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 * Reparent an existing oid.
421 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
423 struct sysctl_oid *oidp;
425 if (oid->oid_parent == parent)
427 oidp = sysctl_find_oidname(oid->oid_name, parent);
430 sysctl_unregister_oid(oid);
431 oid->oid_parent = parent;
432 oid->oid_number = OID_AUTO;
433 sysctl_register_oid(oid);
438 * Register the kernel's oids on startup.
440 SET_DECLARE(sysctl_set, struct sysctl_oid);
443 sysctl_register_all(void *arg)
445 struct sysctl_oid **oidp;
448 SET_FOREACH(oidp, sysctl_set)
449 sysctl_register_oid(*oidp);
451 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
456 * These functions implement a presently undocumented interface
457 * used by the sysctl program to walk the tree, and get the type
458 * so it can print the value.
459 * This interface is under work and consideration, and should probably
460 * be killed with a big axe by the first person who can find the time.
461 * (be aware though, that the proper interface isn't as obvious as it
462 * may seem, there are various conflicting requirements.
464 * {0,0} printf the entire MIB-tree.
465 * {0,1,...} return the name of the "..." OID.
466 * {0,2,...} return the next OID.
467 * {0,3} return the OID of the name in "new"
468 * {0,4,...} return the kind & format info for the "..." OID.
469 * {0,5,...} return the description the "..." OID.
474 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
477 struct sysctl_oid *oidp;
479 SLIST_FOREACH(oidp, l, oid_link) {
484 printf("%d %s ", oidp->oid_number, oidp->oid_name);
487 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
488 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
490 if (oidp->oid_handler)
493 switch (oidp->oid_kind & CTLTYPE) {
496 if (!oidp->oid_handler) {
497 sysctl_sysctl_debug_dump_node(
498 oidp->oid_arg1, i+2);
501 case CTLTYPE_INT: printf(" Int\n"); break;
502 case CTLTYPE_STRING: printf(" String\n"); break;
503 case CTLTYPE_QUAD: printf(" Quad\n"); break;
504 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
505 default: printf("\n");
512 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
516 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
519 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
523 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
524 0, 0, sysctl_sysctl_debug, "-", "");
528 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
530 int *name = (int *) arg1;
531 u_int namelen = arg2;
533 struct sysctl_oid *oid;
534 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
539 snprintf(buf,sizeof(buf),"%d",*name);
541 error = SYSCTL_OUT(req, ".", 1);
543 error = SYSCTL_OUT(req, buf, strlen(buf));
551 SLIST_FOREACH(oid, lsp, oid_link) {
552 if (oid->oid_number != *name)
556 error = SYSCTL_OUT(req, ".", 1);
558 error = SYSCTL_OUT(req, oid->oid_name,
559 strlen(oid->oid_name));
566 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
569 if (oid->oid_handler)
572 lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
577 return (SYSCTL_OUT(req, "", 1));
580 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
583 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
584 int *next, int *len, int level, struct sysctl_oid **oidpp)
586 struct sysctl_oid *oidp;
589 SLIST_FOREACH(oidp, lsp, oid_link) {
590 *next = oidp->oid_number;
593 if (oidp->oid_kind & CTLFLAG_SKIP)
597 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
599 if (oidp->oid_handler)
600 /* We really should call the handler here...*/
602 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
603 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
604 len, level+1, oidpp))
609 if (oidp->oid_number < *name)
612 if (oidp->oid_number > *name) {
613 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
615 if (oidp->oid_handler)
617 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
618 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
619 next+1, len, level+1, oidpp))
623 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
626 if (oidp->oid_handler)
629 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
630 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
631 len, level+1, oidpp))
642 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
644 int *name = (int *) arg1;
645 u_int namelen = arg2;
647 struct sysctl_oid *oid;
648 struct sysctl_oid_list *lsp = &sysctl__children;
649 int newoid[CTL_MAXNAME];
651 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
654 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
658 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
661 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
664 struct sysctl_oid *oidp;
665 struct sysctl_oid_list *lsp = &sysctl__children;
671 p = name + strlen(name) - 1 ;
677 for (p = name; *p && *p != '.'; p++)
683 oidp = SLIST_FIRST(lsp);
685 while (oidp && *len < CTL_MAXNAME) {
686 if (strcmp(name, oidp->oid_name)) {
687 oidp = SLIST_NEXT(oidp, oid_link);
690 *oid++ = oidp->oid_number;
699 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
702 if (oidp->oid_handler)
705 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
706 oidp = SLIST_FIRST(lsp);
708 for (p = name; *p && *p != '.'; p++)
718 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
721 int error, oid[CTL_MAXNAME], len;
722 struct sysctl_oid *op = 0;
726 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
727 return (ENAMETOOLONG);
729 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
731 error = SYSCTL_IN(req, p, req->newlen);
737 p [req->newlen] = '\0';
739 error = name2oid(p, oid, &len, &op);
746 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
750 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
751 sysctl_sysctl_name2oid, "I", "");
754 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
756 struct sysctl_oid *oid;
759 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
765 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
768 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
773 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
776 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
778 struct sysctl_oid *oid;
781 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
787 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
791 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
794 * Default "handler" functions.
798 * Handle an int, signed or unsigned.
800 * a variable: point arg1 at it.
801 * a constant: pass it in arg2.
805 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
807 int tmpout, error = 0;
810 * Attempt to get a coherent snapshot by making a copy of the data.
813 tmpout = *(int *)arg1;
816 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
818 if (error || !req->newptr)
824 error = SYSCTL_IN(req, arg1, sizeof(int));
830 * Based on on sysctl_handle_int() convert milliseconds into ticks.
834 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
838 tt = *(int *)oidp->oid_arg1;
839 s = (int)((int64_t)tt * 1000 / hz);
841 error = sysctl_handle_int(oidp, &s, 0, req);
842 if (error || !req->newptr)
845 tt = (int)((int64_t)s * hz / 1000);
849 *(int *)oidp->oid_arg1 = tt;
855 * Handle a long, signed or unsigned. arg1 points to it.
859 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
868 * Attempt to get a coherent snapshot by making a copy of the data.
872 tmplong = *(long *)arg1;
874 if (req->flags & SCTL_MASK32) {
876 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
879 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
881 if (error || !req->newptr)
885 if (req->flags & SCTL_MASK32) {
886 error = SYSCTL_IN(req, &tmpint, sizeof(int));
887 *(long *)arg1 = (long)tmpint;
890 error = SYSCTL_IN(req, arg1, sizeof(long));
895 * Handle a 64 bit int, signed or unsigned. arg1 points to it.
899 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
905 * Attempt to get a coherent snapshot by making a copy of the data.
909 tmpout = *(uint64_t *)arg1;
910 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
912 if (error || !req->newptr)
915 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
920 * Handle our generic '\0' terminated 'C' string.
922 * a variable string: point arg1 at it, arg2 is max length.
923 * a constant string: point arg1 at it, arg2 is zero.
927 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
934 * Attempt to get a coherent snapshot by copying to a
935 * temporary kernel buffer.
938 outlen = strlen((char *)arg1)+1;
939 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
941 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
942 free(tmparg, M_SYSCTLTMP);
946 error = SYSCTL_OUT(req, tmparg, outlen);
947 free(tmparg, M_SYSCTLTMP);
949 if (error || !req->newptr)
952 if ((req->newlen - req->newidx) >= arg2) {
955 arg2 = (req->newlen - req->newidx);
956 error = SYSCTL_IN(req, arg1, arg2);
957 ((char *)arg1)[arg2] = '\0';
964 * Handle any kind of opaque data.
965 * arg1 points to it, arg2 is the size.
969 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
973 struct sysctl_req req2;
976 * Attempt to get a coherent snapshot, by using the thread
977 * pre-emption counter updated from within mi_switch() to
978 * determine if we were pre-empted during a bcopy() or
979 * copyout(). Make 3 attempts at doing this before giving up.
980 * If we encounter an error, stop immediately.
985 generation = curthread->td_generation;
986 error = SYSCTL_OUT(req, arg1, arg2);
990 if (generation != curthread->td_generation && tries < 3) {
995 error = SYSCTL_IN(req, arg1, arg2);
1001 * Transfer functions to/from kernel space.
1002 * XXX: rather untested at this point
1005 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1011 if (req->oldlen <= req->oldidx)
1014 if (i > req->oldlen - req->oldidx)
1015 i = req->oldlen - req->oldidx;
1017 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1020 if (req->oldptr && i != l)
1026 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1030 if (req->newlen - req->newidx < l)
1032 bcopy((char *)req->newptr + req->newidx, p, l);
1038 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1039 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1042 struct sysctl_req req;
1044 bzero(&req, sizeof req);
1050 req.oldlen = *oldlenp;
1052 req.validlen = req.oldlen;
1059 req.newlen = newlen;
1063 req.oldfunc = sysctl_old_kernel;
1064 req.newfunc = sysctl_new_kernel;
1065 req.lock = REQ_LOCKED;
1069 error = sysctl_root(0, name, namelen, &req);
1071 if (req.lock == REQ_WIRED && req.validlen > 0)
1072 vsunlock(req.oldptr, req.validlen);
1076 if (error && error != ENOMEM)
1080 if (req.oldptr && req.oldidx > req.validlen)
1081 *retval = req.validlen;
1083 *retval = req.oldidx;
1089 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1090 void *new, size_t newlen, size_t *retval, int flags)
1092 int oid[CTL_MAXNAME];
1093 size_t oidlen, plen;
1096 oid[0] = 0; /* sysctl internal magic */
1097 oid[1] = 3; /* name2oid */
1098 oidlen = sizeof(oid);
1100 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1101 (void *)name, strlen(name), &plen, flags);
1105 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1106 new, newlen, retval, flags);
1111 * Transfer function to/from user space.
1114 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1117 size_t i, len, origidx;
1119 origidx = req->oldidx;
1121 if (req->oldptr == NULL)
1124 * If we have not wired the user supplied buffer and we are currently
1125 * holding locks, drop a witness warning, as it's possible that
1126 * write operations to the user page can sleep.
1128 if (req->lock != REQ_WIRED)
1129 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1130 "sysctl_old_user()");
1132 len = req->validlen;
1136 if (i > len - origidx)
1138 error = copyout(p, (char *)req->oldptr + origidx, i);
1148 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1154 if (req->newlen - req->newidx < l)
1156 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1157 "sysctl_new_user()");
1158 error = copyin((char *)req->newptr + req->newidx, p, l);
1164 * Wire the user space destination buffer. If set to a value greater than
1165 * zero, the len parameter limits the maximum amount of wired memory.
1168 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1174 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1176 if (req->lock == REQ_LOCKED && req->oldptr &&
1177 req->oldfunc == sysctl_old_user) {
1178 if (wiredlen != 0) {
1179 ret = vslock(req->oldptr, wiredlen);
1186 * Touch all the wired pages to avoid PTE modified
1187 * bit emulation traps on Alpha while holding locks
1188 * in the sysctl handler.
1190 for (i = (wiredlen + PAGE_SIZE - 1) / PAGE_SIZE,
1191 cp = req->oldptr; i > 0; i--, cp += PAGE_SIZE) {
1192 copyin(cp, &dummy, 1);
1193 copyout(&dummy, cp, 1);
1196 req->lock = REQ_WIRED;
1197 req->validlen = wiredlen;
1203 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1204 int *nindx, struct sysctl_req *req)
1206 struct sysctl_oid *oid;
1209 oid = SLIST_FIRST(&sysctl__children);
1211 while (oid && indx < CTL_MAXNAME) {
1212 if (oid->oid_number == name[indx]) {
1214 if (oid->oid_kind & CTLFLAG_NOLOCK)
1215 req->lock = REQ_UNLOCKED;
1216 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1217 if (oid->oid_handler != NULL ||
1225 (struct sysctl_oid_list *)oid->oid_arg1);
1226 } else if (indx == namelen) {
1235 oid = SLIST_NEXT(oid, oid_link);
1242 * Traverse our tree, and find the right node, execute whatever it points
1243 * to, and return the resulting error code.
1247 sysctl_root(SYSCTL_HANDLER_ARGS)
1249 struct sysctl_oid *oid;
1250 int error, indx, lvl;
1252 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1256 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1258 * You can't call a sysctl when it's a node, but has
1259 * no handler. Inform the user that it's a node.
1260 * The indx may or may not be the same as namelen.
1262 if (oid->oid_handler == NULL)
1266 /* Is this sysctl writable? */
1267 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1270 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1272 /* Is this sysctl sensitive to securelevels? */
1273 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1274 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1275 error = securelevel_gt(req->td->td_ucred, lvl);
1280 /* Is this sysctl writable by only privileged users? */
1281 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1282 if (oid->oid_kind & CTLFLAG_PRISON)
1283 error = priv_check(req->td, PRIV_SYSCTL_WRITEJAIL);
1285 error = priv_check(req->td, PRIV_SYSCTL_WRITE);
1290 if (!oid->oid_handler)
1293 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1294 arg1 = (int *)arg1 + indx;
1297 arg1 = oid->oid_arg1;
1298 arg2 = oid->oid_arg2;
1301 error = mac_check_system_sysctl(req->td->td_ucred, oid, arg1, arg2,
1306 error = oid->oid_handler(oid, arg1, arg2, req);
1311 #ifndef _SYS_SYSPROTO_H_
1312 struct sysctl_args {
1322 __sysctl(struct thread *td, struct sysctl_args *uap)
1324 int error, name[CTL_MAXNAME];
1327 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1330 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1336 error = userland_sysctl(td, name, uap->namelen,
1337 uap->old, uap->oldlenp, 0,
1338 uap->new, uap->newlen, &j, 0);
1339 if (error && error != ENOMEM)
1342 int i = copyout(&j, uap->oldlenp, sizeof(j));
1352 * This is used from various compatibility syscalls too. That's why name
1353 * must be in kernel space.
1356 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1357 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1361 struct sysctl_req req;
1363 bzero(&req, sizeof req);
1370 req.oldlen = *oldlenp;
1372 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1377 req.validlen = req.oldlen;
1380 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1386 if (!useracc(new, newlen, VM_PROT_READ))
1388 req.newlen = newlen;
1392 req.oldfunc = sysctl_old_user;
1393 req.newfunc = sysctl_new_user;
1394 req.lock = REQ_LOCKED;
1401 error = sysctl_root(0, name, namelen, &req);
1402 } while (error == EAGAIN);
1404 if (req.lock == REQ_WIRED && req.validlen > 0)
1405 vsunlock(req.oldptr, req.validlen);
1409 if (error && error != ENOMEM)
1413 if (req.oldptr && req.oldidx > req.validlen)
1414 *retval = req.validlen;
1416 *retval = req.oldidx;
1422 #include <sys/socket.h>
1423 #include <vm/vm_param.h>
1425 #define KINFO_PROC (0<<8)
1426 #define KINFO_RT (1<<8)
1427 #define KINFO_VNODE (2<<8)
1428 #define KINFO_FILE (3<<8)
1429 #define KINFO_METER (4<<8)
1430 #define KINFO_LOADAVG (5<<8)
1431 #define KINFO_CLOCKRATE (6<<8)
1433 /* Non-standard BSDI extension - only present on their 4.3 net-2 releases */
1434 #define KINFO_BSDI_SYSINFO (101<<8)
1437 * XXX this is bloat, but I hope it's better here than on the potentially
1438 * limited kernel stack... -Peter
1442 int bsdi_machine; /* "i386" on BSD/386 */
1443 /* ^^^ this is an offset to the string, relative to the struct start */
1452 int bsdi_ostype; /* "BSD/386" on BSD/386 */
1453 int bsdi_osrelease; /* "1.1" on BSD/386 */
1465 struct timeval pad16;
1466 /* we dont set this, because BSDI's uname used gethostname() instead */
1467 int bsdi_hostname; /* hostname on BSD/386 */
1469 /* the actual string data is appended here */
1474 * this data is appended to the end of the bsdi_si structure during copyout.
1475 * The "char *" offsets are relative to the base of the bsdi_si struct.
1476 * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings
1477 * should not exceed the length of the buffer here... (or else!! :-)
1479 static char bsdi_strings[80]; /* It had better be less than this! */
1481 #ifndef _SYS_SYSPROTO_H_
1482 struct getkerninfo_args {
1490 ogetkerninfo(struct thread *td, struct getkerninfo_args *uap)
1498 switch (uap->op & 0xff00) {
1504 name[3] = (uap->op & 0xff0000) >> 16;
1505 name[4] = uap->op & 0xff;
1507 error = userland_sysctl(td, name, 6, uap->where, uap->size,
1513 name[1] = KERN_VNODE;
1514 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1520 name[1] = KERN_PROC;
1521 name[2] = uap->op & 0xff;
1523 error = userland_sysctl(td, name, 4, uap->where, uap->size,
1529 name[1] = KERN_FILE;
1530 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1537 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1543 name[1] = VM_LOADAVG;
1544 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1548 case KINFO_CLOCKRATE:
1550 name[1] = KERN_CLOCKRATE;
1551 error = userland_sysctl(td, name, 2, uap->where, uap->size,
1555 case KINFO_BSDI_SYSINFO: {
1557 * this is pretty crude, but it's just enough for uname()
1558 * from BSDI's 1.x libc to work.
1560 * *size gives the size of the buffer before the call, and
1561 * the amount of data copied after a successful call.
1562 * If successful, the return value is the amount of data
1563 * available, which can be larger than *size.
1565 * BSDI's 2.x product apparently fails with ENOMEM if *size
1572 bzero((char *)&bsdi_si, sizeof(bsdi_si));
1573 bzero(bsdi_strings, sizeof(bsdi_strings));
1577 bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si);
1581 bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si);
1582 strcpy(s, osrelease);
1585 bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si);
1589 needed = sizeof(bsdi_si) + (s - bsdi_strings);
1591 if ((uap->where == NULL) || (uap->size == NULL)) {
1592 /* process is asking how much buffer to supply.. */
1598 if ((error = copyin(uap->size, &size, sizeof(size))) != 0)
1601 /* if too much buffer supplied, trim it down */
1605 /* how much of the buffer is remaining */
1608 if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0)
1611 /* is there any point in continuing? */
1612 if (left > sizeof(bsdi_si)) {
1613 left -= sizeof(bsdi_si);
1614 error = copyout(&bsdi_strings,
1615 uap->where + sizeof(bsdi_si), left);
1625 td->td_retval[0] = needed ? needed : size;
1627 error = copyout(&size, uap->size, sizeof(size));
1633 #endif /* COMPAT_43 */