1/*-
2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Mike Karels at Berkeley Software Design, Inc.
7 *
8 * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
9 * project, to make these variables more userfriendly.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
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.
22 *
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
33 * SUCH DAMAGE.
34 *
35 * @(#)kern_sysctl.c 8.4 (Berkeley) 4/14/94
36 */
37
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: head/sys/kern/kern_sysctl.c 187656 2009-01-23 22:40:35Z jhb $");
40
41#include "opt_compat.h"
42#include "opt_mac.h"
43
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>
49#include <sys/priv.h>
50#include <sys/proc.h>
51#include <sys/lock.h>
52#include <sys/mutex.h>
53#include <sys/sx.h>
54#include <sys/sysproto.h>
55#include <sys/uio.h>
56#include <sys/vimage.h>
57
58#include <security/mac/mac_framework.h>
59
60#include <vm/vm.h>
61#include <vm/vm_extern.h>
62
63static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
64static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
65static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
66
67/*
68 * Locking - this locks the sysctl tree in memory.
69 */
70static struct sx sysctllock;
71
72#define SYSCTL_LOCK() sx_xlock(&sysctllock)
73#define SYSCTL_UNLOCK() sx_xunlock(&sysctllock)
74#define SYSCTL_LOCK_ASSERT() sx_assert(&sysctllock, SX_XLOCKED)
75#define SYSCTL_INIT() sx_init(&sysctllock, "sysctl lock")
76
77static int sysctl_root(SYSCTL_HANDLER_ARGS);
78
79struct sysctl_oid_list sysctl__children; /* root list */
80
81static struct sysctl_oid *
82sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
83{
84 struct sysctl_oid *oidp;
85
86 SLIST_FOREACH(oidp, list, oid_link) {
87 if (strcmp(oidp->oid_name, name) == 0) {
88 return (oidp);
89 }
90 }
91 return (NULL);
92}
93
94/*
95 * Initialization of the MIB tree.
96 *
97 * Order by number in each list.
98 */
99
100void
101sysctl_register_oid(struct sysctl_oid *oidp)
102{
103 struct sysctl_oid_list *parent = oidp->oid_parent;
104 struct sysctl_oid *p;
105 struct sysctl_oid *q;
106
107 /*
108 * First check if another oid with the same name already
109 * exists in the parent's list.
110 */
111 p = sysctl_find_oidname(oidp->oid_name, parent);
112 if (p != NULL) {
113 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
114 p->oid_refcnt++;
115 return;
116 } else {
117 printf("can't re-use a leaf (%s)!\n", p->oid_name);
118 return;
119 }
120 }
121 /*
122 * If this oid has a number OID_AUTO, give it a number which
123 * is greater than any current oid.
124 * NOTE: DO NOT change the starting value here, change it in
125 * <sys/sysctl.h>, and make sure it is at least 256 to
126 * accomodate e.g. net.inet.raw as a static sysctl node.
127 */
128 if (oidp->oid_number == OID_AUTO) {
129 static int newoid = CTL_AUTO_START;
130
131 oidp->oid_number = newoid++;
132 if (newoid == 0x7fffffff)
133 panic("out of oids");
134 }
135#if 0
136 else if (oidp->oid_number >= CTL_AUTO_START) {
137 /* do not panic; this happens when unregistering sysctl sets */
138 printf("static sysctl oid too high: %d", oidp->oid_number);
139 }
140#endif
141
142 /*
143 * Insert the oid into the parent's list in order.
144 */
145 q = NULL;
146 SLIST_FOREACH(p, parent, oid_link) {
147 if (oidp->oid_number < p->oid_number)
148 break;
149 q = p;
150 }
151 if (q)
152 SLIST_INSERT_AFTER(q, oidp, oid_link);
153 else
154 SLIST_INSERT_HEAD(parent, oidp, oid_link);
155}
156
157void
158sysctl_unregister_oid(struct sysctl_oid *oidp)
159{
160 struct sysctl_oid *p;
161 int error;
162
163 error = ENOENT;
164 if (oidp->oid_number == OID_AUTO) {
165 error = EINVAL;
166 } else {
167 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
168 if (p == oidp) {
169 SLIST_REMOVE(oidp->oid_parent, oidp,
170 sysctl_oid, oid_link);
171 error = 0;
172 break;
173 }
174 }
175 }
176
177 /*
178 * This can happen when a module fails to register and is
179 * being unloaded afterwards. It should not be a panic()
180 * for normal use.
181 */
182 if (error)
183 printf("%s: failed to unregister sysctl\n", __func__);
184}
185
186/* Initialize a new context to keep track of dynamically added sysctls. */
187int
188sysctl_ctx_init(struct sysctl_ctx_list *c)
189{
190
191 if (c == NULL) {
192 return (EINVAL);
193 }
194 TAILQ_INIT(c);
195 return (0);
196}
197
198/* Free the context, and destroy all dynamic oids registered in this context */
199int
200sysctl_ctx_free(struct sysctl_ctx_list *clist)
201{
202 struct sysctl_ctx_entry *e, *e1;
203 int error;
204
205 error = 0;
206 /*
207 * First perform a "dry run" to check if it's ok to remove oids.
208 * XXX FIXME
209 * XXX This algorithm is a hack. But I don't know any
210 * XXX better solution for now...
211 */
212 TAILQ_FOREACH(e, clist, link) {
213 error = sysctl_remove_oid(e->entry, 0, 0);
214 if (error)
215 break;
216 }
217 /*
218 * Restore deregistered entries, either from the end,
219 * or from the place where error occured.
220 * e contains the entry that was not unregistered
221 */
222 if (error)
223 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
224 else
225 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
226 while (e1 != NULL) {
227 sysctl_register_oid(e1->entry);
228 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
229 }
230 if (error)
231 return(EBUSY);
232 /* Now really delete the entries */
233 e = TAILQ_FIRST(clist);
234 while (e != NULL) {
235 e1 = TAILQ_NEXT(e, link);
236 error = sysctl_remove_oid(e->entry, 1, 0);
237 if (error)
238 panic("sysctl_remove_oid: corrupt tree, entry: %s",
239 e->entry->oid_name);
240 free(e, M_SYSCTLOID);
241 e = e1;
242 }
243 return (error);
244}
245
246/* Add an entry to the context */
247struct sysctl_ctx_entry *
248sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
249{
250 struct sysctl_ctx_entry *e;
251
252 if (clist == NULL || oidp == NULL)
253 return(NULL);
254 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
255 e->entry = oidp;
256 TAILQ_INSERT_HEAD(clist, e, link);
257 return (e);
258}
259
260/* Find an entry in the context */
261struct sysctl_ctx_entry *
262sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
263{
264 struct sysctl_ctx_entry *e;
265
266 if (clist == NULL || oidp == NULL)
267 return(NULL);
268 TAILQ_FOREACH(e, clist, link) {
269 if(e->entry == oidp)
270 return(e);
271 }
272 return (e);
273}
274
275/*
276 * Delete an entry from the context.
277 * NOTE: this function doesn't free oidp! You have to remove it
278 * with sysctl_remove_oid().
279 */
280int
281sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
282{
283 struct sysctl_ctx_entry *e;
284
285 if (clist == NULL || oidp == NULL)
286 return (EINVAL);
287 e = sysctl_ctx_entry_find(clist, oidp);
288 if (e != NULL) {
289 TAILQ_REMOVE(clist, e, link);
290 free(e, M_SYSCTLOID);
291 return (0);
292 } else
293 return (ENOENT);
294}
295
296/*
297 * Remove dynamically created sysctl trees.
298 * oidp - top of the tree to be removed
299 * del - if 0 - just deregister, otherwise free up entries as well
300 * recurse - if != 0 traverse the subtree to be deleted
301 */
302int
303sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
304{
305 struct sysctl_oid *p;
306 int error;
307
308 if (oidp == NULL)
309 return(EINVAL);
310 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
311 printf("can't remove non-dynamic nodes!\n");
312 return (EINVAL);
313 }
314 /*
315 * WARNING: normal method to do this should be through
316 * sysctl_ctx_free(). Use recursing as the last resort
317 * method to purge your sysctl tree of leftovers...
318 * However, if some other code still references these nodes,
319 * it will panic.
320 */
321 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
322 if (oidp->oid_refcnt == 1) {
323 SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
324 if (!recurse)
325 return (ENOTEMPTY);
326 error = sysctl_remove_oid(p, del, recurse);
327 if (error)
328 return (error);
329 }
330 if (del)
331 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
332 }
333 }
334 if (oidp->oid_refcnt > 1 ) {
335 oidp->oid_refcnt--;
336 } else {
337 if (oidp->oid_refcnt == 0) {
338 printf("Warning: bad oid_refcnt=%u (%s)!\n",
339 oidp->oid_refcnt, oidp->oid_name);
340 return (EINVAL);
341 }
342 sysctl_unregister_oid(oidp);
343 if (del) {
344 if (oidp->oid_descr)
345 free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
346 free((void *)(uintptr_t)(const void *)oidp->oid_name,
347 M_SYSCTLOID);
348 free(oidp, M_SYSCTLOID);
349 }
350 }
351 return (0);
352}
353
354/*
355 * Create new sysctls at run time.
356 * clist may point to a valid context initialized with sysctl_ctx_init().
357 */
358struct sysctl_oid *
359sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
360 int number, const char *name, int kind, void *arg1, int arg2,
361 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
362{
363 struct sysctl_oid *oidp;
364 ssize_t len;
365 char *newname;
366
367 /* You have to hook up somewhere.. */
368 if (parent == NULL)
369 return(NULL);
370 /* Check if the node already exists, otherwise create it */
371 oidp = sysctl_find_oidname(name, parent);
372 if (oidp != NULL) {
373 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
374 oidp->oid_refcnt++;
375 /* Update the context */
376 if (clist != NULL)
377 sysctl_ctx_entry_add(clist, oidp);
378 return (oidp);
379 } else {
380 printf("can't re-use a leaf (%s)!\n", name);
381 return (NULL);
382 }
383 }
384 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
385 oidp->oid_parent = parent;
386 SLIST_NEXT(oidp, oid_link) = NULL;
387 oidp->oid_number = number;
388 oidp->oid_refcnt = 1;
389 len = strlen(name);
390 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
391 bcopy(name, newname, len + 1);
392 newname[len] = '\0';
393 oidp->oid_name = newname;
394 oidp->oid_handler = handler;
395 oidp->oid_kind = CTLFLAG_DYN | kind;
396 if ((kind & CTLTYPE) == CTLTYPE_NODE) {
397 /* Allocate space for children */
398 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
399 M_SYSCTLOID, M_WAITOK));
400 SLIST_INIT(SYSCTL_CHILDREN(oidp));
401 } else {
402 oidp->oid_arg1 = arg1;
403 oidp->oid_arg2 = arg2;
404 }
405 oidp->oid_fmt = fmt;
406 if (descr) {
407 int len = strlen(descr) + 1;
408 oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
409 if (oidp->oid_descr)
410 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
411 }
412 /* Update the context, if used */
413 if (clist != NULL)
414 sysctl_ctx_entry_add(clist, oidp);
415 /* Register this oid */
416 sysctl_register_oid(oidp);
417 return (oidp);
418}
419
420/*
421 * Rename an existing oid.
422 */
423void
424sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
425{
426 ssize_t len;
427 char *newname;
428 void *oldname;
429
430 oldname = (void *)(uintptr_t)(const void *)oidp->oid_name;
431 len = strlen(name);
432 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
433 bcopy(name, newname, len + 1);
434 newname[len] = '\0';
435 oidp->oid_name = newname;
436 free(oldname, M_SYSCTLOID);
437}
438
439/*
440 * Reparent an existing oid.
441 */
442int
443sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
444{
445 struct sysctl_oid *oidp;
446
447 if (oid->oid_parent == parent)
448 return (0);
449 oidp = sysctl_find_oidname(oid->oid_name, parent);
450 if (oidp != NULL)
451 return (EEXIST);
452 sysctl_unregister_oid(oid);
453 oid->oid_parent = parent;
454 oid->oid_number = OID_AUTO;
455 sysctl_register_oid(oid);
456 return (0);
457}
458
459/*
460 * Register the kernel's oids on startup.
461 */
462SET_DECLARE(sysctl_set, struct sysctl_oid);
463
464static void
465sysctl_register_all(void *arg)
466{
467 struct sysctl_oid **oidp;
468
469 SYSCTL_INIT();
470 SET_FOREACH(oidp, sysctl_set)
471 sysctl_register_oid(*oidp);
472}
473SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
474
475/*
476 * "Staff-functions"
477 *
478 * These functions implement a presently undocumented interface
479 * used by the sysctl program to walk the tree, and get the type
480 * so it can print the value.
481 * This interface is under work and consideration, and should probably
482 * be killed with a big axe by the first person who can find the time.
483 * (be aware though, that the proper interface isn't as obvious as it
484 * may seem, there are various conflicting requirements.
485 *
486 * {0,0} printf the entire MIB-tree.
487 * {0,1,...} return the name of the "..." OID.
488 * {0,2,...} return the next OID.
489 * {0,3} return the OID of the name in "new"
490 * {0,4,...} return the kind & format info for the "..." OID.
491 * {0,5,...} return the description the "..." OID.
492 */
493
494#ifdef SYSCTL_DEBUG
495static void
496sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
497{
498 int k;
499 struct sysctl_oid *oidp;
500
501 SLIST_FOREACH(oidp, l, oid_link) {
502
503 for (k=0; k<i; k++)
504 printf(" ");
505
506 printf("%d %s ", oidp->oid_number, oidp->oid_name);
507
508 printf("%c%c",
509 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
510 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
511
512 if (oidp->oid_handler)
513 printf(" *Handler");
514
515 switch (oidp->oid_kind & CTLTYPE) {
516 case CTLTYPE_NODE:
517 printf(" Node\n");
518 if (!oidp->oid_handler) {
519 sysctl_sysctl_debug_dump_node(
520 oidp->oid_arg1, i+2);
521 }
522 break;
523 case CTLTYPE_INT: printf(" Int\n"); break;
524 case CTLTYPE_STRING: printf(" String\n"); break;
525 case CTLTYPE_QUAD: printf(" Quad\n"); break;
526 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
527 default: printf("\n");
528 }
529
530 }
531}
532
533static int
534sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
535{
536 int error;
537
538 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
539 if (error)
540 return (error);
541 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
542 return (ENOENT);
543}
544
545SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
546 0, 0, sysctl_sysctl_debug, "-", "");
547#endif
548
549static int
550sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
551{
552 int *name = (int *) arg1;
553 u_int namelen = arg2;
554 int error = 0;
555 struct sysctl_oid *oid;
556 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
557 char buf[10];
558
559 while (namelen) {
560 if (!lsp) {
561 snprintf(buf,sizeof(buf),"%d",*name);
562 if (req->oldidx)
563 error = SYSCTL_OUT(req, ".", 1);
564 if (!error)
565 error = SYSCTL_OUT(req, buf, strlen(buf));
566 if (error)
567 return (error);
568 namelen--;
569 name++;
570 continue;
571 }
572 lsp2 = 0;
573 SLIST_FOREACH(oid, lsp, oid_link) {
574 if (oid->oid_number != *name)
575 continue;
576
577 if (req->oldidx)
578 error = SYSCTL_OUT(req, ".", 1);
579 if (!error)
580 error = SYSCTL_OUT(req, oid->oid_name,
581 strlen(oid->oid_name));
582 if (error)
583 return (error);
584
585 namelen--;
586 name++;
587
588 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
589 break;
590
591 if (oid->oid_handler)
592 break;
593
594 lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
595 break;
596 }
597 lsp = lsp2;
598 }
599 return (SYSCTL_OUT(req, "", 1));
600}
601
602static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
603
604static int
605sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
606 int *next, int *len, int level, struct sysctl_oid **oidpp)
607{
608 struct sysctl_oid *oidp;
609
610 *len = level;
611 SLIST_FOREACH(oidp, lsp, oid_link) {
612 *next = oidp->oid_number;
613 *oidpp = oidp;
614
615 if (oidp->oid_kind & CTLFLAG_SKIP)
616 continue;
617
618 if (!namelen) {
619 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
620 return (0);
621 if (oidp->oid_handler)
622 /* We really should call the handler here...*/
623 return (0);
624 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
625 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
626 len, level+1, oidpp))
627 return (0);
628 goto emptynode;
629 }
630
631 if (oidp->oid_number < *name)
632 continue;
633
634 if (oidp->oid_number > *name) {
635 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
636 return (0);
637 if (oidp->oid_handler)
638 return (0);
639 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
640 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
641 next+1, len, level+1, oidpp))
642 return (0);
643 goto next;
644 }
645 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
646 continue;
647
648 if (oidp->oid_handler)
649 continue;
650
651 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
652 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
653 len, level+1, oidpp))
654 return (0);
655 next:
656 namelen = 1;
657 emptynode:
658 *len = level;
659 }
660 return (1);
661}
662
663static int
664sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
665{
666 int *name = (int *) arg1;
667 u_int namelen = arg2;
668 int i, j, error;
669 struct sysctl_oid *oid;
670 struct sysctl_oid_list *lsp = &sysctl__children;
671 int newoid[CTL_MAXNAME];
672
673 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
674 if (i)
675 return (ENOENT);
676 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
677 return (error);
678}
679
680static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
681
682static int
683name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
684{
685 int i;
686 struct sysctl_oid *oidp;
687 struct sysctl_oid_list *lsp = &sysctl__children;
688 char *p;
689
690 SYSCTL_LOCK_ASSERT();
691
692 if (!*name)
693 return (ENOENT);
694
695 p = name + strlen(name) - 1 ;
696 if (*p == '.')
697 *p = '\0';
698
699 *len = 0;
700
701 for (p = name; *p && *p != '.'; p++)
702 ;
703 i = *p;
704 if (i == '.')
705 *p = '\0';
706
707 oidp = SLIST_FIRST(lsp);
708
709 while (oidp && *len < CTL_MAXNAME) {
710 if (strcmp(name, oidp->oid_name)) {
711 oidp = SLIST_NEXT(oidp, oid_link);
712 continue;
713 }
714 *oid++ = oidp->oid_number;
715 (*len)++;
716
717 if (!i) {
718 if (oidpp)
719 *oidpp = oidp;
720 return (0);
721 }
722
723 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
724 break;
725
726 if (oidp->oid_handler)
727 break;
728
729 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
730 oidp = SLIST_FIRST(lsp);
731 name = p+1;
732 for (p = name; *p && *p != '.'; p++)
733 ;
734 i = *p;
735 if (i == '.')
736 *p = '\0';
737 }
738 return (ENOENT);
739}
740
741static int
742sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
743{
744 char *p;
745 int error, oid[CTL_MAXNAME], len;
746 struct sysctl_oid *op = 0;
747
748 SYSCTL_LOCK_ASSERT();
749
750 if (!req->newlen)
751 return (ENOENT);
752 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
753 return (ENAMETOOLONG);
754
755 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
756
757 error = SYSCTL_IN(req, p, req->newlen);
758 if (error) {
759 free(p, M_SYSCTL);
760 return (error);
761 }
762
763 p [req->newlen] = '\0';
764
765 error = name2oid(p, oid, &len, &op);
766
767 free(p, M_SYSCTL);
768
769 if (error)
770 return (error);
771
772 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
773 return (error);
774}
775
776SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
777 sysctl_sysctl_name2oid, "I", "");
778
779static int
780sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
781{
782 struct sysctl_oid *oid;
783 int error;
784
785 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
786 if (error)
787 return (error);
788
789 if (!oid->oid_fmt)
790 return (ENOENT);
791 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
792 if (error)
793 return (error);
794 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
795 return (error);
796}
797
798
799static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
800
801static int
802sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
803{
804 struct sysctl_oid *oid;
805 int error;
806
807 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
808 if (error)
809 return (error);
810
811 if (!oid->oid_descr)
812 return (ENOENT);
813 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
814 return (error);
815}
816
817static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
818
819/*
820 * Default "handler" functions.
821 */
822
823/*
824 * Handle an int, signed or unsigned.
825 * Two cases:
826 * a variable: point arg1 at it.
827 * a constant: pass it in arg2.
828 */
829
830int
831sysctl_handle_int(SYSCTL_HANDLER_ARGS)
832{
833 int tmpout, error = 0;
834
835 /*
836 * Attempt to get a coherent snapshot by making a copy of the data.
837 */
838 if (arg1)
839 tmpout = *(int *)arg1;
840 else
841 tmpout = arg2;
842 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
843
844 if (error || !req->newptr)
845 return (error);
846
847 if (!arg1)
848 error = EPERM;
849 else
850 error = SYSCTL_IN(req, arg1, sizeof(int));
851 return (error);
852}
853
854
855/*
856 * Based on on sysctl_handle_int() convert milliseconds into ticks.
857 */
858
859int
860sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
861{
862 int error, s, tt;
863
864 tt = *(int *)oidp->oid_arg1;
865 s = (int)((int64_t)tt * 1000 / hz);
866
867 error = sysctl_handle_int(oidp, &s, 0, req);
868 if (error || !req->newptr)
869 return (error);
870
871 tt = (int)((int64_t)s * hz / 1000);
872 if (tt < 1)
873 return (EINVAL);
874
875 *(int *)oidp->oid_arg1 = tt;
876 return (0);
877}
878
879
880/*
881 * Handle a long, signed or unsigned. arg1 points to it.
882 */
883
884int
885sysctl_handle_long(SYSCTL_HANDLER_ARGS)
886{
887 int error = 0;
888 long tmplong;
889#ifdef SCTL_MASK32
890 int tmpint;
891#endif
892
893 /*
894 * Attempt to get a coherent snapshot by making a copy of the data.
895 */
896 if (!arg1)
897 return (EINVAL);
898 tmplong = *(long *)arg1;
899#ifdef SCTL_MASK32
900 if (req->flags & SCTL_MASK32) {
901 tmpint = tmplong;
902 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
903 } else
904#endif
905 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
906
907 if (error || !req->newptr)
908 return (error);
909
910#ifdef SCTL_MASK32
911 if (req->flags & SCTL_MASK32) {
912 error = SYSCTL_IN(req, &tmpint, sizeof(int));
913 *(long *)arg1 = (long)tmpint;
914 } else
915#endif
916 error = SYSCTL_IN(req, arg1, sizeof(long));
917 return (error);
918}
919
920/*
921 * Handle a 64 bit int, signed or unsigned. arg1 points to it.
922 */
923
924int
925sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
926{
927 int error = 0;
928 uint64_t tmpout;
929
930 /*
931 * Attempt to get a coherent snapshot by making a copy of the data.
932 */
933 if (!arg1)
934 return (EINVAL);
935 tmpout = *(uint64_t *)arg1;
936 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
937
938 if (error || !req->newptr)
939 return (error);
940
941 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
942 return (error);
943}
944
945/*
946 * Handle our generic '\0' terminated 'C' string.
947 * Two cases:
948 * a variable string: point arg1 at it, arg2 is max length.
949 * a constant string: point arg1 at it, arg2 is zero.
950 */
951
952int
953sysctl_handle_string(SYSCTL_HANDLER_ARGS)
954{
955 int error=0;
956 char *tmparg;
957 size_t outlen;
958
959 /*
960 * Attempt to get a coherent snapshot by copying to a
961 * temporary kernel buffer.
962 */
963retry:
964 outlen = strlen((char *)arg1)+1;
965 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
966
967 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
968 free(tmparg, M_SYSCTLTMP);
969 goto retry;
970 }
971
972 error = SYSCTL_OUT(req, tmparg, outlen);
973 free(tmparg, M_SYSCTLTMP);
974
975 if (error || !req->newptr)
976 return (error);
977
978 if ((req->newlen - req->newidx) >= arg2) {
979 error = EINVAL;
980 } else {
981 arg2 = (req->newlen - req->newidx);
982 error = SYSCTL_IN(req, arg1, arg2);
983 ((char *)arg1)[arg2] = '\0';
984 }
985
986 return (error);
987}
988
989/*
990 * Handle any kind of opaque data.
991 * arg1 points to it, arg2 is the size.
992 */
993
994int
995sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
996{
997 int error, tries;
998 u_int generation;
999 struct sysctl_req req2;
1000
1001 /*
1002 * Attempt to get a coherent snapshot, by using the thread
1003 * pre-emption counter updated from within mi_switch() to
1004 * determine if we were pre-empted during a bcopy() or
1005 * copyout(). Make 3 attempts at doing this before giving up.
1006 * If we encounter an error, stop immediately.
1007 */
1008 tries = 0;
1009 req2 = *req;
1010retry:
1011 generation = curthread->td_generation;
1012 error = SYSCTL_OUT(req, arg1, arg2);
1013 if (error)
1014 return (error);
1015 tries++;
1016 if (generation != curthread->td_generation && tries < 3) {
1017 *req = req2;
1018 goto retry;
1019 }
1020
1021 error = SYSCTL_IN(req, arg1, arg2);
1022
1023 return (error);
1024}
1025
1026/*
1027 * Transfer functions to/from kernel space.
1028 * XXX: rather untested at this point
1029 */
1030static int
1031sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1032{
1033 size_t i = 0;
1034
1035 if (req->oldptr) {
1036 i = l;
1037 if (req->oldlen <= req->oldidx)
1038 i = 0;
1039 else
1040 if (i > req->oldlen - req->oldidx)
1041 i = req->oldlen - req->oldidx;
1042 if (i > 0)
1043 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1044 }
1045 req->oldidx += l;
1046 if (req->oldptr && i != l)
1047 return (ENOMEM);
1048 return (0);
1049}
1050
1051static int
1052sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1053{
1054 if (!req->newptr)
1055 return (0);
1056 if (req->newlen - req->newidx < l)
1057 return (EINVAL);
1058 bcopy((char *)req->newptr + req->newidx, p, l);
1059 req->newidx += l;
1060 return (0);
1061}
1062
1063int
1064kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1065 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1066{
1067 int error = 0;
1068 struct sysctl_req req;
1069
1070 bzero(&req, sizeof req);
1071
1072 req.td = td;
1073 req.flags = flags;
1074
1075 if (oldlenp) {
1076 req.oldlen = *oldlenp;
1077 }
1078 req.validlen = req.oldlen;
1079
1080 if (old) {
1081 req.oldptr= old;
1082 }
1083
1084 if (new != NULL) {
1085 req.newlen = newlen;
1086 req.newptr = new;
1087 }
1088
1089 req.oldfunc = sysctl_old_kernel;
1090 req.newfunc = sysctl_new_kernel;
1091 req.lock = REQ_LOCKED;
1092
1093 SYSCTL_LOCK();
1094 error = sysctl_root(0, name, namelen, &req);
1095 SYSCTL_UNLOCK();
1096
1097 if (req.lock == REQ_WIRED && req.validlen > 0)
1098 vsunlock(req.oldptr, req.validlen);
1099
1100 if (error && error != ENOMEM)
1101 return (error);
1102
1103 if (retval) {
1104 if (req.oldptr && req.oldidx > req.validlen)
1105 *retval = req.validlen;
1106 else
1107 *retval = req.oldidx;
1108 }
1109 return (error);
1110}
1111
1112int
1113kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1114 void *new, size_t newlen, size_t *retval, int flags)
1115{
1116 int oid[CTL_MAXNAME];
1117 size_t oidlen, plen;
1118 int error;
1119
1120 oid[0] = 0; /* sysctl internal magic */
1121 oid[1] = 3; /* name2oid */
1122 oidlen = sizeof(oid);
1123
1124 /*
1125 * XXX: Prone to a possible race condition between lookup and
1126 * execution? Maybe put locking around it?
1127 */
1128
1129 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1130 (void *)name, strlen(name), &plen, flags);
1131 if (error)
1132 return (error);
1133
1134 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1135 new, newlen, retval, flags);
1136 return (error);
1137}
1138
1139/*
1140 * Transfer function to/from user space.
1141 */
1142static int
1143sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1144{
1145 int error = 0;
1146 size_t i, len, origidx;
1147
1148 origidx = req->oldidx;
1149 req->oldidx += l;
1150 if (req->oldptr == NULL)
1151 return (0);
1152 /*
1153 * If we have not wired the user supplied buffer and we are currently
1154 * holding locks, drop a witness warning, as it's possible that
1155 * write operations to the user page can sleep.
1156 */
1157 if (req->lock != REQ_WIRED)
1158 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1159 "sysctl_old_user()");
1160 i = l;
1161 len = req->validlen;
1162 if (len <= origidx)
1163 i = 0;
1164 else {
1165 if (i > len - origidx)
1166 i = len - origidx;
1167 error = copyout(p, (char *)req->oldptr + origidx, i);
1168 }
1169 if (error)
1170 return (error);
1171 if (i < l)
1172 return (ENOMEM);
1173 return (0);
1174}
1175
1176static int
1177sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1178{
1179 int error;
1180
1181 if (!req->newptr)
1182 return (0);
1183 if (req->newlen - req->newidx < l)
1184 return (EINVAL);
1185 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1186 "sysctl_new_user()");
1187 error = copyin((char *)req->newptr + req->newidx, p, l);
1188 req->newidx += l;
1189 return (error);
1190}
1191
1192/*
1193 * Wire the user space destination buffer. If set to a value greater than
1194 * zero, the len parameter limits the maximum amount of wired memory.
1195 */
1196int
1197sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1198{
1199 int ret;
1200 size_t i, wiredlen;
1201 char *cp, dummy;
1202
1203 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1204 ret = 0;
1205 if (req->lock == REQ_LOCKED && req->oldptr &&
1206 req->oldfunc == sysctl_old_user) {
1207 if (wiredlen != 0) {
1208 ret = vslock(req->oldptr, wiredlen);
1209 if (ret != 0) {
1210 if (ret != ENOMEM)
1211 return (ret);
1212 wiredlen = 0;
1213 }
1214 /*
1215 * Touch all the wired pages to avoid PTE modified
1216 * bit emulation traps on Alpha while holding locks
1217 * in the sysctl handler.
1218 */
1219 for (i = (wiredlen + PAGE_SIZE - 1) / PAGE_SIZE,
1220 cp = req->oldptr; i > 0; i--, cp += PAGE_SIZE) {
1221 copyin(cp, &dummy, 1);
1222 copyout(&dummy, cp, 1);
1223 }
1224 }
1225 req->lock = REQ_WIRED;
1226 req->validlen = wiredlen;
1227 }
1228 return (0);
1229}
1230
1231int
1232sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1233 int *nindx, struct sysctl_req *req)
1234{
1235 struct sysctl_oid *oid;
1236 int indx;
1237
1238 oid = SLIST_FIRST(&sysctl__children);
1239 indx = 0;
1240 while (oid && indx < CTL_MAXNAME) {
1241 if (oid->oid_number == name[indx]) {
1242 indx++;
1243 if (oid->oid_kind & CTLFLAG_NOLOCK)
1244 req->lock = REQ_UNLOCKED;
1245 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1246 if (oid->oid_handler != NULL ||
1247 indx == namelen) {
1248 *noid = oid;
1249 if (nindx != NULL)
1250 *nindx = indx;
1251 return (0);
1252 }
1253 oid = SLIST_FIRST(
1254 (struct sysctl_oid_list *)oid->oid_arg1);
1255 } else if (indx == namelen) {
1256 *noid = oid;
1257 if (nindx != NULL)
1258 *nindx = indx;
1259 return (0);
1260 } else {
1261 return (ENOTDIR);
1262 }
1263 } else {
1264 oid = SLIST_NEXT(oid, oid_link);
1265 }
1266 }
1267 return (ENOENT);
1268}
1269
1270/*
1271 * Traverse our tree, and find the right node, execute whatever it points
1272 * to, and return the resulting error code.
1273 */
1274
1275static int
1276sysctl_root(SYSCTL_HANDLER_ARGS)
1277{
1278 struct sysctl_oid *oid;
1279 int error, indx, lvl;
1280
1281 SYSCTL_LOCK_ASSERT();
1282
1283 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1284 if (error)
1285 return (error);
1286
1287 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1288 /*
1289 * You can't call a sysctl when it's a node, but has
1290 * no handler. Inform the user that it's a node.
1291 * The indx may or may not be the same as namelen.
1292 */
1293 if (oid->oid_handler == NULL)
1294 return (EISDIR);
1295 }
1296
1297 /* Is this sysctl writable? */
1298 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1299 return (EPERM);
1300
1301 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1302
1303 /* Is this sysctl sensitive to securelevels? */
1304 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1305 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1306 error = securelevel_gt(req->td->td_ucred, lvl);
1307 if (error)
1308 return (error);
1309 }
1310
1311 /* Is this sysctl writable by only privileged users? */
1312 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1313 if (oid->oid_kind & CTLFLAG_PRISON)
1314 error = priv_check(req->td, PRIV_SYSCTL_WRITEJAIL);
1315 else
1316 error = priv_check(req->td, PRIV_SYSCTL_WRITE);
1317 if (error)
1318 return (error);
1319 }
1320
1321 if (!oid->oid_handler)
1322 return (EINVAL);
1323
1324 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1325 arg1 = (int *)arg1 + indx;
1326 arg2 -= indx;
1327 } else {
1328 arg1 = oid->oid_arg1;
1329 arg2 = oid->oid_arg2;
1330 }
1331#ifdef MAC
1332 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1333 req);
1334 if (error != 0)
1335 return (error);
1336#endif
1337 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1338 mtx_lock(&Giant);
1339 error = oid->oid_handler(oid, arg1, arg2, req);
1340 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1341 mtx_unlock(&Giant);
1342
1343 return (error);
1344}
1345
1346#ifndef _SYS_SYSPROTO_H_
1347struct sysctl_args {
1348 int *name;
1349 u_int namelen;
1350 void *old;
1351 size_t *oldlenp;
1352 void *new;
1353 size_t newlen;
1354};
1355#endif
1356int
1357__sysctl(struct thread *td, struct sysctl_args *uap)
1358{
1359 int error, name[CTL_MAXNAME];
1360 size_t j;
1361
1362 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1363 return (EINVAL);
1364
1365 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1366 if (error)
1367 return (error);
1368
1369 error = userland_sysctl(td, name, uap->namelen,
1370 uap->old, uap->oldlenp, 0,
1371 uap->new, uap->newlen, &j, 0);
1372 if (error && error != ENOMEM)
1373 return (error);
1374 if (uap->oldlenp) {
1375 int i = copyout(&j, uap->oldlenp, sizeof(j));
1376 if (i)
1377 return (i);
1378 }
1379 return (error);
1380}
1381
1382/*
1383 * This is used from various compatibility syscalls too. That's why name
1384 * must be in kernel space.
1385 */
1386int
1387userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1388 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1389 int flags)
1390{
1391 int error = 0;
1392 struct sysctl_req req;
1393
1394 bzero(&req, sizeof req);
1395
1396 req.td = td;
1397 req.flags = flags;
1398
1399 if (oldlenp) {
1400 if (inkernel) {
1401 req.oldlen = *oldlenp;
1402 } else {
1403 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1404 if (error)
1405 return (error);
1406 }
1407 }
1408 req.validlen = req.oldlen;
1409
1410 if (old) {
1411 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1412 return (EFAULT);
1413 req.oldptr= old;
1414 }
1415
1416 if (new != NULL) {
1417 if (!useracc(new, newlen, VM_PROT_READ))
1418 return (EFAULT);
1419 req.newlen = newlen;
1420 req.newptr = new;
1421 }
1422
1423 req.oldfunc = sysctl_old_user;
1424 req.newfunc = sysctl_new_user;
1425 req.lock = REQ_LOCKED;
1426
1427 SYSCTL_LOCK();
1428 CURVNET_SET(TD_TO_VNET(curthread));
1429
1430 for (;;) {
1431 req.oldidx = 0;
1432 req.newidx = 0;
1433 error = sysctl_root(0, name, namelen, &req);
1434 if (error != EAGAIN)
1435 break;
1436 uio_yield();
1437 }
1438
1439 CURVNET_RESTORE();
1440 SYSCTL_UNLOCK();
1441
1442 if (req.lock == REQ_WIRED && req.validlen > 0)
1443 vsunlock(req.oldptr, req.validlen);
1444
1445 if (error && error != ENOMEM)
1446 return (error);
1447
1448 if (retval) {
1449 if (req.oldptr && req.oldidx > req.validlen)
1450 *retval = req.validlen;
1451 else
1452 *retval = req.oldidx;
1453 }
1454 return (error);
1455}
2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Mike Karels at Berkeley Software Design, Inc.
7 *
8 * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
9 * project, to make these variables more userfriendly.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
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.
22 *
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
33 * SUCH DAMAGE.
34 *
35 * @(#)kern_sysctl.c 8.4 (Berkeley) 4/14/94
36 */
37
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: head/sys/kern/kern_sysctl.c 187656 2009-01-23 22:40:35Z jhb $");
40
41#include "opt_compat.h"
42#include "opt_mac.h"
43
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>
49#include <sys/priv.h>
50#include <sys/proc.h>
51#include <sys/lock.h>
52#include <sys/mutex.h>
53#include <sys/sx.h>
54#include <sys/sysproto.h>
55#include <sys/uio.h>
56#include <sys/vimage.h>
57
58#include <security/mac/mac_framework.h>
59
60#include <vm/vm.h>
61#include <vm/vm_extern.h>
62
63static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
64static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
65static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
66
67/*
68 * Locking - this locks the sysctl tree in memory.
69 */
70static struct sx sysctllock;
71
72#define SYSCTL_LOCK() sx_xlock(&sysctllock)
73#define SYSCTL_UNLOCK() sx_xunlock(&sysctllock)
74#define SYSCTL_LOCK_ASSERT() sx_assert(&sysctllock, SX_XLOCKED)
75#define SYSCTL_INIT() sx_init(&sysctllock, "sysctl lock")
76
77static int sysctl_root(SYSCTL_HANDLER_ARGS);
78
79struct sysctl_oid_list sysctl__children; /* root list */
80
81static struct sysctl_oid *
82sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
83{
84 struct sysctl_oid *oidp;
85
86 SLIST_FOREACH(oidp, list, oid_link) {
87 if (strcmp(oidp->oid_name, name) == 0) {
88 return (oidp);
89 }
90 }
91 return (NULL);
92}
93
94/*
95 * Initialization of the MIB tree.
96 *
97 * Order by number in each list.
98 */
99
100void
101sysctl_register_oid(struct sysctl_oid *oidp)
102{
103 struct sysctl_oid_list *parent = oidp->oid_parent;
104 struct sysctl_oid *p;
105 struct sysctl_oid *q;
106
107 /*
108 * First check if another oid with the same name already
109 * exists in the parent's list.
110 */
111 p = sysctl_find_oidname(oidp->oid_name, parent);
112 if (p != NULL) {
113 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
114 p->oid_refcnt++;
115 return;
116 } else {
117 printf("can't re-use a leaf (%s)!\n", p->oid_name);
118 return;
119 }
120 }
121 /*
122 * If this oid has a number OID_AUTO, give it a number which
123 * is greater than any current oid.
124 * NOTE: DO NOT change the starting value here, change it in
125 * <sys/sysctl.h>, and make sure it is at least 256 to
126 * accomodate e.g. net.inet.raw as a static sysctl node.
127 */
128 if (oidp->oid_number == OID_AUTO) {
129 static int newoid = CTL_AUTO_START;
130
131 oidp->oid_number = newoid++;
132 if (newoid == 0x7fffffff)
133 panic("out of oids");
134 }
135#if 0
136 else if (oidp->oid_number >= CTL_AUTO_START) {
137 /* do not panic; this happens when unregistering sysctl sets */
138 printf("static sysctl oid too high: %d", oidp->oid_number);
139 }
140#endif
141
142 /*
143 * Insert the oid into the parent's list in order.
144 */
145 q = NULL;
146 SLIST_FOREACH(p, parent, oid_link) {
147 if (oidp->oid_number < p->oid_number)
148 break;
149 q = p;
150 }
151 if (q)
152 SLIST_INSERT_AFTER(q, oidp, oid_link);
153 else
154 SLIST_INSERT_HEAD(parent, oidp, oid_link);
155}
156
157void
158sysctl_unregister_oid(struct sysctl_oid *oidp)
159{
160 struct sysctl_oid *p;
161 int error;
162
163 error = ENOENT;
164 if (oidp->oid_number == OID_AUTO) {
165 error = EINVAL;
166 } else {
167 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
168 if (p == oidp) {
169 SLIST_REMOVE(oidp->oid_parent, oidp,
170 sysctl_oid, oid_link);
171 error = 0;
172 break;
173 }
174 }
175 }
176
177 /*
178 * This can happen when a module fails to register and is
179 * being unloaded afterwards. It should not be a panic()
180 * for normal use.
181 */
182 if (error)
183 printf("%s: failed to unregister sysctl\n", __func__);
184}
185
186/* Initialize a new context to keep track of dynamically added sysctls. */
187int
188sysctl_ctx_init(struct sysctl_ctx_list *c)
189{
190
191 if (c == NULL) {
192 return (EINVAL);
193 }
194 TAILQ_INIT(c);
195 return (0);
196}
197
198/* Free the context, and destroy all dynamic oids registered in this context */
199int
200sysctl_ctx_free(struct sysctl_ctx_list *clist)
201{
202 struct sysctl_ctx_entry *e, *e1;
203 int error;
204
205 error = 0;
206 /*
207 * First perform a "dry run" to check if it's ok to remove oids.
208 * XXX FIXME
209 * XXX This algorithm is a hack. But I don't know any
210 * XXX better solution for now...
211 */
212 TAILQ_FOREACH(e, clist, link) {
213 error = sysctl_remove_oid(e->entry, 0, 0);
214 if (error)
215 break;
216 }
217 /*
218 * Restore deregistered entries, either from the end,
219 * or from the place where error occured.
220 * e contains the entry that was not unregistered
221 */
222 if (error)
223 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
224 else
225 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
226 while (e1 != NULL) {
227 sysctl_register_oid(e1->entry);
228 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
229 }
230 if (error)
231 return(EBUSY);
232 /* Now really delete the entries */
233 e = TAILQ_FIRST(clist);
234 while (e != NULL) {
235 e1 = TAILQ_NEXT(e, link);
236 error = sysctl_remove_oid(e->entry, 1, 0);
237 if (error)
238 panic("sysctl_remove_oid: corrupt tree, entry: %s",
239 e->entry->oid_name);
240 free(e, M_SYSCTLOID);
241 e = e1;
242 }
243 return (error);
244}
245
246/* Add an entry to the context */
247struct sysctl_ctx_entry *
248sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
249{
250 struct sysctl_ctx_entry *e;
251
252 if (clist == NULL || oidp == NULL)
253 return(NULL);
254 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
255 e->entry = oidp;
256 TAILQ_INSERT_HEAD(clist, e, link);
257 return (e);
258}
259
260/* Find an entry in the context */
261struct sysctl_ctx_entry *
262sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
263{
264 struct sysctl_ctx_entry *e;
265
266 if (clist == NULL || oidp == NULL)
267 return(NULL);
268 TAILQ_FOREACH(e, clist, link) {
269 if(e->entry == oidp)
270 return(e);
271 }
272 return (e);
273}
274
275/*
276 * Delete an entry from the context.
277 * NOTE: this function doesn't free oidp! You have to remove it
278 * with sysctl_remove_oid().
279 */
280int
281sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
282{
283 struct sysctl_ctx_entry *e;
284
285 if (clist == NULL || oidp == NULL)
286 return (EINVAL);
287 e = sysctl_ctx_entry_find(clist, oidp);
288 if (e != NULL) {
289 TAILQ_REMOVE(clist, e, link);
290 free(e, M_SYSCTLOID);
291 return (0);
292 } else
293 return (ENOENT);
294}
295
296/*
297 * Remove dynamically created sysctl trees.
298 * oidp - top of the tree to be removed
299 * del - if 0 - just deregister, otherwise free up entries as well
300 * recurse - if != 0 traverse the subtree to be deleted
301 */
302int
303sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
304{
305 struct sysctl_oid *p;
306 int error;
307
308 if (oidp == NULL)
309 return(EINVAL);
310 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
311 printf("can't remove non-dynamic nodes!\n");
312 return (EINVAL);
313 }
314 /*
315 * WARNING: normal method to do this should be through
316 * sysctl_ctx_free(). Use recursing as the last resort
317 * method to purge your sysctl tree of leftovers...
318 * However, if some other code still references these nodes,
319 * it will panic.
320 */
321 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
322 if (oidp->oid_refcnt == 1) {
323 SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
324 if (!recurse)
325 return (ENOTEMPTY);
326 error = sysctl_remove_oid(p, del, recurse);
327 if (error)
328 return (error);
329 }
330 if (del)
331 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
332 }
333 }
334 if (oidp->oid_refcnt > 1 ) {
335 oidp->oid_refcnt--;
336 } else {
337 if (oidp->oid_refcnt == 0) {
338 printf("Warning: bad oid_refcnt=%u (%s)!\n",
339 oidp->oid_refcnt, oidp->oid_name);
340 return (EINVAL);
341 }
342 sysctl_unregister_oid(oidp);
343 if (del) {
344 if (oidp->oid_descr)
345 free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
346 free((void *)(uintptr_t)(const void *)oidp->oid_name,
347 M_SYSCTLOID);
348 free(oidp, M_SYSCTLOID);
349 }
350 }
351 return (0);
352}
353
354/*
355 * Create new sysctls at run time.
356 * clist may point to a valid context initialized with sysctl_ctx_init().
357 */
358struct sysctl_oid *
359sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
360 int number, const char *name, int kind, void *arg1, int arg2,
361 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
362{
363 struct sysctl_oid *oidp;
364 ssize_t len;
365 char *newname;
366
367 /* You have to hook up somewhere.. */
368 if (parent == NULL)
369 return(NULL);
370 /* Check if the node already exists, otherwise create it */
371 oidp = sysctl_find_oidname(name, parent);
372 if (oidp != NULL) {
373 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
374 oidp->oid_refcnt++;
375 /* Update the context */
376 if (clist != NULL)
377 sysctl_ctx_entry_add(clist, oidp);
378 return (oidp);
379 } else {
380 printf("can't re-use a leaf (%s)!\n", name);
381 return (NULL);
382 }
383 }
384 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
385 oidp->oid_parent = parent;
386 SLIST_NEXT(oidp, oid_link) = NULL;
387 oidp->oid_number = number;
388 oidp->oid_refcnt = 1;
389 len = strlen(name);
390 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
391 bcopy(name, newname, len + 1);
392 newname[len] = '\0';
393 oidp->oid_name = newname;
394 oidp->oid_handler = handler;
395 oidp->oid_kind = CTLFLAG_DYN | kind;
396 if ((kind & CTLTYPE) == CTLTYPE_NODE) {
397 /* Allocate space for children */
398 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
399 M_SYSCTLOID, M_WAITOK));
400 SLIST_INIT(SYSCTL_CHILDREN(oidp));
401 } else {
402 oidp->oid_arg1 = arg1;
403 oidp->oid_arg2 = arg2;
404 }
405 oidp->oid_fmt = fmt;
406 if (descr) {
407 int len = strlen(descr) + 1;
408 oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
409 if (oidp->oid_descr)
410 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
411 }
412 /* Update the context, if used */
413 if (clist != NULL)
414 sysctl_ctx_entry_add(clist, oidp);
415 /* Register this oid */
416 sysctl_register_oid(oidp);
417 return (oidp);
418}
419
420/*
421 * Rename an existing oid.
422 */
423void
424sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
425{
426 ssize_t len;
427 char *newname;
428 void *oldname;
429
430 oldname = (void *)(uintptr_t)(const void *)oidp->oid_name;
431 len = strlen(name);
432 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
433 bcopy(name, newname, len + 1);
434 newname[len] = '\0';
435 oidp->oid_name = newname;
436 free(oldname, M_SYSCTLOID);
437}
438
439/*
440 * Reparent an existing oid.
441 */
442int
443sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
444{
445 struct sysctl_oid *oidp;
446
447 if (oid->oid_parent == parent)
448 return (0);
449 oidp = sysctl_find_oidname(oid->oid_name, parent);
450 if (oidp != NULL)
451 return (EEXIST);
452 sysctl_unregister_oid(oid);
453 oid->oid_parent = parent;
454 oid->oid_number = OID_AUTO;
455 sysctl_register_oid(oid);
456 return (0);
457}
458
459/*
460 * Register the kernel's oids on startup.
461 */
462SET_DECLARE(sysctl_set, struct sysctl_oid);
463
464static void
465sysctl_register_all(void *arg)
466{
467 struct sysctl_oid **oidp;
468
469 SYSCTL_INIT();
470 SET_FOREACH(oidp, sysctl_set)
471 sysctl_register_oid(*oidp);
472}
473SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
474
475/*
476 * "Staff-functions"
477 *
478 * These functions implement a presently undocumented interface
479 * used by the sysctl program to walk the tree, and get the type
480 * so it can print the value.
481 * This interface is under work and consideration, and should probably
482 * be killed with a big axe by the first person who can find the time.
483 * (be aware though, that the proper interface isn't as obvious as it
484 * may seem, there are various conflicting requirements.
485 *
486 * {0,0} printf the entire MIB-tree.
487 * {0,1,...} return the name of the "..." OID.
488 * {0,2,...} return the next OID.
489 * {0,3} return the OID of the name in "new"
490 * {0,4,...} return the kind & format info for the "..." OID.
491 * {0,5,...} return the description the "..." OID.
492 */
493
494#ifdef SYSCTL_DEBUG
495static void
496sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
497{
498 int k;
499 struct sysctl_oid *oidp;
500
501 SLIST_FOREACH(oidp, l, oid_link) {
502
503 for (k=0; k<i; k++)
504 printf(" ");
505
506 printf("%d %s ", oidp->oid_number, oidp->oid_name);
507
508 printf("%c%c",
509 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
510 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
511
512 if (oidp->oid_handler)
513 printf(" *Handler");
514
515 switch (oidp->oid_kind & CTLTYPE) {
516 case CTLTYPE_NODE:
517 printf(" Node\n");
518 if (!oidp->oid_handler) {
519 sysctl_sysctl_debug_dump_node(
520 oidp->oid_arg1, i+2);
521 }
522 break;
523 case CTLTYPE_INT: printf(" Int\n"); break;
524 case CTLTYPE_STRING: printf(" String\n"); break;
525 case CTLTYPE_QUAD: printf(" Quad\n"); break;
526 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
527 default: printf("\n");
528 }
529
530 }
531}
532
533static int
534sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
535{
536 int error;
537
538 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
539 if (error)
540 return (error);
541 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
542 return (ENOENT);
543}
544
545SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
546 0, 0, sysctl_sysctl_debug, "-", "");
547#endif
548
549static int
550sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
551{
552 int *name = (int *) arg1;
553 u_int namelen = arg2;
554 int error = 0;
555 struct sysctl_oid *oid;
556 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
557 char buf[10];
558
559 while (namelen) {
560 if (!lsp) {
561 snprintf(buf,sizeof(buf),"%d",*name);
562 if (req->oldidx)
563 error = SYSCTL_OUT(req, ".", 1);
564 if (!error)
565 error = SYSCTL_OUT(req, buf, strlen(buf));
566 if (error)
567 return (error);
568 namelen--;
569 name++;
570 continue;
571 }
572 lsp2 = 0;
573 SLIST_FOREACH(oid, lsp, oid_link) {
574 if (oid->oid_number != *name)
575 continue;
576
577 if (req->oldidx)
578 error = SYSCTL_OUT(req, ".", 1);
579 if (!error)
580 error = SYSCTL_OUT(req, oid->oid_name,
581 strlen(oid->oid_name));
582 if (error)
583 return (error);
584
585 namelen--;
586 name++;
587
588 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
589 break;
590
591 if (oid->oid_handler)
592 break;
593
594 lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
595 break;
596 }
597 lsp = lsp2;
598 }
599 return (SYSCTL_OUT(req, "", 1));
600}
601
602static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
603
604static int
605sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
606 int *next, int *len, int level, struct sysctl_oid **oidpp)
607{
608 struct sysctl_oid *oidp;
609
610 *len = level;
611 SLIST_FOREACH(oidp, lsp, oid_link) {
612 *next = oidp->oid_number;
613 *oidpp = oidp;
614
615 if (oidp->oid_kind & CTLFLAG_SKIP)
616 continue;
617
618 if (!namelen) {
619 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
620 return (0);
621 if (oidp->oid_handler)
622 /* We really should call the handler here...*/
623 return (0);
624 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
625 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
626 len, level+1, oidpp))
627 return (0);
628 goto emptynode;
629 }
630
631 if (oidp->oid_number < *name)
632 continue;
633
634 if (oidp->oid_number > *name) {
635 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
636 return (0);
637 if (oidp->oid_handler)
638 return (0);
639 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
640 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
641 next+1, len, level+1, oidpp))
642 return (0);
643 goto next;
644 }
645 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
646 continue;
647
648 if (oidp->oid_handler)
649 continue;
650
651 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
652 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
653 len, level+1, oidpp))
654 return (0);
655 next:
656 namelen = 1;
657 emptynode:
658 *len = level;
659 }
660 return (1);
661}
662
663static int
664sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
665{
666 int *name = (int *) arg1;
667 u_int namelen = arg2;
668 int i, j, error;
669 struct sysctl_oid *oid;
670 struct sysctl_oid_list *lsp = &sysctl__children;
671 int newoid[CTL_MAXNAME];
672
673 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
674 if (i)
675 return (ENOENT);
676 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
677 return (error);
678}
679
680static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
681
682static int
683name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
684{
685 int i;
686 struct sysctl_oid *oidp;
687 struct sysctl_oid_list *lsp = &sysctl__children;
688 char *p;
689
690 SYSCTL_LOCK_ASSERT();
691
692 if (!*name)
693 return (ENOENT);
694
695 p = name + strlen(name) - 1 ;
696 if (*p == '.')
697 *p = '\0';
698
699 *len = 0;
700
701 for (p = name; *p && *p != '.'; p++)
702 ;
703 i = *p;
704 if (i == '.')
705 *p = '\0';
706
707 oidp = SLIST_FIRST(lsp);
708
709 while (oidp && *len < CTL_MAXNAME) {
710 if (strcmp(name, oidp->oid_name)) {
711 oidp = SLIST_NEXT(oidp, oid_link);
712 continue;
713 }
714 *oid++ = oidp->oid_number;
715 (*len)++;
716
717 if (!i) {
718 if (oidpp)
719 *oidpp = oidp;
720 return (0);
721 }
722
723 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
724 break;
725
726 if (oidp->oid_handler)
727 break;
728
729 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
730 oidp = SLIST_FIRST(lsp);
731 name = p+1;
732 for (p = name; *p && *p != '.'; p++)
733 ;
734 i = *p;
735 if (i == '.')
736 *p = '\0';
737 }
738 return (ENOENT);
739}
740
741static int
742sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
743{
744 char *p;
745 int error, oid[CTL_MAXNAME], len;
746 struct sysctl_oid *op = 0;
747
748 SYSCTL_LOCK_ASSERT();
749
750 if (!req->newlen)
751 return (ENOENT);
752 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
753 return (ENAMETOOLONG);
754
755 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
756
757 error = SYSCTL_IN(req, p, req->newlen);
758 if (error) {
759 free(p, M_SYSCTL);
760 return (error);
761 }
762
763 p [req->newlen] = '\0';
764
765 error = name2oid(p, oid, &len, &op);
766
767 free(p, M_SYSCTL);
768
769 if (error)
770 return (error);
771
772 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
773 return (error);
774}
775
776SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
777 sysctl_sysctl_name2oid, "I", "");
778
779static int
780sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
781{
782 struct sysctl_oid *oid;
783 int error;
784
785 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
786 if (error)
787 return (error);
788
789 if (!oid->oid_fmt)
790 return (ENOENT);
791 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
792 if (error)
793 return (error);
794 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
795 return (error);
796}
797
798
799static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
800
801static int
802sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
803{
804 struct sysctl_oid *oid;
805 int error;
806
807 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
808 if (error)
809 return (error);
810
811 if (!oid->oid_descr)
812 return (ENOENT);
813 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
814 return (error);
815}
816
817static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
818
819/*
820 * Default "handler" functions.
821 */
822
823/*
824 * Handle an int, signed or unsigned.
825 * Two cases:
826 * a variable: point arg1 at it.
827 * a constant: pass it in arg2.
828 */
829
830int
831sysctl_handle_int(SYSCTL_HANDLER_ARGS)
832{
833 int tmpout, error = 0;
834
835 /*
836 * Attempt to get a coherent snapshot by making a copy of the data.
837 */
838 if (arg1)
839 tmpout = *(int *)arg1;
840 else
841 tmpout = arg2;
842 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
843
844 if (error || !req->newptr)
845 return (error);
846
847 if (!arg1)
848 error = EPERM;
849 else
850 error = SYSCTL_IN(req, arg1, sizeof(int));
851 return (error);
852}
853
854
855/*
856 * Based on on sysctl_handle_int() convert milliseconds into ticks.
857 */
858
859int
860sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
861{
862 int error, s, tt;
863
864 tt = *(int *)oidp->oid_arg1;
865 s = (int)((int64_t)tt * 1000 / hz);
866
867 error = sysctl_handle_int(oidp, &s, 0, req);
868 if (error || !req->newptr)
869 return (error);
870
871 tt = (int)((int64_t)s * hz / 1000);
872 if (tt < 1)
873 return (EINVAL);
874
875 *(int *)oidp->oid_arg1 = tt;
876 return (0);
877}
878
879
880/*
881 * Handle a long, signed or unsigned. arg1 points to it.
882 */
883
884int
885sysctl_handle_long(SYSCTL_HANDLER_ARGS)
886{
887 int error = 0;
888 long tmplong;
889#ifdef SCTL_MASK32
890 int tmpint;
891#endif
892
893 /*
894 * Attempt to get a coherent snapshot by making a copy of the data.
895 */
896 if (!arg1)
897 return (EINVAL);
898 tmplong = *(long *)arg1;
899#ifdef SCTL_MASK32
900 if (req->flags & SCTL_MASK32) {
901 tmpint = tmplong;
902 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
903 } else
904#endif
905 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
906
907 if (error || !req->newptr)
908 return (error);
909
910#ifdef SCTL_MASK32
911 if (req->flags & SCTL_MASK32) {
912 error = SYSCTL_IN(req, &tmpint, sizeof(int));
913 *(long *)arg1 = (long)tmpint;
914 } else
915#endif
916 error = SYSCTL_IN(req, arg1, sizeof(long));
917 return (error);
918}
919
920/*
921 * Handle a 64 bit int, signed or unsigned. arg1 points to it.
922 */
923
924int
925sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
926{
927 int error = 0;
928 uint64_t tmpout;
929
930 /*
931 * Attempt to get a coherent snapshot by making a copy of the data.
932 */
933 if (!arg1)
934 return (EINVAL);
935 tmpout = *(uint64_t *)arg1;
936 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
937
938 if (error || !req->newptr)
939 return (error);
940
941 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
942 return (error);
943}
944
945/*
946 * Handle our generic '\0' terminated 'C' string.
947 * Two cases:
948 * a variable string: point arg1 at it, arg2 is max length.
949 * a constant string: point arg1 at it, arg2 is zero.
950 */
951
952int
953sysctl_handle_string(SYSCTL_HANDLER_ARGS)
954{
955 int error=0;
956 char *tmparg;
957 size_t outlen;
958
959 /*
960 * Attempt to get a coherent snapshot by copying to a
961 * temporary kernel buffer.
962 */
963retry:
964 outlen = strlen((char *)arg1)+1;
965 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
966
967 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
968 free(tmparg, M_SYSCTLTMP);
969 goto retry;
970 }
971
972 error = SYSCTL_OUT(req, tmparg, outlen);
973 free(tmparg, M_SYSCTLTMP);
974
975 if (error || !req->newptr)
976 return (error);
977
978 if ((req->newlen - req->newidx) >= arg2) {
979 error = EINVAL;
980 } else {
981 arg2 = (req->newlen - req->newidx);
982 error = SYSCTL_IN(req, arg1, arg2);
983 ((char *)arg1)[arg2] = '\0';
984 }
985
986 return (error);
987}
988
989/*
990 * Handle any kind of opaque data.
991 * arg1 points to it, arg2 is the size.
992 */
993
994int
995sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
996{
997 int error, tries;
998 u_int generation;
999 struct sysctl_req req2;
1000
1001 /*
1002 * Attempt to get a coherent snapshot, by using the thread
1003 * pre-emption counter updated from within mi_switch() to
1004 * determine if we were pre-empted during a bcopy() or
1005 * copyout(). Make 3 attempts at doing this before giving up.
1006 * If we encounter an error, stop immediately.
1007 */
1008 tries = 0;
1009 req2 = *req;
1010retry:
1011 generation = curthread->td_generation;
1012 error = SYSCTL_OUT(req, arg1, arg2);
1013 if (error)
1014 return (error);
1015 tries++;
1016 if (generation != curthread->td_generation && tries < 3) {
1017 *req = req2;
1018 goto retry;
1019 }
1020
1021 error = SYSCTL_IN(req, arg1, arg2);
1022
1023 return (error);
1024}
1025
1026/*
1027 * Transfer functions to/from kernel space.
1028 * XXX: rather untested at this point
1029 */
1030static int
1031sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1032{
1033 size_t i = 0;
1034
1035 if (req->oldptr) {
1036 i = l;
1037 if (req->oldlen <= req->oldidx)
1038 i = 0;
1039 else
1040 if (i > req->oldlen - req->oldidx)
1041 i = req->oldlen - req->oldidx;
1042 if (i > 0)
1043 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1044 }
1045 req->oldidx += l;
1046 if (req->oldptr && i != l)
1047 return (ENOMEM);
1048 return (0);
1049}
1050
1051static int
1052sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1053{
1054 if (!req->newptr)
1055 return (0);
1056 if (req->newlen - req->newidx < l)
1057 return (EINVAL);
1058 bcopy((char *)req->newptr + req->newidx, p, l);
1059 req->newidx += l;
1060 return (0);
1061}
1062
1063int
1064kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1065 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1066{
1067 int error = 0;
1068 struct sysctl_req req;
1069
1070 bzero(&req, sizeof req);
1071
1072 req.td = td;
1073 req.flags = flags;
1074
1075 if (oldlenp) {
1076 req.oldlen = *oldlenp;
1077 }
1078 req.validlen = req.oldlen;
1079
1080 if (old) {
1081 req.oldptr= old;
1082 }
1083
1084 if (new != NULL) {
1085 req.newlen = newlen;
1086 req.newptr = new;
1087 }
1088
1089 req.oldfunc = sysctl_old_kernel;
1090 req.newfunc = sysctl_new_kernel;
1091 req.lock = REQ_LOCKED;
1092
1093 SYSCTL_LOCK();
1094 error = sysctl_root(0, name, namelen, &req);
1095 SYSCTL_UNLOCK();
1096
1097 if (req.lock == REQ_WIRED && req.validlen > 0)
1098 vsunlock(req.oldptr, req.validlen);
1099
1100 if (error && error != ENOMEM)
1101 return (error);
1102
1103 if (retval) {
1104 if (req.oldptr && req.oldidx > req.validlen)
1105 *retval = req.validlen;
1106 else
1107 *retval = req.oldidx;
1108 }
1109 return (error);
1110}
1111
1112int
1113kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1114 void *new, size_t newlen, size_t *retval, int flags)
1115{
1116 int oid[CTL_MAXNAME];
1117 size_t oidlen, plen;
1118 int error;
1119
1120 oid[0] = 0; /* sysctl internal magic */
1121 oid[1] = 3; /* name2oid */
1122 oidlen = sizeof(oid);
1123
1124 /*
1125 * XXX: Prone to a possible race condition between lookup and
1126 * execution? Maybe put locking around it?
1127 */
1128
1129 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1130 (void *)name, strlen(name), &plen, flags);
1131 if (error)
1132 return (error);
1133
1134 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1135 new, newlen, retval, flags);
1136 return (error);
1137}
1138
1139/*
1140 * Transfer function to/from user space.
1141 */
1142static int
1143sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1144{
1145 int error = 0;
1146 size_t i, len, origidx;
1147
1148 origidx = req->oldidx;
1149 req->oldidx += l;
1150 if (req->oldptr == NULL)
1151 return (0);
1152 /*
1153 * If we have not wired the user supplied buffer and we are currently
1154 * holding locks, drop a witness warning, as it's possible that
1155 * write operations to the user page can sleep.
1156 */
1157 if (req->lock != REQ_WIRED)
1158 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1159 "sysctl_old_user()");
1160 i = l;
1161 len = req->validlen;
1162 if (len <= origidx)
1163 i = 0;
1164 else {
1165 if (i > len - origidx)
1166 i = len - origidx;
1167 error = copyout(p, (char *)req->oldptr + origidx, i);
1168 }
1169 if (error)
1170 return (error);
1171 if (i < l)
1172 return (ENOMEM);
1173 return (0);
1174}
1175
1176static int
1177sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1178{
1179 int error;
1180
1181 if (!req->newptr)
1182 return (0);
1183 if (req->newlen - req->newidx < l)
1184 return (EINVAL);
1185 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1186 "sysctl_new_user()");
1187 error = copyin((char *)req->newptr + req->newidx, p, l);
1188 req->newidx += l;
1189 return (error);
1190}
1191
1192/*
1193 * Wire the user space destination buffer. If set to a value greater than
1194 * zero, the len parameter limits the maximum amount of wired memory.
1195 */
1196int
1197sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1198{
1199 int ret;
1200 size_t i, wiredlen;
1201 char *cp, dummy;
1202
1203 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1204 ret = 0;
1205 if (req->lock == REQ_LOCKED && req->oldptr &&
1206 req->oldfunc == sysctl_old_user) {
1207 if (wiredlen != 0) {
1208 ret = vslock(req->oldptr, wiredlen);
1209 if (ret != 0) {
1210 if (ret != ENOMEM)
1211 return (ret);
1212 wiredlen = 0;
1213 }
1214 /*
1215 * Touch all the wired pages to avoid PTE modified
1216 * bit emulation traps on Alpha while holding locks
1217 * in the sysctl handler.
1218 */
1219 for (i = (wiredlen + PAGE_SIZE - 1) / PAGE_SIZE,
1220 cp = req->oldptr; i > 0; i--, cp += PAGE_SIZE) {
1221 copyin(cp, &dummy, 1);
1222 copyout(&dummy, cp, 1);
1223 }
1224 }
1225 req->lock = REQ_WIRED;
1226 req->validlen = wiredlen;
1227 }
1228 return (0);
1229}
1230
1231int
1232sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1233 int *nindx, struct sysctl_req *req)
1234{
1235 struct sysctl_oid *oid;
1236 int indx;
1237
1238 oid = SLIST_FIRST(&sysctl__children);
1239 indx = 0;
1240 while (oid && indx < CTL_MAXNAME) {
1241 if (oid->oid_number == name[indx]) {
1242 indx++;
1243 if (oid->oid_kind & CTLFLAG_NOLOCK)
1244 req->lock = REQ_UNLOCKED;
1245 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1246 if (oid->oid_handler != NULL ||
1247 indx == namelen) {
1248 *noid = oid;
1249 if (nindx != NULL)
1250 *nindx = indx;
1251 return (0);
1252 }
1253 oid = SLIST_FIRST(
1254 (struct sysctl_oid_list *)oid->oid_arg1);
1255 } else if (indx == namelen) {
1256 *noid = oid;
1257 if (nindx != NULL)
1258 *nindx = indx;
1259 return (0);
1260 } else {
1261 return (ENOTDIR);
1262 }
1263 } else {
1264 oid = SLIST_NEXT(oid, oid_link);
1265 }
1266 }
1267 return (ENOENT);
1268}
1269
1270/*
1271 * Traverse our tree, and find the right node, execute whatever it points
1272 * to, and return the resulting error code.
1273 */
1274
1275static int
1276sysctl_root(SYSCTL_HANDLER_ARGS)
1277{
1278 struct sysctl_oid *oid;
1279 int error, indx, lvl;
1280
1281 SYSCTL_LOCK_ASSERT();
1282
1283 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1284 if (error)
1285 return (error);
1286
1287 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1288 /*
1289 * You can't call a sysctl when it's a node, but has
1290 * no handler. Inform the user that it's a node.
1291 * The indx may or may not be the same as namelen.
1292 */
1293 if (oid->oid_handler == NULL)
1294 return (EISDIR);
1295 }
1296
1297 /* Is this sysctl writable? */
1298 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1299 return (EPERM);
1300
1301 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1302
1303 /* Is this sysctl sensitive to securelevels? */
1304 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1305 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1306 error = securelevel_gt(req->td->td_ucred, lvl);
1307 if (error)
1308 return (error);
1309 }
1310
1311 /* Is this sysctl writable by only privileged users? */
1312 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1313 if (oid->oid_kind & CTLFLAG_PRISON)
1314 error = priv_check(req->td, PRIV_SYSCTL_WRITEJAIL);
1315 else
1316 error = priv_check(req->td, PRIV_SYSCTL_WRITE);
1317 if (error)
1318 return (error);
1319 }
1320
1321 if (!oid->oid_handler)
1322 return (EINVAL);
1323
1324 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1325 arg1 = (int *)arg1 + indx;
1326 arg2 -= indx;
1327 } else {
1328 arg1 = oid->oid_arg1;
1329 arg2 = oid->oid_arg2;
1330 }
1331#ifdef MAC
1332 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1333 req);
1334 if (error != 0)
1335 return (error);
1336#endif
1337 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1338 mtx_lock(&Giant);
1339 error = oid->oid_handler(oid, arg1, arg2, req);
1340 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1341 mtx_unlock(&Giant);
1342
1343 return (error);
1344}
1345
1346#ifndef _SYS_SYSPROTO_H_
1347struct sysctl_args {
1348 int *name;
1349 u_int namelen;
1350 void *old;
1351 size_t *oldlenp;
1352 void *new;
1353 size_t newlen;
1354};
1355#endif
1356int
1357__sysctl(struct thread *td, struct sysctl_args *uap)
1358{
1359 int error, name[CTL_MAXNAME];
1360 size_t j;
1361
1362 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1363 return (EINVAL);
1364
1365 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1366 if (error)
1367 return (error);
1368
1369 error = userland_sysctl(td, name, uap->namelen,
1370 uap->old, uap->oldlenp, 0,
1371 uap->new, uap->newlen, &j, 0);
1372 if (error && error != ENOMEM)
1373 return (error);
1374 if (uap->oldlenp) {
1375 int i = copyout(&j, uap->oldlenp, sizeof(j));
1376 if (i)
1377 return (i);
1378 }
1379 return (error);
1380}
1381
1382/*
1383 * This is used from various compatibility syscalls too. That's why name
1384 * must be in kernel space.
1385 */
1386int
1387userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1388 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1389 int flags)
1390{
1391 int error = 0;
1392 struct sysctl_req req;
1393
1394 bzero(&req, sizeof req);
1395
1396 req.td = td;
1397 req.flags = flags;
1398
1399 if (oldlenp) {
1400 if (inkernel) {
1401 req.oldlen = *oldlenp;
1402 } else {
1403 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1404 if (error)
1405 return (error);
1406 }
1407 }
1408 req.validlen = req.oldlen;
1409
1410 if (old) {
1411 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1412 return (EFAULT);
1413 req.oldptr= old;
1414 }
1415
1416 if (new != NULL) {
1417 if (!useracc(new, newlen, VM_PROT_READ))
1418 return (EFAULT);
1419 req.newlen = newlen;
1420 req.newptr = new;
1421 }
1422
1423 req.oldfunc = sysctl_old_user;
1424 req.newfunc = sysctl_new_user;
1425 req.lock = REQ_LOCKED;
1426
1427 SYSCTL_LOCK();
1428 CURVNET_SET(TD_TO_VNET(curthread));
1429
1430 for (;;) {
1431 req.oldidx = 0;
1432 req.newidx = 0;
1433 error = sysctl_root(0, name, namelen, &req);
1434 if (error != EAGAIN)
1435 break;
1436 uio_yield();
1437 }
1438
1439 CURVNET_RESTORE();
1440 SYSCTL_UNLOCK();
1441
1442 if (req.lock == REQ_WIRED && req.validlen > 0)
1443 vsunlock(req.oldptr, req.validlen);
1444
1445 if (error && error != ENOMEM)
1446 return (error);
1447
1448 if (retval) {
1449 if (req.oldptr && req.oldidx > req.validlen)
1450 *retval = req.validlen;
1451 else
1452 *retval = req.oldidx;
1453 }
1454 return (error);
1455}