198 ret = vslock(req->oldptr, wiredlen);
1205 * Touch all the wired pages to avoid PTE modified
1206 * bit emulation traps on Alpha while holding locks
1207 * in the sysctl handler.
1209 for (i = (wiredlen + PAGE_SIZE - 1) / PAGE_SIZE,
1210 cp = req->oldptr; i > 0; i--, cp += PAGE_SIZE) {
1211 copyin(cp, &dummy, 1);
1212 copyout(&dummy, cp, 1);
1215 req->lock = REQ_WIRED;
1216 req->validlen = wiredlen;
1222 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1223 int *nindx, struct sysctl_req *req)
1225 struct sysctl_oid *oid;
1228 oid = SLIST_FIRST(&sysctl__children);
1230 while (oid && indx < CTL_MAXNAME) {
1231 if (oid->oid_number == name[indx]) {
1233 if (oid->oid_kind & CTLFLAG_NOLOCK)
1234 req->lock = REQ_UNLOCKED;
1235 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1236 if (oid->oid_handler != NULL ||
1244 (struct sysctl_oid_list *)oid->oid_arg1);
1245 } else if (indx == namelen) {
1254 oid = SLIST_NEXT(oid, oid_link);
1261 * Traverse our tree, and find the right node, execute whatever it points
1262 * to, and return the resulting error code.
1266 sysctl_root(SYSCTL_HANDLER_ARGS)
1268 struct sysctl_oid *oid;
1269 int error, indx, lvl;
1271 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1275 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1277 * You can't call a sysctl when it's a node, but has
1278 * no handler. Inform the user that it's a node.
1279 * The indx may or may not be the same as namelen.
1281 if (oid->oid_handler == NULL)
1285 /* Is this sysctl writable? */
1286 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1289 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1291 /* Is this sysctl sensitive to securelevels? */
1292 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1293 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1294 error = securelevel_gt(req->td->td_ucred, lvl);
1299 /* Is this sysctl writable by only privileged users? */
1300 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1301 if (oid->oid_kind & CTLFLAG_PRISON)
1302 error = priv_check(req->td, PRIV_SYSCTL_WRITEJAIL);
1304 error = priv_check(req->td, PRIV_SYSCTL_WRITE);
1309 if (!oid->oid_handler)
1312 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1313 arg1 = (int *)arg1 + indx;
1316 arg1 = oid->oid_arg1;
1317 arg2 = oid->oid_arg2;
1320 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1325 error = oid->oid_handler(oid, arg1, arg2, req);
1330 #ifndef _SYS_SYSPROTO_H_
1331 struct sysctl_args {
1341 __sysctl(struct thread *td, struct sysctl_args *uap)
1343 int error, name[CTL_MAXNAME];
1346 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1349 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1355 error = userland_sysctl(td, name, uap->namelen,
1356 uap->old, uap->oldlenp, 0,
1357 uap->new, uap->newlen, &j, 0);
1358 if (error && error != ENOMEM)
1361 int i = copyout(&j, uap->oldlenp, sizeof(j));
1371 * This is used from various compatibility syscalls too. That's why name
1372 * must be in kernel space.
1375 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1376 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1380 struct sysctl_req req;
1382 bzero(&req, sizeof req);
1389 req.oldlen = *oldlenp;
1391 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1396 req.validlen = req.oldlen;
1399 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1405 if (!useracc(new, newlen, VM_PROT_READ))
1407 req.newlen = newlen;
1411 req.oldfunc = sysctl_old_user;
1412 req.newfunc = sysctl_new_user;
1413 req.lock = REQ_LOCKED;
1420 error = sysctl_root(0, name, namelen, &req);
1421 } while (error == EAGAIN);
1423 if (req.lock == REQ_WIRED && req.validlen > 0)
1424 vsunlock(req.oldptr, req.validlen);
1428 if (error && error != ENOMEM)
1432 if (req.oldptr && req.oldidx > req.validlen)
1433 *retval = req.validlen;
1435 *retval = req.oldidx;
1441 #include <sys/socket.h>
1442 #include <vm/vm_param.h>
1444 #define KINFO_PROC (0<<8)
1445 #define KINFO_RT (1<<8)
1446 #define KINFO_VNODE (2<<8)
1447 #define KINFO_FILE (3<<8)
1448 #define KINFO_METER (4<<8)
1449 #define KINFO_LOADAVG (5<<8)
1450 #define KINFO_CLOCKRATE (6<<8)
1452 /* Non-standard BSDI extension - only present on their 4.3 net-2 releases */
1453 #define KINFO_BSDI_SYSINFO (101<<8)
1456 * XXX this is bloat, but I hope it's better here than on the potentially
1457 * limited kernel stack... -Peter
1461 int bsdi_machine; /* "i386" on BSD/386 */
1462 /* ^^^ this is an offset to the string, relative to the struct start */
1471 int bsdi_ostype; /* "BSD/386" on BSD/386 */
1472 int bsdi_osrelease; /* "1.1" on BSD/386 */
1484 struct timeval pad16;
1485 /* we dont set this, because BSDI's uname used gethostname() instead */
1486 int bsdi_hostname; /* hostname on BSD/386 */
1488 /* the actual string data is appended here */
1493 * this data is appended to the end of the bsdi_si structure during copyout.
1494 * The "char *" offsets are relative to the base of the bsdi_si struct.
1495 * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings
1496 * should not exceed the length of the buffer here... (or else!! :-)
1498 static char bsdi_strings[80]; /* It had better be less than this! */
1500 #ifndef _SYS_SYSPROTO_H_
1501 struct getkerninfo_args {
1509 ogetkerninfo(struct thread *td, struct getkerninfo_args *uap)
1517 switch (uap->op & 0xff00) {
1523 name[3] = (uap->op & 0xff0000) >> 16;
1524 name[4] = uap->op & 0xff;
1526 error = userland_sysctl(td, name, 6, uap->where, uap->size,
1532 name[1] = KERN_VNODE;
1533 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1539 name[1] = KERN_PROC;
1540 name[2] = uap->op & 0xff;
1542 error = userland_sysctl(td, name, 4, uap->where, uap->size,
1548 name[1] = KERN_FILE;
1549 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1556 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1562 name[1] = VM_LOADAVG;
1563 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1567 case KINFO_CLOCKRATE:
1569 name[1] = KERN_CLOCKRATE;
1570 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1574 case KINFO_BSDI_SYSINFO: {
1576 * this is pretty crude, but it's just enough for uname()
1577 * from BSDI's 1.x libc to work.
1579 * *size gives the size of the buffer before the call, and
1580 * the amount of data copied after a successful call.
1581 * If successful, the return value is the amount of data
1582 * available, which can be larger than *size.
1584 * BSDI's 2.x product apparently fails with ENOMEM if *size
1591 bzero((char *)&bsdi_si, sizeof(bsdi_si));
1592 bzero(bsdi_strings, sizeof(bsdi_strings));
1596 bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si);
1600 bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si);
1601 strcpy(s, osrelease);
1604 bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si);
1608 needed = sizeof(bsdi_si) + (s - bsdi_strings);
1610 if ((uap->where == NULL) || (uap->size == NULL)) {
1611 /* process is asking how much buffer to supply.. */
1617 if ((error = copyin(uap->size, &size, sizeof(size))) != 0)
1620 /* if too much buffer supplied, trim it down */
1624 /* how much of the buffer is remaining */
1627 if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0)
1630 /* is there any point in continuing? */
1631 if (left > sizeof(bsdi_si)) {
1632 left -= sizeof(bsdi_si);
1633 error = copyout(&bsdi_strings,
1634 uap->where + sizeof(bsdi_si), left);
1644 td->td_retval[0] = needed ? needed : size;
1646 error = copyout(&size, uap->size, sizeof(size));
1652 #endif /* COMPAT_43 */