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: releng/8.3/sys/kern/kern_sysctl.c 222579 2011-06-01 18:26:59Z mdf $");
40
41 #include "opt_compat.h"
42 #include "opt_ktrace.h"
43
44 #include <sys/param.h>
45 #include <sys/fail.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
48 #include <sys/sysctl.h>
49 #include <sys/malloc.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/jail.h>
53 #include <sys/lock.h>
54 #include <sys/mutex.h>
55 #include <sys/sbuf.h>
56 #include <sys/sx.h>
57 #include <sys/sysproto.h>
58 #include <sys/uio.h>
59 #ifdef KTRACE
60 #include <sys/ktrace.h>
61 #endif
62
63 #include <net/vnet.h>
64
65 #include <security/mac/mac_framework.h>
66
67 #include <vm/vm.h>
68 #include <vm/vm_extern.h>
69
70 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
71 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
72 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
73
74 /*
75 * The sysctllock protects the MIB tree. It also protects sysctl
76 * contexts used with dynamic sysctls. The sysctl_register_oid() and
77 * sysctl_unregister_oid() routines require the sysctllock to already
78 * be held, so the sysctl_lock() and sysctl_unlock() routines are
79 * provided for the few places in the kernel which need to use that
80 * API rather than using the dynamic API. Use of the dynamic API is
81 * strongly encouraged for most code.
82 *
83 * The sysctlmemlock is used to limit the amount of user memory wired for
84 * sysctl requests. This is implemented by serializing any userland
85 * sysctl requests larger than a single page via an exclusive lock.
86 */
87 static struct sx sysctllock;
88 static struct sx sysctlmemlock;
89
90 #define SYSCTL_XLOCK() sx_xlock(&sysctllock)
91 #define SYSCTL_XUNLOCK() sx_xunlock(&sysctllock)
92 #define SYSCTL_ASSERT_XLOCKED() sx_assert(&sysctllock, SA_XLOCKED)
93 #define SYSCTL_INIT() sx_init(&sysctllock, "sysctl lock")
94 #define SYSCTL_SLEEP(ch, wmesg, timo) \
95 sx_sleep(ch, &sysctllock, 0, wmesg, timo)
96
97 static int sysctl_root(SYSCTL_HANDLER_ARGS);
98
99 struct sysctl_oid_list sysctl__children; /* root list */
100
101 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
102 int recurse);
103
104 static struct sysctl_oid *
105 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
106 {
107 struct sysctl_oid *oidp;
108
109 SYSCTL_ASSERT_XLOCKED();
110 SLIST_FOREACH(oidp, list, oid_link) {
111 if (strcmp(oidp->oid_name, name) == 0) {
112 return (oidp);
113 }
114 }
115 return (NULL);
116 }
117
118 /*
119 * Initialization of the MIB tree.
120 *
121 * Order by number in each list.
122 */
123 void
124 sysctl_lock(void)
125 {
126
127 SYSCTL_XLOCK();
128 }
129
130 void
131 sysctl_unlock(void)
132 {
133
134 SYSCTL_XUNLOCK();
135 }
136
137 void
138 sysctl_register_oid(struct sysctl_oid *oidp)
139 {
140 struct sysctl_oid_list *parent = oidp->oid_parent;
141 struct sysctl_oid *p;
142 struct sysctl_oid *q;
143
144 /*
145 * First check if another oid with the same name already
146 * exists in the parent's list.
147 */
148 SYSCTL_ASSERT_XLOCKED();
149 p = sysctl_find_oidname(oidp->oid_name, parent);
150 if (p != NULL) {
151 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
152 p->oid_refcnt++;
153 return;
154 } else {
155 printf("can't re-use a leaf (%s)!\n", p->oid_name);
156 return;
157 }
158 }
159 /*
160 * If this oid has a number OID_AUTO, give it a number which
161 * is greater than any current oid.
162 * NOTE: DO NOT change the starting value here, change it in
163 * <sys/sysctl.h>, and make sure it is at least 256 to
164 * accomodate e.g. net.inet.raw as a static sysctl node.
165 */
166 if (oidp->oid_number == OID_AUTO) {
167 static int newoid = CTL_AUTO_START;
168
169 oidp->oid_number = newoid++;
170 if (newoid == 0x7fffffff)
171 panic("out of oids");
172 }
173 #if 0
174 else if (oidp->oid_number >= CTL_AUTO_START) {
175 /* do not panic; this happens when unregistering sysctl sets */
176 printf("static sysctl oid too high: %d", oidp->oid_number);
177 }
178 #endif
179
180 /*
181 * Insert the oid into the parent's list in order.
182 */
183 q = NULL;
184 SLIST_FOREACH(p, parent, oid_link) {
185 if (oidp->oid_number < p->oid_number)
186 break;
187 q = p;
188 }
189 if (q)
190 SLIST_INSERT_AFTER(q, oidp, oid_link);
191 else
192 SLIST_INSERT_HEAD(parent, oidp, oid_link);
193 }
194
195 void
196 sysctl_unregister_oid(struct sysctl_oid *oidp)
197 {
198 struct sysctl_oid *p;
199 int error;
200
201 SYSCTL_ASSERT_XLOCKED();
202 error = ENOENT;
203 if (oidp->oid_number == OID_AUTO) {
204 error = EINVAL;
205 } else {
206 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
207 if (p == oidp) {
208 SLIST_REMOVE(oidp->oid_parent, oidp,
209 sysctl_oid, oid_link);
210 error = 0;
211 break;
212 }
213 }
214 }
215
216 /*
217 * This can happen when a module fails to register and is
218 * being unloaded afterwards. It should not be a panic()
219 * for normal use.
220 */
221 if (error)
222 printf("%s: failed to unregister sysctl\n", __func__);
223 }
224
225 /* Initialize a new context to keep track of dynamically added sysctls. */
226 int
227 sysctl_ctx_init(struct sysctl_ctx_list *c)
228 {
229
230 if (c == NULL) {
231 return (EINVAL);
232 }
233
234 /*
235 * No locking here, the caller is responsible for not adding
236 * new nodes to a context until after this function has
237 * returned.
238 */
239 TAILQ_INIT(c);
240 return (0);
241 }
242
243 /* Free the context, and destroy all dynamic oids registered in this context */
244 int
245 sysctl_ctx_free(struct sysctl_ctx_list *clist)
246 {
247 struct sysctl_ctx_entry *e, *e1;
248 int error;
249
250 error = 0;
251 /*
252 * First perform a "dry run" to check if it's ok to remove oids.
253 * XXX FIXME
254 * XXX This algorithm is a hack. But I don't know any
255 * XXX better solution for now...
256 */
257 SYSCTL_XLOCK();
258 TAILQ_FOREACH(e, clist, link) {
259 error = sysctl_remove_oid_locked(e->entry, 0, 0);
260 if (error)
261 break;
262 }
263 /*
264 * Restore deregistered entries, either from the end,
265 * or from the place where error occured.
266 * e contains the entry that was not unregistered
267 */
268 if (error)
269 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
270 else
271 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
272 while (e1 != NULL) {
273 sysctl_register_oid(e1->entry);
274 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
275 }
276 if (error) {
277 SYSCTL_XUNLOCK();
278 return(EBUSY);
279 }
280 /* Now really delete the entries */
281 e = TAILQ_FIRST(clist);
282 while (e != NULL) {
283 e1 = TAILQ_NEXT(e, link);
284 error = sysctl_remove_oid_locked(e->entry, 1, 0);
285 if (error)
286 panic("sysctl_remove_oid: corrupt tree, entry: %s",
287 e->entry->oid_name);
288 free(e, M_SYSCTLOID);
289 e = e1;
290 }
291 SYSCTL_XUNLOCK();
292 return (error);
293 }
294
295 /* Add an entry to the context */
296 struct sysctl_ctx_entry *
297 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
298 {
299 struct sysctl_ctx_entry *e;
300
301 SYSCTL_ASSERT_XLOCKED();
302 if (clist == NULL || oidp == NULL)
303 return(NULL);
304 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
305 e->entry = oidp;
306 TAILQ_INSERT_HEAD(clist, e, link);
307 return (e);
308 }
309
310 /* Find an entry in the context */
311 struct sysctl_ctx_entry *
312 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
313 {
314 struct sysctl_ctx_entry *e;
315
316 SYSCTL_ASSERT_XLOCKED();
317 if (clist == NULL || oidp == NULL)
318 return(NULL);
319 TAILQ_FOREACH(e, clist, link) {
320 if(e->entry == oidp)
321 return(e);
322 }
323 return (e);
324 }
325
326 /*
327 * Delete an entry from the context.
328 * NOTE: this function doesn't free oidp! You have to remove it
329 * with sysctl_remove_oid().
330 */
331 int
332 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
333 {
334 struct sysctl_ctx_entry *e;
335
336 if (clist == NULL || oidp == NULL)
337 return (EINVAL);
338 SYSCTL_XLOCK();
339 e = sysctl_ctx_entry_find(clist, oidp);
340 if (e != NULL) {
341 TAILQ_REMOVE(clist, e, link);
342 SYSCTL_XUNLOCK();
343 free(e, M_SYSCTLOID);
344 return (0);
345 } else {
346 SYSCTL_XUNLOCK();
347 return (ENOENT);
348 }
349 }
350
351 /*
352 * Remove dynamically created sysctl trees.
353 * oidp - top of the tree to be removed
354 * del - if 0 - just deregister, otherwise free up entries as well
355 * recurse - if != 0 traverse the subtree to be deleted
356 */
357 int
358 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
359 {
360 int error;
361
362 SYSCTL_XLOCK();
363 error = sysctl_remove_oid_locked(oidp, del, recurse);
364 SYSCTL_XUNLOCK();
365 return (error);
366 }
367
368 static int
369 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
370 {
371 struct sysctl_oid *p;
372 int error;
373
374 SYSCTL_ASSERT_XLOCKED();
375 if (oidp == NULL)
376 return(EINVAL);
377 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
378 printf("can't remove non-dynamic nodes!\n");
379 return (EINVAL);
380 }
381 /*
382 * WARNING: normal method to do this should be through
383 * sysctl_ctx_free(). Use recursing as the last resort
384 * method to purge your sysctl tree of leftovers...
385 * However, if some other code still references these nodes,
386 * it will panic.
387 */
388 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
389 if (oidp->oid_refcnt == 1) {
390 SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
391 if (!recurse)
392 return (ENOTEMPTY);
393 error = sysctl_remove_oid_locked(p, del,
394 recurse);
395 if (error)
396 return (error);
397 }
398 if (del)
399 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
400 }
401 }
402 if (oidp->oid_refcnt > 1 ) {
403 oidp->oid_refcnt--;
404 } else {
405 if (oidp->oid_refcnt == 0) {
406 printf("Warning: bad oid_refcnt=%u (%s)!\n",
407 oidp->oid_refcnt, oidp->oid_name);
408 return (EINVAL);
409 }
410 sysctl_unregister_oid(oidp);
411 if (del) {
412 /*
413 * Wait for all threads running the handler to drain.
414 * This preserves the previous behavior when the
415 * sysctl lock was held across a handler invocation,
416 * and is necessary for module unload correctness.
417 */
418 while (oidp->oid_running > 0) {
419 oidp->oid_kind |= CTLFLAG_DYING;
420 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
421 }
422 if (oidp->oid_descr)
423 free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
424 free((void *)(uintptr_t)(const void *)oidp->oid_name,
425 M_SYSCTLOID);
426 free(oidp, M_SYSCTLOID);
427 }
428 }
429 return (0);
430 }
431
432 /*
433 * Create new sysctls at run time.
434 * clist may point to a valid context initialized with sysctl_ctx_init().
435 */
436 struct sysctl_oid *
437 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
438 int number, const char *name, int kind, void *arg1, int arg2,
439 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
440 {
441 struct sysctl_oid *oidp;
442 ssize_t len;
443 char *newname;
444
445 /* You have to hook up somewhere.. */
446 if (parent == NULL)
447 return(NULL);
448 /* Check if the node already exists, otherwise create it */
449 SYSCTL_XLOCK();
450 oidp = sysctl_find_oidname(name, parent);
451 if (oidp != NULL) {
452 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
453 oidp->oid_refcnt++;
454 /* Update the context */
455 if (clist != NULL)
456 sysctl_ctx_entry_add(clist, oidp);
457 SYSCTL_XUNLOCK();
458 return (oidp);
459 } else {
460 SYSCTL_XUNLOCK();
461 printf("can't re-use a leaf (%s)!\n", name);
462 return (NULL);
463 }
464 }
465 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
466 oidp->oid_parent = parent;
467 SLIST_NEXT(oidp, oid_link) = NULL;
468 oidp->oid_number = number;
469 oidp->oid_refcnt = 1;
470 len = strlen(name);
471 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
472 bcopy(name, newname, len + 1);
473 newname[len] = '\0';
474 oidp->oid_name = newname;
475 oidp->oid_handler = handler;
476 oidp->oid_kind = CTLFLAG_DYN | kind;
477 if ((kind & CTLTYPE) == CTLTYPE_NODE) {
478 /* Allocate space for children */
479 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
480 M_SYSCTLOID, M_WAITOK));
481 SLIST_INIT(SYSCTL_CHILDREN(oidp));
482 } else {
483 oidp->oid_arg1 = arg1;
484 oidp->oid_arg2 = arg2;
485 }
486 oidp->oid_fmt = fmt;
487 if (descr) {
488 int len = strlen(descr) + 1;
489 oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
490 if (oidp->oid_descr)
491 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
492 }
493 /* Update the context, if used */
494 if (clist != NULL)
495 sysctl_ctx_entry_add(clist, oidp);
496 /* Register this oid */
497 sysctl_register_oid(oidp);
498 SYSCTL_XUNLOCK();
499 return (oidp);
500 }
501
502 /*
503 * Rename an existing oid.
504 */
505 void
506 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
507 {
508 ssize_t len;
509 char *newname;
510 void *oldname;
511
512 len = strlen(name);
513 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
514 bcopy(name, newname, len + 1);
515 newname[len] = '\0';
516 SYSCTL_XLOCK();
517 oldname = (void *)(uintptr_t)(const void *)oidp->oid_name;
518 oidp->oid_name = newname;
519 SYSCTL_XUNLOCK();
520 free(oldname, M_SYSCTLOID);
521 }
522
523 /*
524 * Reparent an existing oid.
525 */
526 int
527 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
528 {
529 struct sysctl_oid *oidp;
530
531 SYSCTL_XLOCK();
532 if (oid->oid_parent == parent) {
533 SYSCTL_XUNLOCK();
534 return (0);
535 }
536 oidp = sysctl_find_oidname(oid->oid_name, parent);
537 if (oidp != NULL) {
538 SYSCTL_XUNLOCK();
539 return (EEXIST);
540 }
541 sysctl_unregister_oid(oid);
542 oid->oid_parent = parent;
543 oid->oid_number = OID_AUTO;
544 sysctl_register_oid(oid);
545 SYSCTL_XUNLOCK();
546 return (0);
547 }
548
549 /*
550 * Register the kernel's oids on startup.
551 */
552 SET_DECLARE(sysctl_set, struct sysctl_oid);
553
554 static void
555 sysctl_register_all(void *arg)
556 {
557 struct sysctl_oid **oidp;
558
559 sx_init(&sysctlmemlock, "sysctl mem");
560 SYSCTL_INIT();
561 SYSCTL_XLOCK();
562 SET_FOREACH(oidp, sysctl_set)
563 sysctl_register_oid(*oidp);
564 SYSCTL_XUNLOCK();
565 }
566 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
567
568 /*
569 * "Staff-functions"
570 *
571 * These functions implement a presently undocumented interface
572 * used by the sysctl program to walk the tree, and get the type
573 * so it can print the value.
574 * This interface is under work and consideration, and should probably
575 * be killed with a big axe by the first person who can find the time.
576 * (be aware though, that the proper interface isn't as obvious as it
577 * may seem, there are various conflicting requirements.
578 *
579 * {0,0} printf the entire MIB-tree.
580 * {0,1,...} return the name of the "..." OID.
581 * {0,2,...} return the next OID.
582 * {0,3} return the OID of the name in "new"
583 * {0,4,...} return the kind & format info for the "..." OID.
584 * {0,5,...} return the description the "..." OID.
585 */
586
587 #ifdef SYSCTL_DEBUG
588 static void
589 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
590 {
591 int k;
592 struct sysctl_oid *oidp;
593
594 SYSCTL_ASSERT_XLOCKED();
595 SLIST_FOREACH(oidp, l, oid_link) {
596
597 for (k=0; k<i; k++)
598 printf(" ");
599
600 printf("%d %s ", oidp->oid_number, oidp->oid_name);
601
602 printf("%c%c",
603 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
604 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
605
606 if (oidp->oid_handler)
607 printf(" *Handler");
608
609 switch (oidp->oid_kind & CTLTYPE) {
610 case CTLTYPE_NODE:
611 printf(" Node\n");
612 if (!oidp->oid_handler) {
613 sysctl_sysctl_debug_dump_node(
614 oidp->oid_arg1, i+2);
615 }
616 break;
617 case CTLTYPE_INT: printf(" Int\n"); break;
618 case CTLTYPE_STRING: printf(" String\n"); break;
619 case CTLTYPE_QUAD: printf(" Quad\n"); break;
620 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
621 default: printf("\n");
622 }
623
624 }
625 }
626
627 static int
628 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
629 {
630 int error;
631
632 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
633 if (error)
634 return (error);
635 SYSCTL_XLOCK();
636 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
637 SYSCTL_XUNLOCK();
638 return (ENOENT);
639 }
640
641 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
642 0, 0, sysctl_sysctl_debug, "-", "");
643 #endif
644
645 static int
646 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
647 {
648 int *name = (int *) arg1;
649 u_int namelen = arg2;
650 int error = 0;
651 struct sysctl_oid *oid;
652 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
653 char buf[10];
654
655 SYSCTL_XLOCK();
656 while (namelen) {
657 if (!lsp) {
658 snprintf(buf,sizeof(buf),"%d",*name);
659 if (req->oldidx)
660 error = SYSCTL_OUT(req, ".", 1);
661 if (!error)
662 error = SYSCTL_OUT(req, buf, strlen(buf));
663 if (error)
664 goto out;
665 namelen--;
666 name++;
667 continue;
668 }
669 lsp2 = 0;
670 SLIST_FOREACH(oid, lsp, oid_link) {
671 if (oid->oid_number != *name)
672 continue;
673
674 if (req->oldidx)
675 error = SYSCTL_OUT(req, ".", 1);
676 if (!error)
677 error = SYSCTL_OUT(req, oid->oid_name,
678 strlen(oid->oid_name));
679 if (error)
680 goto out;
681
682 namelen--;
683 name++;
684
685 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
686 break;
687
688 if (oid->oid_handler)
689 break;
690
691 lsp2 = SYSCTL_CHILDREN(oid);
692 break;
693 }
694 lsp = lsp2;
695 }
696 error = SYSCTL_OUT(req, "", 1);
697 out:
698 SYSCTL_XUNLOCK();
699 return (error);
700 }
701
702 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
703
704 static int
705 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
706 int *next, int *len, int level, struct sysctl_oid **oidpp)
707 {
708 struct sysctl_oid *oidp;
709
710 SYSCTL_ASSERT_XLOCKED();
711 *len = level;
712 SLIST_FOREACH(oidp, lsp, oid_link) {
713 *next = oidp->oid_number;
714 *oidpp = oidp;
715
716 if (oidp->oid_kind & CTLFLAG_SKIP)
717 continue;
718
719 if (!namelen) {
720 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
721 return (0);
722 if (oidp->oid_handler)
723 /* We really should call the handler here...*/
724 return (0);
725 lsp = SYSCTL_CHILDREN(oidp);
726 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
727 len, level+1, oidpp))
728 return (0);
729 goto emptynode;
730 }
731
732 if (oidp->oid_number < *name)
733 continue;
734
735 if (oidp->oid_number > *name) {
736 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
737 return (0);
738 if (oidp->oid_handler)
739 return (0);
740 lsp = SYSCTL_CHILDREN(oidp);
741 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
742 next+1, len, level+1, oidpp))
743 return (0);
744 goto next;
745 }
746 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
747 continue;
748
749 if (oidp->oid_handler)
750 continue;
751
752 lsp = SYSCTL_CHILDREN(oidp);
753 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
754 len, level+1, oidpp))
755 return (0);
756 next:
757 namelen = 1;
758 emptynode:
759 *len = level;
760 }
761 return (1);
762 }
763
764 static int
765 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
766 {
767 int *name = (int *) arg1;
768 u_int namelen = arg2;
769 int i, j, error;
770 struct sysctl_oid *oid;
771 struct sysctl_oid_list *lsp = &sysctl__children;
772 int newoid[CTL_MAXNAME];
773
774 SYSCTL_XLOCK();
775 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
776 SYSCTL_XUNLOCK();
777 if (i)
778 return (ENOENT);
779 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
780 return (error);
781 }
782
783 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
784
785 static int
786 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
787 {
788 int i;
789 struct sysctl_oid *oidp;
790 struct sysctl_oid_list *lsp = &sysctl__children;
791 char *p;
792
793 SYSCTL_ASSERT_XLOCKED();
794
795 if (!*name)
796 return (ENOENT);
797
798 p = name + strlen(name) - 1 ;
799 if (*p == '.')
800 *p = '\0';
801
802 *len = 0;
803
804 for (p = name; *p && *p != '.'; p++)
805 ;
806 i = *p;
807 if (i == '.')
808 *p = '\0';
809
810 oidp = SLIST_FIRST(lsp);
811
812 while (oidp && *len < CTL_MAXNAME) {
813 if (strcmp(name, oidp->oid_name)) {
814 oidp = SLIST_NEXT(oidp, oid_link);
815 continue;
816 }
817 *oid++ = oidp->oid_number;
818 (*len)++;
819
820 if (!i) {
821 if (oidpp)
822 *oidpp = oidp;
823 return (0);
824 }
825
826 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
827 break;
828
829 if (oidp->oid_handler)
830 break;
831
832 lsp = SYSCTL_CHILDREN(oidp);
833 oidp = SLIST_FIRST(lsp);
834 name = p+1;
835 for (p = name; *p && *p != '.'; p++)
836 ;
837 i = *p;
838 if (i == '.')
839 *p = '\0';
840 }
841 return (ENOENT);
842 }
843
844 static int
845 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
846 {
847 char *p;
848 int error, oid[CTL_MAXNAME], len;
849 struct sysctl_oid *op = 0;
850
851 if (!req->newlen)
852 return (ENOENT);
853 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
854 return (ENAMETOOLONG);
855
856 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
857
858 error = SYSCTL_IN(req, p, req->newlen);
859 if (error) {
860 free(p, M_SYSCTL);
861 return (error);
862 }
863
864 p [req->newlen] = '\0';
865 len = 0;
866 SYSCTL_XLOCK();
867 error = name2oid(p, oid, &len, &op);
868 SYSCTL_XUNLOCK();
869
870 free(p, M_SYSCTL);
871
872 if (error)
873 return (error);
874
875 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
876 return (error);
877 }
878
879 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_MPSAFE,
880 0, 0, sysctl_sysctl_name2oid, "I", "");
881
882 static int
883 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
884 {
885 struct sysctl_oid *oid;
886 int error;
887
888 SYSCTL_XLOCK();
889 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
890 if (error)
891 goto out;
892
893 if (oid->oid_fmt == NULL) {
894 error = ENOENT;
895 goto out;
896 }
897 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
898 if (error)
899 goto out;
900 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
901 out:
902 SYSCTL_XUNLOCK();
903 return (error);
904 }
905
906
907 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE,
908 sysctl_sysctl_oidfmt, "");
909
910 static int
911 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
912 {
913 struct sysctl_oid *oid;
914 int error;
915
916 SYSCTL_XLOCK();
917 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
918 if (error)
919 goto out;
920
921 if (oid->oid_descr == NULL) {
922 error = ENOENT;
923 goto out;
924 }
925 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
926 out:
927 SYSCTL_XUNLOCK();
928 return (error);
929 }
930
931 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
932
933 /*
934 * Default "handler" functions.
935 */
936
937 /*
938 * Handle an int, signed or unsigned.
939 * Two cases:
940 * a variable: point arg1 at it.
941 * a constant: pass it in arg2.
942 */
943
944 int
945 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
946 {
947 int tmpout, error = 0;
948
949 /*
950 * Attempt to get a coherent snapshot by making a copy of the data.
951 */
952 if (arg1)
953 tmpout = *(int *)arg1;
954 else
955 tmpout = arg2;
956 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
957
958 if (error || !req->newptr)
959 return (error);
960
961 if (!arg1)
962 error = EPERM;
963 else
964 error = SYSCTL_IN(req, arg1, sizeof(int));
965 return (error);
966 }
967
968 /*
969 * Based on on sysctl_handle_int() convert milliseconds into ticks.
970 * Note: this is used by TCP.
971 */
972
973 int
974 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
975 {
976 int error, s, tt;
977
978 tt = *(int *)arg1;
979 s = (int)((int64_t)tt * 1000 / hz);
980
981 error = sysctl_handle_int(oidp, &s, 0, req);
982 if (error || !req->newptr)
983 return (error);
984
985 tt = (int)((int64_t)s * hz / 1000);
986 if (tt < 1)
987 return (EINVAL);
988
989 *(int *)arg1 = tt;
990 return (0);
991 }
992
993
994 /*
995 * Handle a long, signed or unsigned. arg1 points to it.
996 */
997
998 int
999 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1000 {
1001 int error = 0;
1002 long tmplong;
1003 #ifdef SCTL_MASK32
1004 int tmpint;
1005 #endif
1006
1007 /*
1008 * Attempt to get a coherent snapshot by making a copy of the data.
1009 */
1010 if (!arg1)
1011 return (EINVAL);
1012 tmplong = *(long *)arg1;
1013 #ifdef SCTL_MASK32
1014 if (req->flags & SCTL_MASK32) {
1015 tmpint = tmplong;
1016 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1017 } else
1018 #endif
1019 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1020
1021 if (error || !req->newptr)
1022 return (error);
1023
1024 #ifdef SCTL_MASK32
1025 if (req->flags & SCTL_MASK32) {
1026 error = SYSCTL_IN(req, &tmpint, sizeof(int));
1027 *(long *)arg1 = (long)tmpint;
1028 } else
1029 #endif
1030 error = SYSCTL_IN(req, arg1, sizeof(long));
1031 return (error);
1032 }
1033
1034 /*
1035 * Handle a 64 bit int, signed or unsigned. arg1 points to it.
1036 */
1037
1038 int
1039 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
1040 {
1041 int error = 0;
1042 uint64_t tmpout;
1043
1044 /*
1045 * Attempt to get a coherent snapshot by making a copy of the data.
1046 */
1047 if (!arg1)
1048 return (EINVAL);
1049 tmpout = *(uint64_t *)arg1;
1050 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1051
1052 if (error || !req->newptr)
1053 return (error);
1054
1055 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1056 return (error);
1057 }
1058
1059 /*
1060 * Handle our generic '\0' terminated 'C' string.
1061 * Two cases:
1062 * a variable string: point arg1 at it, arg2 is max length.
1063 * a constant string: point arg1 at it, arg2 is zero.
1064 */
1065
1066 int
1067 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1068 {
1069 int error=0;
1070 char *tmparg;
1071 size_t outlen;
1072
1073 /*
1074 * Attempt to get a coherent snapshot by copying to a
1075 * temporary kernel buffer.
1076 */
1077 retry:
1078 outlen = strlen((char *)arg1)+1;
1079 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
1080
1081 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
1082 free(tmparg, M_SYSCTLTMP);
1083 goto retry;
1084 }
1085
1086 error = SYSCTL_OUT(req, tmparg, outlen);
1087 free(tmparg, M_SYSCTLTMP);
1088
1089 if (error || !req->newptr)
1090 return (error);
1091
1092 if ((req->newlen - req->newidx) >= arg2) {
1093 error = EINVAL;
1094 } else {
1095 arg2 = (req->newlen - req->newidx);
1096 error = SYSCTL_IN(req, arg1, arg2);
1097 ((char *)arg1)[arg2] = '\0';
1098 }
1099
1100 return (error);
1101 }
1102
1103 /*
1104 * Handle any kind of opaque data.
1105 * arg1 points to it, arg2 is the size.
1106 */
1107
1108 int
1109 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1110 {
1111 int error, tries;
1112 u_int generation;
1113 struct sysctl_req req2;
1114
1115 /*
1116 * Attempt to get a coherent snapshot, by using the thread
1117 * pre-emption counter updated from within mi_switch() to
1118 * determine if we were pre-empted during a bcopy() or
1119 * copyout(). Make 3 attempts at doing this before giving up.
1120 * If we encounter an error, stop immediately.
1121 */
1122 tries = 0;
1123 req2 = *req;
1124 retry:
1125 generation = curthread->td_generation;
1126 error = SYSCTL_OUT(req, arg1, arg2);
1127 if (error)
1128 return (error);
1129 tries++;
1130 if (generation != curthread->td_generation && tries < 3) {
1131 *req = req2;
1132 goto retry;
1133 }
1134
1135 error = SYSCTL_IN(req, arg1, arg2);
1136
1137 return (error);
1138 }
1139
1140 /*
1141 * Transfer functions to/from kernel space.
1142 * XXX: rather untested at this point
1143 */
1144 static int
1145 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1146 {
1147 size_t i = 0;
1148
1149 if (req->oldptr) {
1150 i = l;
1151 if (req->oldlen <= req->oldidx)
1152 i = 0;
1153 else
1154 if (i > req->oldlen - req->oldidx)
1155 i = req->oldlen - req->oldidx;
1156 if (i > 0)
1157 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1158 }
1159 req->oldidx += l;
1160 if (req->oldptr && i != l)
1161 return (ENOMEM);
1162 return (0);
1163 }
1164
1165 static int
1166 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1167 {
1168 if (!req->newptr)
1169 return (0);
1170 if (req->newlen - req->newidx < l)
1171 return (EINVAL);
1172 bcopy((char *)req->newptr + req->newidx, p, l);
1173 req->newidx += l;
1174 return (0);
1175 }
1176
1177 int
1178 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1179 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1180 {
1181 int error = 0;
1182 struct sysctl_req req;
1183
1184 bzero(&req, sizeof req);
1185
1186 req.td = td;
1187 req.flags = flags;
1188
1189 if (oldlenp) {
1190 req.oldlen = *oldlenp;
1191 }
1192 req.validlen = req.oldlen;
1193
1194 if (old) {
1195 req.oldptr= old;
1196 }
1197
1198 if (new != NULL) {
1199 req.newlen = newlen;
1200 req.newptr = new;
1201 }
1202
1203 req.oldfunc = sysctl_old_kernel;
1204 req.newfunc = sysctl_new_kernel;
1205 req.lock = REQ_LOCKED;
1206
1207 SYSCTL_XLOCK();
1208 error = sysctl_root(0, name, namelen, &req);
1209 SYSCTL_XUNLOCK();
1210
1211 if (req.lock == REQ_WIRED && req.validlen > 0)
1212 vsunlock(req.oldptr, req.validlen);
1213
1214 if (error && error != ENOMEM)
1215 return (error);
1216
1217 if (retval) {
1218 if (req.oldptr && req.oldidx > req.validlen)
1219 *retval = req.validlen;
1220 else
1221 *retval = req.oldidx;
1222 }
1223 return (error);
1224 }
1225
1226 int
1227 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1228 void *new, size_t newlen, size_t *retval, int flags)
1229 {
1230 int oid[CTL_MAXNAME];
1231 size_t oidlen, plen;
1232 int error;
1233
1234 oid[0] = 0; /* sysctl internal magic */
1235 oid[1] = 3; /* name2oid */
1236 oidlen = sizeof(oid);
1237
1238 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1239 (void *)name, strlen(name), &plen, flags);
1240 if (error)
1241 return (error);
1242
1243 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1244 new, newlen, retval, flags);
1245 return (error);
1246 }
1247
1248 /*
1249 * Transfer function to/from user space.
1250 */
1251 static int
1252 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1253 {
1254 int error = 0;
1255 size_t i, len, origidx;
1256
1257 origidx = req->oldidx;
1258 req->oldidx += l;
1259 if (req->oldptr == NULL)
1260 return (0);
1261 /*
1262 * If we have not wired the user supplied buffer and we are currently
1263 * holding locks, drop a witness warning, as it's possible that
1264 * write operations to the user page can sleep.
1265 */
1266 if (req->lock != REQ_WIRED)
1267 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1268 "sysctl_old_user()");
1269 i = l;
1270 len = req->validlen;
1271 if (len <= origidx)
1272 i = 0;
1273 else {
1274 if (i > len - origidx)
1275 i = len - origidx;
1276 error = copyout(p, (char *)req->oldptr + origidx, i);
1277 }
1278 if (error)
1279 return (error);
1280 if (i < l)
1281 return (ENOMEM);
1282 return (0);
1283 }
1284
1285 static int
1286 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1287 {
1288 int error;
1289
1290 if (!req->newptr)
1291 return (0);
1292 if (req->newlen - req->newidx < l)
1293 return (EINVAL);
1294 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1295 "sysctl_new_user()");
1296 error = copyin((char *)req->newptr + req->newidx, p, l);
1297 req->newidx += l;
1298 return (error);
1299 }
1300
1301 /*
1302 * Wire the user space destination buffer. If set to a value greater than
1303 * zero, the len parameter limits the maximum amount of wired memory.
1304 */
1305 int
1306 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1307 {
1308 int ret;
1309 size_t wiredlen;
1310
1311 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1312 ret = 0;
1313 if (req->lock == REQ_LOCKED && req->oldptr &&
1314 req->oldfunc == sysctl_old_user) {
1315 if (wiredlen != 0) {
1316 ret = vslock(req->oldptr, wiredlen);
1317 if (ret != 0) {
1318 if (ret != ENOMEM)
1319 return (ret);
1320 wiredlen = 0;
1321 }
1322 }
1323 req->lock = REQ_WIRED;
1324 req->validlen = wiredlen;
1325 }
1326 return (0);
1327 }
1328
1329 int
1330 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1331 int *nindx, struct sysctl_req *req)
1332 {
1333 struct sysctl_oid_list *lsp;
1334 struct sysctl_oid *oid;
1335 int indx;
1336
1337 SYSCTL_ASSERT_XLOCKED();
1338 lsp = &sysctl__children;
1339 indx = 0;
1340 while (indx < CTL_MAXNAME) {
1341 SLIST_FOREACH(oid, lsp, oid_link) {
1342 if (oid->oid_number == name[indx])
1343 break;
1344 }
1345 if (oid == NULL)
1346 return (ENOENT);
1347
1348 indx++;
1349 if (oid->oid_kind & CTLFLAG_NOLOCK)
1350 req->lock = REQ_UNLOCKED;
1351 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1352 if (oid->oid_handler != NULL || indx == namelen) {
1353 *noid = oid;
1354 if (nindx != NULL)
1355 *nindx = indx;
1356 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1357 ("%s found DYING node %p", __func__, oid));
1358 return (0);
1359 }
1360 lsp = SYSCTL_CHILDREN(oid);
1361 } else if (indx == namelen) {
1362 *noid = oid;
1363 if (nindx != NULL)
1364 *nindx = indx;
1365 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1366 ("%s found DYING node %p", __func__, oid));
1367 return (0);
1368 } else {
1369 return (ENOTDIR);
1370 }
1371 }
1372 return (ENOENT);
1373 }
1374
1375 /*
1376 * Traverse our tree, and find the right node, execute whatever it points
1377 * to, and return the resulting error code.
1378 */
1379
1380 static int
1381 sysctl_root(SYSCTL_HANDLER_ARGS)
1382 {
1383 struct sysctl_oid *oid;
1384 int error, indx, lvl;
1385
1386 SYSCTL_ASSERT_XLOCKED();
1387
1388 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1389 if (error)
1390 return (error);
1391
1392 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1393 /*
1394 * You can't call a sysctl when it's a node, but has
1395 * no handler. Inform the user that it's a node.
1396 * The indx may or may not be the same as namelen.
1397 */
1398 if (oid->oid_handler == NULL)
1399 return (EISDIR);
1400 }
1401
1402 /* Is this sysctl writable? */
1403 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1404 return (EPERM);
1405
1406 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1407
1408 /* Is this sysctl sensitive to securelevels? */
1409 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1410 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1411 error = securelevel_gt(req->td->td_ucred, lvl);
1412 if (error)
1413 return (error);
1414 }
1415
1416 /* Is this sysctl writable by only privileged users? */
1417 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1418 int priv;
1419
1420 if (oid->oid_kind & CTLFLAG_PRISON)
1421 priv = PRIV_SYSCTL_WRITEJAIL;
1422 #ifdef VIMAGE
1423 else if ((oid->oid_kind & CTLFLAG_VNET) &&
1424 prison_owns_vnet(req->td->td_ucred))
1425 priv = PRIV_SYSCTL_WRITEJAIL;
1426 #endif
1427 else
1428 priv = PRIV_SYSCTL_WRITE;
1429 error = priv_check(req->td, priv);
1430 if (error)
1431 return (error);
1432 }
1433
1434 if (!oid->oid_handler)
1435 return (EINVAL);
1436
1437 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1438 arg1 = (int *)arg1 + indx;
1439 arg2 -= indx;
1440 } else {
1441 arg1 = oid->oid_arg1;
1442 arg2 = oid->oid_arg2;
1443 }
1444 #ifdef MAC
1445 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1446 req);
1447 if (error != 0)
1448 return (error);
1449 #endif
1450 oid->oid_running++;
1451 SYSCTL_XUNLOCK();
1452
1453 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1454 mtx_lock(&Giant);
1455 error = oid->oid_handler(oid, arg1, arg2, req);
1456 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1457 mtx_unlock(&Giant);
1458
1459 KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
1460
1461 SYSCTL_XLOCK();
1462 oid->oid_running--;
1463 if (oid->oid_running == 0 && (oid->oid_kind & CTLFLAG_DYING) != 0)
1464 wakeup(&oid->oid_running);
1465 return (error);
1466 }
1467
1468 #ifndef _SYS_SYSPROTO_H_
1469 struct sysctl_args {
1470 int *name;
1471 u_int namelen;
1472 void *old;
1473 size_t *oldlenp;
1474 void *new;
1475 size_t newlen;
1476 };
1477 #endif
1478 int
1479 __sysctl(struct thread *td, struct sysctl_args *uap)
1480 {
1481 int error, i, name[CTL_MAXNAME];
1482 size_t j;
1483
1484 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1485 return (EINVAL);
1486
1487 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1488 if (error)
1489 return (error);
1490
1491 error = userland_sysctl(td, name, uap->namelen,
1492 uap->old, uap->oldlenp, 0,
1493 uap->new, uap->newlen, &j, 0);
1494 if (error && error != ENOMEM)
1495 return (error);
1496 if (uap->oldlenp) {
1497 i = copyout(&j, uap->oldlenp, sizeof(j));
1498 if (i)
1499 return (i);
1500 }
1501 return (error);
1502 }
1503
1504 /*
1505 * This is used from various compatibility syscalls too. That's why name
1506 * must be in kernel space.
1507 */
1508 int
1509 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1510 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1511 int flags)
1512 {
1513 int error = 0, memlocked;
1514 struct sysctl_req req;
1515
1516 bzero(&req, sizeof req);
1517
1518 req.td = td;
1519 req.flags = flags;
1520
1521 if (oldlenp) {
1522 if (inkernel) {
1523 req.oldlen = *oldlenp;
1524 } else {
1525 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1526 if (error)
1527 return (error);
1528 }
1529 }
1530 req.validlen = req.oldlen;
1531
1532 if (old) {
1533 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1534 return (EFAULT);
1535 req.oldptr= old;
1536 }
1537
1538 if (new != NULL) {
1539 if (!useracc(new, newlen, VM_PROT_READ))
1540 return (EFAULT);
1541 req.newlen = newlen;
1542 req.newptr = new;
1543 }
1544
1545 req.oldfunc = sysctl_old_user;
1546 req.newfunc = sysctl_new_user;
1547 req.lock = REQ_LOCKED;
1548
1549 #ifdef KTRACE
1550 if (KTRPOINT(curthread, KTR_SYSCTL))
1551 ktrsysctl(name, namelen);
1552 #endif
1553
1554 if (req.oldlen > PAGE_SIZE) {
1555 memlocked = 1;
1556 sx_xlock(&sysctlmemlock);
1557 } else
1558 memlocked = 0;
1559 CURVNET_SET(TD_TO_VNET(td));
1560
1561 for (;;) {
1562 req.oldidx = 0;
1563 req.newidx = 0;
1564 SYSCTL_XLOCK();
1565 error = sysctl_root(0, name, namelen, &req);
1566 SYSCTL_XUNLOCK();
1567 if (error != EAGAIN)
1568 break;
1569 uio_yield();
1570 }
1571
1572 CURVNET_RESTORE();
1573
1574 if (req.lock == REQ_WIRED && req.validlen > 0)
1575 vsunlock(req.oldptr, req.validlen);
1576 if (memlocked)
1577 sx_xunlock(&sysctlmemlock);
1578
1579 if (error && error != ENOMEM)
1580 return (error);
1581
1582 if (retval) {
1583 if (req.oldptr && req.oldidx > req.validlen)
1584 *retval = req.validlen;
1585 else
1586 *retval = req.oldidx;
1587 }
1588 return (error);
1589 }
1590
1591 /*
1592 * Drain into a sysctl struct. The user buffer should be wired if a page
1593 * fault would cause issue.
1594 */
1595 static int
1596 sbuf_sysctl_drain(void *arg, const char *data, int len)
1597 {
1598 struct sysctl_req *req = arg;
1599 int error;
1600
1601 error = SYSCTL_OUT(req, data, len);
1602 KASSERT(error >= 0, ("Got unexpected negative value %d", error));
1603 return (error == 0 ? len : -error);
1604 }
1605
1606 struct sbuf *
1607 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
1608 struct sysctl_req *req)
1609 {
1610
1611 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN);
1612 sbuf_set_drain(s, sbuf_sysctl_drain, req);
1613 return (s);
1614 }
Cache object: 6067d5fca69a65f207881ae1b2119816
|