]> git.saurik.com Git - apple/xnu.git/blob - bsd/net/rtsock.c
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-6153.81.5.tar.gz
[apple/xnu.git] / bsd / net / rtsock.c
1 /*
2 * Copyright (c) 2000-2019 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28 /*
29 * Copyright (c) 1988, 1991, 1993
30 * The Regents of the University of California. All rights reserved.
31 *
32 * Redistribution and use in source and binary forms, with or without
33 * modification, are permitted provided that the following conditions
34 * are met:
35 * 1. Redistributions of source code must retain the above copyright
36 * notice, this list of conditions and the following disclaimer.
37 * 2. Redistributions in binary form must reproduce the above copyright
38 * notice, this list of conditions and the following disclaimer in the
39 * documentation and/or other materials provided with the distribution.
40 * 3. All advertising materials mentioning features or use of this software
41 * must display the following acknowledgement:
42 * This product includes software developed by the University of
43 * California, Berkeley and its contributors.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)rtsock.c 8.5 (Berkeley) 11/2/94
61 */
62
63 #include <sys/param.h>
64 #include <sys/systm.h>
65 #include <sys/kauth.h>
66 #include <sys/kernel.h>
67 #include <sys/sysctl.h>
68 #include <sys/proc.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/socket.h>
72 #include <sys/socketvar.h>
73 #include <sys/domain.h>
74 #include <sys/protosw.h>
75 #include <sys/syslog.h>
76 #include <sys/mcache.h>
77 #include <kern/locks.h>
78 #include <sys/codesign.h>
79
80 #include <net/if.h>
81 #include <net/route.h>
82 #include <net/dlil.h>
83 #include <net/raw_cb.h>
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/in_arp.h>
87 #include <netinet/ip.h>
88 #include <netinet/ip6.h>
89 #include <netinet6/nd6.h>
90
91 extern struct rtstat rtstat;
92 extern struct domain routedomain_s;
93 static struct domain *routedomain = NULL;
94
95 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
96
97 static struct sockaddr route_dst = { .sa_len = 2, .sa_family = PF_ROUTE, .sa_data = { 0, } };
98 static struct sockaddr route_src = { .sa_len = 2, .sa_family = PF_ROUTE, .sa_data = { 0, } };
99 static struct sockaddr sa_zero = { .sa_len = sizeof(sa_zero), .sa_family = AF_INET, .sa_data = { 0, } };
100
101 struct route_cb {
102 u_int32_t ip_count; /* attached w/ AF_INET */
103 u_int32_t ip6_count; /* attached w/ AF_INET6 */
104 u_int32_t any_count; /* total attached */
105 };
106
107 static struct route_cb route_cb;
108
109 struct walkarg {
110 int w_tmemsize;
111 int w_op, w_arg;
112 caddr_t w_tmem;
113 struct sysctl_req *w_req;
114 };
115
116 static void route_dinit(struct domain *);
117 static int rts_abort(struct socket *);
118 static int rts_attach(struct socket *, int, struct proc *);
119 static int rts_bind(struct socket *, struct sockaddr *, struct proc *);
120 static int rts_connect(struct socket *, struct sockaddr *, struct proc *);
121 static int rts_detach(struct socket *);
122 static int rts_disconnect(struct socket *);
123 static int rts_peeraddr(struct socket *, struct sockaddr **);
124 static int rts_send(struct socket *, int, struct mbuf *, struct sockaddr *,
125 struct mbuf *, struct proc *);
126 static int rts_shutdown(struct socket *);
127 static int rts_sockaddr(struct socket *, struct sockaddr **);
128
129 static int route_output(struct mbuf *, struct socket *);
130 static int rt_setmetrics(u_int32_t, struct rt_metrics *, struct rtentry *);
131 static void rt_getmetrics(struct rtentry *, struct rt_metrics *);
132 static void rt_setif(struct rtentry *, struct sockaddr *, struct sockaddr *,
133 struct sockaddr *, unsigned int);
134 static int rt_xaddrs(caddr_t, caddr_t, struct rt_addrinfo *);
135 static struct mbuf *rt_msg1(int, struct rt_addrinfo *);
136 static int rt_msg2(int, struct rt_addrinfo *, caddr_t, struct walkarg *,
137 kauth_cred_t *);
138 static int sysctl_dumpentry(struct radix_node *rn, void *vw);
139 static int sysctl_dumpentry_ext(struct radix_node *rn, void *vw);
140 static int sysctl_iflist(int af, struct walkarg *w);
141 static int sysctl_iflist2(int af, struct walkarg *w);
142 static int sysctl_rtstat(struct sysctl_req *);
143 static int sysctl_rttrash(struct sysctl_req *);
144 static int sysctl_rtsock SYSCTL_HANDLER_ARGS;
145
146 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_LOCKED,
147 sysctl_rtsock, "");
148
149 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RW | CTLFLAG_LOCKED, 0, "routing");
150
151 /* Align x to 1024 (only power of 2) assuming x is positive */
152 #define ALIGN_BYTES(x) do { \
153 x = P2ALIGN(x, 1024); \
154 } while(0)
155
156 #define ROUNDUP32(a) \
157 ((a) > 0 ? (1 + (((a) - 1) | (sizeof (uint32_t) - 1))) : \
158 sizeof (uint32_t))
159
160 #define ADVANCE32(x, n) \
161 (x += ROUNDUP32((n)->sa_len))
162
163 #define RT_HAS_IFADDR(rt) \
164 ((rt)->rt_ifa != NULL && (rt)->rt_ifa->ifa_addr != NULL)
165
166 /*
167 * It really doesn't make any sense at all for this code to share much
168 * with raw_usrreq.c, since its functionality is so restricted. XXX
169 */
170 static int
171 rts_abort(struct socket *so)
172 {
173 return raw_usrreqs.pru_abort(so);
174 }
175
176 /* pru_accept is EOPNOTSUPP */
177
178 static int
179 rts_attach(struct socket *so, int proto, struct proc *p)
180 {
181 #pragma unused(p)
182 struct rawcb *rp;
183 int error;
184
185 VERIFY(so->so_pcb == NULL);
186
187 MALLOC(rp, struct rawcb *, sizeof(*rp), M_PCB, M_WAITOK | M_ZERO);
188 if (rp == NULL) {
189 return ENOBUFS;
190 }
191
192 so->so_pcb = (caddr_t)rp;
193 /* don't use raw_usrreqs.pru_attach, it checks for SS_PRIV */
194 error = raw_attach(so, proto);
195 rp = sotorawcb(so);
196 if (error) {
197 FREE(rp, M_PCB);
198 so->so_pcb = NULL;
199 so->so_flags |= SOF_PCBCLEARING;
200 return error;
201 }
202
203 switch (rp->rcb_proto.sp_protocol) {
204 case AF_INET:
205 atomic_add_32(&route_cb.ip_count, 1);
206 break;
207 case AF_INET6:
208 atomic_add_32(&route_cb.ip6_count, 1);
209 break;
210 }
211 rp->rcb_faddr = &route_src;
212 atomic_add_32(&route_cb.any_count, 1);
213 /* the socket is already locked when we enter rts_attach */
214 soisconnected(so);
215 so->so_options |= SO_USELOOPBACK;
216 return 0;
217 }
218
219 static int
220 rts_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
221 {
222 return raw_usrreqs.pru_bind(so, nam, p); /* xxx just EINVAL */
223 }
224
225 static int
226 rts_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
227 {
228 return raw_usrreqs.pru_connect(so, nam, p); /* XXX just EINVAL */
229 }
230
231 /* pru_connect2 is EOPNOTSUPP */
232 /* pru_control is EOPNOTSUPP */
233
234 static int
235 rts_detach(struct socket *so)
236 {
237 struct rawcb *rp = sotorawcb(so);
238
239 VERIFY(rp != NULL);
240
241 switch (rp->rcb_proto.sp_protocol) {
242 case AF_INET:
243 atomic_add_32(&route_cb.ip_count, -1);
244 break;
245 case AF_INET6:
246 atomic_add_32(&route_cb.ip6_count, -1);
247 break;
248 }
249 atomic_add_32(&route_cb.any_count, -1);
250 return raw_usrreqs.pru_detach(so);
251 }
252
253 static int
254 rts_disconnect(struct socket *so)
255 {
256 return raw_usrreqs.pru_disconnect(so);
257 }
258
259 /* pru_listen is EOPNOTSUPP */
260
261 static int
262 rts_peeraddr(struct socket *so, struct sockaddr **nam)
263 {
264 return raw_usrreqs.pru_peeraddr(so, nam);
265 }
266
267 /* pru_rcvd is EOPNOTSUPP */
268 /* pru_rcvoob is EOPNOTSUPP */
269
270 static int
271 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
272 struct mbuf *control, struct proc *p)
273 {
274 return raw_usrreqs.pru_send(so, flags, m, nam, control, p);
275 }
276
277 /* pru_sense is null */
278
279 static int
280 rts_shutdown(struct socket *so)
281 {
282 return raw_usrreqs.pru_shutdown(so);
283 }
284
285 static int
286 rts_sockaddr(struct socket *so, struct sockaddr **nam)
287 {
288 return raw_usrreqs.pru_sockaddr(so, nam);
289 }
290
291 static struct pr_usrreqs route_usrreqs = {
292 .pru_abort = rts_abort,
293 .pru_attach = rts_attach,
294 .pru_bind = rts_bind,
295 .pru_connect = rts_connect,
296 .pru_detach = rts_detach,
297 .pru_disconnect = rts_disconnect,
298 .pru_peeraddr = rts_peeraddr,
299 .pru_send = rts_send,
300 .pru_shutdown = rts_shutdown,
301 .pru_sockaddr = rts_sockaddr,
302 .pru_sosend = sosend,
303 .pru_soreceive = soreceive,
304 };
305
306 /*ARGSUSED*/
307 static int
308 route_output(struct mbuf *m, struct socket *so)
309 {
310 struct rt_msghdr *rtm = NULL;
311 struct rtentry *rt = NULL;
312 struct rtentry *saved_nrt = NULL;
313 struct radix_node_head *rnh;
314 struct rt_addrinfo info;
315 int len, error = 0;
316 sa_family_t dst_sa_family = 0;
317 struct ifnet *ifp = NULL;
318 struct sockaddr_in dst_in, gate_in;
319 int sendonlytoself = 0;
320 unsigned int ifscope = IFSCOPE_NONE;
321 struct rawcb *rp = NULL;
322 boolean_t is_router = FALSE;
323 #define senderr(e) { error = (e); goto flush; }
324 if (m == NULL || ((m->m_len < sizeof(intptr_t)) &&
325 (m = m_pullup(m, sizeof(intptr_t))) == NULL)) {
326 return ENOBUFS;
327 }
328 VERIFY(m->m_flags & M_PKTHDR);
329
330 /*
331 * Unlock the socket (but keep a reference) it won't be
332 * accessed until raw_input appends to it.
333 */
334 socket_unlock(so, 0);
335 lck_mtx_lock(rnh_lock);
336
337 len = m->m_pkthdr.len;
338 if (len < sizeof(*rtm) ||
339 len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
340 info.rti_info[RTAX_DST] = NULL;
341 senderr(EINVAL);
342 }
343 R_Malloc(rtm, struct rt_msghdr *, len);
344 if (rtm == NULL) {
345 info.rti_info[RTAX_DST] = NULL;
346 senderr(ENOBUFS);
347 }
348 m_copydata(m, 0, len, (caddr_t)rtm);
349 if (rtm->rtm_version != RTM_VERSION) {
350 info.rti_info[RTAX_DST] = NULL;
351 senderr(EPROTONOSUPPORT);
352 }
353
354 /*
355 * Silent version of RTM_GET for Reachabiltiy APIs. We may change
356 * all RTM_GETs to be silent in the future, so this is private for now.
357 */
358 if (rtm->rtm_type == RTM_GET_SILENT) {
359 if (!(so->so_options & SO_USELOOPBACK)) {
360 senderr(EINVAL);
361 }
362 sendonlytoself = 1;
363 rtm->rtm_type = RTM_GET;
364 }
365
366 /*
367 * Perform permission checking, only privileged sockets
368 * may perform operations other than RTM_GET
369 */
370 if (rtm->rtm_type != RTM_GET && !(so->so_state & SS_PRIV)) {
371 info.rti_info[RTAX_DST] = NULL;
372 senderr(EPERM);
373 }
374
375 rtm->rtm_pid = proc_selfpid();
376 info.rti_addrs = rtm->rtm_addrs;
377 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
378 info.rti_info[RTAX_DST] = NULL;
379 senderr(EINVAL);
380 }
381 if (info.rti_info[RTAX_DST] == NULL ||
382 info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
383 (info.rti_info[RTAX_GATEWAY] != NULL &&
384 info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
385 senderr(EINVAL);
386 }
387
388 if (info.rti_info[RTAX_DST]->sa_family == AF_INET &&
389 info.rti_info[RTAX_DST]->sa_len != sizeof(struct sockaddr_in)) {
390 /* At minimum, we need up to sin_addr */
391 if (info.rti_info[RTAX_DST]->sa_len <
392 offsetof(struct sockaddr_in, sin_zero)) {
393 senderr(EINVAL);
394 }
395 bzero(&dst_in, sizeof(dst_in));
396 dst_in.sin_len = sizeof(dst_in);
397 dst_in.sin_family = AF_INET;
398 dst_in.sin_port = SIN(info.rti_info[RTAX_DST])->sin_port;
399 dst_in.sin_addr = SIN(info.rti_info[RTAX_DST])->sin_addr;
400 info.rti_info[RTAX_DST] = (struct sockaddr *)&dst_in;
401 dst_sa_family = info.rti_info[RTAX_DST]->sa_family;
402 } else if (info.rti_info[RTAX_DST]->sa_family == AF_INET6 &&
403 info.rti_info[RTAX_DST]->sa_len < sizeof(struct sockaddr_in6)) {
404 senderr(EINVAL);
405 }
406
407 if (info.rti_info[RTAX_GATEWAY] != NULL) {
408 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_INET &&
409 info.rti_info[RTAX_GATEWAY]->sa_len != sizeof(struct sockaddr_in)) {
410 /* At minimum, we need up to sin_addr */
411 if (info.rti_info[RTAX_GATEWAY]->sa_len <
412 offsetof(struct sockaddr_in, sin_zero)) {
413 senderr(EINVAL);
414 }
415 bzero(&gate_in, sizeof(gate_in));
416 gate_in.sin_len = sizeof(gate_in);
417 gate_in.sin_family = AF_INET;
418 gate_in.sin_port = SIN(info.rti_info[RTAX_GATEWAY])->sin_port;
419 gate_in.sin_addr = SIN(info.rti_info[RTAX_GATEWAY])->sin_addr;
420 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gate_in;
421 } else if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_INET6 &&
422 info.rti_info[RTAX_GATEWAY]->sa_len < sizeof(struct sockaddr_in6)) {
423 senderr(EINVAL);
424 }
425 }
426
427 if (info.rti_info[RTAX_GENMASK]) {
428 struct radix_node *t;
429 t = rn_addmask((caddr_t)info.rti_info[RTAX_GENMASK], 0, 1);
430 if (t != NULL && Bcmp(info.rti_info[RTAX_GENMASK],
431 t->rn_key, *(u_char *)info.rti_info[RTAX_GENMASK]) == 0) {
432 info.rti_info[RTAX_GENMASK] =
433 (struct sockaddr *)(t->rn_key);
434 } else {
435 senderr(ENOBUFS);
436 }
437 }
438
439 /*
440 * If RTF_IFSCOPE flag is set, then rtm_index specifies the scope.
441 */
442 if (rtm->rtm_flags & RTF_IFSCOPE) {
443 if (info.rti_info[RTAX_DST]->sa_family != AF_INET &&
444 info.rti_info[RTAX_DST]->sa_family != AF_INET6) {
445 senderr(EINVAL);
446 }
447 ifscope = rtm->rtm_index;
448 }
449 /*
450 * Block changes on INTCOPROC interfaces.
451 */
452 if (ifscope) {
453 unsigned int intcoproc_scope = 0;
454 ifnet_head_lock_shared();
455 TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
456 if (IFNET_IS_INTCOPROC(ifp)) {
457 intcoproc_scope = ifp->if_index;
458 break;
459 }
460 }
461 ifnet_head_done();
462 if (intcoproc_scope == ifscope && current_proc()->p_pid != 0) {
463 senderr(EINVAL);
464 }
465 }
466
467 /*
468 * RTF_PROXY can only be set internally from within the kernel.
469 */
470 if (rtm->rtm_flags & RTF_PROXY) {
471 senderr(EINVAL);
472 }
473
474 /*
475 * For AF_INET, always zero out the embedded scope ID. If this is
476 * a scoped request, it must be done explicitly by setting RTF_IFSCOPE
477 * flag and the corresponding rtm_index value. This is to prevent
478 * false interpretation of the scope ID because it's using the sin_zero
479 * field, which might not be properly cleared by the requestor.
480 */
481 if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
482 sin_set_ifscope(info.rti_info[RTAX_DST], IFSCOPE_NONE);
483 }
484 if (info.rti_info[RTAX_GATEWAY] != NULL &&
485 info.rti_info[RTAX_GATEWAY]->sa_family == AF_INET) {
486 sin_set_ifscope(info.rti_info[RTAX_GATEWAY], IFSCOPE_NONE);
487 }
488 switch (rtm->rtm_type) {
489 case RTM_ADD:
490 if (info.rti_info[RTAX_GATEWAY] == NULL) {
491 senderr(EINVAL);
492 }
493
494 error = rtrequest_scoped_locked(RTM_ADD,
495 info.rti_info[RTAX_DST], info.rti_info[RTAX_GATEWAY],
496 info.rti_info[RTAX_NETMASK], rtm->rtm_flags, &saved_nrt,
497 ifscope);
498 if (error == 0 && saved_nrt != NULL) {
499 RT_LOCK(saved_nrt);
500 /*
501 * If the route request specified an interface with
502 * IFA and/or IFP, we set the requested interface on
503 * the route with rt_setif. It would be much better
504 * to do this inside rtrequest, but that would
505 * require passing the desired interface, in some
506 * form, to rtrequest. Since rtrequest is called in
507 * so many places (roughly 40 in our source), adding
508 * a parameter is to much for us to swallow; this is
509 * something for the FreeBSD developers to tackle.
510 * Instead, we let rtrequest compute whatever
511 * interface it wants, then come in behind it and
512 * stick in the interface that we really want. This
513 * works reasonably well except when rtrequest can't
514 * figure out what interface to use (with
515 * ifa_withroute) and returns ENETUNREACH. Ideally
516 * it shouldn't matter if rtrequest can't figure out
517 * the interface if we're going to explicitly set it
518 * ourselves anyway. But practically we can't
519 * recover here because rtrequest will not do any of
520 * the work necessary to add the route if it can't
521 * find an interface. As long as there is a default
522 * route that leads to some interface, rtrequest will
523 * find an interface, so this problem should be
524 * rarely encountered.
525 * dwiggins@bbn.com
526 */
527 rt_setif(saved_nrt,
528 info.rti_info[RTAX_IFP], info.rti_info[RTAX_IFA],
529 info.rti_info[RTAX_GATEWAY], ifscope);
530 (void)rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, saved_nrt);
531 saved_nrt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
532 saved_nrt->rt_rmx.rmx_locks |=
533 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
534 saved_nrt->rt_genmask = info.rti_info[RTAX_GENMASK];
535 RT_REMREF_LOCKED(saved_nrt);
536 RT_UNLOCK(saved_nrt);
537 }
538 break;
539
540 case RTM_DELETE:
541 error = rtrequest_scoped_locked(RTM_DELETE,
542 info.rti_info[RTAX_DST], info.rti_info[RTAX_GATEWAY],
543 info.rti_info[RTAX_NETMASK], rtm->rtm_flags, &saved_nrt,
544 ifscope);
545 if (error == 0) {
546 rt = saved_nrt;
547 RT_LOCK(rt);
548 goto report;
549 }
550 break;
551
552 case RTM_GET:
553 case RTM_CHANGE:
554 case RTM_LOCK:
555 rnh = rt_tables[info.rti_info[RTAX_DST]->sa_family];
556 if (rnh == NULL) {
557 senderr(EAFNOSUPPORT);
558 }
559 /*
560 * Lookup the best match based on the key-mask pair;
561 * callee adds a reference and checks for root node.
562 */
563 rt = rt_lookup(TRUE, info.rti_info[RTAX_DST],
564 info.rti_info[RTAX_NETMASK], rnh, ifscope);
565 if (rt == NULL) {
566 senderr(ESRCH);
567 }
568 RT_LOCK(rt);
569
570 /*
571 * Holding rnh_lock here prevents the possibility of
572 * ifa from changing (e.g. in_ifinit), so it is safe
573 * to access its ifa_addr (down below) without locking.
574 */
575 switch (rtm->rtm_type) {
576 case RTM_GET: {
577 kauth_cred_t cred;
578 kauth_cred_t* credp;
579 struct ifaddr *ifa2;
580 report:
581 cred = kauth_cred_proc_ref(current_proc());
582 credp = &cred;
583
584 ifa2 = NULL;
585 RT_LOCK_ASSERT_HELD(rt);
586 info.rti_info[RTAX_DST] = rt_key(rt);
587 dst_sa_family = info.rti_info[RTAX_DST]->sa_family;
588 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
589 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
590 info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
591 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
592 ifp = rt->rt_ifp;
593 if (ifp != NULL) {
594 ifnet_lock_shared(ifp);
595 ifa2 = ifp->if_lladdr;
596 info.rti_info[RTAX_IFP] =
597 ifa2->ifa_addr;
598 IFA_ADDREF(ifa2);
599 ifnet_lock_done(ifp);
600 info.rti_info[RTAX_IFA] =
601 rt->rt_ifa->ifa_addr;
602 rtm->rtm_index = ifp->if_index;
603 } else {
604 info.rti_info[RTAX_IFP] = NULL;
605 info.rti_info[RTAX_IFA] = NULL;
606 }
607 } else if ((ifp = rt->rt_ifp) != NULL) {
608 rtm->rtm_index = ifp->if_index;
609 }
610 if (ifa2 != NULL) {
611 IFA_LOCK(ifa2);
612 }
613 len = rt_msg2(rtm->rtm_type, &info, NULL, NULL, credp);
614 if (ifa2 != NULL) {
615 IFA_UNLOCK(ifa2);
616 }
617 struct rt_msghdr *out_rtm;
618 R_Malloc(out_rtm, struct rt_msghdr *, len);
619 if (out_rtm == NULL) {
620 RT_UNLOCK(rt);
621 if (ifa2 != NULL) {
622 IFA_REMREF(ifa2);
623 }
624 senderr(ENOBUFS);
625 }
626 Bcopy(rtm, out_rtm, sizeof(struct rt_msghdr));
627 if (ifa2 != NULL) {
628 IFA_LOCK(ifa2);
629 }
630 (void) rt_msg2(out_rtm->rtm_type, &info, (caddr_t)out_rtm,
631 NULL, &cred);
632 if (ifa2 != NULL) {
633 IFA_UNLOCK(ifa2);
634 }
635 R_Free(rtm);
636 rtm = out_rtm;
637 rtm->rtm_flags = rt->rt_flags;
638 rt_getmetrics(rt, &rtm->rtm_rmx);
639 rtm->rtm_addrs = info.rti_addrs;
640 if (ifa2 != NULL) {
641 IFA_REMREF(ifa2);
642 }
643
644 kauth_cred_unref(&cred);
645 break;
646 }
647
648 case RTM_CHANGE:
649 is_router = (rt->rt_flags & RTF_ROUTER) ? TRUE : FALSE;
650
651 if (info.rti_info[RTAX_GATEWAY] != NULL &&
652 (error = rt_setgate(rt, rt_key(rt),
653 info.rti_info[RTAX_GATEWAY]))) {
654 int tmp = error;
655 RT_UNLOCK(rt);
656 senderr(tmp);
657 }
658 /*
659 * If they tried to change things but didn't specify
660 * the required gateway, then just use the old one.
661 * This can happen if the user tries to change the
662 * flags on the default route without changing the
663 * default gateway. Changing flags still doesn't work.
664 */
665 if ((rt->rt_flags & RTF_GATEWAY) &&
666 info.rti_info[RTAX_GATEWAY] == NULL) {
667 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
668 }
669
670 /*
671 * On Darwin, we call rt_setif which contains the
672 * equivalent to the code found at this very spot
673 * in BSD.
674 */
675 rt_setif(rt,
676 info.rti_info[RTAX_IFP], info.rti_info[RTAX_IFA],
677 info.rti_info[RTAX_GATEWAY], ifscope);
678
679 if ((error = rt_setmetrics(rtm->rtm_inits,
680 &rtm->rtm_rmx, rt))) {
681 int tmp = error;
682 RT_UNLOCK(rt);
683 senderr(tmp);
684 }
685 if (info.rti_info[RTAX_GENMASK]) {
686 rt->rt_genmask = info.rti_info[RTAX_GENMASK];
687 }
688
689 /*
690 * Enqueue work item to invoke callback for this route entry
691 * This may not be needed always, but for now issue it anytime
692 * RTM_CHANGE gets called.
693 */
694 route_event_enqueue_nwk_wq_entry(rt, NULL, ROUTE_ENTRY_REFRESH, NULL, TRUE);
695 /*
696 * If the route is for a router, walk the tree to send refresh
697 * event to protocol cloned entries
698 */
699 if (is_router) {
700 struct route_event rt_ev;
701 route_event_init(&rt_ev, rt, NULL, ROUTE_ENTRY_REFRESH);
702 RT_UNLOCK(rt);
703 (void) rnh->rnh_walktree(rnh, route_event_walktree, (void *)&rt_ev);
704 RT_LOCK(rt);
705 }
706 /* FALLTHRU */
707 case RTM_LOCK:
708 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
709 rt->rt_rmx.rmx_locks |=
710 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
711 break;
712 }
713 RT_UNLOCK(rt);
714 break;
715 default:
716 senderr(EOPNOTSUPP);
717 }
718 flush:
719 if (rtm != NULL) {
720 if (error) {
721 rtm->rtm_errno = error;
722 } else {
723 rtm->rtm_flags |= RTF_DONE;
724 }
725 }
726 if (rt != NULL) {
727 RT_LOCK_ASSERT_NOTHELD(rt);
728 rtfree_locked(rt);
729 }
730 lck_mtx_unlock(rnh_lock);
731
732 /* relock the socket now */
733 socket_lock(so, 0);
734 /*
735 * Check to see if we don't want our own messages.
736 */
737 if (!(so->so_options & SO_USELOOPBACK)) {
738 if (route_cb.any_count <= 1) {
739 if (rtm != NULL) {
740 R_Free(rtm);
741 }
742 m_freem(m);
743 return error;
744 }
745 /* There is another listener, so construct message */
746 rp = sotorawcb(so);
747 }
748 if (rtm != NULL) {
749 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
750 if (m->m_pkthdr.len < rtm->rtm_msglen) {
751 m_freem(m);
752 m = NULL;
753 } else if (m->m_pkthdr.len > rtm->rtm_msglen) {
754 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
755 }
756 R_Free(rtm);
757 }
758 if (sendonlytoself && m != NULL) {
759 error = 0;
760 if (sbappendaddr(&so->so_rcv, &route_src, m,
761 NULL, &error) != 0) {
762 sorwakeup(so);
763 }
764 if (error) {
765 return error;
766 }
767 } else {
768 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
769 if (rp != NULL) {
770 rp->rcb_proto.sp_family = 0; /* Avoid us */
771 }
772 if (dst_sa_family != 0) {
773 route_proto.sp_protocol = dst_sa_family;
774 }
775 if (m != NULL) {
776 socket_unlock(so, 0);
777 raw_input(m, &route_proto, &route_src, &route_dst);
778 socket_lock(so, 0);
779 }
780 if (rp != NULL) {
781 rp->rcb_proto.sp_family = PF_ROUTE;
782 }
783 }
784 return error;
785 }
786
787 void
788 rt_setexpire(struct rtentry *rt, uint64_t expiry)
789 {
790 /* set both rt_expire and rmx_expire */
791 rt->rt_expire = expiry;
792 if (expiry) {
793 rt->rt_rmx.rmx_expire = expiry + rt->base_calendartime -
794 rt->base_uptime;
795 } else {
796 rt->rt_rmx.rmx_expire = 0;
797 }
798 }
799
800 static int
801 rt_setmetrics(u_int32_t which, struct rt_metrics *in, struct rtentry *out)
802 {
803 if (!(which & RTV_REFRESH_HOST)) {
804 struct timeval caltime;
805 getmicrotime(&caltime);
806 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->e;
807 metric(RTV_RPIPE, rmx_recvpipe);
808 metric(RTV_SPIPE, rmx_sendpipe);
809 metric(RTV_SSTHRESH, rmx_ssthresh);
810 metric(RTV_RTT, rmx_rtt);
811 metric(RTV_RTTVAR, rmx_rttvar);
812 metric(RTV_HOPCOUNT, rmx_hopcount);
813 metric(RTV_MTU, rmx_mtu);
814 metric(RTV_EXPIRE, rmx_expire);
815 #undef metric
816 if (out->rt_rmx.rmx_expire > 0) {
817 /* account for system time change */
818 getmicrotime(&caltime);
819 out->base_calendartime +=
820 NET_CALCULATE_CLOCKSKEW(caltime,
821 out->base_calendartime,
822 net_uptime(), out->base_uptime);
823 rt_setexpire(out,
824 out->rt_rmx.rmx_expire -
825 out->base_calendartime +
826 out->base_uptime);
827 } else {
828 rt_setexpire(out, 0);
829 }
830
831 VERIFY(out->rt_expire == 0 || out->rt_rmx.rmx_expire != 0);
832 VERIFY(out->rt_expire != 0 || out->rt_rmx.rmx_expire == 0);
833 } else {
834 /* Only RTV_REFRESH_HOST must be set */
835 if ((which & ~RTV_REFRESH_HOST) ||
836 (out->rt_flags & RTF_STATIC) ||
837 !(out->rt_flags & RTF_LLINFO)) {
838 return EINVAL;
839 }
840
841 if (out->rt_llinfo_refresh == NULL) {
842 return ENOTSUP;
843 }
844
845 out->rt_llinfo_refresh(out);
846 }
847 return 0;
848 }
849
850 static void
851 rt_getmetrics(struct rtentry *in, struct rt_metrics *out)
852 {
853 struct timeval caltime;
854
855 VERIFY(in->rt_expire == 0 || in->rt_rmx.rmx_expire != 0);
856 VERIFY(in->rt_expire != 0 || in->rt_rmx.rmx_expire == 0);
857
858 *out = in->rt_rmx;
859
860 if (in->rt_expire != 0) {
861 /* account for system time change */
862 getmicrotime(&caltime);
863
864 in->base_calendartime +=
865 NET_CALCULATE_CLOCKSKEW(caltime,
866 in->base_calendartime, net_uptime(), in->base_uptime);
867
868 out->rmx_expire = in->base_calendartime +
869 in->rt_expire - in->base_uptime;
870 } else {
871 out->rmx_expire = 0;
872 }
873 }
874
875 /*
876 * Set route's interface given info.rti_info[RTAX_IFP],
877 * info.rti_info[RTAX_IFA], and gateway.
878 */
879 static void
880 rt_setif(struct rtentry *rt, struct sockaddr *Ifpaddr, struct sockaddr *Ifaaddr,
881 struct sockaddr *Gate, unsigned int ifscope)
882 {
883 struct ifaddr *ifa = NULL;
884 struct ifnet *ifp = NULL;
885 void (*ifa_rtrequest)(int, struct rtentry *, struct sockaddr *);
886
887 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_OWNED);
888
889 RT_LOCK_ASSERT_HELD(rt);
890
891 /* Don't update a defunct route */
892 if (rt->rt_flags & RTF_CONDEMNED) {
893 return;
894 }
895
896 /* Add an extra ref for ourselves */
897 RT_ADDREF_LOCKED(rt);
898
899 /* Become a regular mutex, just in case */
900 RT_CONVERT_LOCK(rt);
901
902 /*
903 * New gateway could require new ifaddr, ifp; flags may also
904 * be different; ifp may be specified by ll sockaddr when
905 * protocol address is ambiguous.
906 */
907 if (Ifpaddr && (ifa = ifa_ifwithnet_scoped(Ifpaddr, ifscope)) &&
908 (ifp = ifa->ifa_ifp) && (Ifaaddr || Gate)) {
909 IFA_REMREF(ifa);
910 ifa = ifaof_ifpforaddr(Ifaaddr ? Ifaaddr : Gate, ifp);
911 } else {
912 if (ifa != NULL) {
913 IFA_REMREF(ifa);
914 ifa = NULL;
915 }
916 if (Ifpaddr && (ifp = if_withname(Ifpaddr))) {
917 if (Gate) {
918 ifa = ifaof_ifpforaddr(Gate, ifp);
919 } else {
920 ifnet_lock_shared(ifp);
921 ifa = TAILQ_FIRST(&ifp->if_addrhead);
922 if (ifa != NULL) {
923 IFA_ADDREF(ifa);
924 }
925 ifnet_lock_done(ifp);
926 }
927 } else if (Ifaaddr &&
928 (ifa = ifa_ifwithaddr_scoped(Ifaaddr, ifscope))) {
929 ifp = ifa->ifa_ifp;
930 } else if (Gate != NULL) {
931 /*
932 * Safe to drop rt_lock and use rt_key, since holding
933 * rnh_lock here prevents another thread from calling
934 * rt_setgate() on this route. We cannot hold the
935 * lock across ifa_ifwithroute since the lookup done
936 * by that routine may point to the same route.
937 */
938 RT_UNLOCK(rt);
939 if ((ifa = ifa_ifwithroute_scoped_locked(rt->rt_flags,
940 rt_key(rt), Gate, ifscope)) != NULL) {
941 ifp = ifa->ifa_ifp;
942 }
943 RT_LOCK(rt);
944 /* Don't update a defunct route */
945 if (rt->rt_flags & RTF_CONDEMNED) {
946 if (ifa != NULL) {
947 IFA_REMREF(ifa);
948 }
949 /* Release extra ref */
950 RT_REMREF_LOCKED(rt);
951 return;
952 }
953 }
954 }
955
956 /* trigger route cache reevaluation */
957 if (rt_key(rt)->sa_family == AF_INET) {
958 routegenid_inet_update();
959 }
960 #if INET6
961 else if (rt_key(rt)->sa_family == AF_INET6) {
962 routegenid_inet6_update();
963 }
964 #endif /* INET6 */
965
966 if (ifa != NULL) {
967 struct ifaddr *oifa = rt->rt_ifa;
968 if (oifa != ifa) {
969 if (oifa != NULL) {
970 IFA_LOCK_SPIN(oifa);
971 ifa_rtrequest = oifa->ifa_rtrequest;
972 IFA_UNLOCK(oifa);
973 if (ifa_rtrequest != NULL) {
974 ifa_rtrequest(RTM_DELETE, rt, Gate);
975 }
976 }
977 rtsetifa(rt, ifa);
978
979 if (rt->rt_ifp != ifp) {
980 /*
981 * Purge any link-layer info caching.
982 */
983 if (rt->rt_llinfo_purge != NULL) {
984 rt->rt_llinfo_purge(rt);
985 }
986
987 /*
988 * Adjust route ref count for the interfaces.
989 */
990 if (rt->rt_if_ref_fn != NULL) {
991 rt->rt_if_ref_fn(ifp, 1);
992 rt->rt_if_ref_fn(rt->rt_ifp, -1);
993 }
994 }
995 rt->rt_ifp = ifp;
996 /*
997 * If this is the (non-scoped) default route, record
998 * the interface index used for the primary ifscope.
999 */
1000 if (rt_primary_default(rt, rt_key(rt))) {
1001 set_primary_ifscope(rt_key(rt)->sa_family,
1002 rt->rt_ifp->if_index);
1003 }
1004 /*
1005 * If rmx_mtu is not locked, update it
1006 * to the MTU used by the new interface.
1007 */
1008 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1009 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
1010 if (rt_key(rt)->sa_family == AF_INET &&
1011 INTF_ADJUST_MTU_FOR_CLAT46(ifp)) {
1012 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp);
1013 /* Further adjust the size for CLAT46 expansion */
1014 rt->rt_rmx.rmx_mtu -= CLAT46_HDR_EXPANSION_OVERHD;
1015 }
1016 }
1017
1018 if (rt->rt_ifa != NULL) {
1019 IFA_LOCK_SPIN(rt->rt_ifa);
1020 ifa_rtrequest = rt->rt_ifa->ifa_rtrequest;
1021 IFA_UNLOCK(rt->rt_ifa);
1022 if (ifa_rtrequest != NULL) {
1023 ifa_rtrequest(RTM_ADD, rt, Gate);
1024 }
1025 }
1026 IFA_REMREF(ifa);
1027 /* Release extra ref */
1028 RT_REMREF_LOCKED(rt);
1029 return;
1030 }
1031 IFA_REMREF(ifa);
1032 ifa = NULL;
1033 }
1034
1035 /* XXX: to reset gateway to correct value, at RTM_CHANGE */
1036 if (rt->rt_ifa != NULL) {
1037 IFA_LOCK_SPIN(rt->rt_ifa);
1038 ifa_rtrequest = rt->rt_ifa->ifa_rtrequest;
1039 IFA_UNLOCK(rt->rt_ifa);
1040 if (ifa_rtrequest != NULL) {
1041 ifa_rtrequest(RTM_ADD, rt, Gate);
1042 }
1043 }
1044
1045 /*
1046 * Workaround for local address routes pointing to the loopback
1047 * interface added by configd, until <rdar://problem/12970142>.
1048 */
1049 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
1050 (rt->rt_flags & RTF_HOST) && rt->rt_ifa->ifa_ifp == rt->rt_ifp) {
1051 ifa = ifa_ifwithaddr(rt_key(rt));
1052 if (ifa != NULL) {
1053 if (ifa != rt->rt_ifa) {
1054 rtsetifa(rt, ifa);
1055 }
1056 IFA_REMREF(ifa);
1057 }
1058 }
1059
1060 /* Release extra ref */
1061 RT_REMREF_LOCKED(rt);
1062 }
1063
1064 /*
1065 * Extract the addresses of the passed sockaddrs.
1066 * Do a little sanity checking so as to avoid bad memory references.
1067 * This data is derived straight from userland.
1068 */
1069 static int
1070 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1071 {
1072 struct sockaddr *sa;
1073 int i;
1074
1075 bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
1076 for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
1077 if ((rtinfo->rti_addrs & (1 << i)) == 0) {
1078 continue;
1079 }
1080 sa = (struct sockaddr *)cp;
1081 /*
1082 * It won't fit.
1083 */
1084 if ((cp + sa->sa_len) > cplim) {
1085 return EINVAL;
1086 }
1087 if (sa->sa_len > sizeof(struct sockaddr_storage)) {
1088 return EINVAL;
1089 }
1090 /*
1091 * there are no more.. quit now
1092 * If there are more bits, they are in error.
1093 * I've seen this. route(1) can evidently generate these.
1094 * This causes kernel to core dump.
1095 * for compatibility, If we see this, point to a safe address.
1096 */
1097 if (sa->sa_len == 0) {
1098 rtinfo->rti_info[i] = &sa_zero;
1099 return 0; /* should be EINVAL but for compat */
1100 }
1101 /* accept it */
1102 rtinfo->rti_info[i] = sa;
1103 ADVANCE32(cp, sa);
1104 }
1105 return 0;
1106 }
1107
1108 static struct mbuf *
1109 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1110 {
1111 struct rt_msghdr *rtm;
1112 struct mbuf *m;
1113 int i;
1114 int len, dlen, off;
1115
1116 switch (type) {
1117 case RTM_DELADDR:
1118 case RTM_NEWADDR:
1119 len = sizeof(struct ifa_msghdr);
1120 break;
1121
1122 case RTM_DELMADDR:
1123 case RTM_NEWMADDR:
1124 len = sizeof(struct ifma_msghdr);
1125 break;
1126
1127 case RTM_IFINFO:
1128 len = sizeof(struct if_msghdr);
1129 break;
1130
1131 default:
1132 len = sizeof(struct rt_msghdr);
1133 }
1134 m = m_gethdr(M_DONTWAIT, MT_DATA);
1135 if (m && len > MHLEN) {
1136 MCLGET(m, M_DONTWAIT);
1137 if (!(m->m_flags & M_EXT)) {
1138 m_free(m);
1139 m = NULL;
1140 }
1141 }
1142 if (m == NULL) {
1143 return NULL;
1144 }
1145 m->m_pkthdr.len = m->m_len = len;
1146 m->m_pkthdr.rcvif = NULL;
1147 rtm = mtod(m, struct rt_msghdr *);
1148 bzero((caddr_t)rtm, len);
1149 off = len;
1150 for (i = 0; i < RTAX_MAX; i++) {
1151 struct sockaddr *sa, *hint;
1152 uint8_t ssbuf[SOCK_MAXADDRLEN + 1];
1153
1154 /*
1155 * Make sure to accomodate the largest possible size of sa_len.
1156 */
1157 _CASSERT(sizeof(ssbuf) == (SOCK_MAXADDRLEN + 1));
1158
1159 if ((sa = rtinfo->rti_info[i]) == NULL) {
1160 continue;
1161 }
1162
1163 switch (i) {
1164 case RTAX_DST:
1165 case RTAX_NETMASK:
1166 if ((hint = rtinfo->rti_info[RTAX_DST]) == NULL) {
1167 hint = rtinfo->rti_info[RTAX_IFA];
1168 }
1169
1170 /* Scrub away any trace of embedded interface scope */
1171 sa = rtm_scrub(type, i, hint, sa, &ssbuf,
1172 sizeof(ssbuf), NULL);
1173 break;
1174
1175 default:
1176 break;
1177 }
1178
1179 rtinfo->rti_addrs |= (1 << i);
1180 dlen = sa->sa_len;
1181 m_copyback(m, off, dlen, (caddr_t)sa);
1182 len = off + dlen;
1183 off += ROUNDUP32(dlen);
1184 }
1185 if (m->m_pkthdr.len != len) {
1186 m_freem(m);
1187 return NULL;
1188 }
1189 rtm->rtm_msglen = len;
1190 rtm->rtm_version = RTM_VERSION;
1191 rtm->rtm_type = type;
1192 return m;
1193 }
1194
1195 static int
1196 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w,
1197 kauth_cred_t* credp)
1198 {
1199 int i;
1200 int len, dlen, rlen, second_time = 0;
1201 caddr_t cp0;
1202
1203 rtinfo->rti_addrs = 0;
1204 again:
1205 switch (type) {
1206 case RTM_DELADDR:
1207 case RTM_NEWADDR:
1208 len = sizeof(struct ifa_msghdr);
1209 break;
1210
1211 case RTM_DELMADDR:
1212 case RTM_NEWMADDR:
1213 len = sizeof(struct ifma_msghdr);
1214 break;
1215
1216 case RTM_IFINFO:
1217 len = sizeof(struct if_msghdr);
1218 break;
1219
1220 case RTM_IFINFO2:
1221 len = sizeof(struct if_msghdr2);
1222 break;
1223
1224 case RTM_NEWMADDR2:
1225 len = sizeof(struct ifma_msghdr2);
1226 break;
1227
1228 case RTM_GET_EXT:
1229 len = sizeof(struct rt_msghdr_ext);
1230 break;
1231
1232 case RTM_GET2:
1233 len = sizeof(struct rt_msghdr2);
1234 break;
1235
1236 default:
1237 len = sizeof(struct rt_msghdr);
1238 }
1239 cp0 = cp;
1240 if (cp0) {
1241 cp += len;
1242 }
1243 for (i = 0; i < RTAX_MAX; i++) {
1244 struct sockaddr *sa, *hint;
1245 uint8_t ssbuf[SOCK_MAXADDRLEN + 1];
1246
1247 /*
1248 * Make sure to accomodate the largest possible size of sa_len.
1249 */
1250 _CASSERT(sizeof(ssbuf) == (SOCK_MAXADDRLEN + 1));
1251
1252 if ((sa = rtinfo->rti_info[i]) == NULL) {
1253 continue;
1254 }
1255
1256 switch (i) {
1257 case RTAX_DST:
1258 case RTAX_NETMASK:
1259 if ((hint = rtinfo->rti_info[RTAX_DST]) == NULL) {
1260 hint = rtinfo->rti_info[RTAX_IFA];
1261 }
1262
1263 /* Scrub away any trace of embedded interface scope */
1264 sa = rtm_scrub(type, i, hint, sa, &ssbuf,
1265 sizeof(ssbuf), NULL);
1266 break;
1267 case RTAX_GATEWAY:
1268 case RTAX_IFP:
1269 sa = rtm_scrub(type, i, NULL, sa, &ssbuf,
1270 sizeof(ssbuf), credp);
1271 break;
1272
1273 default:
1274 break;
1275 }
1276
1277 rtinfo->rti_addrs |= (1 << i);
1278 dlen = sa->sa_len;
1279 rlen = ROUNDUP32(dlen);
1280 if (cp) {
1281 bcopy((caddr_t)sa, cp, (size_t)dlen);
1282 if (dlen != rlen) {
1283 bzero(cp + dlen, rlen - dlen);
1284 }
1285 cp += rlen;
1286 }
1287 len += rlen;
1288 }
1289 if (cp == NULL && w != NULL && !second_time) {
1290 struct walkarg *rw = w;
1291
1292 if (rw->w_req != NULL) {
1293 if (rw->w_tmemsize < len) {
1294 if (rw->w_tmem != NULL) {
1295 FREE(rw->w_tmem, M_RTABLE);
1296 }
1297 rw->w_tmem = _MALLOC(len, M_RTABLE, M_ZERO | M_WAITOK);
1298 if (rw->w_tmem != NULL) {
1299 rw->w_tmemsize = len;
1300 }
1301 }
1302 if (rw->w_tmem != NULL) {
1303 cp = rw->w_tmem;
1304 second_time = 1;
1305 goto again;
1306 }
1307 }
1308 }
1309 if (cp) {
1310 struct rt_msghdr *rtm = (struct rt_msghdr *)(void *)cp0;
1311
1312 rtm->rtm_version = RTM_VERSION;
1313 rtm->rtm_type = type;
1314 rtm->rtm_msglen = len;
1315 }
1316 return len;
1317 }
1318
1319 /*
1320 * This routine is called to generate a message from the routing
1321 * socket indicating that a redirect has occurred, a routing lookup
1322 * has failed, or that a protocol has detected timeouts to a particular
1323 * destination.
1324 */
1325 void
1326 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1327 {
1328 struct rt_msghdr *rtm;
1329 struct mbuf *m;
1330 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1331 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1332
1333 if (route_cb.any_count == 0) {
1334 return;
1335 }
1336 m = rt_msg1(type, rtinfo);
1337 if (m == NULL) {
1338 return;
1339 }
1340 rtm = mtod(m, struct rt_msghdr *);
1341 rtm->rtm_flags = RTF_DONE | flags;
1342 rtm->rtm_errno = error;
1343 rtm->rtm_addrs = rtinfo->rti_addrs;
1344 route_proto.sp_family = sa ? sa->sa_family : 0;
1345 raw_input(m, &route_proto, &route_src, &route_dst);
1346 }
1347
1348 /*
1349 * This routine is called to generate a message from the routing
1350 * socket indicating that the status of a network interface has changed.
1351 */
1352 void
1353 rt_ifmsg(struct ifnet *ifp)
1354 {
1355 struct if_msghdr *ifm;
1356 struct mbuf *m;
1357 struct rt_addrinfo info;
1358 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1359
1360 if (route_cb.any_count == 0) {
1361 return;
1362 }
1363 bzero((caddr_t)&info, sizeof(info));
1364 m = rt_msg1(RTM_IFINFO, &info);
1365 if (m == NULL) {
1366 return;
1367 }
1368 ifm = mtod(m, struct if_msghdr *);
1369 ifm->ifm_index = ifp->if_index;
1370 ifm->ifm_flags = (u_short)ifp->if_flags;
1371 if_data_internal_to_if_data(ifp, &ifp->if_data, &ifm->ifm_data);
1372 ifm->ifm_addrs = 0;
1373 raw_input(m, &route_proto, &route_src, &route_dst);
1374 }
1375
1376 /*
1377 * This is called to generate messages from the routing socket
1378 * indicating a network interface has had addresses associated with it.
1379 * if we ever reverse the logic and replace messages TO the routing
1380 * socket indicate a request to configure interfaces, then it will
1381 * be unnecessary as the routing socket will automatically generate
1382 * copies of it.
1383 *
1384 * Since this is coming from the interface, it is expected that the
1385 * interface will be locked. Caller must hold rnh_lock and rt_lock.
1386 */
1387 void
1388 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1389 {
1390 struct rt_addrinfo info;
1391 struct sockaddr *sa = 0;
1392 int pass;
1393 struct mbuf *m = 0;
1394 struct ifnet *ifp = ifa->ifa_ifp;
1395 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1396
1397 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_OWNED);
1398 RT_LOCK_ASSERT_HELD(rt);
1399
1400 if (route_cb.any_count == 0) {
1401 return;
1402 }
1403
1404 /* Become a regular mutex, just in case */
1405 RT_CONVERT_LOCK(rt);
1406 for (pass = 1; pass < 3; pass++) {
1407 bzero((caddr_t)&info, sizeof(info));
1408 if ((cmd == RTM_ADD && pass == 1) ||
1409 (cmd == RTM_DELETE && pass == 2)) {
1410 struct ifa_msghdr *ifam;
1411 int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1412
1413 /* Lock ifp for if_lladdr */
1414 ifnet_lock_shared(ifp);
1415 IFA_LOCK(ifa);
1416 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1417 /*
1418 * Holding ifnet lock here prevents the link address
1419 * from changing contents, so no need to hold its
1420 * lock. The link address is always present; it's
1421 * never freed.
1422 */
1423 info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr;
1424 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1425 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1426 if ((m = rt_msg1(ncmd, &info)) == NULL) {
1427 IFA_UNLOCK(ifa);
1428 ifnet_lock_done(ifp);
1429 continue;
1430 }
1431 IFA_UNLOCK(ifa);
1432 ifnet_lock_done(ifp);
1433 ifam = mtod(m, struct ifa_msghdr *);
1434 ifam->ifam_index = ifp->if_index;
1435 IFA_LOCK_SPIN(ifa);
1436 ifam->ifam_metric = ifa->ifa_metric;
1437 ifam->ifam_flags = ifa->ifa_flags;
1438 IFA_UNLOCK(ifa);
1439 ifam->ifam_addrs = info.rti_addrs;
1440 }
1441 if ((cmd == RTM_ADD && pass == 2) ||
1442 (cmd == RTM_DELETE && pass == 1)) {
1443 struct rt_msghdr *rtm;
1444
1445 if (rt == NULL) {
1446 continue;
1447 }
1448 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1449 info.rti_info[RTAX_DST] = sa = rt_key(rt);
1450 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1451 if ((m = rt_msg1(cmd, &info)) == NULL) {
1452 continue;
1453 }
1454 rtm = mtod(m, struct rt_msghdr *);
1455 rtm->rtm_index = ifp->if_index;
1456 rtm->rtm_flags |= rt->rt_flags;
1457 rtm->rtm_errno = error;
1458 rtm->rtm_addrs = info.rti_addrs;
1459 }
1460 route_proto.sp_protocol = sa ? sa->sa_family : 0;
1461 raw_input(m, &route_proto, &route_src, &route_dst);
1462 }
1463 }
1464
1465 /*
1466 * This is the analogue to the rt_newaddrmsg which performs the same
1467 * function but for multicast group memberhips. This is easier since
1468 * there is no route state to worry about.
1469 */
1470 void
1471 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1472 {
1473 struct rt_addrinfo info;
1474 struct mbuf *m = 0;
1475 struct ifnet *ifp = ifma->ifma_ifp;
1476 struct ifma_msghdr *ifmam;
1477 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1478
1479 if (route_cb.any_count == 0) {
1480 return;
1481 }
1482
1483 /* Lock ifp for if_lladdr */
1484 ifnet_lock_shared(ifp);
1485 bzero((caddr_t)&info, sizeof(info));
1486 IFMA_LOCK(ifma);
1487 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1488 /* lladdr doesn't need lock */
1489 info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr;
1490
1491 /*
1492 * If a link-layer address is present, present it as a ``gateway''
1493 * (similarly to how ARP entries, e.g., are presented).
1494 */
1495 info.rti_info[RTAX_GATEWAY] = (ifma->ifma_ll != NULL) ?
1496 ifma->ifma_ll->ifma_addr : NULL;
1497 if ((m = rt_msg1(cmd, &info)) == NULL) {
1498 IFMA_UNLOCK(ifma);
1499 ifnet_lock_done(ifp);
1500 return;
1501 }
1502 ifmam = mtod(m, struct ifma_msghdr *);
1503 ifmam->ifmam_index = ifp->if_index;
1504 ifmam->ifmam_addrs = info.rti_addrs;
1505 route_proto.sp_protocol = ifma->ifma_addr->sa_family;
1506 IFMA_UNLOCK(ifma);
1507 ifnet_lock_done(ifp);
1508 raw_input(m, &route_proto, &route_src, &route_dst);
1509 }
1510
1511 const char *
1512 rtm2str(int cmd)
1513 {
1514 const char *c = "RTM_?";
1515
1516 switch (cmd) {
1517 case RTM_ADD:
1518 c = "RTM_ADD";
1519 break;
1520 case RTM_DELETE:
1521 c = "RTM_DELETE";
1522 break;
1523 case RTM_CHANGE:
1524 c = "RTM_CHANGE";
1525 break;
1526 case RTM_GET:
1527 c = "RTM_GET";
1528 break;
1529 case RTM_LOSING:
1530 c = "RTM_LOSING";
1531 break;
1532 case RTM_REDIRECT:
1533 c = "RTM_REDIRECT";
1534 break;
1535 case RTM_MISS:
1536 c = "RTM_MISS";
1537 break;
1538 case RTM_LOCK:
1539 c = "RTM_LOCK";
1540 break;
1541 case RTM_OLDADD:
1542 c = "RTM_OLDADD";
1543 break;
1544 case RTM_OLDDEL:
1545 c = "RTM_OLDDEL";
1546 break;
1547 case RTM_RESOLVE:
1548 c = "RTM_RESOLVE";
1549 break;
1550 case RTM_NEWADDR:
1551 c = "RTM_NEWADDR";
1552 break;
1553 case RTM_DELADDR:
1554 c = "RTM_DELADDR";
1555 break;
1556 case RTM_IFINFO:
1557 c = "RTM_IFINFO";
1558 break;
1559 case RTM_NEWMADDR:
1560 c = "RTM_NEWMADDR";
1561 break;
1562 case RTM_DELMADDR:
1563 c = "RTM_DELMADDR";
1564 break;
1565 case RTM_GET_SILENT:
1566 c = "RTM_GET_SILENT";
1567 break;
1568 case RTM_IFINFO2:
1569 c = "RTM_IFINFO2";
1570 break;
1571 case RTM_NEWMADDR2:
1572 c = "RTM_NEWMADDR2";
1573 break;
1574 case RTM_GET2:
1575 c = "RTM_GET2";
1576 break;
1577 case RTM_GET_EXT:
1578 c = "RTM_GET_EXT";
1579 break;
1580 }
1581
1582 return c;
1583 }
1584
1585 /*
1586 * This is used in dumping the kernel table via sysctl().
1587 */
1588 static int
1589 sysctl_dumpentry(struct radix_node *rn, void *vw)
1590 {
1591 struct walkarg *w = vw;
1592 struct rtentry *rt = (struct rtentry *)rn;
1593 int error = 0, size;
1594 struct rt_addrinfo info;
1595 kauth_cred_t cred;
1596 kauth_cred_t *credp;
1597
1598 cred = kauth_cred_proc_ref(current_proc());
1599 credp = &cred;
1600
1601 RT_LOCK(rt);
1602 if ((w->w_op == NET_RT_FLAGS || w->w_op == NET_RT_FLAGS_PRIV) &&
1603 !(rt->rt_flags & w->w_arg)) {
1604 goto done;
1605 }
1606
1607 /*
1608 * If the matching route has RTF_LLINFO set, then we can skip scrubbing the MAC
1609 * only if the outgoing interface is not loopback and the process has entitlement
1610 * for neighbor cache read.
1611 */
1612 if (w->w_op == NET_RT_FLAGS_PRIV && (rt->rt_flags & RTF_LLINFO)) {
1613 if (rt->rt_ifp != lo_ifp &&
1614 (route_op_entitlement_check(NULL, cred, ROUTE_OP_READ, TRUE) == 0)) {
1615 credp = NULL;
1616 }
1617 }
1618
1619 bzero((caddr_t)&info, sizeof(info));
1620 info.rti_info[RTAX_DST] = rt_key(rt);
1621 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1622 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1623 info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1624 if (RT_HAS_IFADDR(rt)) {
1625 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1626 }
1627
1628 if (w->w_op != NET_RT_DUMP2) {
1629 size = rt_msg2(RTM_GET, &info, NULL, w, credp);
1630 if (w->w_req != NULL && w->w_tmem != NULL) {
1631 struct rt_msghdr *rtm =
1632 (struct rt_msghdr *)(void *)w->w_tmem;
1633
1634 rtm->rtm_flags = rt->rt_flags;
1635 rtm->rtm_use = rt->rt_use;
1636 rt_getmetrics(rt, &rtm->rtm_rmx);
1637 rtm->rtm_index = rt->rt_ifp->if_index;
1638 rtm->rtm_pid = 0;
1639 rtm->rtm_seq = 0;
1640 rtm->rtm_errno = 0;
1641 rtm->rtm_addrs = info.rti_addrs;
1642 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1643 }
1644 } else {
1645 size = rt_msg2(RTM_GET2, &info, NULL, w, credp);
1646 if (w->w_req != NULL && w->w_tmem != NULL) {
1647 struct rt_msghdr2 *rtm =
1648 (struct rt_msghdr2 *)(void *)w->w_tmem;
1649
1650 rtm->rtm_flags = rt->rt_flags;
1651 rtm->rtm_use = rt->rt_use;
1652 rt_getmetrics(rt, &rtm->rtm_rmx);
1653 rtm->rtm_index = rt->rt_ifp->if_index;
1654 rtm->rtm_refcnt = rt->rt_refcnt;
1655 if (rt->rt_parent) {
1656 rtm->rtm_parentflags = rt->rt_parent->rt_flags;
1657 } else {
1658 rtm->rtm_parentflags = 0;
1659 }
1660 rtm->rtm_reserved = 0;
1661 rtm->rtm_addrs = info.rti_addrs;
1662 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1663 }
1664 }
1665
1666 done:
1667 RT_UNLOCK(rt);
1668 kauth_cred_unref(&cred);
1669 return error;
1670 }
1671
1672 /*
1673 * This is used for dumping extended information from route entries.
1674 */
1675 static int
1676 sysctl_dumpentry_ext(struct radix_node *rn, void *vw)
1677 {
1678 struct walkarg *w = vw;
1679 struct rtentry *rt = (struct rtentry *)rn;
1680 int error = 0, size;
1681 struct rt_addrinfo info;
1682 kauth_cred_t cred;
1683
1684 cred = kauth_cred_proc_ref(current_proc());
1685
1686 RT_LOCK(rt);
1687 if (w->w_op == NET_RT_DUMPX_FLAGS && !(rt->rt_flags & w->w_arg)) {
1688 goto done;
1689 }
1690 bzero(&info, sizeof(info));
1691 info.rti_info[RTAX_DST] = rt_key(rt);
1692 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1693 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1694 info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1695
1696 size = rt_msg2(RTM_GET_EXT, &info, NULL, w, &cred);
1697 if (w->w_req != NULL && w->w_tmem != NULL) {
1698 struct rt_msghdr_ext *ertm =
1699 (struct rt_msghdr_ext *)(void *)w->w_tmem;
1700
1701 ertm->rtm_flags = rt->rt_flags;
1702 ertm->rtm_use = rt->rt_use;
1703 rt_getmetrics(rt, &ertm->rtm_rmx);
1704 ertm->rtm_index = rt->rt_ifp->if_index;
1705 ertm->rtm_pid = 0;
1706 ertm->rtm_seq = 0;
1707 ertm->rtm_errno = 0;
1708 ertm->rtm_addrs = info.rti_addrs;
1709 if (rt->rt_llinfo_get_ri == NULL) {
1710 bzero(&ertm->rtm_ri, sizeof(ertm->rtm_ri));
1711 ertm->rtm_ri.ri_rssi = IFNET_RSSI_UNKNOWN;
1712 ertm->rtm_ri.ri_lqm = IFNET_LQM_THRESH_OFF;
1713 ertm->rtm_ri.ri_npm = IFNET_NPM_THRESH_UNKNOWN;
1714 } else {
1715 rt->rt_llinfo_get_ri(rt, &ertm->rtm_ri);
1716 }
1717 error = SYSCTL_OUT(w->w_req, (caddr_t)ertm, size);
1718 }
1719
1720 done:
1721 RT_UNLOCK(rt);
1722 kauth_cred_unref(&cred);
1723 return error;
1724 }
1725
1726 /*
1727 * rdar://9307819
1728 * To avoid to call copyout() while holding locks and to cause problems
1729 * in the paging path, sysctl_iflist() and sysctl_iflist2() contstruct
1730 * the list in two passes. In the first pass we compute the total
1731 * length of the data we are going to copyout, then we release
1732 * all locks to allocate a temporary buffer that gets filled
1733 * in the second pass.
1734 *
1735 * Note that we are verifying the assumption that _MALLOC returns a buffer
1736 * that is at least 32 bits aligned and that the messages and addresses are
1737 * 32 bits aligned.
1738 */
1739 static int
1740 sysctl_iflist(int af, struct walkarg *w)
1741 {
1742 struct ifnet *ifp;
1743 struct ifaddr *ifa;
1744 struct rt_addrinfo info;
1745 int len = 0, error = 0;
1746 int pass = 0;
1747 int total_len = 0, current_len = 0;
1748 char *total_buffer = NULL, *cp = NULL;
1749 kauth_cred_t cred;
1750
1751 cred = kauth_cred_proc_ref(current_proc());
1752
1753 bzero((caddr_t)&info, sizeof(info));
1754
1755 for (pass = 0; pass < 2; pass++) {
1756 ifnet_head_lock_shared();
1757
1758 TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1759 if (error) {
1760 break;
1761 }
1762 if (w->w_arg && w->w_arg != ifp->if_index) {
1763 continue;
1764 }
1765 ifnet_lock_shared(ifp);
1766 /*
1767 * Holding ifnet lock here prevents the link address
1768 * from changing contents, so no need to hold the ifa
1769 * lock. The link address is always present; it's
1770 * never freed.
1771 */
1772 ifa = ifp->if_lladdr;
1773 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1774 len = rt_msg2(RTM_IFINFO, &info, NULL, NULL, &cred);
1775 if (pass == 0) {
1776 total_len += len;
1777 } else {
1778 struct if_msghdr *ifm;
1779
1780 if (current_len + len > total_len) {
1781 ifnet_lock_done(ifp);
1782 error = ENOBUFS;
1783 break;
1784 }
1785 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1786 len = rt_msg2(RTM_IFINFO, &info,
1787 (caddr_t)cp, NULL, &cred);
1788 info.rti_info[RTAX_IFP] = NULL;
1789
1790 ifm = (struct if_msghdr *)(void *)cp;
1791 ifm->ifm_index = ifp->if_index;
1792 ifm->ifm_flags = (u_short)ifp->if_flags;
1793 if_data_internal_to_if_data(ifp, &ifp->if_data,
1794 &ifm->ifm_data);
1795 ifm->ifm_addrs = info.rti_addrs;
1796 /*
1797 * <rdar://problem/32940901>
1798 * Round bytes only for non-platform
1799 */
1800 if (!csproc_get_platform_binary(w->w_req->p)) {
1801 ALIGN_BYTES(ifm->ifm_data.ifi_ibytes);
1802 ALIGN_BYTES(ifm->ifm_data.ifi_obytes);
1803 }
1804
1805 cp += len;
1806 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1807 current_len += len;
1808 }
1809 while ((ifa = ifa->ifa_link.tqe_next) != NULL) {
1810 IFA_LOCK(ifa);
1811 if (af && af != ifa->ifa_addr->sa_family) {
1812 IFA_UNLOCK(ifa);
1813 continue;
1814 }
1815 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1816 (((struct in6_ifaddr *)ifa)->ia6_flags &
1817 IN6_IFF_CLAT46) != 0) {
1818 IFA_UNLOCK(ifa);
1819 continue;
1820 }
1821 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1822 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1823 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1824 len = rt_msg2(RTM_NEWADDR, &info, NULL, NULL,
1825 &cred);
1826 if (pass == 0) {
1827 total_len += len;
1828 } else {
1829 struct ifa_msghdr *ifam;
1830
1831 if (current_len + len > total_len) {
1832 IFA_UNLOCK(ifa);
1833 error = ENOBUFS;
1834 break;
1835 }
1836 len = rt_msg2(RTM_NEWADDR, &info,
1837 (caddr_t)cp, NULL, &cred);
1838
1839 ifam = (struct ifa_msghdr *)(void *)cp;
1840 ifam->ifam_index =
1841 ifa->ifa_ifp->if_index;
1842 ifam->ifam_flags = ifa->ifa_flags;
1843 ifam->ifam_metric = ifa->ifa_metric;
1844 ifam->ifam_addrs = info.rti_addrs;
1845
1846 cp += len;
1847 VERIFY(IS_P2ALIGNED(cp,
1848 sizeof(u_int32_t)));
1849 current_len += len;
1850 }
1851 IFA_UNLOCK(ifa);
1852 }
1853 ifnet_lock_done(ifp);
1854 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1855 info.rti_info[RTAX_BRD] = NULL;
1856 }
1857
1858 ifnet_head_done();
1859
1860 if (error != 0) {
1861 if (error == ENOBUFS) {
1862 printf("%s: current_len (%d) + len (%d) > "
1863 "total_len (%d)\n", __func__, current_len,
1864 len, total_len);
1865 }
1866 break;
1867 }
1868
1869 if (pass == 0) {
1870 /* Better to return zero length buffer than ENOBUFS */
1871 if (total_len == 0) {
1872 total_len = 1;
1873 }
1874 total_len += total_len >> 3;
1875 total_buffer = _MALLOC(total_len, M_RTABLE,
1876 M_ZERO | M_WAITOK);
1877 if (total_buffer == NULL) {
1878 printf("%s: _MALLOC(%d) failed\n", __func__,
1879 total_len);
1880 error = ENOBUFS;
1881 break;
1882 }
1883 cp = total_buffer;
1884 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1885 } else {
1886 error = SYSCTL_OUT(w->w_req, total_buffer, current_len);
1887 if (error) {
1888 break;
1889 }
1890 }
1891 }
1892
1893 if (total_buffer != NULL) {
1894 _FREE(total_buffer, M_RTABLE);
1895 }
1896
1897 kauth_cred_unref(&cred);
1898 return error;
1899 }
1900
1901 static int
1902 sysctl_iflist2(int af, struct walkarg *w)
1903 {
1904 struct ifnet *ifp;
1905 struct ifaddr *ifa;
1906 struct rt_addrinfo info;
1907 int len = 0, error = 0;
1908 int pass = 0;
1909 int total_len = 0, current_len = 0;
1910 char *total_buffer = NULL, *cp = NULL;
1911 kauth_cred_t cred;
1912
1913 cred = kauth_cred_proc_ref(current_proc());
1914
1915 bzero((caddr_t)&info, sizeof(info));
1916
1917 for (pass = 0; pass < 2; pass++) {
1918 struct ifmultiaddr *ifma;
1919
1920 ifnet_head_lock_shared();
1921
1922 TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1923 if (error) {
1924 break;
1925 }
1926 if (w->w_arg && w->w_arg != ifp->if_index) {
1927 continue;
1928 }
1929 ifnet_lock_shared(ifp);
1930 /*
1931 * Holding ifnet lock here prevents the link address
1932 * from changing contents, so no need to hold the ifa
1933 * lock. The link address is always present; it's
1934 * never freed.
1935 */
1936 ifa = ifp->if_lladdr;
1937 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1938 len = rt_msg2(RTM_IFINFO2, &info, NULL, NULL, &cred);
1939 if (pass == 0) {
1940 total_len += len;
1941 } else {
1942 struct if_msghdr2 *ifm;
1943
1944 if (current_len + len > total_len) {
1945 ifnet_lock_done(ifp);
1946 error = ENOBUFS;
1947 break;
1948 }
1949 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1950 len = rt_msg2(RTM_IFINFO2, &info,
1951 (caddr_t)cp, NULL, &cred);
1952 info.rti_info[RTAX_IFP] = NULL;
1953
1954 ifm = (struct if_msghdr2 *)(void *)cp;
1955 ifm->ifm_addrs = info.rti_addrs;
1956 ifm->ifm_flags = (u_short)ifp->if_flags;
1957 ifm->ifm_index = ifp->if_index;
1958 ifm->ifm_snd_len = IFCQ_LEN(&ifp->if_snd);
1959 ifm->ifm_snd_maxlen = IFCQ_MAXLEN(&ifp->if_snd);
1960 ifm->ifm_snd_drops =
1961 ifp->if_snd.ifcq_dropcnt.packets;
1962 ifm->ifm_timer = ifp->if_timer;
1963 if_data_internal_to_if_data64(ifp,
1964 &ifp->if_data, &ifm->ifm_data);
1965 /*
1966 * <rdar://problem/32940901>
1967 * Round bytes only for non-platform
1968 */
1969 if (!csproc_get_platform_binary(w->w_req->p)) {
1970 ALIGN_BYTES(ifm->ifm_data.ifi_ibytes);
1971 ALIGN_BYTES(ifm->ifm_data.ifi_obytes);
1972 }
1973
1974 cp += len;
1975 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1976 current_len += len;
1977 }
1978 while ((ifa = ifa->ifa_link.tqe_next) != NULL) {
1979 IFA_LOCK(ifa);
1980 if (af && af != ifa->ifa_addr->sa_family) {
1981 IFA_UNLOCK(ifa);
1982 continue;
1983 }
1984 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1985 (((struct in6_ifaddr *)ifa)->ia6_flags &
1986 IN6_IFF_CLAT46) != 0) {
1987 IFA_UNLOCK(ifa);
1988 continue;
1989 }
1990
1991 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1992 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1993 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1994 len = rt_msg2(RTM_NEWADDR, &info, NULL, NULL,
1995 &cred);
1996 if (pass == 0) {
1997 total_len += len;
1998 } else {
1999 struct ifa_msghdr *ifam;
2000
2001 if (current_len + len > total_len) {
2002 IFA_UNLOCK(ifa);
2003 error = ENOBUFS;
2004 break;
2005 }
2006 len = rt_msg2(RTM_NEWADDR, &info,
2007 (caddr_t)cp, NULL, &cred);
2008
2009 ifam = (struct ifa_msghdr *)(void *)cp;
2010 ifam->ifam_index =
2011 ifa->ifa_ifp->if_index;
2012 ifam->ifam_flags = ifa->ifa_flags;
2013 ifam->ifam_metric = ifa->ifa_metric;
2014 ifam->ifam_addrs = info.rti_addrs;
2015
2016 cp += len;
2017 VERIFY(IS_P2ALIGNED(cp,
2018 sizeof(u_int32_t)));
2019 current_len += len;
2020 }
2021 IFA_UNLOCK(ifa);
2022 }
2023 if (error) {
2024 ifnet_lock_done(ifp);
2025 break;
2026 }
2027
2028 for (ifma = LIST_FIRST(&ifp->if_multiaddrs);
2029 ifma != NULL; ifma = LIST_NEXT(ifma, ifma_link)) {
2030 struct ifaddr *ifa0;
2031
2032 IFMA_LOCK(ifma);
2033 if (af && af != ifma->ifma_addr->sa_family) {
2034 IFMA_UNLOCK(ifma);
2035 continue;
2036 }
2037 bzero((caddr_t)&info, sizeof(info));
2038 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2039 /*
2040 * Holding ifnet lock here prevents the link
2041 * address from changing contents, so no need
2042 * to hold the ifa0 lock. The link address is
2043 * always present; it's never freed.
2044 */
2045 ifa0 = ifp->if_lladdr;
2046 info.rti_info[RTAX_IFP] = ifa0->ifa_addr;
2047 if (ifma->ifma_ll != NULL) {
2048 info.rti_info[RTAX_GATEWAY] =
2049 ifma->ifma_ll->ifma_addr;
2050 }
2051 len = rt_msg2(RTM_NEWMADDR2, &info, NULL, NULL,
2052 &cred);
2053 if (pass == 0) {
2054 total_len += len;
2055 } else {
2056 struct ifma_msghdr2 *ifmam;
2057
2058 if (current_len + len > total_len) {
2059 IFMA_UNLOCK(ifma);
2060 error = ENOBUFS;
2061 break;
2062 }
2063 len = rt_msg2(RTM_NEWMADDR2, &info,
2064 (caddr_t)cp, NULL, &cred);
2065
2066 ifmam =
2067 (struct ifma_msghdr2 *)(void *)cp;
2068 ifmam->ifmam_addrs = info.rti_addrs;
2069 ifmam->ifmam_flags = 0;
2070 ifmam->ifmam_index =
2071 ifma->ifma_ifp->if_index;
2072 ifmam->ifmam_refcount =
2073 ifma->ifma_reqcnt;
2074
2075 cp += len;
2076 VERIFY(IS_P2ALIGNED(cp,
2077 sizeof(u_int32_t)));
2078 current_len += len;
2079 }
2080 IFMA_UNLOCK(ifma);
2081 }
2082 ifnet_lock_done(ifp);
2083 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
2084 info.rti_info[RTAX_BRD] = NULL;
2085 }
2086 ifnet_head_done();
2087
2088 if (error) {
2089 if (error == ENOBUFS) {
2090 printf("%s: current_len (%d) + len (%d) > "
2091 "total_len (%d)\n", __func__, current_len,
2092 len, total_len);
2093 }
2094 break;
2095 }
2096
2097 if (pass == 0) {
2098 /* Better to return zero length buffer than ENOBUFS */
2099 if (total_len == 0) {
2100 total_len = 1;
2101 }
2102 total_len += total_len >> 3;
2103 total_buffer = _MALLOC(total_len, M_RTABLE,
2104 M_ZERO | M_WAITOK);
2105 if (total_buffer == NULL) {
2106 printf("%s: _MALLOC(%d) failed\n", __func__,
2107 total_len);
2108 error = ENOBUFS;
2109 break;
2110 }
2111 cp = total_buffer;
2112 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
2113 } else {
2114 error = SYSCTL_OUT(w->w_req, total_buffer, current_len);
2115 if (error) {
2116 break;
2117 }
2118 }
2119 }
2120
2121 if (total_buffer != NULL) {
2122 _FREE(total_buffer, M_RTABLE);
2123 }
2124
2125 kauth_cred_unref(&cred);
2126 return error;
2127 }
2128
2129
2130 static int
2131 sysctl_rtstat(struct sysctl_req *req)
2132 {
2133 return SYSCTL_OUT(req, &rtstat, sizeof(struct rtstat));
2134 }
2135
2136 static int
2137 sysctl_rttrash(struct sysctl_req *req)
2138 {
2139 return SYSCTL_OUT(req, &rttrash, sizeof(rttrash));
2140 }
2141
2142 static int
2143 sysctl_rtsock SYSCTL_HANDLER_ARGS
2144 {
2145 #pragma unused(oidp)
2146 int *name = (int *)arg1;
2147 u_int namelen = arg2;
2148 struct radix_node_head *rnh;
2149 int i, error = EINVAL;
2150 u_char af;
2151 struct walkarg w;
2152
2153 name++;
2154 namelen--;
2155 if (req->newptr) {
2156 return EPERM;
2157 }
2158 if (namelen != 3) {
2159 return EINVAL;
2160 }
2161 af = name[0];
2162 Bzero(&w, sizeof(w));
2163 w.w_op = name[1];
2164 w.w_arg = name[2];
2165 w.w_req = req;
2166
2167 switch (w.w_op) {
2168 case NET_RT_DUMP:
2169 case NET_RT_DUMP2:
2170 case NET_RT_FLAGS:
2171 case NET_RT_FLAGS_PRIV:
2172 lck_mtx_lock(rnh_lock);
2173 for (i = 1; i <= AF_MAX; i++) {
2174 if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
2175 (error = rnh->rnh_walktree(rnh,
2176 sysctl_dumpentry, &w))) {
2177 break;
2178 }
2179 }
2180 lck_mtx_unlock(rnh_lock);
2181 break;
2182 case NET_RT_DUMPX:
2183 case NET_RT_DUMPX_FLAGS:
2184 lck_mtx_lock(rnh_lock);
2185 for (i = 1; i <= AF_MAX; i++) {
2186 if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
2187 (error = rnh->rnh_walktree(rnh,
2188 sysctl_dumpentry_ext, &w))) {
2189 break;
2190 }
2191 }
2192 lck_mtx_unlock(rnh_lock);
2193 break;
2194 case NET_RT_IFLIST:
2195 error = sysctl_iflist(af, &w);
2196 break;
2197 case NET_RT_IFLIST2:
2198 error = sysctl_iflist2(af, &w);
2199 break;
2200 case NET_RT_STAT:
2201 error = sysctl_rtstat(req);
2202 break;
2203 case NET_RT_TRASH:
2204 error = sysctl_rttrash(req);
2205 break;
2206 }
2207 if (w.w_tmem != NULL) {
2208 FREE(w.w_tmem, M_RTABLE);
2209 }
2210 return error;
2211 }
2212
2213 /*
2214 * Definitions of protocols supported in the ROUTE domain.
2215 */
2216 static struct protosw routesw[] = {
2217 {
2218 .pr_type = SOCK_RAW,
2219 .pr_protocol = 0,
2220 .pr_flags = PR_ATOMIC | PR_ADDR,
2221 .pr_output = route_output,
2222 .pr_ctlinput = raw_ctlinput,
2223 .pr_init = raw_init,
2224 .pr_usrreqs = &route_usrreqs,
2225 }
2226 };
2227
2228 static int route_proto_count = (sizeof(routesw) / sizeof(struct protosw));
2229
2230 struct domain routedomain_s = {
2231 .dom_family = PF_ROUTE,
2232 .dom_name = "route",
2233 .dom_init = route_dinit,
2234 };
2235
2236 static void
2237 route_dinit(struct domain *dp)
2238 {
2239 struct protosw *pr;
2240 int i;
2241
2242 VERIFY(!(dp->dom_flags & DOM_INITIALIZED));
2243 VERIFY(routedomain == NULL);
2244
2245 routedomain = dp;
2246
2247 for (i = 0, pr = &routesw[0]; i < route_proto_count; i++, pr++) {
2248 net_add_proto(pr, dp, 1);
2249 }
2250
2251 route_init();
2252 }
mirror of XNU