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/10.4/sys/kern/kern_sysctl.c 316127 2017-03-29 08:00:11Z ngie $");
40
41 #include "opt_capsicum.h"
42 #include "opt_compat.h"
43 #include "opt_ktrace.h"
44
45 #include <sys/param.h>
46 #include <sys/fail.h>
47 #include <sys/systm.h>
48 #include <sys/capsicum.h>
49 #include <sys/kernel.h>
50 #include <sys/sysctl.h>
51 #include <sys/malloc.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/jail.h>
55 #include <sys/lock.h>
56 #include <sys/mutex.h>
57 #include <sys/sbuf.h>
58 #include <sys/sx.h>
59 #include <sys/sysproto.h>
60 #include <sys/uio.h>
61 #ifdef KTRACE
62 #include <sys/ktrace.h>
63 #endif
64
65 #include <net/vnet.h>
66
67 #include <security/mac/mac_framework.h>
68
69 #include <vm/vm.h>
70 #include <vm/vm_extern.h>
71
72 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
73 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
74 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
75
76 /*
77 * The sysctllock protects the MIB tree. It also protects sysctl
78 * contexts used with dynamic sysctls. The sysctl_register_oid() and
79 * sysctl_unregister_oid() routines require the sysctllock to already
80 * be held, so the sysctl_lock() and sysctl_unlock() routines are
81 * provided for the few places in the kernel which need to use that
82 * API rather than using the dynamic API. Use of the dynamic API is
83 * strongly encouraged for most code.
84 *
85 * The sysctlmemlock is used to limit the amount of user memory wired for
86 * sysctl requests. This is implemented by serializing any userland
87 * sysctl requests larger than a single page via an exclusive lock.
88 */
89 static struct sx sysctllock;
90 static struct sx sysctlmemlock;
91
92 #define SYSCTL_XLOCK() sx_xlock(&sysctllock)
93 #define SYSCTL_XUNLOCK() sx_xunlock(&sysctllock)
94 #define SYSCTL_ASSERT_XLOCKED() sx_assert(&sysctllock, SA_XLOCKED)
95 #define SYSCTL_INIT() sx_init(&sysctllock, "sysctl lock")
96 #define SYSCTL_SLEEP(ch, wmesg, timo) \
97 sx_sleep(ch, &sysctllock, 0, wmesg, timo)
98
99 static int sysctl_root(SYSCTL_HANDLER_ARGS);
100
101 struct sysctl_oid_list sysctl__children; /* root list */
102
103 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
104 int recurse);
105
106 static struct sysctl_oid *
107 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
108 {
109 struct sysctl_oid *oidp;
110
111 SYSCTL_ASSERT_XLOCKED();
112 SLIST_FOREACH(oidp, list, oid_link) {
113 if (strcmp(oidp->oid_name, name) == 0) {
114 return (oidp);
115 }
116 }
117 return (NULL);
118 }
119
120 /*
121 * Initialization of the MIB tree.
122 *
123 * Order by number in each list.
124 */
125 void
126 sysctl_lock(void)
127 {
128
129 SYSCTL_XLOCK();
130 }
131
132 void
133 sysctl_unlock(void)
134 {
135
136 SYSCTL_XUNLOCK();
137 }
138
139 void
140 sysctl_register_oid(struct sysctl_oid *oidp)
141 {
142 struct sysctl_oid_list *parent = oidp->oid_parent;
143 struct sysctl_oid *p;
144 struct sysctl_oid *q;
145 int oid_number;
146 int timeout = 2;
147
148 /*
149 * First check if another oid with the same name already
150 * exists in the parent's list.
151 */
152 SYSCTL_ASSERT_XLOCKED();
153 p = sysctl_find_oidname(oidp->oid_name, parent);
154 if (p != NULL) {
155 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
156 p->oid_refcnt++;
157 return;
158 } else {
159 printf("can't re-use a leaf (%s)!\n", p->oid_name);
160 return;
161 }
162 }
163 /* get current OID number */
164 oid_number = oidp->oid_number;
165
166 #if (OID_AUTO >= 0)
167 #error "OID_AUTO is expected to be a negative value"
168 #endif
169 /*
170 * Any negative OID number qualifies as OID_AUTO. Valid OID
171 * numbers should always be positive.
172 *
173 * NOTE: DO NOT change the starting value here, change it in
174 * <sys/sysctl.h>, and make sure it is at least 256 to
175 * accommodate e.g. net.inet.raw as a static sysctl node.
176 */
177 if (oid_number < 0) {
178 static int newoid;
179
180 /*
181 * By decrementing the next OID number we spend less
182 * time inserting the OIDs into a sorted list.
183 */
184 if (--newoid < CTL_AUTO_START)
185 newoid = 0x7fffffff;
186
187 oid_number = newoid;
188 }
189
190 /*
191 * Insert the OID into the parent's list sorted by OID number.
192 */
193 retry:
194 q = NULL;
195 SLIST_FOREACH(p, parent, oid_link) {
196 /* check if the current OID number is in use */
197 if (oid_number == p->oid_number) {
198 /* get the next valid OID number */
199 if (oid_number < CTL_AUTO_START ||
200 oid_number == 0x7fffffff) {
201 /* wraparound - restart */
202 oid_number = CTL_AUTO_START;
203 /* don't loop forever */
204 if (!timeout--)
205 panic("sysctl: Out of OID numbers\n");
206 goto retry;
207 } else {
208 oid_number++;
209 }
210 } else if (oid_number < p->oid_number)
211 break;
212 q = p;
213 }
214 /* check for non-auto OID number collision */
215 if (oidp->oid_number >= 0 && oidp->oid_number < CTL_AUTO_START &&
216 oid_number >= CTL_AUTO_START) {
217 printf("sysctl: OID number(%d) is already in use for '%s'\n",
218 oidp->oid_number, oidp->oid_name);
219 }
220 /* update the OID number, if any */
221 oidp->oid_number = oid_number;
222 if (q != NULL)
223 SLIST_INSERT_AFTER(q, oidp, oid_link);
224 else
225 SLIST_INSERT_HEAD(parent, oidp, oid_link);
226 }
227
228 void
229 sysctl_unregister_oid(struct sysctl_oid *oidp)
230 {
231 struct sysctl_oid *p;
232 int error;
233
234 SYSCTL_ASSERT_XLOCKED();
235 error = ENOENT;
236 if (oidp->oid_number == OID_AUTO) {
237 error = EINVAL;
238 } else {
239 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
240 if (p == oidp) {
241 SLIST_REMOVE(oidp->oid_parent, oidp,
242 sysctl_oid, oid_link);
243 error = 0;
244 break;
245 }
246 }
247 }
248
249 /*
250 * This can happen when a module fails to register and is
251 * being unloaded afterwards. It should not be a panic()
252 * for normal use.
253 */
254 if (error)
255 printf("%s: failed to unregister sysctl\n", __func__);
256 }
257
258 /* Initialize a new context to keep track of dynamically added sysctls. */
259 int
260 sysctl_ctx_init(struct sysctl_ctx_list *c)
261 {
262
263 if (c == NULL) {
264 return (EINVAL);
265 }
266
267 /*
268 * No locking here, the caller is responsible for not adding
269 * new nodes to a context until after this function has
270 * returned.
271 */
272 TAILQ_INIT(c);
273 return (0);
274 }
275
276 /* Free the context, and destroy all dynamic oids registered in this context */
277 int
278 sysctl_ctx_free(struct sysctl_ctx_list *clist)
279 {
280 struct sysctl_ctx_entry *e, *e1;
281 int error;
282
283 error = 0;
284 /*
285 * First perform a "dry run" to check if it's ok to remove oids.
286 * XXX FIXME
287 * XXX This algorithm is a hack. But I don't know any
288 * XXX better solution for now...
289 */
290 SYSCTL_XLOCK();
291 TAILQ_FOREACH(e, clist, link) {
292 error = sysctl_remove_oid_locked(e->entry, 0, 0);
293 if (error)
294 break;
295 }
296 /*
297 * Restore deregistered entries, either from the end,
298 * or from the place where error occurred.
299 * e contains the entry that was not unregistered
300 */
301 if (error)
302 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
303 else
304 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
305 while (e1 != NULL) {
306 sysctl_register_oid(e1->entry);
307 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
308 }
309 if (error) {
310 SYSCTL_XUNLOCK();
311 return(EBUSY);
312 }
313 /* Now really delete the entries */
314 e = TAILQ_FIRST(clist);
315 while (e != NULL) {
316 e1 = TAILQ_NEXT(e, link);
317 error = sysctl_remove_oid_locked(e->entry, 1, 0);
318 if (error)
319 panic("sysctl_remove_oid: corrupt tree, entry: %s",
320 e->entry->oid_name);
321 free(e, M_SYSCTLOID);
322 e = e1;
323 }
324 SYSCTL_XUNLOCK();
325 return (error);
326 }
327
328 /* Add an entry to the context */
329 struct sysctl_ctx_entry *
330 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
331 {
332 struct sysctl_ctx_entry *e;
333
334 SYSCTL_ASSERT_XLOCKED();
335 if (clist == NULL || oidp == NULL)
336 return(NULL);
337 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
338 e->entry = oidp;
339 TAILQ_INSERT_HEAD(clist, e, link);
340 return (e);
341 }
342
343 /* Find an entry in the context */
344 struct sysctl_ctx_entry *
345 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
346 {
347 struct sysctl_ctx_entry *e;
348
349 SYSCTL_ASSERT_XLOCKED();
350 if (clist == NULL || oidp == NULL)
351 return(NULL);
352 TAILQ_FOREACH(e, clist, link) {
353 if(e->entry == oidp)
354 return(e);
355 }
356 return (e);
357 }
358
359 /*
360 * Delete an entry from the context.
361 * NOTE: this function doesn't free oidp! You have to remove it
362 * with sysctl_remove_oid().
363 */
364 int
365 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
366 {
367 struct sysctl_ctx_entry *e;
368
369 if (clist == NULL || oidp == NULL)
370 return (EINVAL);
371 SYSCTL_XLOCK();
372 e = sysctl_ctx_entry_find(clist, oidp);
373 if (e != NULL) {
374 TAILQ_REMOVE(clist, e, link);
375 SYSCTL_XUNLOCK();
376 free(e, M_SYSCTLOID);
377 return (0);
378 } else {
379 SYSCTL_XUNLOCK();
380 return (ENOENT);
381 }
382 }
383
384 /*
385 * Remove dynamically created sysctl trees.
386 * oidp - top of the tree to be removed
387 * del - if 0 - just deregister, otherwise free up entries as well
388 * recurse - if != 0 traverse the subtree to be deleted
389 */
390 int
391 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
392 {
393 int error;
394
395 SYSCTL_XLOCK();
396 error = sysctl_remove_oid_locked(oidp, del, recurse);
397 SYSCTL_XUNLOCK();
398 return (error);
399 }
400
401 int
402 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
403 int del, int recurse)
404 {
405 struct sysctl_oid *p, *tmp;
406 int error;
407
408 error = ENOENT;
409 SYSCTL_XLOCK();
410 SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
411 if (strcmp(p->oid_name, name) == 0) {
412 error = sysctl_remove_oid_locked(p, del, recurse);
413 break;
414 }
415 }
416 SYSCTL_XUNLOCK();
417
418 return (error);
419 }
420
421
422 static int
423 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
424 {
425 struct sysctl_oid *p, *tmp;
426 int error;
427
428 SYSCTL_ASSERT_XLOCKED();
429 if (oidp == NULL)
430 return(EINVAL);
431 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
432 printf("Warning: can't remove non-dynamic nodes (%s)!\n",
433 oidp->oid_name);
434 return (EINVAL);
435 }
436 /*
437 * WARNING: normal method to do this should be through
438 * sysctl_ctx_free(). Use recursing as the last resort
439 * method to purge your sysctl tree of leftovers...
440 * However, if some other code still references these nodes,
441 * it will panic.
442 */
443 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
444 if (oidp->oid_refcnt == 1) {
445 SLIST_FOREACH_SAFE(p,
446 SYSCTL_CHILDREN(oidp), oid_link, tmp) {
447 if (!recurse) {
448 printf("Warning: failed attempt to "
449 "remove oid %s with child %s\n",
450 oidp->oid_name, p->oid_name);
451 return (ENOTEMPTY);
452 }
453 error = sysctl_remove_oid_locked(p, del,
454 recurse);
455 if (error)
456 return (error);
457 }
458 if (del)
459 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
460 }
461 }
462 if (oidp->oid_refcnt > 1 ) {
463 oidp->oid_refcnt--;
464 } else {
465 if (oidp->oid_refcnt == 0) {
466 printf("Warning: bad oid_refcnt=%u (%s)!\n",
467 oidp->oid_refcnt, oidp->oid_name);
468 return (EINVAL);
469 }
470 sysctl_unregister_oid(oidp);
471 if (del) {
472 /*
473 * Wait for all threads running the handler to drain.
474 * This preserves the previous behavior when the
475 * sysctl lock was held across a handler invocation,
476 * and is necessary for module unload correctness.
477 */
478 while (oidp->oid_running > 0) {
479 oidp->oid_kind |= CTLFLAG_DYING;
480 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
481 }
482 if (oidp->oid_descr)
483 free(__DECONST(char *, oidp->oid_descr),
484 M_SYSCTLOID);
485 free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
486 free(oidp, M_SYSCTLOID);
487 }
488 }
489 return (0);
490 }
491 /*
492 * Create new sysctls at run time.
493 * clist may point to a valid context initialized with sysctl_ctx_init().
494 */
495 struct sysctl_oid *
496 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
497 int number, const char *name, int kind, void *arg1, intptr_t arg2,
498 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
499 {
500 struct sysctl_oid *oidp;
501
502 /* You have to hook up somewhere.. */
503 if (parent == NULL)
504 return(NULL);
505 /* Check if the node already exists, otherwise create it */
506 SYSCTL_XLOCK();
507 oidp = sysctl_find_oidname(name, parent);
508 if (oidp != NULL) {
509 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
510 oidp->oid_refcnt++;
511 /* Update the context */
512 if (clist != NULL)
513 sysctl_ctx_entry_add(clist, oidp);
514 SYSCTL_XUNLOCK();
515 return (oidp);
516 } else {
517 SYSCTL_XUNLOCK();
518 printf("can't re-use a leaf (%s)!\n", name);
519 return (NULL);
520 }
521 }
522 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
523 oidp->oid_parent = parent;
524 SLIST_NEXT(oidp, oid_link) = NULL;
525 oidp->oid_number = number;
526 oidp->oid_refcnt = 1;
527 oidp->oid_name = strdup(name, M_SYSCTLOID);
528 oidp->oid_handler = handler;
529 oidp->oid_kind = CTLFLAG_DYN | kind;
530 if ((kind & CTLTYPE) == CTLTYPE_NODE) {
531 /* Allocate space for children */
532 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
533 M_SYSCTLOID, M_WAITOK));
534 SLIST_INIT(SYSCTL_CHILDREN(oidp));
535 oidp->oid_arg2 = arg2;
536 } else {
537 oidp->oid_arg1 = arg1;
538 oidp->oid_arg2 = arg2;
539 }
540 oidp->oid_fmt = fmt;
541 if (descr)
542 oidp->oid_descr = strdup(descr, M_SYSCTLOID);
543 /* Update the context, if used */
544 if (clist != NULL)
545 sysctl_ctx_entry_add(clist, oidp);
546 /* Register this oid */
547 sysctl_register_oid(oidp);
548 SYSCTL_XUNLOCK();
549 return (oidp);
550 }
551
552 /*
553 * Rename an existing oid.
554 */
555 void
556 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
557 {
558 char *newname;
559 char *oldname;
560
561 newname = strdup(name, M_SYSCTLOID);
562 SYSCTL_XLOCK();
563 oldname = __DECONST(char *, oidp->oid_name);
564 oidp->oid_name = newname;
565 SYSCTL_XUNLOCK();
566 free(oldname, M_SYSCTLOID);
567 }
568
569 /*
570 * Reparent an existing oid.
571 */
572 int
573 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
574 {
575 struct sysctl_oid *oidp;
576
577 SYSCTL_XLOCK();
578 if (oid->oid_parent == parent) {
579 SYSCTL_XUNLOCK();
580 return (0);
581 }
582 oidp = sysctl_find_oidname(oid->oid_name, parent);
583 if (oidp != NULL) {
584 SYSCTL_XUNLOCK();
585 return (EEXIST);
586 }
587 sysctl_unregister_oid(oid);
588 oid->oid_parent = parent;
589 oid->oid_number = OID_AUTO;
590 sysctl_register_oid(oid);
591 SYSCTL_XUNLOCK();
592 return (0);
593 }
594
595 /*
596 * Register the kernel's oids on startup.
597 */
598 SET_DECLARE(sysctl_set, struct sysctl_oid);
599
600 static void
601 sysctl_register_all(void *arg)
602 {
603 struct sysctl_oid **oidp;
604
605 sx_init(&sysctlmemlock, "sysctl mem");
606 SYSCTL_INIT();
607 SYSCTL_XLOCK();
608 SET_FOREACH(oidp, sysctl_set)
609 sysctl_register_oid(*oidp);
610 SYSCTL_XUNLOCK();
611 }
612 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
613
614 /*
615 * "Staff-functions"
616 *
617 * These functions implement a presently undocumented interface
618 * used by the sysctl program to walk the tree, and get the type
619 * so it can print the value.
620 * This interface is under work and consideration, and should probably
621 * be killed with a big axe by the first person who can find the time.
622 * (be aware though, that the proper interface isn't as obvious as it
623 * may seem, there are various conflicting requirements.
624 *
625 * {0,0} printf the entire MIB-tree.
626 * {0,1,...} return the name of the "..." OID.
627 * {0,2,...} return the next OID.
628 * {0,3} return the OID of the name in "new"
629 * {0,4,...} return the kind & format info for the "..." OID.
630 * {0,5,...} return the description the "..." OID.
631 */
632
633 #ifdef SYSCTL_DEBUG
634 static void
635 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
636 {
637 int k;
638 struct sysctl_oid *oidp;
639
640 SYSCTL_ASSERT_XLOCKED();
641 SLIST_FOREACH(oidp, l, oid_link) {
642
643 for (k=0; k<i; k++)
644 printf(" ");
645
646 printf("%d %s ", oidp->oid_number, oidp->oid_name);
647
648 printf("%c%c",
649 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
650 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
651
652 if (oidp->oid_handler)
653 printf(" *Handler");
654
655 switch (oidp->oid_kind & CTLTYPE) {
656 case CTLTYPE_NODE:
657 printf(" Node\n");
658 if (!oidp->oid_handler) {
659 sysctl_sysctl_debug_dump_node(
660 oidp->oid_arg1, i+2);
661 }
662 break;
663 case CTLTYPE_INT: printf(" Int\n"); break;
664 case CTLTYPE_UINT: printf(" u_int\n"); break;
665 case CTLTYPE_LONG: printf(" Long\n"); break;
666 case CTLTYPE_ULONG: printf(" u_long\n"); break;
667 case CTLTYPE_STRING: printf(" String\n"); break;
668 case CTLTYPE_U64: printf(" uint64_t\n"); break;
669 case CTLTYPE_S64: printf(" int64_t\n"); break;
670 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
671 default: printf("\n");
672 }
673
674 }
675 }
676
677 static int
678 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
679 {
680 int error;
681
682 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
683 if (error)
684 return (error);
685 SYSCTL_XLOCK();
686 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
687 SYSCTL_XUNLOCK();
688 return (ENOENT);
689 }
690
691 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
692 0, 0, sysctl_sysctl_debug, "-", "");
693 #endif
694
695 static int
696 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
697 {
698 int *name = (int *) arg1;
699 u_int namelen = arg2;
700 int error = 0;
701 struct sysctl_oid *oid;
702 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
703 char buf[10];
704
705 SYSCTL_XLOCK();
706 while (namelen) {
707 if (!lsp) {
708 snprintf(buf,sizeof(buf),"%d",*name);
709 if (req->oldidx)
710 error = SYSCTL_OUT(req, ".", 1);
711 if (!error)
712 error = SYSCTL_OUT(req, buf, strlen(buf));
713 if (error)
714 goto out;
715 namelen--;
716 name++;
717 continue;
718 }
719 lsp2 = 0;
720 SLIST_FOREACH(oid, lsp, oid_link) {
721 if (oid->oid_number != *name)
722 continue;
723
724 if (req->oldidx)
725 error = SYSCTL_OUT(req, ".", 1);
726 if (!error)
727 error = SYSCTL_OUT(req, oid->oid_name,
728 strlen(oid->oid_name));
729 if (error)
730 goto out;
731
732 namelen--;
733 name++;
734
735 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
736 break;
737
738 if (oid->oid_handler)
739 break;
740
741 lsp2 = SYSCTL_CHILDREN(oid);
742 break;
743 }
744 lsp = lsp2;
745 }
746 error = SYSCTL_OUT(req, "", 1);
747 out:
748 SYSCTL_XUNLOCK();
749 return (error);
750 }
751
752 /*
753 * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
754 * capability mode.
755 */
756 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD | CTLFLAG_CAPRD,
757 sysctl_sysctl_name, "");
758
759 static int
760 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
761 int *next, int *len, int level, struct sysctl_oid **oidpp)
762 {
763 struct sysctl_oid *oidp;
764
765 SYSCTL_ASSERT_XLOCKED();
766 *len = level;
767 SLIST_FOREACH(oidp, lsp, oid_link) {
768 *next = oidp->oid_number;
769 *oidpp = oidp;
770
771 if (oidp->oid_kind & CTLFLAG_SKIP)
772 continue;
773
774 if (!namelen) {
775 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
776 return (0);
777 if (oidp->oid_handler)
778 /* We really should call the handler here...*/
779 return (0);
780 lsp = SYSCTL_CHILDREN(oidp);
781 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
782 len, level+1, oidpp))
783 return (0);
784 goto emptynode;
785 }
786
787 if (oidp->oid_number < *name)
788 continue;
789
790 if (oidp->oid_number > *name) {
791 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
792 return (0);
793 if (oidp->oid_handler)
794 return (0);
795 lsp = SYSCTL_CHILDREN(oidp);
796 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
797 next+1, len, level+1, oidpp))
798 return (0);
799 goto next;
800 }
801 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
802 continue;
803
804 if (oidp->oid_handler)
805 continue;
806
807 lsp = SYSCTL_CHILDREN(oidp);
808 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
809 len, level+1, oidpp))
810 return (0);
811 next:
812 namelen = 1;
813 emptynode:
814 *len = level;
815 }
816 return (1);
817 }
818
819 static int
820 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
821 {
822 int *name = (int *) arg1;
823 u_int namelen = arg2;
824 int i, j, error;
825 struct sysctl_oid *oid;
826 struct sysctl_oid_list *lsp = &sysctl__children;
827 int newoid[CTL_MAXNAME];
828
829 SYSCTL_XLOCK();
830 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
831 SYSCTL_XUNLOCK();
832 if (i)
833 return (ENOENT);
834 error = SYSCTL_OUT(req, newoid, j * sizeof (int));
835 return (error);
836 }
837
838 /*
839 * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
840 * capability mode.
841 */
842 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD | CTLFLAG_CAPRD,
843 sysctl_sysctl_next, "");
844
845 static int
846 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
847 {
848 struct sysctl_oid *oidp;
849 struct sysctl_oid_list *lsp = &sysctl__children;
850 char *p;
851
852 SYSCTL_ASSERT_XLOCKED();
853
854 for (*len = 0; *len < CTL_MAXNAME;) {
855 p = strsep(&name, ".");
856
857 oidp = SLIST_FIRST(lsp);
858 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
859 if (oidp == NULL)
860 return (ENOENT);
861 if (strcmp(p, oidp->oid_name) == 0)
862 break;
863 }
864 *oid++ = oidp->oid_number;
865 (*len)++;
866
867 if (name == NULL || *name == '\0') {
868 if (oidpp)
869 *oidpp = oidp;
870 return (0);
871 }
872
873 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
874 break;
875
876 if (oidp->oid_handler)
877 break;
878
879 lsp = SYSCTL_CHILDREN(oidp);
880 }
881 return (ENOENT);
882 }
883
884 static int
885 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
886 {
887 char *p;
888 int error, oid[CTL_MAXNAME], len = 0;
889 struct sysctl_oid *op = 0;
890
891 if (!req->newlen)
892 return (ENOENT);
893 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
894 return (ENAMETOOLONG);
895
896 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
897
898 error = SYSCTL_IN(req, p, req->newlen);
899 if (error) {
900 free(p, M_SYSCTL);
901 return (error);
902 }
903
904 p [req->newlen] = '\0';
905
906 SYSCTL_XLOCK();
907 error = name2oid(p, oid, &len, &op);
908 SYSCTL_XUNLOCK();
909
910 free(p, M_SYSCTL);
911
912 if (error)
913 return (error);
914
915 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
916 return (error);
917 }
918
919 /*
920 * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
921 * capability mode.
922 */
923 SYSCTL_PROC(_sysctl, 3, name2oid,
924 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE
925 | CTLFLAG_CAPRW, 0, 0, sysctl_sysctl_name2oid, "I", "");
926
927 static int
928 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
929 {
930 struct sysctl_oid *oid;
931 int error;
932
933 SYSCTL_XLOCK();
934 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
935 if (error)
936 goto out;
937
938 if (oid->oid_fmt == NULL) {
939 error = ENOENT;
940 goto out;
941 }
942 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
943 if (error)
944 goto out;
945 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
946 out:
947 SYSCTL_XUNLOCK();
948 return (error);
949 }
950
951
952 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
953 sysctl_sysctl_oidfmt, "");
954
955 static int
956 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
957 {
958 struct sysctl_oid *oid;
959 int error;
960
961 SYSCTL_XLOCK();
962 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
963 if (error)
964 goto out;
965
966 if (oid->oid_descr == NULL) {
967 error = ENOENT;
968 goto out;
969 }
970 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
971 out:
972 SYSCTL_XUNLOCK();
973 return (error);
974 }
975
976 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD|CTLFLAG_CAPRD,
977 sysctl_sysctl_oiddescr, "");
978
979 /*
980 * Default "handler" functions.
981 */
982
983 /*
984 * Handle an int, signed or unsigned.
985 * Two cases:
986 * a variable: point arg1 at it.
987 * a constant: pass it in arg2.
988 */
989
990 int
991 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
992 {
993 int tmpout, error = 0;
994
995 /*
996 * Attempt to get a coherent snapshot by making a copy of the data.
997 */
998 if (arg1)
999 tmpout = *(int *)arg1;
1000 else
1001 tmpout = arg2;
1002 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1003
1004 if (error || !req->newptr)
1005 return (error);
1006
1007 if (!arg1)
1008 error = EPERM;
1009 else
1010 error = SYSCTL_IN(req, arg1, sizeof(int));
1011 return (error);
1012 }
1013
1014 /*
1015 * Based on on sysctl_handle_int() convert milliseconds into ticks.
1016 * Note: this is used by TCP.
1017 */
1018
1019 int
1020 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1021 {
1022 int error, s, tt;
1023
1024 tt = *(int *)arg1;
1025 s = (int)((int64_t)tt * 1000 / hz);
1026
1027 error = sysctl_handle_int(oidp, &s, 0, req);
1028 if (error || !req->newptr)
1029 return (error);
1030
1031 tt = (int)((int64_t)s * hz / 1000);
1032 if (tt < 1)
1033 return (EINVAL);
1034
1035 *(int *)arg1 = tt;
1036 return (0);
1037 }
1038
1039
1040 /*
1041 * Handle a long, signed or unsigned.
1042 * Two cases:
1043 * a variable: point arg1 at it.
1044 * a constant: pass it in arg2.
1045 */
1046
1047 int
1048 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1049 {
1050 int error = 0;
1051 long tmplong;
1052 #ifdef SCTL_MASK32
1053 int tmpint;
1054 #endif
1055
1056 /*
1057 * Attempt to get a coherent snapshot by making a copy of the data.
1058 */
1059 if (arg1)
1060 tmplong = *(long *)arg1;
1061 else
1062 tmplong = arg2;
1063 #ifdef SCTL_MASK32
1064 if (req->flags & SCTL_MASK32) {
1065 tmpint = tmplong;
1066 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1067 } else
1068 #endif
1069 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1070
1071 if (error || !req->newptr)
1072 return (error);
1073
1074 if (!arg1)
1075 error = EPERM;
1076 #ifdef SCTL_MASK32
1077 else if (req->flags & SCTL_MASK32) {
1078 error = SYSCTL_IN(req, &tmpint, sizeof(int));
1079 *(long *)arg1 = (long)tmpint;
1080 }
1081 #endif
1082 else
1083 error = SYSCTL_IN(req, arg1, sizeof(long));
1084 return (error);
1085 }
1086
1087 /*
1088 * Handle a 64 bit int, signed or unsigned.
1089 * Two cases:
1090 * a variable: point arg1 at it.
1091 * a constant: pass it in arg2.
1092 */
1093 int
1094 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1095 {
1096 int error = 0;
1097 uint64_t tmpout;
1098
1099 /*
1100 * Attempt to get a coherent snapshot by making a copy of the data.
1101 */
1102 if (arg1)
1103 tmpout = *(uint64_t *)arg1;
1104 else
1105 tmpout = arg2;
1106 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1107
1108 if (error || !req->newptr)
1109 return (error);
1110
1111 if (!arg1)
1112 error = EPERM;
1113 else
1114 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1115 return (error);
1116 }
1117
1118 /*
1119 * Handle our generic '\0' terminated 'C' string.
1120 * Two cases:
1121 * a variable string: point arg1 at it, arg2 is max length.
1122 * a constant string: point arg1 at it, arg2 is zero.
1123 */
1124
1125 int
1126 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1127 {
1128 int error=0;
1129 char *tmparg;
1130 size_t outlen;
1131
1132 /*
1133 * Attempt to get a coherent snapshot by copying to a
1134 * temporary kernel buffer.
1135 */
1136 retry:
1137 outlen = strlen((char *)arg1)+1;
1138 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
1139
1140 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
1141 free(tmparg, M_SYSCTLTMP);
1142 goto retry;
1143 }
1144
1145 error = SYSCTL_OUT(req, tmparg, outlen);
1146 free(tmparg, M_SYSCTLTMP);
1147
1148 if (error || !req->newptr)
1149 return (error);
1150
1151 if ((req->newlen - req->newidx) >= arg2) {
1152 error = EINVAL;
1153 } else {
1154 arg2 = (req->newlen - req->newidx);
1155 error = SYSCTL_IN(req, arg1, arg2);
1156 ((char *)arg1)[arg2] = '\0';
1157 }
1158
1159 return (error);
1160 }
1161
1162 /*
1163 * Handle any kind of opaque data.
1164 * arg1 points to it, arg2 is the size.
1165 */
1166
1167 int
1168 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1169 {
1170 int error, tries;
1171 u_int generation;
1172 struct sysctl_req req2;
1173
1174 /*
1175 * Attempt to get a coherent snapshot, by using the thread
1176 * pre-emption counter updated from within mi_switch() to
1177 * determine if we were pre-empted during a bcopy() or
1178 * copyout(). Make 3 attempts at doing this before giving up.
1179 * If we encounter an error, stop immediately.
1180 */
1181 tries = 0;
1182 req2 = *req;
1183 retry:
1184 generation = curthread->td_generation;
1185 error = SYSCTL_OUT(req, arg1, arg2);
1186 if (error)
1187 return (error);
1188 tries++;
1189 if (generation != curthread->td_generation && tries < 3) {
1190 *req = req2;
1191 goto retry;
1192 }
1193
1194 error = SYSCTL_IN(req, arg1, arg2);
1195
1196 return (error);
1197 }
1198
1199 /*
1200 * Transfer functions to/from kernel space.
1201 * XXX: rather untested at this point
1202 */
1203 static int
1204 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1205 {
1206 size_t i = 0;
1207
1208 if (req->oldptr) {
1209 i = l;
1210 if (req->oldlen <= req->oldidx)
1211 i = 0;
1212 else
1213 if (i > req->oldlen - req->oldidx)
1214 i = req->oldlen - req->oldidx;
1215 if (i > 0)
1216 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1217 }
1218 req->oldidx += l;
1219 if (req->oldptr && i != l)
1220 return (ENOMEM);
1221 return (0);
1222 }
1223
1224 static int
1225 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1226 {
1227 if (!req->newptr)
1228 return (0);
1229 if (req->newlen - req->newidx < l)
1230 return (EINVAL);
1231 bcopy((char *)req->newptr + req->newidx, p, l);
1232 req->newidx += l;
1233 return (0);
1234 }
1235
1236 int
1237 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1238 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1239 {
1240 int error = 0;
1241 struct sysctl_req req;
1242
1243 bzero(&req, sizeof req);
1244
1245 req.td = td;
1246 req.flags = flags;
1247
1248 if (oldlenp) {
1249 req.oldlen = *oldlenp;
1250 }
1251 req.validlen = req.oldlen;
1252
1253 if (old) {
1254 req.oldptr= old;
1255 }
1256
1257 if (new != NULL) {
1258 req.newlen = newlen;
1259 req.newptr = new;
1260 }
1261
1262 req.oldfunc = sysctl_old_kernel;
1263 req.newfunc = sysctl_new_kernel;
1264 req.lock = REQ_UNWIRED;
1265
1266 SYSCTL_XLOCK();
1267 error = sysctl_root(0, name, namelen, &req);
1268 SYSCTL_XUNLOCK();
1269
1270 if (req.lock == REQ_WIRED && req.validlen > 0)
1271 vsunlock(req.oldptr, req.validlen);
1272
1273 if (error && error != ENOMEM)
1274 return (error);
1275
1276 if (retval) {
1277 if (req.oldptr && req.oldidx > req.validlen)
1278 *retval = req.validlen;
1279 else
1280 *retval = req.oldidx;
1281 }
1282 return (error);
1283 }
1284
1285 int
1286 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1287 void *new, size_t newlen, size_t *retval, int flags)
1288 {
1289 int oid[CTL_MAXNAME];
1290 size_t oidlen, plen;
1291 int error;
1292
1293 oid[0] = 0; /* sysctl internal magic */
1294 oid[1] = 3; /* name2oid */
1295 oidlen = sizeof(oid);
1296
1297 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1298 (void *)name, strlen(name), &plen, flags);
1299 if (error)
1300 return (error);
1301
1302 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1303 new, newlen, retval, flags);
1304 return (error);
1305 }
1306
1307 /*
1308 * Transfer function to/from user space.
1309 */
1310 static int
1311 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1312 {
1313 size_t i, len, origidx;
1314 int error;
1315
1316 origidx = req->oldidx;
1317 req->oldidx += l;
1318 if (req->oldptr == NULL)
1319 return (0);
1320 /*
1321 * If we have not wired the user supplied buffer and we are currently
1322 * holding locks, drop a witness warning, as it's possible that
1323 * write operations to the user page can sleep.
1324 */
1325 if (req->lock != REQ_WIRED)
1326 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1327 "sysctl_old_user()");
1328 i = l;
1329 len = req->validlen;
1330 if (len <= origidx)
1331 i = 0;
1332 else {
1333 if (i > len - origidx)
1334 i = len - origidx;
1335 if (req->lock == REQ_WIRED) {
1336 error = copyout_nofault(p, (char *)req->oldptr +
1337 origidx, i);
1338 } else
1339 error = copyout(p, (char *)req->oldptr + origidx, i);
1340 if (error != 0)
1341 return (error);
1342 }
1343 if (i < l)
1344 return (ENOMEM);
1345 return (0);
1346 }
1347
1348 static int
1349 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1350 {
1351 int error;
1352
1353 if (!req->newptr)
1354 return (0);
1355 if (req->newlen - req->newidx < l)
1356 return (EINVAL);
1357 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1358 "sysctl_new_user()");
1359 error = copyin((char *)req->newptr + req->newidx, p, l);
1360 req->newidx += l;
1361 return (error);
1362 }
1363
1364 /*
1365 * Wire the user space destination buffer. If set to a value greater than
1366 * zero, the len parameter limits the maximum amount of wired memory.
1367 */
1368 int
1369 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1370 {
1371 int ret;
1372 size_t wiredlen;
1373
1374 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1375 ret = 0;
1376 if (req->lock != REQ_WIRED && req->oldptr &&
1377 req->oldfunc == sysctl_old_user) {
1378 if (wiredlen != 0) {
1379 ret = vslock(req->oldptr, wiredlen);
1380 if (ret != 0) {
1381 if (ret != ENOMEM)
1382 return (ret);
1383 wiredlen = 0;
1384 }
1385 }
1386 req->lock = REQ_WIRED;
1387 req->validlen = wiredlen;
1388 }
1389 return (0);
1390 }
1391
1392 int
1393 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1394 int *nindx, struct sysctl_req *req)
1395 {
1396 struct sysctl_oid_list *lsp;
1397 struct sysctl_oid *oid;
1398 int indx;
1399
1400 SYSCTL_ASSERT_XLOCKED();
1401 lsp = &sysctl__children;
1402 indx = 0;
1403 while (indx < CTL_MAXNAME) {
1404 SLIST_FOREACH(oid, lsp, oid_link) {
1405 if (oid->oid_number == name[indx])
1406 break;
1407 }
1408 if (oid == NULL)
1409 return (ENOENT);
1410
1411 indx++;
1412 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1413 if (oid->oid_handler != NULL || indx == namelen) {
1414 *noid = oid;
1415 if (nindx != NULL)
1416 *nindx = indx;
1417 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1418 ("%s found DYING node %p", __func__, oid));
1419 return (0);
1420 }
1421 lsp = SYSCTL_CHILDREN(oid);
1422 } else if (indx == namelen) {
1423 *noid = oid;
1424 if (nindx != NULL)
1425 *nindx = indx;
1426 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
1427 ("%s found DYING node %p", __func__, oid));
1428 return (0);
1429 } else {
1430 return (ENOTDIR);
1431 }
1432 }
1433 return (ENOENT);
1434 }
1435
1436 /*
1437 * Traverse our tree, and find the right node, execute whatever it points
1438 * to, and return the resulting error code.
1439 */
1440
1441 static int
1442 sysctl_root(SYSCTL_HANDLER_ARGS)
1443 {
1444 struct sysctl_oid *oid;
1445 int error, indx, lvl;
1446
1447 SYSCTL_ASSERT_XLOCKED();
1448
1449 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1450 if (error)
1451 return (error);
1452
1453 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1454 /*
1455 * You can't call a sysctl when it's a node, but has
1456 * no handler. Inform the user that it's a node.
1457 * The indx may or may not be the same as namelen.
1458 */
1459 if (oid->oid_handler == NULL)
1460 return (EISDIR);
1461 }
1462
1463 /* Is this sysctl writable? */
1464 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1465 return (EPERM);
1466
1467 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1468
1469 #ifdef CAPABILITY_MODE
1470 /*
1471 * If the process is in capability mode, then don't permit reading or
1472 * writing unless specifically granted for the node.
1473 */
1474 if (IN_CAPABILITY_MODE(req->td)) {
1475 if (req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD))
1476 return (EPERM);
1477 if (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))
1478 return (EPERM);
1479 }
1480 #endif
1481
1482 /* Is this sysctl sensitive to securelevels? */
1483 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1484 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1485 error = securelevel_gt(req->td->td_ucred, lvl);
1486 if (error)
1487 return (error);
1488 }
1489
1490 /* Is this sysctl writable by only privileged users? */
1491 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1492 int priv;
1493
1494 if (oid->oid_kind & CTLFLAG_PRISON)
1495 priv = PRIV_SYSCTL_WRITEJAIL;
1496 #ifdef VIMAGE
1497 else if ((oid->oid_kind & CTLFLAG_VNET) &&
1498 prison_owns_vnet(req->td->td_ucred))
1499 priv = PRIV_SYSCTL_WRITEJAIL;
1500 #endif
1501 else
1502 priv = PRIV_SYSCTL_WRITE;
1503 error = priv_check(req->td, priv);
1504 if (error)
1505 return (error);
1506 }
1507
1508 if (!oid->oid_handler)
1509 return (EINVAL);
1510
1511 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1512 arg1 = (int *)arg1 + indx;
1513 arg2 -= indx;
1514 } else {
1515 arg1 = oid->oid_arg1;
1516 arg2 = oid->oid_arg2;
1517 }
1518 #ifdef MAC
1519 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1520 req);
1521 if (error != 0)
1522 return (error);
1523 #endif
1524 oid->oid_running++;
1525 SYSCTL_XUNLOCK();
1526 #ifdef VIMAGE
1527 if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
1528 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
1529 #endif
1530 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1531 mtx_lock(&Giant);
1532 error = oid->oid_handler(oid, arg1, arg2, req);
1533 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
1534 mtx_unlock(&Giant);
1535
1536 KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
1537
1538 SYSCTL_XLOCK();
1539 oid->oid_running--;
1540 if (oid->oid_running == 0 && (oid->oid_kind & CTLFLAG_DYING) != 0)
1541 wakeup(&oid->oid_running);
1542 return (error);
1543 }
1544
1545 #ifndef _SYS_SYSPROTO_H_
1546 struct sysctl_args {
1547 int *name;
1548 u_int namelen;
1549 void *old;
1550 size_t *oldlenp;
1551 void *new;
1552 size_t newlen;
1553 };
1554 #endif
1555 int
1556 sys___sysctl(struct thread *td, struct sysctl_args *uap)
1557 {
1558 int error, i, name[CTL_MAXNAME];
1559 size_t j;
1560
1561 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1562 return (EINVAL);
1563
1564 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1565 if (error)
1566 return (error);
1567
1568 error = userland_sysctl(td, name, uap->namelen,
1569 uap->old, uap->oldlenp, 0,
1570 uap->new, uap->newlen, &j, 0);
1571 if (error && error != ENOMEM)
1572 return (error);
1573 if (uap->oldlenp) {
1574 i = copyout(&j, uap->oldlenp, sizeof(j));
1575 if (i)
1576 return (i);
1577 }
1578 return (error);
1579 }
1580
1581 /*
1582 * This is used from various compatibility syscalls too. That's why name
1583 * must be in kernel space.
1584 */
1585 int
1586 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1587 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1588 int flags)
1589 {
1590 int error = 0, memlocked;
1591 struct sysctl_req req;
1592
1593 bzero(&req, sizeof req);
1594
1595 req.td = td;
1596 req.flags = flags;
1597
1598 if (oldlenp) {
1599 if (inkernel) {
1600 req.oldlen = *oldlenp;
1601 } else {
1602 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1603 if (error)
1604 return (error);
1605 }
1606 }
1607 req.validlen = req.oldlen;
1608
1609 if (old) {
1610 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1611 return (EFAULT);
1612 req.oldptr= old;
1613 }
1614
1615 if (new != NULL) {
1616 if (!useracc(new, newlen, VM_PROT_READ))
1617 return (EFAULT);
1618 req.newlen = newlen;
1619 req.newptr = new;
1620 }
1621
1622 req.oldfunc = sysctl_old_user;
1623 req.newfunc = sysctl_new_user;
1624 req.lock = REQ_UNWIRED;
1625
1626 #ifdef KTRACE
1627 if (KTRPOINT(curthread, KTR_SYSCTL))
1628 ktrsysctl(name, namelen);
1629 #endif
1630
1631 if (req.oldlen > PAGE_SIZE) {
1632 memlocked = 1;
1633 sx_xlock(&sysctlmemlock);
1634 } else
1635 memlocked = 0;
1636 CURVNET_SET(TD_TO_VNET(td));
1637
1638 for (;;) {
1639 req.oldidx = 0;
1640 req.newidx = 0;
1641 SYSCTL_XLOCK();
1642 error = sysctl_root(0, name, namelen, &req);
1643 SYSCTL_XUNLOCK();
1644 if (error != EAGAIN)
1645 break;
1646 kern_yield(PRI_USER);
1647 }
1648
1649 CURVNET_RESTORE();
1650
1651 if (req.lock == REQ_WIRED && req.validlen > 0)
1652 vsunlock(req.oldptr, req.validlen);
1653 if (memlocked)
1654 sx_xunlock(&sysctlmemlock);
1655
1656 if (error && error != ENOMEM)
1657 return (error);
1658
1659 if (retval) {
1660 if (req.oldptr && req.oldidx > req.validlen)
1661 *retval = req.validlen;
1662 else
1663 *retval = req.oldidx;
1664 }
1665 return (error);
1666 }
1667
1668 /*
1669 * Drain into a sysctl struct. The user buffer should be wired if a page
1670 * fault would cause issue.
1671 */
1672 static int
1673 sbuf_sysctl_drain(void *arg, const char *data, int len)
1674 {
1675 struct sysctl_req *req = arg;
1676 int error;
1677
1678 error = SYSCTL_OUT(req, data, len);
1679 KASSERT(error >= 0, ("Got unexpected negative value %d", error));
1680 return (error == 0 ? len : -error);
1681 }
1682
1683 struct sbuf *
1684 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
1685 struct sysctl_req *req)
1686 {
1687
1688 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN);
1689 sbuf_set_drain(s, sbuf_sysctl_drain, req);
1690 return (s);
1691 }
Cache object: 75db78c3cc0a210f29764cf3fa519080
|