1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Mike Karels at Berkeley Software Design, Inc.
9 *
10 * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
11 * project, to make these variables more userfriendly.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * @(#)kern_sysctl.c 8.4 (Berkeley) 4/14/94
38 */
39
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42
43 #include "opt_capsicum.h"
44 #include "opt_ktrace.h"
45
46 #include <sys/param.h>
47 #include <sys/fail.h>
48 #include <sys/systm.h>
49 #include <sys/capsicum.h>
50 #include <sys/kernel.h>
51 #include <sys/sysctl.h>
52 #include <sys/malloc.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/jail.h>
56 #include <sys/lock.h>
57 #include <sys/mutex.h>
58 #include <sys/rmlock.h>
59 #include <sys/sbuf.h>
60 #include <sys/sx.h>
61 #include <sys/sysproto.h>
62 #include <sys/uio.h>
63 #ifdef KTRACE
64 #include <sys/ktrace.h>
65 #endif
66
67 #include <net/vnet.h>
68
69 #include <security/mac/mac_framework.h>
70
71 #include <vm/vm.h>
72 #include <vm/vm_extern.h>
73
74 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
75 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
76 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
77
78 /*
79 * The sysctllock protects the MIB tree. It also protects sysctl
80 * contexts used with dynamic sysctls. The sysctl_register_oid() and
81 * sysctl_unregister_oid() routines require the sysctllock to already
82 * be held, so the sysctl_wlock() and sysctl_wunlock() routines are
83 * provided for the few places in the kernel which need to use that
84 * API rather than using the dynamic API. Use of the dynamic API is
85 * strongly encouraged for most code.
86 *
87 * The sysctlmemlock is used to limit the amount of user memory wired for
88 * sysctl requests. This is implemented by serializing any userland
89 * sysctl requests larger than a single page via an exclusive lock.
90 */
91 static struct rmlock sysctllock;
92 static struct sx __exclusive_cache_line sysctlmemlock;
93
94 #define SYSCTL_WLOCK() rm_wlock(&sysctllock)
95 #define SYSCTL_WUNLOCK() rm_wunlock(&sysctllock)
96 #define SYSCTL_RLOCK(tracker) rm_rlock(&sysctllock, (tracker))
97 #define SYSCTL_RUNLOCK(tracker) rm_runlock(&sysctllock, (tracker))
98 #define SYSCTL_WLOCKED() rm_wowned(&sysctllock)
99 #define SYSCTL_ASSERT_LOCKED() rm_assert(&sysctllock, RA_LOCKED)
100 #define SYSCTL_ASSERT_WLOCKED() rm_assert(&sysctllock, RA_WLOCKED)
101 #define SYSCTL_ASSERT_RLOCKED() rm_assert(&sysctllock, RA_RLOCKED)
102 #define SYSCTL_INIT() rm_init_flags(&sysctllock, "sysctl lock", \
103 RM_SLEEPABLE)
104 #define SYSCTL_SLEEP(ch, wmesg, timo) \
105 rm_sleep(ch, &sysctllock, 0, wmesg, timo)
106
107 static int sysctl_root(SYSCTL_HANDLER_ARGS);
108
109 /* Root list */
110 struct sysctl_oid_list sysctl__children = SLIST_HEAD_INITIALIZER(&sysctl__children);
111
112 static char* sysctl_escape_name(const char*);
113 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
114 int recurse);
115 static int sysctl_old_kernel(struct sysctl_req *, const void *, size_t);
116 static int sysctl_new_kernel(struct sysctl_req *, void *, size_t);
117
118 static struct sysctl_oid *
119 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
120 {
121 struct sysctl_oid *oidp;
122
123 SYSCTL_ASSERT_LOCKED();
124 SLIST_FOREACH(oidp, list, oid_link) {
125 if (strcmp(oidp->oid_name, name) == 0) {
126 return (oidp);
127 }
128 }
129 return (NULL);
130 }
131
132 /*
133 * Initialization of the MIB tree.
134 *
135 * Order by number in each list.
136 */
137 void
138 sysctl_wlock(void)
139 {
140
141 SYSCTL_WLOCK();
142 }
143
144 void
145 sysctl_wunlock(void)
146 {
147
148 SYSCTL_WUNLOCK();
149 }
150
151 static int
152 sysctl_root_handler_locked(struct sysctl_oid *oid, void *arg1, intmax_t arg2,
153 struct sysctl_req *req, struct rm_priotracker *tracker)
154 {
155 int error;
156
157 if (oid->oid_kind & CTLFLAG_DYN)
158 atomic_add_int(&oid->oid_running, 1);
159
160 if (tracker != NULL)
161 SYSCTL_RUNLOCK(tracker);
162 else
163 SYSCTL_WUNLOCK();
164
165 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
166 mtx_lock(&Giant);
167 error = oid->oid_handler(oid, arg1, arg2, req);
168 if (!(oid->oid_kind & CTLFLAG_MPSAFE))
169 mtx_unlock(&Giant);
170
171 KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
172
173 if (tracker != NULL)
174 SYSCTL_RLOCK(tracker);
175 else
176 SYSCTL_WLOCK();
177
178 if (oid->oid_kind & CTLFLAG_DYN) {
179 if (atomic_fetchadd_int(&oid->oid_running, -1) == 1 &&
180 (oid->oid_kind & CTLFLAG_DYING) != 0)
181 wakeup(&oid->oid_running);
182 }
183
184 return (error);
185 }
186
187 static void
188 sysctl_load_tunable_by_oid_locked(struct sysctl_oid *oidp)
189 {
190 struct sysctl_req req;
191 struct sysctl_oid *curr;
192 char *penv = NULL;
193 char path[96];
194 ssize_t rem = sizeof(path);
195 ssize_t len;
196 uint8_t data[512] __aligned(sizeof(uint64_t));
197 int size;
198 int error;
199
200 path[--rem] = 0;
201
202 for (curr = oidp; curr != NULL; curr = SYSCTL_PARENT(curr)) {
203 len = strlen(curr->oid_name);
204 rem -= len;
205 if (curr != oidp)
206 rem -= 1;
207 if (rem < 0) {
208 printf("OID path exceeds %d bytes\n", (int)sizeof(path));
209 return;
210 }
211 memcpy(path + rem, curr->oid_name, len);
212 if (curr != oidp)
213 path[rem + len] = '.';
214 }
215
216 memset(&req, 0, sizeof(req));
217
218 req.td = curthread;
219 req.oldfunc = sysctl_old_kernel;
220 req.newfunc = sysctl_new_kernel;
221 req.lock = REQ_UNWIRED;
222
223 switch (oidp->oid_kind & CTLTYPE) {
224 case CTLTYPE_INT:
225 if (getenv_array(path + rem, data, sizeof(data), &size,
226 sizeof(int), GETENV_SIGNED) == 0)
227 return;
228 req.newlen = size;
229 req.newptr = data;
230 break;
231 case CTLTYPE_UINT:
232 if (getenv_array(path + rem, data, sizeof(data), &size,
233 sizeof(int), GETENV_UNSIGNED) == 0)
234 return;
235 req.newlen = size;
236 req.newptr = data;
237 break;
238 case CTLTYPE_LONG:
239 if (getenv_array(path + rem, data, sizeof(data), &size,
240 sizeof(long), GETENV_SIGNED) == 0)
241 return;
242 req.newlen = size;
243 req.newptr = data;
244 break;
245 case CTLTYPE_ULONG:
246 if (getenv_array(path + rem, data, sizeof(data), &size,
247 sizeof(long), GETENV_UNSIGNED) == 0)
248 return;
249 req.newlen = size;
250 req.newptr = data;
251 break;
252 case CTLTYPE_S8:
253 if (getenv_array(path + rem, data, sizeof(data), &size,
254 sizeof(int8_t), GETENV_SIGNED) == 0)
255 return;
256 req.newlen = size;
257 req.newptr = data;
258 break;
259 case CTLTYPE_S16:
260 if (getenv_array(path + rem, data, sizeof(data), &size,
261 sizeof(int16_t), GETENV_SIGNED) == 0)
262 return;
263 req.newlen = size;
264 req.newptr = data;
265 break;
266 case CTLTYPE_S32:
267 if (getenv_array(path + rem, data, sizeof(data), &size,
268 sizeof(int32_t), GETENV_SIGNED) == 0)
269 return;
270 req.newlen = size;
271 req.newptr = data;
272 break;
273 case CTLTYPE_S64:
274 if (getenv_array(path + rem, data, sizeof(data), &size,
275 sizeof(int64_t), GETENV_SIGNED) == 0)
276 return;
277 req.newlen = size;
278 req.newptr = data;
279 break;
280 case CTLTYPE_U8:
281 if (getenv_array(path + rem, data, sizeof(data), &size,
282 sizeof(uint8_t), GETENV_UNSIGNED) == 0)
283 return;
284 req.newlen = size;
285 req.newptr = data;
286 break;
287 case CTLTYPE_U16:
288 if (getenv_array(path + rem, data, sizeof(data), &size,
289 sizeof(uint16_t), GETENV_UNSIGNED) == 0)
290 return;
291 req.newlen = size;
292 req.newptr = data;
293 break;
294 case CTLTYPE_U32:
295 if (getenv_array(path + rem, data, sizeof(data), &size,
296 sizeof(uint32_t), GETENV_UNSIGNED) == 0)
297 return;
298 req.newlen = size;
299 req.newptr = data;
300 break;
301 case CTLTYPE_U64:
302 if (getenv_array(path + rem, data, sizeof(data), &size,
303 sizeof(uint64_t), GETENV_UNSIGNED) == 0)
304 return;
305 req.newlen = size;
306 req.newptr = data;
307 break;
308 case CTLTYPE_STRING:
309 penv = kern_getenv(path + rem);
310 if (penv == NULL)
311 return;
312 req.newlen = strlen(penv);
313 req.newptr = penv;
314 break;
315 default:
316 return;
317 }
318 error = sysctl_root_handler_locked(oidp, oidp->oid_arg1,
319 oidp->oid_arg2, &req, NULL);
320 if (error != 0)
321 printf("Setting sysctl %s failed: %d\n", path + rem, error);
322 if (penv != NULL)
323 freeenv(penv);
324 }
325
326 static int
327 sbuf_printf_drain(void *arg __unused, const char *data, int len)
328 {
329
330 return (printf("%.*s", len, data));
331 }
332
333 /*
334 * Locate the path to a given oid. Returns the length of the resulting path,
335 * or -1 if the oid was not found. nodes must have room for CTL_MAXNAME
336 * elements and be NULL initialized.
337 */
338 static int
339 sysctl_search_oid(struct sysctl_oid **nodes, struct sysctl_oid *needle)
340 {
341 int indx;
342
343 SYSCTL_ASSERT_LOCKED();
344 indx = 0;
345 while (indx < CTL_MAXNAME && indx >= 0) {
346 if (nodes[indx] == NULL && indx == 0)
347 nodes[indx] = SLIST_FIRST(&sysctl__children);
348 else if (nodes[indx] == NULL)
349 nodes[indx] = SLIST_FIRST(&nodes[indx - 1]->oid_children);
350 else
351 nodes[indx] = SLIST_NEXT(nodes[indx], oid_link);
352
353 if (nodes[indx] == needle)
354 return (indx + 1);
355
356 if (nodes[indx] == NULL) {
357 indx--;
358 continue;
359 }
360
361 if ((nodes[indx]->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
362 indx++;
363 continue;
364 }
365 }
366 return (-1);
367 }
368
369 static void
370 sysctl_warn_reuse(const char *func, struct sysctl_oid *leaf)
371 {
372 struct sysctl_oid *nodes[CTL_MAXNAME];
373 char buf[128];
374 struct sbuf sb;
375 int rc, i;
376
377 (void)sbuf_new(&sb, buf, sizeof(buf), SBUF_FIXEDLEN | SBUF_INCLUDENUL);
378 sbuf_set_drain(&sb, sbuf_printf_drain, NULL);
379
380 sbuf_printf(&sb, "%s: can't re-use a leaf (", __func__);
381
382 memset(nodes, 0, sizeof(nodes));
383 rc = sysctl_search_oid(nodes, leaf);
384 if (rc > 0) {
385 for (i = 0; i < rc; i++)
386 sbuf_printf(&sb, "%s%.*s", nodes[i]->oid_name,
387 i != (rc - 1), ".");
388 } else {
389 sbuf_printf(&sb, "%s", leaf->oid_name);
390 }
391 sbuf_printf(&sb, ")!\n");
392
393 (void)sbuf_finish(&sb);
394 }
395
396 #ifdef SYSCTL_DEBUG
397 static int
398 sysctl_reuse_test(SYSCTL_HANDLER_ARGS)
399 {
400 struct rm_priotracker tracker;
401
402 SYSCTL_RLOCK(&tracker);
403 sysctl_warn_reuse(__func__, oidp);
404 SYSCTL_RUNLOCK(&tracker);
405 return (0);
406 }
407 SYSCTL_PROC(_sysctl, 0, reuse_test, CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE,
408 0, 0, sysctl_reuse_test, "-", "");
409 #endif
410
411 void
412 sysctl_register_oid(struct sysctl_oid *oidp)
413 {
414 struct sysctl_oid_list *parent = oidp->oid_parent;
415 struct sysctl_oid *p;
416 struct sysctl_oid *q;
417 int oid_number;
418 int timeout = 2;
419
420 /*
421 * First check if another oid with the same name already
422 * exists in the parent's list.
423 */
424 SYSCTL_ASSERT_WLOCKED();
425 p = sysctl_find_oidname(oidp->oid_name, parent);
426 if (p != NULL) {
427 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
428 p->oid_refcnt++;
429 return;
430 } else {
431 sysctl_warn_reuse(__func__, p);
432 return;
433 }
434 }
435 /* get current OID number */
436 oid_number = oidp->oid_number;
437
438 #if (OID_AUTO >= 0)
439 #error "OID_AUTO is expected to be a negative value"
440 #endif
441 /*
442 * Any negative OID number qualifies as OID_AUTO. Valid OID
443 * numbers should always be positive.
444 *
445 * NOTE: DO NOT change the starting value here, change it in
446 * <sys/sysctl.h>, and make sure it is at least 256 to
447 * accommodate e.g. net.inet.raw as a static sysctl node.
448 */
449 if (oid_number < 0) {
450 static int newoid;
451
452 /*
453 * By decrementing the next OID number we spend less
454 * time inserting the OIDs into a sorted list.
455 */
456 if (--newoid < CTL_AUTO_START)
457 newoid = 0x7fffffff;
458
459 oid_number = newoid;
460 }
461
462 /*
463 * Insert the OID into the parent's list sorted by OID number.
464 */
465 retry:
466 q = NULL;
467 SLIST_FOREACH(p, parent, oid_link) {
468 /* check if the current OID number is in use */
469 if (oid_number == p->oid_number) {
470 /* get the next valid OID number */
471 if (oid_number < CTL_AUTO_START ||
472 oid_number == 0x7fffffff) {
473 /* wraparound - restart */
474 oid_number = CTL_AUTO_START;
475 /* don't loop forever */
476 if (!timeout--)
477 panic("sysctl: Out of OID numbers\n");
478 goto retry;
479 } else {
480 oid_number++;
481 }
482 } else if (oid_number < p->oid_number)
483 break;
484 q = p;
485 }
486 /* check for non-auto OID number collision */
487 if (oidp->oid_number >= 0 && oidp->oid_number < CTL_AUTO_START &&
488 oid_number >= CTL_AUTO_START) {
489 printf("sysctl: OID number(%d) is already in use for '%s'\n",
490 oidp->oid_number, oidp->oid_name);
491 }
492 /* update the OID number, if any */
493 oidp->oid_number = oid_number;
494 if (q != NULL)
495 SLIST_INSERT_AFTER(q, oidp, oid_link);
496 else
497 SLIST_INSERT_HEAD(parent, oidp, oid_link);
498
499 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE &&
500 #ifdef VIMAGE
501 (oidp->oid_kind & CTLFLAG_VNET) == 0 &&
502 #endif
503 (oidp->oid_kind & CTLFLAG_TUN) != 0 &&
504 (oidp->oid_kind & CTLFLAG_NOFETCH) == 0) {
505 /* only fetch value once */
506 oidp->oid_kind |= CTLFLAG_NOFETCH;
507 /* try to fetch value from kernel environment */
508 sysctl_load_tunable_by_oid_locked(oidp);
509 }
510 }
511
512 void
513 sysctl_register_disabled_oid(struct sysctl_oid *oidp)
514 {
515
516 /*
517 * Mark the leaf as dormant if it's not to be immediately enabled.
518 * We do not disable nodes as they can be shared between modules
519 * and it is always safe to access a node.
520 */
521 KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
522 ("internal flag is set in oid_kind"));
523 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
524 oidp->oid_kind |= CTLFLAG_DORMANT;
525 sysctl_register_oid(oidp);
526 }
527
528 void
529 sysctl_enable_oid(struct sysctl_oid *oidp)
530 {
531
532 SYSCTL_ASSERT_WLOCKED();
533 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
534 KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
535 ("sysctl node is marked as dormant"));
536 return;
537 }
538 KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) != 0,
539 ("enabling already enabled sysctl oid"));
540 oidp->oid_kind &= ~CTLFLAG_DORMANT;
541 }
542
543 void
544 sysctl_unregister_oid(struct sysctl_oid *oidp)
545 {
546 struct sysctl_oid *p;
547 int error;
548
549 SYSCTL_ASSERT_WLOCKED();
550 if (oidp->oid_number == OID_AUTO) {
551 error = EINVAL;
552 } else {
553 error = ENOENT;
554 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
555 if (p == oidp) {
556 SLIST_REMOVE(oidp->oid_parent, oidp,
557 sysctl_oid, oid_link);
558 error = 0;
559 break;
560 }
561 }
562 }
563
564 /*
565 * This can happen when a module fails to register and is
566 * being unloaded afterwards. It should not be a panic()
567 * for normal use.
568 */
569 if (error) {
570 printf("%s: failed(%d) to unregister sysctl(%s)\n",
571 __func__, error, oidp->oid_name);
572 }
573 }
574
575 /* Initialize a new context to keep track of dynamically added sysctls. */
576 int
577 sysctl_ctx_init(struct sysctl_ctx_list *c)
578 {
579
580 if (c == NULL) {
581 return (EINVAL);
582 }
583
584 /*
585 * No locking here, the caller is responsible for not adding
586 * new nodes to a context until after this function has
587 * returned.
588 */
589 TAILQ_INIT(c);
590 return (0);
591 }
592
593 /* Free the context, and destroy all dynamic oids registered in this context */
594 int
595 sysctl_ctx_free(struct sysctl_ctx_list *clist)
596 {
597 struct sysctl_ctx_entry *e, *e1;
598 int error;
599
600 error = 0;
601 /*
602 * First perform a "dry run" to check if it's ok to remove oids.
603 * XXX FIXME
604 * XXX This algorithm is a hack. But I don't know any
605 * XXX better solution for now...
606 */
607 SYSCTL_WLOCK();
608 TAILQ_FOREACH(e, clist, link) {
609 error = sysctl_remove_oid_locked(e->entry, 0, 0);
610 if (error)
611 break;
612 }
613 /*
614 * Restore deregistered entries, either from the end,
615 * or from the place where error occurred.
616 * e contains the entry that was not unregistered
617 */
618 if (error)
619 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
620 else
621 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
622 while (e1 != NULL) {
623 sysctl_register_oid(e1->entry);
624 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
625 }
626 if (error) {
627 SYSCTL_WUNLOCK();
628 return(EBUSY);
629 }
630 /* Now really delete the entries */
631 e = TAILQ_FIRST(clist);
632 while (e != NULL) {
633 e1 = TAILQ_NEXT(e, link);
634 error = sysctl_remove_oid_locked(e->entry, 1, 0);
635 if (error)
636 panic("sysctl_remove_oid: corrupt tree, entry: %s",
637 e->entry->oid_name);
638 free(e, M_SYSCTLOID);
639 e = e1;
640 }
641 SYSCTL_WUNLOCK();
642 return (error);
643 }
644
645 /* Add an entry to the context */
646 struct sysctl_ctx_entry *
647 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
648 {
649 struct sysctl_ctx_entry *e;
650
651 SYSCTL_ASSERT_WLOCKED();
652 if (clist == NULL || oidp == NULL)
653 return(NULL);
654 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
655 e->entry = oidp;
656 TAILQ_INSERT_HEAD(clist, e, link);
657 return (e);
658 }
659
660 /* Find an entry in the context */
661 struct sysctl_ctx_entry *
662 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
663 {
664 struct sysctl_ctx_entry *e;
665
666 SYSCTL_ASSERT_WLOCKED();
667 if (clist == NULL || oidp == NULL)
668 return(NULL);
669 TAILQ_FOREACH(e, clist, link) {
670 if(e->entry == oidp)
671 return(e);
672 }
673 return (e);
674 }
675
676 /*
677 * Delete an entry from the context.
678 * NOTE: this function doesn't free oidp! You have to remove it
679 * with sysctl_remove_oid().
680 */
681 int
682 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
683 {
684 struct sysctl_ctx_entry *e;
685
686 if (clist == NULL || oidp == NULL)
687 return (EINVAL);
688 SYSCTL_WLOCK();
689 e = sysctl_ctx_entry_find(clist, oidp);
690 if (e != NULL) {
691 TAILQ_REMOVE(clist, e, link);
692 SYSCTL_WUNLOCK();
693 free(e, M_SYSCTLOID);
694 return (0);
695 } else {
696 SYSCTL_WUNLOCK();
697 return (ENOENT);
698 }
699 }
700
701 /*
702 * Remove dynamically created sysctl trees.
703 * oidp - top of the tree to be removed
704 * del - if 0 - just deregister, otherwise free up entries as well
705 * recurse - if != 0 traverse the subtree to be deleted
706 */
707 int
708 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
709 {
710 int error;
711
712 SYSCTL_WLOCK();
713 error = sysctl_remove_oid_locked(oidp, del, recurse);
714 SYSCTL_WUNLOCK();
715 return (error);
716 }
717
718 int
719 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
720 int del, int recurse)
721 {
722 struct sysctl_oid *p, *tmp;
723 int error;
724
725 error = ENOENT;
726 SYSCTL_WLOCK();
727 SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
728 if (strcmp(p->oid_name, name) == 0) {
729 error = sysctl_remove_oid_locked(p, del, recurse);
730 break;
731 }
732 }
733 SYSCTL_WUNLOCK();
734
735 return (error);
736 }
737
738 /*
739 * Duplicate the provided string, escaping any illegal characters. The result
740 * must be freed when no longer in use.
741 *
742 * The list of illegal characters is ".".
743 */
744 static char*
745 sysctl_escape_name(const char* orig)
746 {
747 int i, s = 0, d = 0, nillegals = 0;
748 char *new;
749
750 /* First count the number of illegal characters */
751 for (i = 0; orig[i] != '\0'; i++) {
752 if (orig[i] == '.')
753 nillegals++;
754 }
755
756 /* Allocate storage for new string */
757 new = malloc(i + 2 * nillegals + 1, M_SYSCTLOID, M_WAITOK);
758
759 /* Copy the name, escaping characters as we go */
760 while (orig[s] != '\0') {
761 if (orig[s] == '.') {
762 /* %25 is the hexadecimal representation of '.' */
763 new[d++] = '%';
764 new[d++] = '2';
765 new[d++] = '5';
766 s++;
767 } else {
768 new[d++] = orig[s++];
769 }
770 }
771
772 /* Finally, nul-terminate */
773 new[d] = '\0';
774
775 return (new);
776 }
777
778 static int
779 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
780 {
781 struct sysctl_oid *p, *tmp;
782 int error;
783
784 SYSCTL_ASSERT_WLOCKED();
785 if (oidp == NULL)
786 return(EINVAL);
787 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
788 printf("Warning: can't remove non-dynamic nodes (%s)!\n",
789 oidp->oid_name);
790 return (EINVAL);
791 }
792 /*
793 * WARNING: normal method to do this should be through
794 * sysctl_ctx_free(). Use recursing as the last resort
795 * method to purge your sysctl tree of leftovers...
796 * However, if some other code still references these nodes,
797 * it will panic.
798 */
799 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
800 if (oidp->oid_refcnt == 1) {
801 SLIST_FOREACH_SAFE(p,
802 SYSCTL_CHILDREN(oidp), oid_link, tmp) {
803 if (!recurse) {
804 printf("Warning: failed attempt to "
805 "remove oid %s with child %s\n",
806 oidp->oid_name, p->oid_name);
807 return (ENOTEMPTY);
808 }
809 error = sysctl_remove_oid_locked(p, del,
810 recurse);
811 if (error)
812 return (error);
813 }
814 }
815 }
816 if (oidp->oid_refcnt > 1 ) {
817 oidp->oid_refcnt--;
818 } else {
819 if (oidp->oid_refcnt == 0) {
820 printf("Warning: bad oid_refcnt=%u (%s)!\n",
821 oidp->oid_refcnt, oidp->oid_name);
822 return (EINVAL);
823 }
824 sysctl_unregister_oid(oidp);
825 if (del) {
826 /*
827 * Wait for all threads running the handler to drain.
828 * This preserves the previous behavior when the
829 * sysctl lock was held across a handler invocation,
830 * and is necessary for module unload correctness.
831 */
832 while (oidp->oid_running > 0) {
833 oidp->oid_kind |= CTLFLAG_DYING;
834 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
835 }
836 if (oidp->oid_descr)
837 free(__DECONST(char *, oidp->oid_descr),
838 M_SYSCTLOID);
839 if (oidp->oid_label)
840 free(__DECONST(char *, oidp->oid_label),
841 M_SYSCTLOID);
842 free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
843 free(oidp, M_SYSCTLOID);
844 }
845 }
846 return (0);
847 }
848 /*
849 * Create new sysctls at run time.
850 * clist may point to a valid context initialized with sysctl_ctx_init().
851 */
852 struct sysctl_oid *
853 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
854 int number, const char *name, int kind, void *arg1, intmax_t arg2,
855 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr,
856 const char *label)
857 {
858 struct sysctl_oid *oidp;
859 char *escaped;
860
861 /* You have to hook up somewhere.. */
862 if (parent == NULL)
863 return(NULL);
864 escaped = sysctl_escape_name(name);
865 /* Check if the node already exists, otherwise create it */
866 SYSCTL_WLOCK();
867 oidp = sysctl_find_oidname(escaped, parent);
868 if (oidp != NULL) {
869 free(escaped, M_SYSCTLOID);
870 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
871 oidp->oid_refcnt++;
872 /* Update the context */
873 if (clist != NULL)
874 sysctl_ctx_entry_add(clist, oidp);
875 SYSCTL_WUNLOCK();
876 return (oidp);
877 } else {
878 sysctl_warn_reuse(__func__, oidp);
879 SYSCTL_WUNLOCK();
880 return (NULL);
881 }
882 }
883 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
884 oidp->oid_parent = parent;
885 SLIST_INIT(&oidp->oid_children);
886 oidp->oid_number = number;
887 oidp->oid_refcnt = 1;
888 oidp->oid_name = escaped;
889 oidp->oid_handler = handler;
890 oidp->oid_kind = CTLFLAG_DYN | kind;
891 oidp->oid_arg1 = arg1;
892 oidp->oid_arg2 = arg2;
893 oidp->oid_fmt = fmt;
894 if (descr != NULL)
895 oidp->oid_descr = strdup(descr, M_SYSCTLOID);
896 if (label != NULL)
897 oidp->oid_label = strdup(label, M_SYSCTLOID);
898 /* Update the context, if used */
899 if (clist != NULL)
900 sysctl_ctx_entry_add(clist, oidp);
901 /* Register this oid */
902 sysctl_register_oid(oidp);
903 SYSCTL_WUNLOCK();
904 return (oidp);
905 }
906
907 /*
908 * Rename an existing oid.
909 */
910 void
911 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
912 {
913 char *newname;
914 char *oldname;
915
916 newname = strdup(name, M_SYSCTLOID);
917 SYSCTL_WLOCK();
918 oldname = __DECONST(char *, oidp->oid_name);
919 oidp->oid_name = newname;
920 SYSCTL_WUNLOCK();
921 free(oldname, M_SYSCTLOID);
922 }
923
924 /*
925 * Reparent an existing oid.
926 */
927 int
928 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
929 {
930 struct sysctl_oid *oidp;
931
932 SYSCTL_WLOCK();
933 if (oid->oid_parent == parent) {
934 SYSCTL_WUNLOCK();
935 return (0);
936 }
937 oidp = sysctl_find_oidname(oid->oid_name, parent);
938 if (oidp != NULL) {
939 SYSCTL_WUNLOCK();
940 return (EEXIST);
941 }
942 sysctl_unregister_oid(oid);
943 oid->oid_parent = parent;
944 oid->oid_number = OID_AUTO;
945 sysctl_register_oid(oid);
946 SYSCTL_WUNLOCK();
947 return (0);
948 }
949
950 /*
951 * Register the kernel's oids on startup.
952 */
953 SET_DECLARE(sysctl_set, struct sysctl_oid);
954
955 static void
956 sysctl_register_all(void *arg)
957 {
958 struct sysctl_oid **oidp;
959
960 sx_init(&sysctlmemlock, "sysctl mem");
961 SYSCTL_INIT();
962 SYSCTL_WLOCK();
963 SET_FOREACH(oidp, sysctl_set)
964 sysctl_register_oid(*oidp);
965 SYSCTL_WUNLOCK();
966 }
967 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_FIRST, sysctl_register_all, NULL);
968
969 /*
970 * "Staff-functions"
971 *
972 * These functions implement a presently undocumented interface
973 * used by the sysctl program to walk the tree, and get the type
974 * so it can print the value.
975 * This interface is under work and consideration, and should probably
976 * be killed with a big axe by the first person who can find the time.
977 * (be aware though, that the proper interface isn't as obvious as it
978 * may seem, there are various conflicting requirements.
979 *
980 * {CTL_SYSCTL, CTL_SYSCTL_DEBUG} printf the entire MIB-tree.
981 * {CTL_SYSCTL, CTL_SYSCTL_NAME, ...} return the name of the "..."
982 * OID.
983 * {CTL_SYSCTL, CTL_SYSCTL_NEXT, ...} return the next OID, honoring
984 * CTLFLAG_SKIP.
985 * {CTL_SYSCTL, CTL_SYSCTL_NAME2OID} return the OID of the name in
986 * "new"
987 * {CTL_SYSCTL, CTL_SYSCTL_OIDFMT, ...} return the kind & format info
988 * for the "..." OID.
989 * {CTL_SYSCTL, CTL_SYSCTL_OIDDESCR, ...} return the description of the
990 * "..." OID.
991 * {CTL_SYSCTL, CTL_SYSCTL_OIDLABEL, ...} return the aggregation label of
992 * the "..." OID.
993 * {CTL_SYSCTL, CTL_SYSCTL_NEXTNOSKIP, ...} return the next OID, ignoring
994 * CTLFLAG_SKIP.
995 */
996
997 #ifdef SYSCTL_DEBUG
998 static void
999 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
1000 {
1001 int k;
1002 struct sysctl_oid *oidp;
1003
1004 SYSCTL_ASSERT_LOCKED();
1005 SLIST_FOREACH(oidp, l, oid_link) {
1006
1007 for (k=0; k<i; k++)
1008 printf(" ");
1009
1010 printf("%d %s ", oidp->oid_number, oidp->oid_name);
1011
1012 printf("%c%c",
1013 oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
1014 oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
1015
1016 if (oidp->oid_handler)
1017 printf(" *Handler");
1018
1019 switch (oidp->oid_kind & CTLTYPE) {
1020 case CTLTYPE_NODE:
1021 printf(" Node\n");
1022 if (!oidp->oid_handler) {
1023 sysctl_sysctl_debug_dump_node(
1024 SYSCTL_CHILDREN(oidp), i + 2);
1025 }
1026 break;
1027 case CTLTYPE_INT: printf(" Int\n"); break;
1028 case CTLTYPE_UINT: printf(" u_int\n"); break;
1029 case CTLTYPE_LONG: printf(" Long\n"); break;
1030 case CTLTYPE_ULONG: printf(" u_long\n"); break;
1031 case CTLTYPE_STRING: printf(" String\n"); break;
1032 case CTLTYPE_S8: printf(" int8_t\n"); break;
1033 case CTLTYPE_S16: printf(" int16_t\n"); break;
1034 case CTLTYPE_S32: printf(" int32_t\n"); break;
1035 case CTLTYPE_S64: printf(" int64_t\n"); break;
1036 case CTLTYPE_U8: printf(" uint8_t\n"); break;
1037 case CTLTYPE_U16: printf(" uint16_t\n"); break;
1038 case CTLTYPE_U32: printf(" uint32_t\n"); break;
1039 case CTLTYPE_U64: printf(" uint64_t\n"); break;
1040 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
1041 default: printf("\n");
1042 }
1043
1044 }
1045 }
1046
1047 static int
1048 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
1049 {
1050 struct rm_priotracker tracker;
1051 int error;
1052
1053 error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
1054 if (error)
1055 return (error);
1056 SYSCTL_RLOCK(&tracker);
1057 sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
1058 SYSCTL_RUNLOCK(&tracker);
1059 return (ENOENT);
1060 }
1061
1062 SYSCTL_PROC(_sysctl, CTL_SYSCTL_DEBUG, debug, CTLTYPE_STRING | CTLFLAG_RD |
1063 CTLFLAG_MPSAFE, 0, 0, sysctl_sysctl_debug, "-", "");
1064 #endif
1065
1066 static int
1067 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
1068 {
1069 int *name = (int *) arg1;
1070 u_int namelen = arg2;
1071 int error;
1072 struct sysctl_oid *oid;
1073 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
1074 struct rm_priotracker tracker;
1075 char buf[10];
1076
1077 error = sysctl_wire_old_buffer(req, 0);
1078 if (error)
1079 return (error);
1080
1081 SYSCTL_RLOCK(&tracker);
1082 while (namelen) {
1083 if (!lsp) {
1084 snprintf(buf,sizeof(buf),"%d",*name);
1085 if (req->oldidx)
1086 error = SYSCTL_OUT(req, ".", 1);
1087 if (!error)
1088 error = SYSCTL_OUT(req, buf, strlen(buf));
1089 if (error)
1090 goto out;
1091 namelen--;
1092 name++;
1093 continue;
1094 }
1095 lsp2 = NULL;
1096 SLIST_FOREACH(oid, lsp, oid_link) {
1097 if (oid->oid_number != *name)
1098 continue;
1099
1100 if (req->oldidx)
1101 error = SYSCTL_OUT(req, ".", 1);
1102 if (!error)
1103 error = SYSCTL_OUT(req, oid->oid_name,
1104 strlen(oid->oid_name));
1105 if (error)
1106 goto out;
1107
1108 namelen--;
1109 name++;
1110
1111 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1112 break;
1113
1114 if (oid->oid_handler)
1115 break;
1116
1117 lsp2 = SYSCTL_CHILDREN(oid);
1118 break;
1119 }
1120 lsp = lsp2;
1121 }
1122 error = SYSCTL_OUT(req, "", 1);
1123 out:
1124 SYSCTL_RUNLOCK(&tracker);
1125 return (error);
1126 }
1127
1128 /*
1129 * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
1130 * capability mode.
1131 */
1132 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NAME, name, CTLFLAG_RD |
1133 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_name, "");
1134
1135 /*
1136 * Walk the sysctl subtree at lsp until we find the given name,
1137 * and return the next name in order by oid_number.
1138 */
1139 static int
1140 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
1141 int *next, int *len, int level, bool honor_skip)
1142 {
1143 struct sysctl_oid *oidp;
1144
1145 SYSCTL_ASSERT_LOCKED();
1146 *len = level;
1147 SLIST_FOREACH(oidp, lsp, oid_link) {
1148 *next = oidp->oid_number;
1149
1150 if ((oidp->oid_kind & CTLFLAG_DORMANT) != 0)
1151 continue;
1152
1153 if (honor_skip && (oidp->oid_kind & CTLFLAG_SKIP) != 0)
1154 continue;
1155
1156 if (namelen == 0) {
1157 /*
1158 * We have reached a node with a full name match and are
1159 * looking for the next oid in its children.
1160 *
1161 * For CTL_SYSCTL_NEXTNOSKIP we are done.
1162 *
1163 * For CTL_SYSCTL_NEXT we skip CTLTYPE_NODE (unless it
1164 * has a handler) and move on to the children.
1165 */
1166 if (!honor_skip)
1167 return (0);
1168 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1169 return (0);
1170 if (oidp->oid_handler)
1171 return (0);
1172 lsp = SYSCTL_CHILDREN(oidp);
1173 if (!sysctl_sysctl_next_ls(lsp, NULL, 0, next + 1, len,
1174 level + 1, honor_skip))
1175 return (0);
1176 /*
1177 * There were no useable children in this node.
1178 * Continue searching for the next oid at this level.
1179 */
1180 goto emptynode;
1181 }
1182
1183 /*
1184 * No match yet. Continue seeking the given name.
1185 *
1186 * We are iterating in order by oid_number, so skip oids lower
1187 * than the one we are looking for.
1188 *
1189 * When the current oid_number is higher than the one we seek,
1190 * that means we have reached the next oid in the sequence and
1191 * should return it.
1192 *
1193 * If the oid_number matches the name at this level then we
1194 * have to find a node to continue searching at the next level.
1195 */
1196 if (oidp->oid_number < *name)
1197 continue;
1198 if (oidp->oid_number > *name) {
1199 /*
1200 * We have reached the next oid.
1201 *
1202 * For CTL_SYSCTL_NEXTNOSKIP we are done.
1203 *
1204 * For CTL_SYSCTL_NEXT we skip CTLTYPE_NODE (unless it
1205 * has a handler) and move on to the children.
1206 */
1207 if (!honor_skip)
1208 return (0);
1209 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1210 return (0);
1211 if (oidp->oid_handler)
1212 return (0);
1213 lsp = SYSCTL_CHILDREN(oidp);
1214 if (!sysctl_sysctl_next_ls(lsp, name + 1, namelen - 1,
1215 next + 1, len, level + 1, honor_skip))
1216 return (0);
1217 goto next;
1218 }
1219 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1220 continue;
1221 if (oidp->oid_handler)
1222 continue;
1223 lsp = SYSCTL_CHILDREN(oidp);
1224 if (!sysctl_sysctl_next_ls(lsp, name + 1, namelen - 1,
1225 next + 1, len, level + 1, honor_skip))
1226 return (0);
1227 next:
1228 /*
1229 * There were no useable children in this node.
1230 * Continue searching for the next oid at the root level.
1231 */
1232 namelen = 1;
1233 emptynode:
1234 /* Reset len in case a failed recursive call changed it. */
1235 *len = level;
1236 }
1237 return (ENOENT);
1238 }
1239
1240 static int
1241 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
1242 {
1243 int *name = (int *) arg1;
1244 u_int namelen = arg2;
1245 int len, error;
1246 struct sysctl_oid_list *lsp = &sysctl__children;
1247 struct rm_priotracker tracker;
1248 int next[CTL_MAXNAME];
1249
1250 SYSCTL_RLOCK(&tracker);
1251 error = sysctl_sysctl_next_ls(lsp, name, namelen, next, &len, 1,
1252 oidp->oid_number == CTL_SYSCTL_NEXT);
1253 SYSCTL_RUNLOCK(&tracker);
1254 if (error)
1255 return (error);
1256 error = SYSCTL_OUT(req, next, len * sizeof (int));
1257 return (error);
1258 }
1259
1260 /*
1261 * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
1262 * capability mode.
1263 */
1264 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NEXT, next, CTLFLAG_RD |
1265 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_next, "");
1266
1267 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_NEXTNOSKIP, nextnoskip, CTLFLAG_RD |
1268 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_next, "");
1269
1270 static int
1271 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
1272 {
1273 struct sysctl_oid *oidp;
1274 struct sysctl_oid_list *lsp = &sysctl__children;
1275 char *p;
1276
1277 SYSCTL_ASSERT_LOCKED();
1278
1279 for (*len = 0; *len < CTL_MAXNAME;) {
1280 p = strsep(&name, ".");
1281
1282 oidp = SLIST_FIRST(lsp);
1283 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
1284 if (oidp == NULL)
1285 return (ENOENT);
1286 if (strcmp(p, oidp->oid_name) == 0)
1287 break;
1288 }
1289 *oid++ = oidp->oid_number;
1290 (*len)++;
1291
1292 if (name == NULL || *name == '\0') {
1293 if (oidpp)
1294 *oidpp = oidp;
1295 return (0);
1296 }
1297
1298 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
1299 break;
1300
1301 if (oidp->oid_handler)
1302 break;
1303
1304 lsp = SYSCTL_CHILDREN(oidp);
1305 }
1306 return (ENOENT);
1307 }
1308
1309 static int
1310 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
1311 {
1312 char *p;
1313 int error, oid[CTL_MAXNAME], len = 0;
1314 struct sysctl_oid *op = NULL;
1315 struct rm_priotracker tracker;
1316 char buf[32];
1317
1318 if (!req->newlen)
1319 return (ENOENT);
1320 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */
1321 return (ENAMETOOLONG);
1322
1323 p = buf;
1324 if (req->newlen >= sizeof(buf))
1325 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
1326
1327 error = SYSCTL_IN(req, p, req->newlen);
1328 if (error) {
1329 if (p != buf)
1330 free(p, M_SYSCTL);
1331 return (error);
1332 }
1333
1334 p [req->newlen] = '\0';
1335
1336 SYSCTL_RLOCK(&tracker);
1337 error = name2oid(p, oid, &len, &op);
1338 SYSCTL_RUNLOCK(&tracker);
1339
1340 if (p != buf)
1341 free(p, M_SYSCTL);
1342
1343 if (error)
1344 return (error);
1345
1346 error = SYSCTL_OUT(req, oid, len * sizeof *oid);
1347 return (error);
1348 }
1349
1350 /*
1351 * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
1352 * capability mode.
1353 */
1354 SYSCTL_PROC(_sysctl, CTL_SYSCTL_NAME2OID, name2oid, CTLTYPE_INT | CTLFLAG_RW |
1355 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE | CTLFLAG_CAPRW, 0, 0,
1356 sysctl_sysctl_name2oid, "I", "");
1357
1358 static int
1359 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
1360 {
1361 struct sysctl_oid *oid;
1362 struct rm_priotracker tracker;
1363 int error;
1364
1365 error = sysctl_wire_old_buffer(req, 0);
1366 if (error)
1367 return (error);
1368
1369 SYSCTL_RLOCK(&tracker);
1370 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1371 if (error)
1372 goto out;
1373
1374 if (oid->oid_fmt == NULL) {
1375 error = ENOENT;
1376 goto out;
1377 }
1378 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
1379 if (error)
1380 goto out;
1381 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
1382 out:
1383 SYSCTL_RUNLOCK(&tracker);
1384 return (error);
1385 }
1386
1387
1388 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDFMT, oidfmt, CTLFLAG_RD |
1389 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidfmt, "");
1390
1391 static int
1392 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
1393 {
1394 struct sysctl_oid *oid;
1395 struct rm_priotracker tracker;
1396 int error;
1397
1398 error = sysctl_wire_old_buffer(req, 0);
1399 if (error)
1400 return (error);
1401
1402 SYSCTL_RLOCK(&tracker);
1403 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1404 if (error)
1405 goto out;
1406
1407 if (oid->oid_descr == NULL) {
1408 error = ENOENT;
1409 goto out;
1410 }
1411 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
1412 out:
1413 SYSCTL_RUNLOCK(&tracker);
1414 return (error);
1415 }
1416
1417 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDDESCR, oiddescr, CTLFLAG_RD |
1418 CTLFLAG_MPSAFE|CTLFLAG_CAPRD, sysctl_sysctl_oiddescr, "");
1419
1420 static int
1421 sysctl_sysctl_oidlabel(SYSCTL_HANDLER_ARGS)
1422 {
1423 struct sysctl_oid *oid;
1424 struct rm_priotracker tracker;
1425 int error;
1426
1427 error = sysctl_wire_old_buffer(req, 0);
1428 if (error)
1429 return (error);
1430
1431 SYSCTL_RLOCK(&tracker);
1432 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
1433 if (error)
1434 goto out;
1435
1436 if (oid->oid_label == NULL) {
1437 error = ENOENT;
1438 goto out;
1439 }
1440 error = SYSCTL_OUT(req, oid->oid_label, strlen(oid->oid_label) + 1);
1441 out:
1442 SYSCTL_RUNLOCK(&tracker);
1443 return (error);
1444 }
1445
1446 static SYSCTL_NODE(_sysctl, CTL_SYSCTL_OIDLABEL, oidlabel, CTLFLAG_RD |
1447 CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidlabel, "");
1448
1449 /*
1450 * Default "handler" functions.
1451 */
1452
1453 /*
1454 * Handle a bool.
1455 * Two cases:
1456 * a variable: point arg1 at it.
1457 * a constant: pass it in arg2.
1458 */
1459
1460 int
1461 sysctl_handle_bool(SYSCTL_HANDLER_ARGS)
1462 {
1463 uint8_t temp;
1464 int error;
1465
1466 /*
1467 * Attempt to get a coherent snapshot by making a copy of the data.
1468 */
1469 if (arg1)
1470 temp = *(bool *)arg1 ? 1 : 0;
1471 else
1472 temp = arg2 ? 1 : 0;
1473
1474 error = SYSCTL_OUT(req, &temp, sizeof(temp));
1475 if (error || !req->newptr)
1476 return (error);
1477
1478 if (!arg1)
1479 error = EPERM;
1480 else {
1481 error = SYSCTL_IN(req, &temp, sizeof(temp));
1482 if (!error)
1483 *(bool *)arg1 = temp ? 1 : 0;
1484 }
1485 return (error);
1486 }
1487
1488 /*
1489 * Handle an int8_t, signed or unsigned.
1490 * Two cases:
1491 * a variable: point arg1 at it.
1492 * a constant: pass it in arg2.
1493 */
1494
1495 int
1496 sysctl_handle_8(SYSCTL_HANDLER_ARGS)
1497 {
1498 int8_t tmpout;
1499 int error = 0;
1500
1501 /*
1502 * Attempt to get a coherent snapshot by making a copy of the data.
1503 */
1504 if (arg1)
1505 tmpout = *(int8_t *)arg1;
1506 else
1507 tmpout = arg2;
1508 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1509
1510 if (error || !req->newptr)
1511 return (error);
1512
1513 if (!arg1)
1514 error = EPERM;
1515 else
1516 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1517 return (error);
1518 }
1519
1520 /*
1521 * Handle an int16_t, signed or unsigned.
1522 * Two cases:
1523 * a variable: point arg1 at it.
1524 * a constant: pass it in arg2.
1525 */
1526
1527 int
1528 sysctl_handle_16(SYSCTL_HANDLER_ARGS)
1529 {
1530 int16_t tmpout;
1531 int error = 0;
1532
1533 /*
1534 * Attempt to get a coherent snapshot by making a copy of the data.
1535 */
1536 if (arg1)
1537 tmpout = *(int16_t *)arg1;
1538 else
1539 tmpout = arg2;
1540 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1541
1542 if (error || !req->newptr)
1543 return (error);
1544
1545 if (!arg1)
1546 error = EPERM;
1547 else
1548 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1549 return (error);
1550 }
1551
1552 /*
1553 * Handle an int32_t, signed or unsigned.
1554 * Two cases:
1555 * a variable: point arg1 at it.
1556 * a constant: pass it in arg2.
1557 */
1558
1559 int
1560 sysctl_handle_32(SYSCTL_HANDLER_ARGS)
1561 {
1562 int32_t tmpout;
1563 int error = 0;
1564
1565 /*
1566 * Attempt to get a coherent snapshot by making a copy of the data.
1567 */
1568 if (arg1)
1569 tmpout = *(int32_t *)arg1;
1570 else
1571 tmpout = arg2;
1572 error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
1573
1574 if (error || !req->newptr)
1575 return (error);
1576
1577 if (!arg1)
1578 error = EPERM;
1579 else
1580 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
1581 return (error);
1582 }
1583
1584 /*
1585 * Handle an int, signed or unsigned.
1586 * Two cases:
1587 * a variable: point arg1 at it.
1588 * a constant: pass it in arg2.
1589 */
1590
1591 int
1592 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
1593 {
1594 int tmpout, error = 0;
1595
1596 /*
1597 * Attempt to get a coherent snapshot by making a copy of the data.
1598 */
1599 if (arg1)
1600 tmpout = *(int *)arg1;
1601 else
1602 tmpout = arg2;
1603 error = SYSCTL_OUT(req, &tmpout, sizeof(int));
1604
1605 if (error || !req->newptr)
1606 return (error);
1607
1608 if (!arg1)
1609 error = EPERM;
1610 else
1611 error = SYSCTL_IN(req, arg1, sizeof(int));
1612 return (error);
1613 }
1614
1615 /*
1616 * Based on on sysctl_handle_int() convert milliseconds into ticks.
1617 * Note: this is used by TCP.
1618 */
1619
1620 int
1621 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
1622 {
1623 int error, s, tt;
1624
1625 tt = *(int *)arg1;
1626 s = (int)((int64_t)tt * 1000 / hz);
1627
1628 error = sysctl_handle_int(oidp, &s, 0, req);
1629 if (error || !req->newptr)
1630 return (error);
1631
1632 tt = (int)((int64_t)s * hz / 1000);
1633 if (tt < 1)
1634 return (EINVAL);
1635
1636 *(int *)arg1 = tt;
1637 return (0);
1638 }
1639
1640
1641 /*
1642 * Handle a long, signed or unsigned.
1643 * Two cases:
1644 * a variable: point arg1 at it.
1645 * a constant: pass it in arg2.
1646 */
1647
1648 int
1649 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
1650 {
1651 int error = 0;
1652 long tmplong;
1653 #ifdef SCTL_MASK32
1654 int tmpint;
1655 #endif
1656
1657 /*
1658 * Attempt to get a coherent snapshot by making a copy of the data.
1659 */
1660 if (arg1)
1661 tmplong = *(long *)arg1;
1662 else
1663 tmplong = arg2;
1664 #ifdef SCTL_MASK32
1665 if (req->flags & SCTL_MASK32) {
1666 tmpint = tmplong;
1667 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
1668 } else
1669 #endif
1670 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
1671
1672 if (error || !req->newptr)
1673 return (error);
1674
1675 if (!arg1)
1676 error = EPERM;
1677 #ifdef SCTL_MASK32
1678 else if (req->flags & SCTL_MASK32) {
1679 error = SYSCTL_IN(req, &tmpint, sizeof(int));
1680 *(long *)arg1 = (long)tmpint;
1681 }
1682 #endif
1683 else
1684 error = SYSCTL_IN(req, arg1, sizeof(long));
1685 return (error);
1686 }
1687
1688 /*
1689 * Handle a 64 bit int, signed or unsigned.
1690 * Two cases:
1691 * a variable: point arg1 at it.
1692 * a constant: pass it in arg2.
1693 */
1694 int
1695 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
1696 {
1697 int error = 0;
1698 uint64_t tmpout;
1699
1700 /*
1701 * Attempt to get a coherent snapshot by making a copy of the data.
1702 */
1703 if (arg1)
1704 tmpout = *(uint64_t *)arg1;
1705 else
1706 tmpout = arg2;
1707 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
1708
1709 if (error || !req->newptr)
1710 return (error);
1711
1712 if (!arg1)
1713 error = EPERM;
1714 else
1715 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
1716 return (error);
1717 }
1718
1719 /*
1720 * Handle our generic '\0' terminated 'C' string.
1721 * Two cases:
1722 * a variable string: point arg1 at it, arg2 is max length.
1723 * a constant string: point arg1 at it, arg2 is zero.
1724 */
1725
1726 int
1727 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
1728 {
1729 size_t outlen;
1730 int error = 0, ro_string = 0;
1731
1732 /*
1733 * A zero-length buffer indicates a fixed size read-only
1734 * string:
1735 */
1736 if (arg2 == 0) {
1737 arg2 = strlen((char *)arg1) + 1;
1738 ro_string = 1;
1739 }
1740
1741 if (req->oldptr != NULL) {
1742 char *tmparg;
1743
1744 if (ro_string) {
1745 tmparg = arg1;
1746 } else {
1747 /* try to make a coherent snapshot of the string */
1748 tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK);
1749 memcpy(tmparg, arg1, arg2);
1750 }
1751
1752 outlen = strnlen(tmparg, arg2 - 1) + 1;
1753 error = SYSCTL_OUT(req, tmparg, outlen);
1754
1755 if (!ro_string)
1756 free(tmparg, M_SYSCTLTMP);
1757 } else {
1758 outlen = strnlen((char *)arg1, arg2 - 1) + 1;
1759 error = SYSCTL_OUT(req, NULL, outlen);
1760 }
1761 if (error || !req->newptr)
1762 return (error);
1763
1764 if ((req->newlen - req->newidx) >= arg2) {
1765 error = EINVAL;
1766 } else {
1767 arg2 = (req->newlen - req->newidx);
1768 error = SYSCTL_IN(req, arg1, arg2);
1769 ((char *)arg1)[arg2] = '\0';
1770 }
1771 return (error);
1772 }
1773
1774 /*
1775 * Handle any kind of opaque data.
1776 * arg1 points to it, arg2 is the size.
1777 */
1778
1779 int
1780 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
1781 {
1782 int error, tries;
1783 u_int generation;
1784 struct sysctl_req req2;
1785
1786 /*
1787 * Attempt to get a coherent snapshot, by using the thread
1788 * pre-emption counter updated from within mi_switch() to
1789 * determine if we were pre-empted during a bcopy() or
1790 * copyout(). Make 3 attempts at doing this before giving up.
1791 * If we encounter an error, stop immediately.
1792 */
1793 tries = 0;
1794 req2 = *req;
1795 retry:
1796 generation = curthread->td_generation;
1797 error = SYSCTL_OUT(req, arg1, arg2);
1798 if (error)
1799 return (error);
1800 tries++;
1801 if (generation != curthread->td_generation && tries < 3) {
1802 *req = req2;
1803 goto retry;
1804 }
1805
1806 error = SYSCTL_IN(req, arg1, arg2);
1807
1808 return (error);
1809 }
1810
1811 /*
1812 * Convert seconds to a struct timeval. Intended for use with
1813 * intervals and thus does not permit negative seconds.
1814 */
1815 int
1816 sysctl_sec_to_timeval(SYSCTL_HANDLER_ARGS)
1817 {
1818 struct timeval *tv;
1819 int error, secs;
1820
1821 tv = arg1;
1822 secs = tv->tv_sec;
1823
1824 error = sysctl_handle_int(oidp, &secs, 0, req);
1825 if (error || req->newptr == NULL)
1826 return (error);
1827
1828 if (secs < 0)
1829 return (EINVAL);
1830 tv->tv_sec = secs;
1831
1832 return (0);
1833 }
1834
1835 /*
1836 * Transfer functions to/from kernel space.
1837 * XXX: rather untested at this point
1838 */
1839 static int
1840 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1841 {
1842 size_t i = 0;
1843
1844 if (req->oldptr) {
1845 i = l;
1846 if (req->oldlen <= req->oldidx)
1847 i = 0;
1848 else
1849 if (i > req->oldlen - req->oldidx)
1850 i = req->oldlen - req->oldidx;
1851 if (i > 0)
1852 bcopy(p, (char *)req->oldptr + req->oldidx, i);
1853 }
1854 req->oldidx += l;
1855 if (req->oldptr && i != l)
1856 return (ENOMEM);
1857 return (0);
1858 }
1859
1860 static int
1861 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1862 {
1863 if (!req->newptr)
1864 return (0);
1865 if (req->newlen - req->newidx < l)
1866 return (EINVAL);
1867 bcopy((char *)req->newptr + req->newidx, p, l);
1868 req->newidx += l;
1869 return (0);
1870 }
1871
1872 int
1873 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1874 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1875 {
1876 int error = 0;
1877 struct sysctl_req req;
1878
1879 bzero(&req, sizeof req);
1880
1881 req.td = td;
1882 req.flags = flags;
1883
1884 if (oldlenp) {
1885 req.oldlen = *oldlenp;
1886 }
1887 req.validlen = req.oldlen;
1888
1889 if (old) {
1890 req.oldptr= old;
1891 }
1892
1893 if (new != NULL) {
1894 req.newlen = newlen;
1895 req.newptr = new;
1896 }
1897
1898 req.oldfunc = sysctl_old_kernel;
1899 req.newfunc = sysctl_new_kernel;
1900 req.lock = REQ_UNWIRED;
1901
1902 error = sysctl_root(0, name, namelen, &req);
1903
1904 if (req.lock == REQ_WIRED && req.validlen > 0)
1905 vsunlock(req.oldptr, req.validlen);
1906
1907 if (error && error != ENOMEM)
1908 return (error);
1909
1910 if (retval) {
1911 if (req.oldptr && req.oldidx > req.validlen)
1912 *retval = req.validlen;
1913 else
1914 *retval = req.oldidx;
1915 }
1916 return (error);
1917 }
1918
1919 int
1920 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1921 void *new, size_t newlen, size_t *retval, int flags)
1922 {
1923 int oid[CTL_MAXNAME];
1924 size_t oidlen, plen;
1925 int error;
1926
1927 oid[0] = CTL_SYSCTL;
1928 oid[1] = CTL_SYSCTL_NAME2OID;
1929 oidlen = sizeof(oid);
1930
1931 error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1932 (void *)name, strlen(name), &plen, flags);
1933 if (error)
1934 return (error);
1935
1936 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1937 new, newlen, retval, flags);
1938 return (error);
1939 }
1940
1941 /*
1942 * Transfer function to/from user space.
1943 */
1944 static int
1945 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1946 {
1947 size_t i, len, origidx;
1948 int error;
1949
1950 origidx = req->oldidx;
1951 req->oldidx += l;
1952 if (req->oldptr == NULL)
1953 return (0);
1954 /*
1955 * If we have not wired the user supplied buffer and we are currently
1956 * holding locks, drop a witness warning, as it's possible that
1957 * write operations to the user page can sleep.
1958 */
1959 if (req->lock != REQ_WIRED)
1960 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1961 "sysctl_old_user()");
1962 i = l;
1963 len = req->validlen;
1964 if (len <= origidx)
1965 i = 0;
1966 else {
1967 if (i > len - origidx)
1968 i = len - origidx;
1969 if (req->lock == REQ_WIRED) {
1970 error = copyout_nofault(p, (char *)req->oldptr +
1971 origidx, i);
1972 } else
1973 error = copyout(p, (char *)req->oldptr + origidx, i);
1974 if (error != 0)
1975 return (error);
1976 }
1977 if (i < l)
1978 return (ENOMEM);
1979 return (0);
1980 }
1981
1982 static int
1983 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1984 {
1985 int error;
1986
1987 if (!req->newptr)
1988 return (0);
1989 if (req->newlen - req->newidx < l)
1990 return (EINVAL);
1991 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1992 "sysctl_new_user()");
1993 error = copyin((char *)req->newptr + req->newidx, p, l);
1994 req->newidx += l;
1995 return (error);
1996 }
1997
1998 /*
1999 * Wire the user space destination buffer. If set to a value greater than
2000 * zero, the len parameter limits the maximum amount of wired memory.
2001 */
2002 int
2003 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
2004 {
2005 int ret;
2006 size_t wiredlen;
2007
2008 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
2009 ret = 0;
2010 if (req->lock != REQ_WIRED && req->oldptr &&
2011 req->oldfunc == sysctl_old_user) {
2012 if (wiredlen != 0) {
2013 ret = vslock(req->oldptr, wiredlen);
2014 if (ret != 0) {
2015 if (ret != ENOMEM)
2016 return (ret);
2017 wiredlen = 0;
2018 }
2019 }
2020 req->lock = REQ_WIRED;
2021 req->validlen = wiredlen;
2022 }
2023 return (0);
2024 }
2025
2026 int
2027 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
2028 int *nindx, struct sysctl_req *req)
2029 {
2030 struct sysctl_oid_list *lsp;
2031 struct sysctl_oid *oid;
2032 int indx;
2033
2034 SYSCTL_ASSERT_LOCKED();
2035 lsp = &sysctl__children;
2036 indx = 0;
2037 while (indx < CTL_MAXNAME) {
2038 SLIST_FOREACH(oid, lsp, oid_link) {
2039 if (oid->oid_number == name[indx])
2040 break;
2041 }
2042 if (oid == NULL)
2043 return (ENOENT);
2044
2045 indx++;
2046 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2047 if (oid->oid_handler != NULL || indx == namelen) {
2048 *noid = oid;
2049 if (nindx != NULL)
2050 *nindx = indx;
2051 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
2052 ("%s found DYING node %p", __func__, oid));
2053 return (0);
2054 }
2055 lsp = SYSCTL_CHILDREN(oid);
2056 } else if (indx == namelen) {
2057 if ((oid->oid_kind & CTLFLAG_DORMANT) != 0)
2058 return (ENOENT);
2059 *noid = oid;
2060 if (nindx != NULL)
2061 *nindx = indx;
2062 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
2063 ("%s found DYING node %p", __func__, oid));
2064 return (0);
2065 } else {
2066 return (ENOTDIR);
2067 }
2068 }
2069 return (ENOENT);
2070 }
2071
2072 /*
2073 * Traverse our tree, and find the right node, execute whatever it points
2074 * to, and return the resulting error code.
2075 */
2076
2077 static int
2078 sysctl_root(SYSCTL_HANDLER_ARGS)
2079 {
2080 struct sysctl_oid *oid;
2081 struct rm_priotracker tracker;
2082 int error, indx, lvl;
2083
2084 SYSCTL_RLOCK(&tracker);
2085
2086 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
2087 if (error)
2088 goto out;
2089
2090 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2091 /*
2092 * You can't call a sysctl when it's a node, but has
2093 * no handler. Inform the user that it's a node.
2094 * The indx may or may not be the same as namelen.
2095 */
2096 if (oid->oid_handler == NULL) {
2097 error = EISDIR;
2098 goto out;
2099 }
2100 }
2101
2102 /* Is this sysctl writable? */
2103 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) {
2104 error = EPERM;
2105 goto out;
2106 }
2107
2108 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
2109
2110 #ifdef CAPABILITY_MODE
2111 /*
2112 * If the process is in capability mode, then don't permit reading or
2113 * writing unless specifically granted for the node.
2114 */
2115 if (IN_CAPABILITY_MODE(req->td)) {
2116 if ((req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) ||
2117 (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))) {
2118 error = EPERM;
2119 goto out;
2120 }
2121 }
2122 #endif
2123
2124 /* Is this sysctl sensitive to securelevels? */
2125 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
2126 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
2127 error = securelevel_gt(req->td->td_ucred, lvl);
2128 if (error)
2129 goto out;
2130 }
2131
2132 /* Is this sysctl writable by only privileged users? */
2133 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
2134 int priv;
2135
2136 if (oid->oid_kind & CTLFLAG_PRISON)
2137 priv = PRIV_SYSCTL_WRITEJAIL;
2138 #ifdef VIMAGE
2139 else if ((oid->oid_kind & CTLFLAG_VNET) &&
2140 prison_owns_vnet(req->td->td_ucred))
2141 priv = PRIV_SYSCTL_WRITEJAIL;
2142 #endif
2143 else
2144 priv = PRIV_SYSCTL_WRITE;
2145 error = priv_check(req->td, priv);
2146 if (error)
2147 goto out;
2148 }
2149
2150 if (!oid->oid_handler) {
2151 error = EINVAL;
2152 goto out;
2153 }
2154
2155 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
2156 arg1 = (int *)arg1 + indx;
2157 arg2 -= indx;
2158 } else {
2159 arg1 = oid->oid_arg1;
2160 arg2 = oid->oid_arg2;
2161 }
2162 #ifdef MAC
2163 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
2164 req);
2165 if (error != 0)
2166 goto out;
2167 #endif
2168 #ifdef VIMAGE
2169 if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
2170 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
2171 #endif
2172 error = sysctl_root_handler_locked(oid, arg1, arg2, req, &tracker);
2173
2174 out:
2175 SYSCTL_RUNLOCK(&tracker);
2176 return (error);
2177 }
2178
2179 #ifndef _SYS_SYSPROTO_H_
2180 struct sysctl_args {
2181 int *name;
2182 u_int namelen;
2183 void *old;
2184 size_t *oldlenp;
2185 void *new;
2186 size_t newlen;
2187 };
2188 #endif
2189 int
2190 sys___sysctl(struct thread *td, struct sysctl_args *uap)
2191 {
2192 int error, i, name[CTL_MAXNAME];
2193 size_t j;
2194
2195 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
2196 return (EINVAL);
2197
2198 error = copyin(uap->name, &name, uap->namelen * sizeof(int));
2199 if (error)
2200 return (error);
2201
2202 error = userland_sysctl(td, name, uap->namelen,
2203 uap->old, uap->oldlenp, 0,
2204 uap->new, uap->newlen, &j, 0);
2205 if (error && error != ENOMEM)
2206 return (error);
2207 if (uap->oldlenp) {
2208 i = copyout(&j, uap->oldlenp, sizeof(j));
2209 if (i)
2210 return (i);
2211 }
2212 return (error);
2213 }
2214
2215 int
2216 kern___sysctlbyname(struct thread *td, const char *oname, size_t namelen,
2217 void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval,
2218 int flags, bool inkernel)
2219 {
2220 int oid[CTL_MAXNAME];
2221 char namebuf[16];
2222 char *name;
2223 size_t oidlen;
2224 int error;
2225
2226 if (namelen > MAXPATHLEN || namelen == 0)
2227 return (EINVAL);
2228 name = namebuf;
2229 if (namelen > sizeof(namebuf))
2230 name = malloc(namelen, M_SYSCTL, M_WAITOK);
2231 error = copyin(oname, name, namelen);
2232 if (error != 0)
2233 goto out;
2234
2235 oid[0] = CTL_SYSCTL;
2236 oid[1] = CTL_SYSCTL_NAME2OID;
2237 oidlen = sizeof(oid);
2238 error = kernel_sysctl(td, oid, 2, oid, &oidlen, (void *)name, namelen,
2239 retval, flags);
2240 if (error != 0)
2241 goto out;
2242 error = userland_sysctl(td, oid, *retval / sizeof(int), old, oldlenp,
2243 inkernel, new, newlen, retval, flags);
2244
2245 out:
2246 if (namelen > sizeof(namebuf))
2247 free(name, M_SYSCTL);
2248 return (error);
2249 }
2250
2251 #ifndef _SYS_SYSPROTO_H_
2252 struct __sysctlbyname_args {
2253 const char *name;
2254 size_t namelen;
2255 void *old;
2256 size_t *oldlenp;
2257 void *new;
2258 size_t newlen;
2259 };
2260 #endif
2261 int
2262 sys___sysctlbyname(struct thread *td, struct __sysctlbyname_args *uap)
2263 {
2264 size_t rv;
2265 int error;
2266
2267 error = kern___sysctlbyname(td, uap->name, uap->namelen, uap->old,
2268 uap->oldlenp, uap->new, uap->newlen, &rv, 0, 0);
2269 if (error != 0)
2270 return (error);
2271 if (uap->oldlenp != NULL)
2272 error = copyout(&rv, uap->oldlenp, sizeof(rv));
2273
2274 return (error);
2275 }
2276
2277 /*
2278 * This is used from various compatibility syscalls too. That's why name
2279 * must be in kernel space.
2280 */
2281 int
2282 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
2283 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
2284 int flags)
2285 {
2286 int error = 0, memlocked;
2287 struct sysctl_req req;
2288
2289 bzero(&req, sizeof req);
2290
2291 req.td = td;
2292 req.flags = flags;
2293
2294 if (oldlenp) {
2295 if (inkernel) {
2296 req.oldlen = *oldlenp;
2297 } else {
2298 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
2299 if (error)
2300 return (error);
2301 }
2302 }
2303 req.validlen = req.oldlen;
2304 req.oldptr = old;
2305
2306 if (new != NULL) {
2307 req.newlen = newlen;
2308 req.newptr = new;
2309 }
2310
2311 req.oldfunc = sysctl_old_user;
2312 req.newfunc = sysctl_new_user;
2313 req.lock = REQ_UNWIRED;
2314
2315 #ifdef KTRACE
2316 if (KTRPOINT(curthread, KTR_SYSCTL))
2317 ktrsysctl(name, namelen);
2318 #endif
2319 memlocked = 0;
2320 if (req.oldptr && req.oldlen > 4 * PAGE_SIZE) {
2321 memlocked = 1;
2322 sx_xlock(&sysctlmemlock);
2323 }
2324 CURVNET_SET(TD_TO_VNET(td));
2325
2326 for (;;) {
2327 req.oldidx = 0;
2328 req.newidx = 0;
2329 error = sysctl_root(0, name, namelen, &req);
2330 if (error != EAGAIN)
2331 break;
2332 kern_yield(PRI_USER);
2333 }
2334
2335 CURVNET_RESTORE();
2336
2337 if (req.lock == REQ_WIRED && req.validlen > 0)
2338 vsunlock(req.oldptr, req.validlen);
2339 if (memlocked)
2340 sx_xunlock(&sysctlmemlock);
2341
2342 if (error && error != ENOMEM)
2343 return (error);
2344
2345 if (retval) {
2346 if (req.oldptr && req.oldidx > req.validlen)
2347 *retval = req.validlen;
2348 else
2349 *retval = req.oldidx;
2350 }
2351 return (error);
2352 }
2353
2354 /*
2355 * Drain into a sysctl struct. The user buffer should be wired if a page
2356 * fault would cause issue.
2357 */
2358 static int
2359 sbuf_sysctl_drain(void *arg, const char *data, int len)
2360 {
2361 struct sysctl_req *req = arg;
2362 int error;
2363
2364 error = SYSCTL_OUT(req, data, len);
2365 KASSERT(error >= 0, ("Got unexpected negative value %d", error));
2366 return (error == 0 ? len : -error);
2367 }
2368
2369 struct sbuf *
2370 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
2371 struct sysctl_req *req)
2372 {
2373
2374 /* Supply a default buffer size if none given. */
2375 if (buf == NULL && length == 0)
2376 length = 64;
2377 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
2378 sbuf_set_drain(s, sbuf_sysctl_drain, req);
2379 return (s);
2380 }
Cache object: 687330fb8f3c1ca0f1230a1bae4f208b
|