]> git.saurik.com Git - apple/xnu.git/blob - bsd/net/rtsock.c
projects / apple / xnu.git / blob
? search:


a8b2601593c7397ac9d3339cecad3917be078234
[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 /*
1088 * there are no more.. quit now
1089 * If there are more bits, they are in error.
1090 * I've seen this. route(1) can evidently generate these.
1091 * This causes kernel to core dump.
1092 * for compatibility, If we see this, point to a safe address.
1093 */
1094 if (sa->sa_len == 0) {
1095 rtinfo->rti_info[i] = &sa_zero;
1096 return 0; /* should be EINVAL but for compat */
1097 }
1098 /* accept it */
1099 rtinfo->rti_info[i] = sa;
1100 ADVANCE32(cp, sa);
1101 }
1102 return 0;
1103 }
1104
1105 static struct mbuf *
1106 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1107 {
1108 struct rt_msghdr *rtm;
1109 struct mbuf *m;
1110 int i;
1111 int len, dlen, off;
1112
1113 switch (type) {
1114 case RTM_DELADDR:
1115 case RTM_NEWADDR:
1116 len = sizeof(struct ifa_msghdr);
1117 break;
1118
1119 case RTM_DELMADDR:
1120 case RTM_NEWMADDR:
1121 len = sizeof(struct ifma_msghdr);
1122 break;
1123
1124 case RTM_IFINFO:
1125 len = sizeof(struct if_msghdr);
1126 break;
1127
1128 default:
1129 len = sizeof(struct rt_msghdr);
1130 }
1131 m = m_gethdr(M_DONTWAIT, MT_DATA);
1132 if (m && len > MHLEN) {
1133 MCLGET(m, M_DONTWAIT);
1134 if (!(m->m_flags & M_EXT)) {
1135 m_free(m);
1136 m = NULL;
1137 }
1138 }
1139 if (m == NULL) {
1140 return NULL;
1141 }
1142 m->m_pkthdr.len = m->m_len = len;
1143 m->m_pkthdr.rcvif = NULL;
1144 rtm = mtod(m, struct rt_msghdr *);
1145 bzero((caddr_t)rtm, len);
1146 off = len;
1147 for (i = 0; i < RTAX_MAX; i++) {
1148 struct sockaddr *sa, *hint;
1149 uint8_t ssbuf[SOCK_MAXADDRLEN + 1];
1150
1151 /*
1152 * Make sure to accomodate the largest possible size of sa_len.
1153 */
1154 _CASSERT(sizeof(ssbuf) == (SOCK_MAXADDRLEN + 1));
1155
1156 if ((sa = rtinfo->rti_info[i]) == NULL) {
1157 continue;
1158 }
1159
1160 switch (i) {
1161 case RTAX_DST:
1162 case RTAX_NETMASK:
1163 if ((hint = rtinfo->rti_info[RTAX_DST]) == NULL) {
1164 hint = rtinfo->rti_info[RTAX_IFA];
1165 }
1166
1167 /* Scrub away any trace of embedded interface scope */
1168 sa = rtm_scrub(type, i, hint, sa, &ssbuf,
1169 sizeof(ssbuf), NULL);
1170 break;
1171
1172 default:
1173 break;
1174 }
1175
1176 rtinfo->rti_addrs |= (1 << i);
1177 dlen = sa->sa_len;
1178 m_copyback(m, off, dlen, (caddr_t)sa);
1179 len = off + dlen;
1180 off += ROUNDUP32(dlen);
1181 }
1182 if (m->m_pkthdr.len != len) {
1183 m_freem(m);
1184 return NULL;
1185 }
1186 rtm->rtm_msglen = len;
1187 rtm->rtm_version = RTM_VERSION;
1188 rtm->rtm_type = type;
1189 return m;
1190 }
1191
1192 static int
1193 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w,
1194 kauth_cred_t* credp)
1195 {
1196 int i;
1197 int len, dlen, rlen, second_time = 0;
1198 caddr_t cp0;
1199
1200 rtinfo->rti_addrs = 0;
1201 again:
1202 switch (type) {
1203 case RTM_DELADDR:
1204 case RTM_NEWADDR:
1205 len = sizeof(struct ifa_msghdr);
1206 break;
1207
1208 case RTM_DELMADDR:
1209 case RTM_NEWMADDR:
1210 len = sizeof(struct ifma_msghdr);
1211 break;
1212
1213 case RTM_IFINFO:
1214 len = sizeof(struct if_msghdr);
1215 break;
1216
1217 case RTM_IFINFO2:
1218 len = sizeof(struct if_msghdr2);
1219 break;
1220
1221 case RTM_NEWMADDR2:
1222 len = sizeof(struct ifma_msghdr2);
1223 break;
1224
1225 case RTM_GET_EXT:
1226 len = sizeof(struct rt_msghdr_ext);
1227 break;
1228
1229 case RTM_GET2:
1230 len = sizeof(struct rt_msghdr2);
1231 break;
1232
1233 default:
1234 len = sizeof(struct rt_msghdr);
1235 }
1236 cp0 = cp;
1237 if (cp0) {
1238 cp += len;
1239 }
1240 for (i = 0; i < RTAX_MAX; i++) {
1241 struct sockaddr *sa, *hint;
1242 uint8_t ssbuf[SOCK_MAXADDRLEN + 1];
1243
1244 /*
1245 * Make sure to accomodate the largest possible size of sa_len.
1246 */
1247 _CASSERT(sizeof(ssbuf) == (SOCK_MAXADDRLEN + 1));
1248
1249 if ((sa = rtinfo->rti_info[i]) == NULL) {
1250 continue;
1251 }
1252
1253 switch (i) {
1254 case RTAX_DST:
1255 case RTAX_NETMASK:
1256 if ((hint = rtinfo->rti_info[RTAX_DST]) == NULL) {
1257 hint = rtinfo->rti_info[RTAX_IFA];
1258 }
1259
1260 /* Scrub away any trace of embedded interface scope */
1261 sa = rtm_scrub(type, i, hint, sa, &ssbuf,
1262 sizeof(ssbuf), NULL);
1263 break;
1264 case RTAX_GATEWAY:
1265 case RTAX_IFP:
1266 sa = rtm_scrub(type, i, NULL, sa, &ssbuf,
1267 sizeof(ssbuf), credp);
1268 break;
1269
1270 default:
1271 break;
1272 }
1273
1274 rtinfo->rti_addrs |= (1 << i);
1275 dlen = sa->sa_len;
1276 rlen = ROUNDUP32(dlen);
1277 if (cp) {
1278 bcopy((caddr_t)sa, cp, (size_t)dlen);
1279 if (dlen != rlen) {
1280 bzero(cp + dlen, rlen - dlen);
1281 }
1282 cp += rlen;
1283 }
1284 len += rlen;
1285 }
1286 if (cp == NULL && w != NULL && !second_time) {
1287 struct walkarg *rw = w;
1288
1289 if (rw->w_req != NULL) {
1290 if (rw->w_tmemsize < len) {
1291 if (rw->w_tmem != NULL) {
1292 FREE(rw->w_tmem, M_RTABLE);
1293 }
1294 rw->w_tmem = _MALLOC(len, M_RTABLE, M_ZERO | M_WAITOK);
1295 if (rw->w_tmem != NULL) {
1296 rw->w_tmemsize = len;
1297 }
1298 }
1299 if (rw->w_tmem != NULL) {
1300 cp = rw->w_tmem;
1301 second_time = 1;
1302 goto again;
1303 }
1304 }
1305 }
1306 if (cp) {
1307 struct rt_msghdr *rtm = (struct rt_msghdr *)(void *)cp0;
1308
1309 rtm->rtm_version = RTM_VERSION;
1310 rtm->rtm_type = type;
1311 rtm->rtm_msglen = len;
1312 }
1313 return len;
1314 }
1315
1316 /*
1317 * This routine is called to generate a message from the routing
1318 * socket indicating that a redirect has occurred, a routing lookup
1319 * has failed, or that a protocol has detected timeouts to a particular
1320 * destination.
1321 */
1322 void
1323 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1324 {
1325 struct rt_msghdr *rtm;
1326 struct mbuf *m;
1327 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1328 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1329
1330 if (route_cb.any_count == 0) {
1331 return;
1332 }
1333 m = rt_msg1(type, rtinfo);
1334 if (m == NULL) {
1335 return;
1336 }
1337 rtm = mtod(m, struct rt_msghdr *);
1338 rtm->rtm_flags = RTF_DONE | flags;
1339 rtm->rtm_errno = error;
1340 rtm->rtm_addrs = rtinfo->rti_addrs;
1341 route_proto.sp_family = sa ? sa->sa_family : 0;
1342 raw_input(m, &route_proto, &route_src, &route_dst);
1343 }
1344
1345 /*
1346 * This routine is called to generate a message from the routing
1347 * socket indicating that the status of a network interface has changed.
1348 */
1349 void
1350 rt_ifmsg(struct ifnet *ifp)
1351 {
1352 struct if_msghdr *ifm;
1353 struct mbuf *m;
1354 struct rt_addrinfo info;
1355 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1356
1357 if (route_cb.any_count == 0) {
1358 return;
1359 }
1360 bzero((caddr_t)&info, sizeof(info));
1361 m = rt_msg1(RTM_IFINFO, &info);
1362 if (m == NULL) {
1363 return;
1364 }
1365 ifm = mtod(m, struct if_msghdr *);
1366 ifm->ifm_index = ifp->if_index;
1367 ifm->ifm_flags = (u_short)ifp->if_flags;
1368 if_data_internal_to_if_data(ifp, &ifp->if_data, &ifm->ifm_data);
1369 ifm->ifm_addrs = 0;
1370 raw_input(m, &route_proto, &route_src, &route_dst);
1371 }
1372
1373 /*
1374 * This is called to generate messages from the routing socket
1375 * indicating a network interface has had addresses associated with it.
1376 * if we ever reverse the logic and replace messages TO the routing
1377 * socket indicate a request to configure interfaces, then it will
1378 * be unnecessary as the routing socket will automatically generate
1379 * copies of it.
1380 *
1381 * Since this is coming from the interface, it is expected that the
1382 * interface will be locked. Caller must hold rnh_lock and rt_lock.
1383 */
1384 void
1385 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1386 {
1387 struct rt_addrinfo info;
1388 struct sockaddr *sa = 0;
1389 int pass;
1390 struct mbuf *m = 0;
1391 struct ifnet *ifp = ifa->ifa_ifp;
1392 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1393
1394 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_OWNED);
1395 RT_LOCK_ASSERT_HELD(rt);
1396
1397 if (route_cb.any_count == 0) {
1398 return;
1399 }
1400
1401 /* Become a regular mutex, just in case */
1402 RT_CONVERT_LOCK(rt);
1403 for (pass = 1; pass < 3; pass++) {
1404 bzero((caddr_t)&info, sizeof(info));
1405 if ((cmd == RTM_ADD && pass == 1) ||
1406 (cmd == RTM_DELETE && pass == 2)) {
1407 struct ifa_msghdr *ifam;
1408 int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1409
1410 /* Lock ifp for if_lladdr */
1411 ifnet_lock_shared(ifp);
1412 IFA_LOCK(ifa);
1413 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1414 /*
1415 * Holding ifnet lock here prevents the link address
1416 * from changing contents, so no need to hold its
1417 * lock. The link address is always present; it's
1418 * never freed.
1419 */
1420 info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr;
1421 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1422 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1423 if ((m = rt_msg1(ncmd, &info)) == NULL) {
1424 IFA_UNLOCK(ifa);
1425 ifnet_lock_done(ifp);
1426 continue;
1427 }
1428 IFA_UNLOCK(ifa);
1429 ifnet_lock_done(ifp);
1430 ifam = mtod(m, struct ifa_msghdr *);
1431 ifam->ifam_index = ifp->if_index;
1432 IFA_LOCK_SPIN(ifa);
1433 ifam->ifam_metric = ifa->ifa_metric;
1434 ifam->ifam_flags = ifa->ifa_flags;
1435 IFA_UNLOCK(ifa);
1436 ifam->ifam_addrs = info.rti_addrs;
1437 }
1438 if ((cmd == RTM_ADD && pass == 2) ||
1439 (cmd == RTM_DELETE && pass == 1)) {
1440 struct rt_msghdr *rtm;
1441
1442 if (rt == NULL) {
1443 continue;
1444 }
1445 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1446 info.rti_info[RTAX_DST] = sa = rt_key(rt);
1447 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1448 if ((m = rt_msg1(cmd, &info)) == NULL) {
1449 continue;
1450 }
1451 rtm = mtod(m, struct rt_msghdr *);
1452 rtm->rtm_index = ifp->if_index;
1453 rtm->rtm_flags |= rt->rt_flags;
1454 rtm->rtm_errno = error;
1455 rtm->rtm_addrs = info.rti_addrs;
1456 }
1457 route_proto.sp_protocol = sa ? sa->sa_family : 0;
1458 raw_input(m, &route_proto, &route_src, &route_dst);
1459 }
1460 }
1461
1462 /*
1463 * This is the analogue to the rt_newaddrmsg which performs the same
1464 * function but for multicast group memberhips. This is easier since
1465 * there is no route state to worry about.
1466 */
1467 void
1468 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1469 {
1470 struct rt_addrinfo info;
1471 struct mbuf *m = 0;
1472 struct ifnet *ifp = ifma->ifma_ifp;
1473 struct ifma_msghdr *ifmam;
1474 struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1475
1476 if (route_cb.any_count == 0) {
1477 return;
1478 }
1479
1480 /* Lock ifp for if_lladdr */
1481 ifnet_lock_shared(ifp);
1482 bzero((caddr_t)&info, sizeof(info));
1483 IFMA_LOCK(ifma);
1484 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1485 /* lladdr doesn't need lock */
1486 info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr;
1487
1488 /*
1489 * If a link-layer address is present, present it as a ``gateway''
1490 * (similarly to how ARP entries, e.g., are presented).
1491 */
1492 info.rti_info[RTAX_GATEWAY] = (ifma->ifma_ll != NULL) ?
1493 ifma->ifma_ll->ifma_addr : NULL;
1494 if ((m = rt_msg1(cmd, &info)) == NULL) {
1495 IFMA_UNLOCK(ifma);
1496 ifnet_lock_done(ifp);
1497 return;
1498 }
1499 ifmam = mtod(m, struct ifma_msghdr *);
1500 ifmam->ifmam_index = ifp->if_index;
1501 ifmam->ifmam_addrs = info.rti_addrs;
1502 route_proto.sp_protocol = ifma->ifma_addr->sa_family;
1503 IFMA_UNLOCK(ifma);
1504 ifnet_lock_done(ifp);
1505 raw_input(m, &route_proto, &route_src, &route_dst);
1506 }
1507
1508 const char *
1509 rtm2str(int cmd)
1510 {
1511 const char *c = "RTM_?";
1512
1513 switch (cmd) {
1514 case RTM_ADD:
1515 c = "RTM_ADD";
1516 break;
1517 case RTM_DELETE:
1518 c = "RTM_DELETE";
1519 break;
1520 case RTM_CHANGE:
1521 c = "RTM_CHANGE";
1522 break;
1523 case RTM_GET:
1524 c = "RTM_GET";
1525 break;
1526 case RTM_LOSING:
1527 c = "RTM_LOSING";
1528 break;
1529 case RTM_REDIRECT:
1530 c = "RTM_REDIRECT";
1531 break;
1532 case RTM_MISS:
1533 c = "RTM_MISS";
1534 break;
1535 case RTM_LOCK:
1536 c = "RTM_LOCK";
1537 break;
1538 case RTM_OLDADD:
1539 c = "RTM_OLDADD";
1540 break;
1541 case RTM_OLDDEL:
1542 c = "RTM_OLDDEL";
1543 break;
1544 case RTM_RESOLVE:
1545 c = "RTM_RESOLVE";
1546 break;
1547 case RTM_NEWADDR:
1548 c = "RTM_NEWADDR";
1549 break;
1550 case RTM_DELADDR:
1551 c = "RTM_DELADDR";
1552 break;
1553 case RTM_IFINFO:
1554 c = "RTM_IFINFO";
1555 break;
1556 case RTM_NEWMADDR:
1557 c = "RTM_NEWMADDR";
1558 break;
1559 case RTM_DELMADDR:
1560 c = "RTM_DELMADDR";
1561 break;
1562 case RTM_GET_SILENT:
1563 c = "RTM_GET_SILENT";
1564 break;
1565 case RTM_IFINFO2:
1566 c = "RTM_IFINFO2";
1567 break;
1568 case RTM_NEWMADDR2:
1569 c = "RTM_NEWMADDR2";
1570 break;
1571 case RTM_GET2:
1572 c = "RTM_GET2";
1573 break;
1574 case RTM_GET_EXT:
1575 c = "RTM_GET_EXT";
1576 break;
1577 }
1578
1579 return c;
1580 }
1581
1582 /*
1583 * This is used in dumping the kernel table via sysctl().
1584 */
1585 static int
1586 sysctl_dumpentry(struct radix_node *rn, void *vw)
1587 {
1588 struct walkarg *w = vw;
1589 struct rtentry *rt = (struct rtentry *)rn;
1590 int error = 0, size;
1591 struct rt_addrinfo info;
1592 kauth_cred_t cred;
1593 kauth_cred_t *credp;
1594
1595 cred = kauth_cred_proc_ref(current_proc());
1596 credp = &cred;
1597
1598 RT_LOCK(rt);
1599 if ((w->w_op == NET_RT_FLAGS || w->w_op == NET_RT_FLAGS_PRIV) &&
1600 !(rt->rt_flags & w->w_arg)) {
1601 goto done;
1602 }
1603
1604 /*
1605 * If the matching route has RTF_LLINFO set, then we can skip scrubbing the MAC
1606 * only if the outgoing interface is not loopback and the process has entitlement
1607 * for neighbor cache read.
1608 */
1609 if (w->w_op == NET_RT_FLAGS_PRIV && (rt->rt_flags & RTF_LLINFO)) {
1610 if (rt->rt_ifp != lo_ifp &&
1611 (route_op_entitlement_check(NULL, cred, ROUTE_OP_READ, TRUE) == 0)) {
1612 credp = NULL;
1613 }
1614 }
1615
1616 bzero((caddr_t)&info, sizeof(info));
1617 info.rti_info[RTAX_DST] = rt_key(rt);
1618 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1619 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1620 info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1621 if (RT_HAS_IFADDR(rt)) {
1622 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1623 }
1624
1625 if (w->w_op != NET_RT_DUMP2) {
1626 size = rt_msg2(RTM_GET, &info, NULL, w, credp);
1627 if (w->w_req != NULL && w->w_tmem != NULL) {
1628 struct rt_msghdr *rtm =
1629 (struct rt_msghdr *)(void *)w->w_tmem;
1630
1631 rtm->rtm_flags = rt->rt_flags;
1632 rtm->rtm_use = rt->rt_use;
1633 rt_getmetrics(rt, &rtm->rtm_rmx);
1634 rtm->rtm_index = rt->rt_ifp->if_index;
1635 rtm->rtm_pid = 0;
1636 rtm->rtm_seq = 0;
1637 rtm->rtm_errno = 0;
1638 rtm->rtm_addrs = info.rti_addrs;
1639 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1640 }
1641 } else {
1642 size = rt_msg2(RTM_GET2, &info, NULL, w, credp);
1643 if (w->w_req != NULL && w->w_tmem != NULL) {
1644 struct rt_msghdr2 *rtm =
1645 (struct rt_msghdr2 *)(void *)w->w_tmem;
1646
1647 rtm->rtm_flags = rt->rt_flags;
1648 rtm->rtm_use = rt->rt_use;
1649 rt_getmetrics(rt, &rtm->rtm_rmx);
1650 rtm->rtm_index = rt->rt_ifp->if_index;
1651 rtm->rtm_refcnt = rt->rt_refcnt;
1652 if (rt->rt_parent) {
1653 rtm->rtm_parentflags = rt->rt_parent->rt_flags;
1654 } else {
1655 rtm->rtm_parentflags = 0;
1656 }
1657 rtm->rtm_reserved = 0;
1658 rtm->rtm_addrs = info.rti_addrs;
1659 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1660 }
1661 }
1662
1663 done:
1664 RT_UNLOCK(rt);
1665 kauth_cred_unref(&cred);
1666 return error;
1667 }
1668
1669 /*
1670 * This is used for dumping extended information from route entries.
1671 */
1672 static int
1673 sysctl_dumpentry_ext(struct radix_node *rn, void *vw)
1674 {
1675 struct walkarg *w = vw;
1676 struct rtentry *rt = (struct rtentry *)rn;
1677 int error = 0, size;
1678 struct rt_addrinfo info;
1679 kauth_cred_t cred;
1680
1681 cred = kauth_cred_proc_ref(current_proc());
1682
1683 RT_LOCK(rt);
1684 if (w->w_op == NET_RT_DUMPX_FLAGS && !(rt->rt_flags & w->w_arg)) {
1685 goto done;
1686 }
1687 bzero(&info, sizeof(info));
1688 info.rti_info[RTAX_DST] = rt_key(rt);
1689 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1690 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1691 info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1692
1693 size = rt_msg2(RTM_GET_EXT, &info, NULL, w, &cred);
1694 if (w->w_req != NULL && w->w_tmem != NULL) {
1695 struct rt_msghdr_ext *ertm =
1696 (struct rt_msghdr_ext *)(void *)w->w_tmem;
1697
1698 ertm->rtm_flags = rt->rt_flags;
1699 ertm->rtm_use = rt->rt_use;
1700 rt_getmetrics(rt, &ertm->rtm_rmx);
1701 ertm->rtm_index = rt->rt_ifp->if_index;
1702 ertm->rtm_pid = 0;
1703 ertm->rtm_seq = 0;
1704 ertm->rtm_errno = 0;
1705 ertm->rtm_addrs = info.rti_addrs;
1706 if (rt->rt_llinfo_get_ri == NULL) {
1707 bzero(&ertm->rtm_ri, sizeof(ertm->rtm_ri));
1708 ertm->rtm_ri.ri_rssi = IFNET_RSSI_UNKNOWN;
1709 ertm->rtm_ri.ri_lqm = IFNET_LQM_THRESH_OFF;
1710 ertm->rtm_ri.ri_npm = IFNET_NPM_THRESH_UNKNOWN;
1711 } else {
1712 rt->rt_llinfo_get_ri(rt, &ertm->rtm_ri);
1713 }
1714 error = SYSCTL_OUT(w->w_req, (caddr_t)ertm, size);
1715 }
1716
1717 done:
1718 RT_UNLOCK(rt);
1719 kauth_cred_unref(&cred);
1720 return error;
1721 }
1722
1723 /*
1724 * rdar://9307819
1725 * To avoid to call copyout() while holding locks and to cause problems
1726 * in the paging path, sysctl_iflist() and sysctl_iflist2() contstruct
1727 * the list in two passes. In the first pass we compute the total
1728 * length of the data we are going to copyout, then we release
1729 * all locks to allocate a temporary buffer that gets filled
1730 * in the second pass.
1731 *
1732 * Note that we are verifying the assumption that _MALLOC returns a buffer
1733 * that is at least 32 bits aligned and that the messages and addresses are
1734 * 32 bits aligned.
1735 */
1736 static int
1737 sysctl_iflist(int af, struct walkarg *w)
1738 {
1739 struct ifnet *ifp;
1740 struct ifaddr *ifa;
1741 struct rt_addrinfo info;
1742 int len = 0, error = 0;
1743 int pass = 0;
1744 int total_len = 0, current_len = 0;
1745 char *total_buffer = NULL, *cp = NULL;
1746 kauth_cred_t cred;
1747
1748 cred = kauth_cred_proc_ref(current_proc());
1749
1750 bzero((caddr_t)&info, sizeof(info));
1751
1752 for (pass = 0; pass < 2; pass++) {
1753 ifnet_head_lock_shared();
1754
1755 TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1756 if (error) {
1757 break;
1758 }
1759 if (w->w_arg && w->w_arg != ifp->if_index) {
1760 continue;
1761 }
1762 ifnet_lock_shared(ifp);
1763 /*
1764 * Holding ifnet lock here prevents the link address
1765 * from changing contents, so no need to hold the ifa
1766 * lock. The link address is always present; it's
1767 * never freed.
1768 */
1769 ifa = ifp->if_lladdr;
1770 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1771 len = rt_msg2(RTM_IFINFO, &info, NULL, NULL, &cred);
1772 if (pass == 0) {
1773 total_len += len;
1774 } else {
1775 struct if_msghdr *ifm;
1776
1777 if (current_len + len > total_len) {
1778 ifnet_lock_done(ifp);
1779 error = ENOBUFS;
1780 break;
1781 }
1782 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1783 len = rt_msg2(RTM_IFINFO, &info,
1784 (caddr_t)cp, NULL, &cred);
1785 info.rti_info[RTAX_IFP] = NULL;
1786
1787 ifm = (struct if_msghdr *)(void *)cp;
1788 ifm->ifm_index = ifp->if_index;
1789 ifm->ifm_flags = (u_short)ifp->if_flags;
1790 if_data_internal_to_if_data(ifp, &ifp->if_data,
1791 &ifm->ifm_data);
1792 ifm->ifm_addrs = info.rti_addrs;
1793 /*
1794 * <rdar://problem/32940901>
1795 * Round bytes only for non-platform
1796 */
1797 if (!csproc_get_platform_binary(w->w_req->p)) {
1798 ALIGN_BYTES(ifm->ifm_data.ifi_ibytes);
1799 ALIGN_BYTES(ifm->ifm_data.ifi_obytes);
1800 }
1801
1802 cp += len;
1803 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1804 current_len += len;
1805 }
1806 while ((ifa = ifa->ifa_link.tqe_next) != NULL) {
1807 IFA_LOCK(ifa);
1808 if (af && af != ifa->ifa_addr->sa_family) {
1809 IFA_UNLOCK(ifa);
1810 continue;
1811 }
1812 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1813 (((struct in6_ifaddr *)ifa)->ia6_flags &
1814 IN6_IFF_CLAT46) != 0) {
1815 IFA_UNLOCK(ifa);
1816 continue;
1817 }
1818 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1819 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1820 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1821 len = rt_msg2(RTM_NEWADDR, &info, NULL, NULL,
1822 &cred);
1823 if (pass == 0) {
1824 total_len += len;
1825 } else {
1826 struct ifa_msghdr *ifam;
1827
1828 if (current_len + len > total_len) {
1829 IFA_UNLOCK(ifa);
1830 error = ENOBUFS;
1831 break;
1832 }
1833 len = rt_msg2(RTM_NEWADDR, &info,
1834 (caddr_t)cp, NULL, &cred);
1835
1836 ifam = (struct ifa_msghdr *)(void *)cp;
1837 ifam->ifam_index =
1838 ifa->ifa_ifp->if_index;
1839 ifam->ifam_flags = ifa->ifa_flags;
1840 ifam->ifam_metric = ifa->ifa_metric;
1841 ifam->ifam_addrs = info.rti_addrs;
1842
1843 cp += len;
1844 VERIFY(IS_P2ALIGNED(cp,
1845 sizeof(u_int32_t)));
1846 current_len += len;
1847 }
1848 IFA_UNLOCK(ifa);
1849 }
1850 ifnet_lock_done(ifp);
1851 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1852 info.rti_info[RTAX_BRD] = NULL;
1853 }
1854
1855 ifnet_head_done();
1856
1857 if (error != 0) {
1858 if (error == ENOBUFS) {
1859 printf("%s: current_len (%d) + len (%d) > "
1860 "total_len (%d)\n", __func__, current_len,
1861 len, total_len);
1862 }
1863 break;
1864 }
1865
1866 if (pass == 0) {
1867 /* Better to return zero length buffer than ENOBUFS */
1868 if (total_len == 0) {
1869 total_len = 1;
1870 }
1871 total_len += total_len >> 3;
1872 total_buffer = _MALLOC(total_len, M_RTABLE,
1873 M_ZERO | M_WAITOK);
1874 if (total_buffer == NULL) {
1875 printf("%s: _MALLOC(%d) failed\n", __func__,
1876 total_len);
1877 error = ENOBUFS;
1878 break;
1879 }
1880 cp = total_buffer;
1881 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1882 } else {
1883 error = SYSCTL_OUT(w->w_req, total_buffer, current_len);
1884 if (error) {
1885 break;
1886 }
1887 }
1888 }
1889
1890 if (total_buffer != NULL) {
1891 _FREE(total_buffer, M_RTABLE);
1892 }
1893
1894 kauth_cred_unref(&cred);
1895 return error;
1896 }
1897
1898 static int
1899 sysctl_iflist2(int af, struct walkarg *w)
1900 {
1901 struct ifnet *ifp;
1902 struct ifaddr *ifa;
1903 struct rt_addrinfo info;
1904 int len = 0, error = 0;
1905 int pass = 0;
1906 int total_len = 0, current_len = 0;
1907 char *total_buffer = NULL, *cp = NULL;
1908 kauth_cred_t cred;
1909
1910 cred = kauth_cred_proc_ref(current_proc());
1911
1912 bzero((caddr_t)&info, sizeof(info));
1913
1914 for (pass = 0; pass < 2; pass++) {
1915 struct ifmultiaddr *ifma;
1916
1917 ifnet_head_lock_shared();
1918
1919 TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1920 if (error) {
1921 break;
1922 }
1923 if (w->w_arg && w->w_arg != ifp->if_index) {
1924 continue;
1925 }
1926 ifnet_lock_shared(ifp);
1927 /*
1928 * Holding ifnet lock here prevents the link address
1929 * from changing contents, so no need to hold the ifa
1930 * lock. The link address is always present; it's
1931 * never freed.
1932 */
1933 ifa = ifp->if_lladdr;
1934 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1935 len = rt_msg2(RTM_IFINFO2, &info, NULL, NULL, &cred);
1936 if (pass == 0) {
1937 total_len += len;
1938 } else {
1939 struct if_msghdr2 *ifm;
1940
1941 if (current_len + len > total_len) {
1942 ifnet_lock_done(ifp);
1943 error = ENOBUFS;
1944 break;
1945 }
1946 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1947 len = rt_msg2(RTM_IFINFO2, &info,
1948 (caddr_t)cp, NULL, &cred);
1949 info.rti_info[RTAX_IFP] = NULL;
1950
1951 ifm = (struct if_msghdr2 *)(void *)cp;
1952 ifm->ifm_addrs = info.rti_addrs;
1953 ifm->ifm_flags = (u_short)ifp->if_flags;
1954 ifm->ifm_index = ifp->if_index;
1955 ifm->ifm_snd_len = IFCQ_LEN(&ifp->if_snd);
1956 ifm->ifm_snd_maxlen = IFCQ_MAXLEN(&ifp->if_snd);
1957 ifm->ifm_snd_drops =
1958 ifp->if_snd.ifcq_dropcnt.packets;
1959 ifm->ifm_timer = ifp->if_timer;
1960 if_data_internal_to_if_data64(ifp,
1961 &ifp->if_data, &ifm->ifm_data);
1962 /*
1963 * <rdar://problem/32940901>
1964 * Round bytes only for non-platform
1965 */
1966 if (!csproc_get_platform_binary(w->w_req->p)) {
1967 ALIGN_BYTES(ifm->ifm_data.ifi_ibytes);
1968 ALIGN_BYTES(ifm->ifm_data.ifi_obytes);
1969 }
1970
1971 cp += len;
1972 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1973 current_len += len;
1974 }
1975 while ((ifa = ifa->ifa_link.tqe_next) != NULL) {
1976 IFA_LOCK(ifa);
1977 if (af && af != ifa->ifa_addr->sa_family) {
1978 IFA_UNLOCK(ifa);
1979 continue;
1980 }
1981 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1982 (((struct in6_ifaddr *)ifa)->ia6_flags &
1983 IN6_IFF_CLAT46) != 0) {
1984 IFA_UNLOCK(ifa);
1985 continue;
1986 }
1987
1988 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1989 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1990 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1991 len = rt_msg2(RTM_NEWADDR, &info, NULL, NULL,
1992 &cred);
1993 if (pass == 0) {
1994 total_len += len;
1995 } else {
1996 struct ifa_msghdr *ifam;
1997
1998 if (current_len + len > total_len) {
1999 IFA_UNLOCK(ifa);
2000 error = ENOBUFS;
2001 break;
2002 }
2003 len = rt_msg2(RTM_NEWADDR, &info,
2004 (caddr_t)cp, NULL, &cred);
2005
2006 ifam = (struct ifa_msghdr *)(void *)cp;
2007 ifam->ifam_index =
2008 ifa->ifa_ifp->if_index;
2009 ifam->ifam_flags = ifa->ifa_flags;
2010 ifam->ifam_metric = ifa->ifa_metric;
2011 ifam->ifam_addrs = info.rti_addrs;
2012
2013 cp += len;
2014 VERIFY(IS_P2ALIGNED(cp,
2015 sizeof(u_int32_t)));
2016 current_len += len;
2017 }
2018 IFA_UNLOCK(ifa);
2019 }
2020 if (error) {
2021 ifnet_lock_done(ifp);
2022 break;
2023 }
2024
2025 for (ifma = LIST_FIRST(&ifp->if_multiaddrs);
2026 ifma != NULL; ifma = LIST_NEXT(ifma, ifma_link)) {
2027 struct ifaddr *ifa0;
2028
2029 IFMA_LOCK(ifma);
2030 if (af && af != ifma->ifma_addr->sa_family) {
2031 IFMA_UNLOCK(ifma);
2032 continue;
2033 }
2034 bzero((caddr_t)&info, sizeof(info));
2035 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2036 /*
2037 * Holding ifnet lock here prevents the link
2038 * address from changing contents, so no need
2039 * to hold the ifa0 lock. The link address is
2040 * always present; it's never freed.
2041 */
2042 ifa0 = ifp->if_lladdr;
2043 info.rti_info[RTAX_IFP] = ifa0->ifa_addr;
2044 if (ifma->ifma_ll != NULL) {
2045 info.rti_info[RTAX_GATEWAY] =
2046 ifma->ifma_ll->ifma_addr;
2047 }
2048 len = rt_msg2(RTM_NEWMADDR2, &info, NULL, NULL,
2049 &cred);
2050 if (pass == 0) {
2051 total_len += len;
2052 } else {
2053 struct ifma_msghdr2 *ifmam;
2054
2055 if (current_len + len > total_len) {
2056 IFMA_UNLOCK(ifma);
2057 error = ENOBUFS;
2058 break;
2059 }
2060 len = rt_msg2(RTM_NEWMADDR2, &info,
2061 (caddr_t)cp, NULL, &cred);
2062
2063 ifmam =
2064 (struct ifma_msghdr2 *)(void *)cp;
2065 ifmam->ifmam_addrs = info.rti_addrs;
2066 ifmam->ifmam_flags = 0;
2067 ifmam->ifmam_index =
2068 ifma->ifma_ifp->if_index;
2069 ifmam->ifmam_refcount =
2070 ifma->ifma_reqcnt;
2071
2072 cp += len;
2073 VERIFY(IS_P2ALIGNED(cp,
2074 sizeof(u_int32_t)));
2075 current_len += len;
2076 }
2077 IFMA_UNLOCK(ifma);
2078 }
2079 ifnet_lock_done(ifp);
2080 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
2081 info.rti_info[RTAX_BRD] = NULL;
2082 }
2083 ifnet_head_done();
2084
2085 if (error) {
2086 if (error == ENOBUFS) {
2087 printf("%s: current_len (%d) + len (%d) > "
2088 "total_len (%d)\n", __func__, current_len,
2089 len, total_len);
2090 }
2091 break;
2092 }
2093
2094 if (pass == 0) {
2095 /* Better to return zero length buffer than ENOBUFS */
2096 if (total_len == 0) {
2097 total_len = 1;
2098 }
2099 total_len += total_len >> 3;
2100 total_buffer = _MALLOC(total_len, M_RTABLE,
2101 M_ZERO | M_WAITOK);
2102 if (total_buffer == NULL) {
2103 printf("%s: _MALLOC(%d) failed\n", __func__,
2104 total_len);
2105 error = ENOBUFS;
2106 break;
2107 }
2108 cp = total_buffer;
2109 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
2110 } else {
2111 error = SYSCTL_OUT(w->w_req, total_buffer, current_len);
2112 if (error) {
2113 break;
2114 }
2115 }
2116 }
2117
2118 if (total_buffer != NULL) {
2119 _FREE(total_buffer, M_RTABLE);
2120 }
2121
2122 kauth_cred_unref(&cred);
2123 return error;
2124 }
2125
2126
2127 static int
2128 sysctl_rtstat(struct sysctl_req *req)
2129 {
2130 return SYSCTL_OUT(req, &rtstat, sizeof(struct rtstat));
2131 }
2132
2133 static int
2134 sysctl_rttrash(struct sysctl_req *req)
2135 {
2136 return SYSCTL_OUT(req, &rttrash, sizeof(rttrash));
2137 }
2138
2139 static int
2140 sysctl_rtsock SYSCTL_HANDLER_ARGS
2141 {
2142 #pragma unused(oidp)
2143 int *name = (int *)arg1;
2144 u_int namelen = arg2;
2145 struct radix_node_head *rnh;
2146 int i, error = EINVAL;
2147 u_char af;
2148 struct walkarg w;
2149
2150 name++;
2151 namelen--;
2152 if (req->newptr) {
2153 return EPERM;
2154 }
2155 if (namelen != 3) {
2156 return EINVAL;
2157 }
2158 af = name[0];
2159 Bzero(&w, sizeof(w));
2160 w.w_op = name[1];
2161 w.w_arg = name[2];
2162 w.w_req = req;
2163
2164 switch (w.w_op) {
2165 case NET_RT_DUMP:
2166 case NET_RT_DUMP2:
2167 case NET_RT_FLAGS:
2168 case NET_RT_FLAGS_PRIV:
2169 lck_mtx_lock(rnh_lock);
2170 for (i = 1; i <= AF_MAX; i++) {
2171 if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
2172 (error = rnh->rnh_walktree(rnh,
2173 sysctl_dumpentry, &w))) {
2174 break;
2175 }
2176 }
2177 lck_mtx_unlock(rnh_lock);
2178 break;
2179 case NET_RT_DUMPX:
2180 case NET_RT_DUMPX_FLAGS:
2181 lck_mtx_lock(rnh_lock);
2182 for (i = 1; i <= AF_MAX; i++) {
2183 if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
2184 (error = rnh->rnh_walktree(rnh,
2185 sysctl_dumpentry_ext, &w))) {
2186 break;
2187 }
2188 }
2189 lck_mtx_unlock(rnh_lock);
2190 break;
2191 case NET_RT_IFLIST:
2192 error = sysctl_iflist(af, &w);
2193 break;
2194 case NET_RT_IFLIST2:
2195 error = sysctl_iflist2(af, &w);
2196 break;
2197 case NET_RT_STAT:
2198 error = sysctl_rtstat(req);
2199 break;
2200 case NET_RT_TRASH:
2201 error = sysctl_rttrash(req);
2202 break;
2203 }
2204 if (w.w_tmem != NULL) {
2205 FREE(w.w_tmem, M_RTABLE);
2206 }
2207 return error;
2208 }
2209
2210 /*
2211 * Definitions of protocols supported in the ROUTE domain.
2212 */
2213 static struct protosw routesw[] = {
2214 {
2215 .pr_type = SOCK_RAW,
2216 .pr_protocol = 0,
2217 .pr_flags = PR_ATOMIC | PR_ADDR,
2218 .pr_output = route_output,
2219 .pr_ctlinput = raw_ctlinput,
2220 .pr_init = raw_init,
2221 .pr_usrreqs = &route_usrreqs,
2222 }
2223 };
2224
2225 static int route_proto_count = (sizeof(routesw) / sizeof(struct protosw));
2226
2227 struct domain routedomain_s = {
2228 .dom_family = PF_ROUTE,
2229 .dom_name = "route",
2230 .dom_init = route_dinit,
2231 };
2232
2233 static void
2234 route_dinit(struct domain *dp)
2235 {
2236 struct protosw *pr;
2237 int i;
2238
2239 VERIFY(!(dp->dom_flags & DOM_INITIALIZED));
2240 VERIFY(routedomain == NULL);
2241
2242 routedomain = dp;
2243
2244 for (i = 0, pr = &routesw[0]; i < route_proto_count; i++, pr++) {
2245 net_add_proto(pr, dp, 1);
2246 }
2247
2248 route_init();
2249 }
mirror of XNU