xref: /xnu-10002.41.9/bsd/net/rtsock.c (revision 699cd48037512bf4380799317ca44ca453c82f57)
1 /*
2  * Copyright (c) 2000-2021 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 #include <IOKit/IOBSD.h>
92 
93 extern struct rtstat rtstat;
94 extern struct domain routedomain_s;
95 static struct domain *routedomain = NULL;
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(u_char, struct rt_addrinfo *);
136 static int rt_msg2(u_char, 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 = (uint32_t)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
rts_abort(struct socket * so)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
rts_attach(struct socket * so,int proto,struct proc * p)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 	rp = kalloc_type(struct rawcb, Z_WAITOK_ZERO_NOFAIL);
188 	so->so_pcb = (caddr_t)rp;
189 	/* don't use raw_usrreqs.pru_attach, it checks for SS_PRIV */
190 	error = raw_attach(so, proto);
191 	rp = sotorawcb(so);
192 	if (error) {
193 		kfree_type(struct rawcb, rp);
194 		so->so_pcb = NULL;
195 		so->so_flags |= SOF_PCBCLEARING;
196 		return error;
197 	}
198 
199 	switch (rp->rcb_proto.sp_protocol) {
200 	case AF_INET:
201 		os_atomic_inc(&route_cb.ip_count, relaxed);
202 		break;
203 	case AF_INET6:
204 		os_atomic_inc(&route_cb.ip6_count, relaxed);
205 		break;
206 	}
207 	rp->rcb_faddr = &route_src;
208 	os_atomic_inc(&route_cb.any_count, relaxed);
209 	/* the socket is already locked when we enter rts_attach */
210 	soisconnected(so);
211 	so->so_options |= SO_USELOOPBACK;
212 	return 0;
213 }
214 
215 static int
rts_bind(struct socket * so,struct sockaddr * nam,struct proc * p)216 rts_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
217 {
218 	return raw_usrreqs.pru_bind(so, nam, p); /* xxx just EINVAL */
219 }
220 
221 static int
rts_connect(struct socket * so,struct sockaddr * nam,struct proc * p)222 rts_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
223 {
224 	return raw_usrreqs.pru_connect(so, nam, p); /* XXX just EINVAL */
225 }
226 
227 /* pru_connect2 is EOPNOTSUPP */
228 /* pru_control is EOPNOTSUPP */
229 
230 static int
rts_detach(struct socket * so)231 rts_detach(struct socket *so)
232 {
233 	struct rawcb *rp = sotorawcb(so);
234 
235 	VERIFY(rp != NULL);
236 
237 	switch (rp->rcb_proto.sp_protocol) {
238 	case AF_INET:
239 		os_atomic_dec(&route_cb.ip_count, relaxed);
240 		break;
241 	case AF_INET6:
242 		os_atomic_dec(&route_cb.ip6_count, relaxed);
243 		break;
244 	}
245 	os_atomic_dec(&route_cb.any_count, relaxed);
246 	return raw_usrreqs.pru_detach(so);
247 }
248 
249 static int
rts_disconnect(struct socket * so)250 rts_disconnect(struct socket *so)
251 {
252 	return raw_usrreqs.pru_disconnect(so);
253 }
254 
255 /* pru_listen is EOPNOTSUPP */
256 
257 static int
rts_peeraddr(struct socket * so,struct sockaddr ** nam)258 rts_peeraddr(struct socket *so, struct sockaddr **nam)
259 {
260 	return raw_usrreqs.pru_peeraddr(so, nam);
261 }
262 
263 /* pru_rcvd is EOPNOTSUPP */
264 /* pru_rcvoob is EOPNOTSUPP */
265 
266 static int
rts_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)267 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
268     struct mbuf *control, struct proc *p)
269 {
270 	return raw_usrreqs.pru_send(so, flags, m, nam, control, p);
271 }
272 
273 /* pru_sense is null */
274 
275 static int
rts_shutdown(struct socket * so)276 rts_shutdown(struct socket *so)
277 {
278 	return raw_usrreqs.pru_shutdown(so);
279 }
280 
281 static int
rts_sockaddr(struct socket * so,struct sockaddr ** nam)282 rts_sockaddr(struct socket *so, struct sockaddr **nam)
283 {
284 	return raw_usrreqs.pru_sockaddr(so, nam);
285 }
286 
287 static struct pr_usrreqs route_usrreqs = {
288 	.pru_abort =            rts_abort,
289 	.pru_attach =           rts_attach,
290 	.pru_bind =             rts_bind,
291 	.pru_connect =          rts_connect,
292 	.pru_detach =           rts_detach,
293 	.pru_disconnect =       rts_disconnect,
294 	.pru_peeraddr =         rts_peeraddr,
295 	.pru_send =             rts_send,
296 	.pru_shutdown =         rts_shutdown,
297 	.pru_sockaddr =         rts_sockaddr,
298 	.pru_sosend =           sosend,
299 	.pru_soreceive =        soreceive,
300 };
301 
302 /*ARGSUSED*/
303 static int
route_output(struct mbuf * m,struct socket * so)304 route_output(struct mbuf *m, struct socket *so)
305 {
306 	struct rt_msghdr *rtm = NULL;
307 	size_t rtm_len = 0;
308 	struct rtentry *rt = NULL;
309 	struct rtentry *saved_nrt = NULL;
310 	struct radix_node_head *rnh;
311 	struct rt_addrinfo info;
312 	int len, error = 0;
313 	sa_family_t dst_sa_family = 0;
314 	struct ifnet *ifp = NULL;
315 	struct sockaddr_in dst_in, gate_in;
316 	int sendonlytoself = 0;
317 	unsigned int ifscope = IFSCOPE_NONE;
318 	struct rawcb *rp = NULL;
319 	boolean_t is_router = FALSE;
320 #define senderr(e) { error = (e); goto flush; }
321 	if (m == NULL || ((m->m_len < sizeof(intptr_t)) &&
322 	    (m = m_pullup(m, sizeof(intptr_t))) == NULL)) {
323 		return ENOBUFS;
324 	}
325 	VERIFY(m->m_flags & M_PKTHDR);
326 
327 	/*
328 	 * Unlock the socket (but keep a reference) it won't be
329 	 * accessed until raw_input appends to it.
330 	 */
331 	socket_unlock(so, 0);
332 	lck_mtx_lock(rnh_lock);
333 
334 	len = m->m_pkthdr.len;
335 	if (len < sizeof(*rtm) ||
336 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
337 		info.rti_info[RTAX_DST] = NULL;
338 		senderr(EINVAL);
339 	}
340 	rtm = kalloc_data(len, Z_WAITOK);
341 	if (rtm == NULL) {
342 		info.rti_info[RTAX_DST] = NULL;
343 		senderr(ENOBUFS);
344 	}
345 	rtm_len = (size_t)len;
346 	m_copydata(m, 0, len, (caddr_t)rtm);
347 	if (rtm->rtm_version != RTM_VERSION) {
348 		info.rti_info[RTAX_DST] = NULL;
349 		senderr(EPROTONOSUPPORT);
350 	}
351 
352 	/*
353 	 * Silent version of RTM_GET for Reachabiltiy APIs. We may change
354 	 * all RTM_GETs to be silent in the future, so this is private for now.
355 	 */
356 	if (rtm->rtm_type == RTM_GET_SILENT) {
357 		if (!(so->so_options & SO_USELOOPBACK)) {
358 			senderr(EINVAL);
359 		}
360 		sendonlytoself = 1;
361 		rtm->rtm_type = RTM_GET;
362 	}
363 
364 	/*
365 	 * Perform permission checking, only privileged sockets
366 	 * may perform operations other than RTM_GET
367 	 */
368 	if (rtm->rtm_type != RTM_GET && !(so->so_state & SS_PRIV)) {
369 		info.rti_info[RTAX_DST] = NULL;
370 		senderr(EPERM);
371 	}
372 
373 	rtm->rtm_pid = proc_selfpid();
374 	info.rti_addrs = rtm->rtm_addrs;
375 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
376 		info.rti_info[RTAX_DST] = NULL;
377 		senderr(EINVAL);
378 	}
379 	if (info.rti_info[RTAX_DST] == NULL ||
380 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
381 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
382 	    info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
383 		senderr(EINVAL);
384 	}
385 
386 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET &&
387 	    info.rti_info[RTAX_DST]->sa_len != sizeof(struct sockaddr_in)) {
388 		/* At minimum, we need up to sin_addr */
389 		if (info.rti_info[RTAX_DST]->sa_len <
390 		    offsetof(struct sockaddr_in, sin_zero)) {
391 			senderr(EINVAL);
392 		}
393 		bzero(&dst_in, sizeof(dst_in));
394 		dst_in.sin_len = sizeof(dst_in);
395 		dst_in.sin_family = AF_INET;
396 		dst_in.sin_port = SIN(info.rti_info[RTAX_DST])->sin_port;
397 		dst_in.sin_addr = SIN(info.rti_info[RTAX_DST])->sin_addr;
398 		info.rti_info[RTAX_DST] = (struct sockaddr *)&dst_in;
399 		dst_sa_family = info.rti_info[RTAX_DST]->sa_family;
400 	} else if (info.rti_info[RTAX_DST]->sa_family == AF_INET6 &&
401 	    info.rti_info[RTAX_DST]->sa_len < sizeof(struct sockaddr_in6)) {
402 		senderr(EINVAL);
403 	}
404 
405 	if (info.rti_info[RTAX_GATEWAY] != NULL) {
406 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_INET &&
407 		    info.rti_info[RTAX_GATEWAY]->sa_len != sizeof(struct sockaddr_in)) {
408 			/* At minimum, we need up to sin_addr */
409 			if (info.rti_info[RTAX_GATEWAY]->sa_len <
410 			    offsetof(struct sockaddr_in, sin_zero)) {
411 				senderr(EINVAL);
412 			}
413 			bzero(&gate_in, sizeof(gate_in));
414 			gate_in.sin_len = sizeof(gate_in);
415 			gate_in.sin_family = AF_INET;
416 			gate_in.sin_port = SIN(info.rti_info[RTAX_GATEWAY])->sin_port;
417 			gate_in.sin_addr = SIN(info.rti_info[RTAX_GATEWAY])->sin_addr;
418 			info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gate_in;
419 		} else if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_INET6 &&
420 		    info.rti_info[RTAX_GATEWAY]->sa_len < sizeof(struct sockaddr_in6)) {
421 			senderr(EINVAL);
422 		}
423 	}
424 
425 	if (info.rti_info[RTAX_GENMASK]) {
426 		struct radix_node *t;
427 		t = rn_addmask((caddr_t)info.rti_info[RTAX_GENMASK], 0, 1);
428 		if (t != NULL && Bcmp(info.rti_info[RTAX_GENMASK],
429 		    t->rn_key, *(u_char *)info.rti_info[RTAX_GENMASK]) == 0) {
430 			info.rti_info[RTAX_GENMASK] =
431 			    (struct sockaddr *)(t->rn_key);
432 		} else {
433 			senderr(ENOBUFS);
434 		}
435 	}
436 
437 	/*
438 	 * If RTF_IFSCOPE flag is set, then rtm_index specifies the scope.
439 	 */
440 	if (rtm->rtm_flags & RTF_IFSCOPE) {
441 		if (info.rti_info[RTAX_DST]->sa_family != AF_INET &&
442 		    info.rti_info[RTAX_DST]->sa_family != AF_INET6) {
443 			senderr(EINVAL);
444 		}
445 		ifscope = rtm->rtm_index;
446 	}
447 	/*
448 	 * Block changes on INTCOPROC interfaces.
449 	 */
450 	if (ifscope != IFSCOPE_NONE) {
451 		unsigned int intcoproc_scope = 0;
452 		ifnet_head_lock_shared();
453 		TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
454 			if (IFNET_IS_INTCOPROC(ifp)) {
455 				intcoproc_scope = ifp->if_index;
456 				break;
457 			}
458 		}
459 		ifnet_head_done();
460 		if (intcoproc_scope == ifscope && proc_getpid(current_proc()) != 0) {
461 			senderr(EINVAL);
462 		}
463 	}
464 	/*
465 	 * Require entitlement to change management interfaces
466 	 */
467 	if (management_control_unrestricted == false && if_management_interface_check_needed == true &&
468 	    ifscope != IFSCOPE_NONE && proc_getpid(current_proc()) != 0) {
469 		bool is_management = false;
470 
471 		ifnet_head_lock_shared();
472 		if (IF_INDEX_IN_RANGE(ifscope)) {
473 			if (IFNET_IS_MANAGEMENT(ifindex2ifnet[ifscope])) {
474 				is_management = true;
475 			}
476 		}
477 		ifnet_head_done();
478 
479 		if (is_management && !IOCurrentTaskHasEntitlement(MANAGEMENT_CONTROL_ENTITLEMENT)) {
480 			senderr(EINVAL);
481 		}
482 	}
483 
484 	/*
485 	 * RTF_PROXY can only be set internally from within the kernel.
486 	 */
487 	if (rtm->rtm_flags & RTF_PROXY) {
488 		senderr(EINVAL);
489 	}
490 
491 	/*
492 	 * For AF_INET, always zero out the embedded scope ID.  If this is
493 	 * a scoped request, it must be done explicitly by setting RTF_IFSCOPE
494 	 * flag and the corresponding rtm_index value.  This is to prevent
495 	 * false interpretation of the scope ID because it's using the sin_zero
496 	 * field, which might not be properly cleared by the requestor.
497 	 */
498 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
499 		sin_set_ifscope(info.rti_info[RTAX_DST], IFSCOPE_NONE);
500 	}
501 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
502 	    info.rti_info[RTAX_GATEWAY]->sa_family == AF_INET) {
503 		sin_set_ifscope(info.rti_info[RTAX_GATEWAY], IFSCOPE_NONE);
504 	}
505 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET6 &&
506 	    IN6_IS_SCOPE_EMBED(&SIN6(info.rti_info[RTAX_DST])->sin6_addr) &&
507 	    !IN6_IS_ADDR_UNICAST_BASED_MULTICAST(&SIN6(info.rti_info[RTAX_DST])->sin6_addr) &&
508 	    SIN6(info.rti_info[RTAX_DST])->sin6_scope_id == 0) {
509 		SIN6(info.rti_info[RTAX_DST])->sin6_scope_id = ntohs(SIN6(info.rti_info[RTAX_DST])->sin6_addr.s6_addr16[1]);
510 		SIN6(info.rti_info[RTAX_DST])->sin6_addr.s6_addr16[1] = 0;
511 	}
512 
513 	switch (rtm->rtm_type) {
514 	case RTM_ADD:
515 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
516 			senderr(EINVAL);
517 		}
518 
519 		error = rtrequest_scoped_locked(RTM_ADD,
520 		    info.rti_info[RTAX_DST], info.rti_info[RTAX_GATEWAY],
521 		    info.rti_info[RTAX_NETMASK], rtm->rtm_flags, &saved_nrt,
522 		    ifscope);
523 		if (error == 0 && saved_nrt != NULL) {
524 			RT_LOCK(saved_nrt);
525 			/*
526 			 * If the route request specified an interface with
527 			 * IFA and/or IFP, we set the requested interface on
528 			 * the route with rt_setif.  It would be much better
529 			 * to do this inside rtrequest, but that would
530 			 * require passing the desired interface, in some
531 			 * form, to rtrequest.  Since rtrequest is called in
532 			 * so many places (roughly 40 in our source), adding
533 			 * a parameter is to much for us to swallow; this is
534 			 * something for the FreeBSD developers to tackle.
535 			 * Instead, we let rtrequest compute whatever
536 			 * interface it wants, then come in behind it and
537 			 * stick in the interface that we really want.  This
538 			 * works reasonably well except when rtrequest can't
539 			 * figure out what interface to use (with
540 			 * ifa_withroute) and returns ENETUNREACH.  Ideally
541 			 * it shouldn't matter if rtrequest can't figure out
542 			 * the interface if we're going to explicitly set it
543 			 * ourselves anyway.  But practically we can't
544 			 * recover here because rtrequest will not do any of
545 			 * the work necessary to add the route if it can't
546 			 * find an interface.  As long as there is a default
547 			 * route that leads to some interface, rtrequest will
548 			 * find an interface, so this problem should be
549 			 * rarely encountered.
550 			 * [email protected]
551 			 */
552 			rt_setif(saved_nrt,
553 			    info.rti_info[RTAX_IFP], info.rti_info[RTAX_IFA],
554 			    info.rti_info[RTAX_GATEWAY], ifscope);
555 			(void)rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, saved_nrt);
556 			saved_nrt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
557 			saved_nrt->rt_rmx.rmx_locks |=
558 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
559 			saved_nrt->rt_genmask = info.rti_info[RTAX_GENMASK];
560 			RT_REMREF_LOCKED(saved_nrt);
561 			RT_UNLOCK(saved_nrt);
562 		}
563 		break;
564 
565 	case RTM_DELETE:
566 		error = rtrequest_scoped_locked(RTM_DELETE,
567 		    info.rti_info[RTAX_DST], info.rti_info[RTAX_GATEWAY],
568 		    info.rti_info[RTAX_NETMASK], rtm->rtm_flags, &saved_nrt,
569 		    ifscope);
570 		if (error == 0) {
571 			rt = saved_nrt;
572 			RT_LOCK(rt);
573 			goto report;
574 		}
575 		break;
576 
577 	case RTM_GET:
578 	case RTM_CHANGE:
579 	case RTM_LOCK:
580 		rnh = rt_tables[info.rti_info[RTAX_DST]->sa_family];
581 		if (rnh == NULL) {
582 			senderr(EAFNOSUPPORT);
583 		}
584 		/*
585 		 * Lookup the best match based on the key-mask pair;
586 		 * callee adds a reference and checks for root node.
587 		 */
588 		rt = rt_lookup(TRUE, info.rti_info[RTAX_DST],
589 		    info.rti_info[RTAX_NETMASK], rnh, ifscope);
590 		if (rt == NULL) {
591 			senderr(ESRCH);
592 		}
593 		RT_LOCK(rt);
594 
595 		/*
596 		 * Holding rnh_lock here prevents the possibility of
597 		 * ifa from changing (e.g. in_ifinit), so it is safe
598 		 * to access its ifa_addr (down below) without locking.
599 		 */
600 		switch (rtm->rtm_type) {
601 		case RTM_GET: {
602 			kauth_cred_t cred;
603 			kauth_cred_t* credp;
604 			struct ifaddr *ifa2;
605 report:
606 			cred = kauth_cred_proc_ref(current_proc());
607 			credp = &cred;
608 
609 			ifa2 = NULL;
610 			RT_LOCK_ASSERT_HELD(rt);
611 			info.rti_info[RTAX_DST] = rt_key(rt);
612 			dst_sa_family = info.rti_info[RTAX_DST]->sa_family;
613 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
614 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
615 			info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
616 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
617 				ifp = rt->rt_ifp;
618 				if (ifp != NULL) {
619 					ifnet_lock_shared(ifp);
620 					ifa2 = ifp->if_lladdr;
621 					info.rti_info[RTAX_IFP] =
622 					    ifa2->ifa_addr;
623 					IFA_ADDREF(ifa2);
624 					ifnet_lock_done(ifp);
625 					info.rti_info[RTAX_IFA] =
626 					    rt->rt_ifa->ifa_addr;
627 					rtm->rtm_index = ifp->if_index;
628 				} else {
629 					info.rti_info[RTAX_IFP] = NULL;
630 					info.rti_info[RTAX_IFA] = NULL;
631 				}
632 			} else if ((ifp = rt->rt_ifp) != NULL) {
633 				rtm->rtm_index = ifp->if_index;
634 			}
635 			if (ifa2 != NULL) {
636 				IFA_LOCK(ifa2);
637 			}
638 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL, credp);
639 			if (ifa2 != NULL) {
640 				IFA_UNLOCK(ifa2);
641 			}
642 			struct rt_msghdr *out_rtm;
643 			out_rtm = kalloc_data(len, Z_WAITOK);
644 			if (out_rtm == NULL) {
645 				RT_UNLOCK(rt);
646 				if (ifa2 != NULL) {
647 					IFA_REMREF(ifa2);
648 				}
649 				senderr(ENOBUFS);
650 			}
651 			Bcopy(rtm, out_rtm, sizeof(struct rt_msghdr));
652 			if (ifa2 != NULL) {
653 				IFA_LOCK(ifa2);
654 			}
655 			(void) rt_msg2(out_rtm->rtm_type, &info, (caddr_t)out_rtm,
656 			    NULL, &cred);
657 			if (ifa2 != NULL) {
658 				IFA_UNLOCK(ifa2);
659 			}
660 			kfree_data(rtm, rtm_len);
661 			rtm = out_rtm;
662 			rtm_len = len;
663 			rtm->rtm_flags = rt->rt_flags;
664 			rt_getmetrics(rt, &rtm->rtm_rmx);
665 			rtm->rtm_addrs = info.rti_addrs;
666 			if (ifa2 != NULL) {
667 				IFA_REMREF(ifa2);
668 			}
669 
670 			kauth_cred_unref(&cred);
671 			break;
672 		}
673 
674 		case RTM_CHANGE:
675 			is_router = (rt->rt_flags & RTF_ROUTER) ? TRUE : FALSE;
676 
677 			if (info.rti_info[RTAX_GATEWAY] != NULL &&
678 			    (error = rt_setgate(rt, rt_key(rt),
679 			    info.rti_info[RTAX_GATEWAY]))) {
680 				int tmp = error;
681 				RT_UNLOCK(rt);
682 				senderr(tmp);
683 			}
684 			/*
685 			 * If they tried to change things but didn't specify
686 			 * the required gateway, then just use the old one.
687 			 * This can happen if the user tries to change the
688 			 * flags on the default route without changing the
689 			 * default gateway. Changing flags still doesn't work.
690 			 */
691 			if ((rt->rt_flags & RTF_GATEWAY) &&
692 			    info.rti_info[RTAX_GATEWAY] == NULL) {
693 				info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
694 			}
695 
696 			/*
697 			 * On Darwin, we call rt_setif which contains the
698 			 * equivalent to the code found at this very spot
699 			 * in BSD.
700 			 */
701 			rt_setif(rt,
702 			    info.rti_info[RTAX_IFP], info.rti_info[RTAX_IFA],
703 			    info.rti_info[RTAX_GATEWAY], ifscope);
704 
705 			if ((error = rt_setmetrics(rtm->rtm_inits,
706 			    &rtm->rtm_rmx, rt))) {
707 				int tmp = error;
708 				RT_UNLOCK(rt);
709 				senderr(tmp);
710 			}
711 			if (info.rti_info[RTAX_GENMASK]) {
712 				rt->rt_genmask = info.rti_info[RTAX_GENMASK];
713 			}
714 
715 			/*
716 			 * Enqueue work item to invoke callback for this route entry
717 			 * This may not be needed always, but for now issue it anytime
718 			 * RTM_CHANGE gets called.
719 			 */
720 			route_event_enqueue_nwk_wq_entry(rt, NULL, ROUTE_ENTRY_REFRESH, NULL, TRUE);
721 			/*
722 			 * If the route is for a router, walk the tree to send refresh
723 			 * event to protocol cloned entries
724 			 */
725 			if (is_router) {
726 				struct route_event rt_ev;
727 				route_event_init(&rt_ev, rt, NULL, ROUTE_ENTRY_REFRESH);
728 				RT_UNLOCK(rt);
729 				(void) rnh->rnh_walktree(rnh, route_event_walktree, (void *)&rt_ev);
730 				RT_LOCK(rt);
731 			}
732 			OS_FALLTHROUGH;
733 		case RTM_LOCK:
734 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
735 			rt->rt_rmx.rmx_locks |=
736 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
737 			break;
738 		}
739 		RT_UNLOCK(rt);
740 		break;
741 	default:
742 		senderr(EOPNOTSUPP);
743 	}
744 flush:
745 	if (rtm != NULL) {
746 		if (error) {
747 			rtm->rtm_errno = error;
748 		} else {
749 			rtm->rtm_flags |= RTF_DONE;
750 		}
751 	}
752 	if (rt != NULL) {
753 		RT_LOCK_ASSERT_NOTHELD(rt);
754 		rtfree_locked(rt);
755 	}
756 	lck_mtx_unlock(rnh_lock);
757 
758 	/* relock the socket now */
759 	socket_lock(so, 0);
760 	/*
761 	 * Check to see if we don't want our own messages.
762 	 */
763 	if (!(so->so_options & SO_USELOOPBACK)) {
764 		if (route_cb.any_count <= 1) {
765 			kfree_data(rtm, rtm_len);
766 			m_freem(m);
767 			return error;
768 		}
769 		/* There is another listener, so construct message */
770 		rp = sotorawcb(so);
771 	}
772 	if (rtm != NULL) {
773 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
774 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
775 			m_freem(m);
776 			m = NULL;
777 		} else if (m->m_pkthdr.len > rtm->rtm_msglen) {
778 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
779 		}
780 		kfree_data(rtm, rtm_len);
781 	}
782 	if (sendonlytoself && m != NULL) {
783 		error = 0;
784 		if (sbappendaddr(&so->so_rcv, &route_src, m,
785 		    NULL, &error) != 0) {
786 			sorwakeup(so);
787 		}
788 		if (error) {
789 			return error;
790 		}
791 	} else {
792 		struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
793 		if (rp != NULL) {
794 			rp->rcb_proto.sp_family = 0; /* Avoid us */
795 		}
796 		if (dst_sa_family != 0) {
797 			route_proto.sp_protocol = dst_sa_family;
798 		}
799 		if (m != NULL) {
800 			socket_unlock(so, 0);
801 			raw_input(m, &route_proto, &route_src, &route_dst);
802 			socket_lock(so, 0);
803 		}
804 		if (rp != NULL) {
805 			rp->rcb_proto.sp_family = PF_ROUTE;
806 		}
807 	}
808 	return error;
809 }
810 
811 void
rt_setexpire(struct rtentry * rt,uint64_t expiry)812 rt_setexpire(struct rtentry *rt, uint64_t expiry)
813 {
814 	/* set both rt_expire and rmx_expire */
815 	rt->rt_expire = expiry;
816 	if (expiry) {
817 		rt->rt_rmx.rmx_expire =
818 		    (int32_t)(expiry + rt->base_calendartime -
819 		    rt->base_uptime);
820 	} else {
821 		rt->rt_rmx.rmx_expire = 0;
822 	}
823 }
824 
825 static int
rt_setmetrics(u_int32_t which,struct rt_metrics * in,struct rtentry * out)826 rt_setmetrics(u_int32_t which, struct rt_metrics *in, struct rtentry *out)
827 {
828 	if (!(which & RTV_REFRESH_HOST)) {
829 		struct timeval caltime;
830 		getmicrotime(&caltime);
831 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->e;
832 		metric(RTV_RPIPE, rmx_recvpipe);
833 		metric(RTV_SPIPE, rmx_sendpipe);
834 		metric(RTV_SSTHRESH, rmx_ssthresh);
835 		metric(RTV_RTT, rmx_rtt);
836 		metric(RTV_RTTVAR, rmx_rttvar);
837 		metric(RTV_HOPCOUNT, rmx_hopcount);
838 		metric(RTV_MTU, rmx_mtu);
839 		metric(RTV_EXPIRE, rmx_expire);
840 #undef metric
841 		if (out->rt_rmx.rmx_expire > 0) {
842 			/* account for system time change */
843 			getmicrotime(&caltime);
844 			out->base_calendartime +=
845 			    NET_CALCULATE_CLOCKSKEW(caltime,
846 			    out->base_calendartime,
847 			    net_uptime(), out->base_uptime);
848 			rt_setexpire(out,
849 			    out->rt_rmx.rmx_expire -
850 			    out->base_calendartime +
851 			    out->base_uptime);
852 		} else {
853 			rt_setexpire(out, 0);
854 		}
855 
856 		VERIFY(out->rt_expire == 0 || out->rt_rmx.rmx_expire != 0);
857 		VERIFY(out->rt_expire != 0 || out->rt_rmx.rmx_expire == 0);
858 	} else {
859 		/* Only RTV_REFRESH_HOST must be set */
860 		if ((which & ~RTV_REFRESH_HOST) ||
861 		    (out->rt_flags & RTF_STATIC) ||
862 		    !(out->rt_flags & RTF_LLINFO)) {
863 			return EINVAL;
864 		}
865 
866 		if (out->rt_llinfo_refresh == NULL) {
867 			return ENOTSUP;
868 		}
869 
870 		out->rt_llinfo_refresh(out);
871 	}
872 	return 0;
873 }
874 
875 static void
rt_getmetrics(struct rtentry * in,struct rt_metrics * out)876 rt_getmetrics(struct rtentry *in, struct rt_metrics *out)
877 {
878 	struct timeval caltime;
879 
880 	VERIFY(in->rt_expire == 0 || in->rt_rmx.rmx_expire != 0);
881 	VERIFY(in->rt_expire != 0 || in->rt_rmx.rmx_expire == 0);
882 
883 	*out = in->rt_rmx;
884 
885 	if (in->rt_expire != 0) {
886 		/* account for system time change */
887 		getmicrotime(&caltime);
888 
889 		in->base_calendartime +=
890 		    NET_CALCULATE_CLOCKSKEW(caltime,
891 		    in->base_calendartime, net_uptime(), in->base_uptime);
892 
893 		out->rmx_expire = (int32_t)(in->base_calendartime +
894 		    in->rt_expire - in->base_uptime);
895 	} else {
896 		out->rmx_expire = 0;
897 	}
898 }
899 
900 /*
901  * Set route's interface given info.rti_info[RTAX_IFP],
902  * info.rti_info[RTAX_IFA], and gateway.
903  */
904 static void
rt_setif(struct rtentry * rt,struct sockaddr * Ifpaddr,struct sockaddr * Ifaaddr,struct sockaddr * Gate,unsigned int ifscope)905 rt_setif(struct rtentry *rt, struct sockaddr *Ifpaddr, struct sockaddr *Ifaaddr,
906     struct sockaddr *Gate, unsigned int ifscope)
907 {
908 	struct ifaddr *ifa = NULL;
909 	struct ifnet *ifp = NULL;
910 	void (*ifa_rtrequest)(int, struct rtentry *, struct sockaddr *);
911 
912 	LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_OWNED);
913 
914 	RT_LOCK_ASSERT_HELD(rt);
915 
916 	/* Don't update a defunct route */
917 	if (rt->rt_flags & RTF_CONDEMNED) {
918 		return;
919 	}
920 
921 	/* Add an extra ref for ourselves */
922 	RT_ADDREF_LOCKED(rt);
923 
924 	/* Become a regular mutex, just in case */
925 	RT_CONVERT_LOCK(rt);
926 
927 	/*
928 	 * New gateway could require new ifaddr, ifp; flags may also
929 	 * be different; ifp may be specified by ll sockaddr when
930 	 * protocol address is ambiguous.
931 	 */
932 	if (Ifpaddr && (ifa = ifa_ifwithnet_scoped(Ifpaddr, ifscope)) &&
933 	    (ifp = ifa->ifa_ifp) && (Ifaaddr || Gate)) {
934 		IFA_REMREF(ifa);
935 		ifa = ifaof_ifpforaddr(Ifaaddr ? Ifaaddr : Gate, ifp);
936 	} else {
937 		if (ifa != NULL) {
938 			IFA_REMREF(ifa);
939 			ifa = NULL;
940 		}
941 		if (Ifpaddr && (ifp = if_withname(Ifpaddr))) {
942 			if (Gate) {
943 				ifa = ifaof_ifpforaddr(Gate, ifp);
944 			} else {
945 				ifnet_lock_shared(ifp);
946 				ifa = TAILQ_FIRST(&ifp->if_addrhead);
947 				if (ifa != NULL) {
948 					IFA_ADDREF(ifa);
949 				}
950 				ifnet_lock_done(ifp);
951 			}
952 		} else if (Ifaaddr &&
953 		    (ifa = ifa_ifwithaddr_scoped(Ifaaddr, ifscope))) {
954 			ifp = ifa->ifa_ifp;
955 		} else if (Gate != NULL) {
956 			/*
957 			 * Safe to drop rt_lock and use rt_key, since holding
958 			 * rnh_lock here prevents another thread from calling
959 			 * rt_setgate() on this route.  We cannot hold the
960 			 * lock across ifa_ifwithroute since the lookup done
961 			 * by that routine may point to the same route.
962 			 */
963 			RT_UNLOCK(rt);
964 			if ((ifa = ifa_ifwithroute_scoped_locked(rt->rt_flags,
965 			    rt_key(rt), Gate, ifscope)) != NULL) {
966 				ifp = ifa->ifa_ifp;
967 			}
968 			RT_LOCK(rt);
969 			/* Don't update a defunct route */
970 			if (rt->rt_flags & RTF_CONDEMNED) {
971 				if (ifa != NULL) {
972 					IFA_REMREF(ifa);
973 				}
974 				/* Release extra ref */
975 				RT_REMREF_LOCKED(rt);
976 				return;
977 			}
978 		}
979 	}
980 
981 	/* trigger route cache reevaluation */
982 	if (rt_key(rt)->sa_family == AF_INET) {
983 		routegenid_inet_update();
984 	} else if (rt_key(rt)->sa_family == AF_INET6) {
985 		routegenid_inet6_update();
986 	}
987 
988 	if (ifa != NULL) {
989 		struct ifaddr *oifa = rt->rt_ifa;
990 		if (oifa != ifa) {
991 			if (oifa != NULL) {
992 				IFA_LOCK_SPIN(oifa);
993 				ifa_rtrequest = oifa->ifa_rtrequest;
994 				IFA_UNLOCK(oifa);
995 				if (ifa_rtrequest != NULL) {
996 					ifa_rtrequest(RTM_DELETE, rt, Gate);
997 				}
998 			}
999 			rtsetifa(rt, ifa);
1000 
1001 			if (rt->rt_ifp != ifp) {
1002 				/*
1003 				 * Purge any link-layer info caching.
1004 				 */
1005 				if (rt->rt_llinfo_purge != NULL) {
1006 					rt->rt_llinfo_purge(rt);
1007 				}
1008 
1009 				/*
1010 				 * Adjust route ref count for the interfaces.
1011 				 */
1012 				if (rt->rt_if_ref_fn != NULL) {
1013 					rt->rt_if_ref_fn(ifp, 1);
1014 					rt->rt_if_ref_fn(rt->rt_ifp, -1);
1015 				}
1016 			}
1017 			rt->rt_ifp = ifp;
1018 			/*
1019 			 * If this is the (non-scoped) default route, record
1020 			 * the interface index used for the primary ifscope.
1021 			 */
1022 			if (rt_primary_default(rt, rt_key(rt))) {
1023 				set_primary_ifscope(rt_key(rt)->sa_family,
1024 				    rt->rt_ifp->if_index);
1025 			}
1026 			/*
1027 			 * If rmx_mtu is not locked, update it
1028 			 * to the MTU used by the new interface.
1029 			 */
1030 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1031 				rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
1032 				if (rt_key(rt)->sa_family == AF_INET &&
1033 				    INTF_ADJUST_MTU_FOR_CLAT46(ifp)) {
1034 					rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp);
1035 					/* Further adjust the size for CLAT46 expansion */
1036 					rt->rt_rmx.rmx_mtu -= CLAT46_HDR_EXPANSION_OVERHD;
1037 				}
1038 			}
1039 
1040 			if (rt->rt_ifa != NULL) {
1041 				IFA_LOCK_SPIN(rt->rt_ifa);
1042 				ifa_rtrequest = rt->rt_ifa->ifa_rtrequest;
1043 				IFA_UNLOCK(rt->rt_ifa);
1044 				if (ifa_rtrequest != NULL) {
1045 					ifa_rtrequest(RTM_ADD, rt, Gate);
1046 				}
1047 			}
1048 			IFA_REMREF(ifa);
1049 			/* Release extra ref */
1050 			RT_REMREF_LOCKED(rt);
1051 			return;
1052 		}
1053 		IFA_REMREF(ifa);
1054 		ifa = NULL;
1055 	}
1056 
1057 	/* XXX: to reset gateway to correct value, at RTM_CHANGE */
1058 	if (rt->rt_ifa != NULL) {
1059 		IFA_LOCK_SPIN(rt->rt_ifa);
1060 		ifa_rtrequest = rt->rt_ifa->ifa_rtrequest;
1061 		IFA_UNLOCK(rt->rt_ifa);
1062 		if (ifa_rtrequest != NULL) {
1063 			ifa_rtrequest(RTM_ADD, rt, Gate);
1064 		}
1065 	}
1066 
1067 	/*
1068 	 * Workaround for local address routes pointing to the loopback
1069 	 * interface added by configd, until <rdar://problem/12970142>.
1070 	 */
1071 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
1072 	    (rt->rt_flags & RTF_HOST) && rt->rt_ifa->ifa_ifp == rt->rt_ifp) {
1073 		ifa = ifa_ifwithaddr(rt_key(rt));
1074 		if (ifa != NULL) {
1075 			if (ifa != rt->rt_ifa) {
1076 				rtsetifa(rt, ifa);
1077 			}
1078 			IFA_REMREF(ifa);
1079 		}
1080 	}
1081 
1082 	/* Release extra ref */
1083 	RT_REMREF_LOCKED(rt);
1084 }
1085 
1086 /*
1087  * Extract the addresses of the passed sockaddrs.
1088  * Do a little sanity checking so as to avoid bad memory references.
1089  * This data is derived straight from userland.
1090  */
1091 static int
rt_xaddrs(caddr_t cp,caddr_t cplim,struct rt_addrinfo * rtinfo)1092 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1093 {
1094 	struct sockaddr *sa;
1095 	int i;
1096 
1097 	bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
1098 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
1099 		if ((rtinfo->rti_addrs & (1 << i)) == 0) {
1100 			continue;
1101 		}
1102 		sa = (struct sockaddr *)cp;
1103 		/*
1104 		 * It won't fit.
1105 		 */
1106 		if ((cp + sa->sa_len) > cplim) {
1107 			return EINVAL;
1108 		}
1109 		if (sa->sa_len > sizeof(struct sockaddr_storage)) {
1110 			return EINVAL;
1111 		}
1112 		/*
1113 		 * there are no more.. quit now
1114 		 * If there are more bits, they are in error.
1115 		 * I've seen this. route(1) can evidently generate these.
1116 		 * This causes kernel to core dump.
1117 		 * for compatibility, If we see this, point to a safe address.
1118 		 */
1119 		if (sa->sa_len == 0) {
1120 			rtinfo->rti_info[i] = &sa_zero;
1121 			return 0; /* should be EINVAL but for compat */
1122 		}
1123 		if (sa->sa_len < offsetof(struct sockaddr, sa_data)) {
1124 			return EINVAL;
1125 		}
1126 		/* accept it */
1127 		rtinfo->rti_info[i] = sa;
1128 		ADVANCE32(cp, sa);
1129 	}
1130 	return 0;
1131 }
1132 
1133 static struct mbuf *
rt_msg1(u_char type,struct rt_addrinfo * rtinfo)1134 rt_msg1(u_char type, struct rt_addrinfo *rtinfo)
1135 {
1136 	struct rt_msghdr *rtm;
1137 	struct mbuf *m;
1138 	int i;
1139 	int len, dlen, off;
1140 
1141 	switch (type) {
1142 	case RTM_DELADDR:
1143 	case RTM_NEWADDR:
1144 		len = sizeof(struct ifa_msghdr);
1145 		break;
1146 
1147 	case RTM_DELMADDR:
1148 	case RTM_NEWMADDR:
1149 		len = sizeof(struct ifma_msghdr);
1150 		break;
1151 
1152 	case RTM_IFINFO:
1153 		len = sizeof(struct if_msghdr);
1154 		break;
1155 
1156 	default:
1157 		len = sizeof(struct rt_msghdr);
1158 	}
1159 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1160 	if (m && len > MHLEN) {
1161 		MCLGET(m, M_DONTWAIT);
1162 		if (!(m->m_flags & M_EXT)) {
1163 			m_free(m);
1164 			m = NULL;
1165 		}
1166 	}
1167 	if (m == NULL) {
1168 		return NULL;
1169 	}
1170 	m->m_pkthdr.len = m->m_len = len;
1171 	m->m_pkthdr.rcvif = NULL;
1172 	rtm = mtod(m, struct rt_msghdr *);
1173 	bzero((caddr_t)rtm, len);
1174 	off = len;
1175 	for (i = 0; i < RTAX_MAX; i++) {
1176 		struct sockaddr *sa, *hint;
1177 		uint8_t ssbuf[SOCK_MAXADDRLEN + 1];
1178 
1179 		/*
1180 		 * Make sure to accomodate the largest possible size of sa_len.
1181 		 */
1182 		_CASSERT(sizeof(ssbuf) == (SOCK_MAXADDRLEN + 1));
1183 
1184 		if ((sa = rtinfo->rti_info[i]) == NULL) {
1185 			continue;
1186 		}
1187 
1188 		switch (i) {
1189 		case RTAX_DST:
1190 		case RTAX_NETMASK:
1191 			if ((hint = rtinfo->rti_info[RTAX_DST]) == NULL) {
1192 				hint = rtinfo->rti_info[RTAX_IFA];
1193 			}
1194 
1195 			/* Scrub away any trace of embedded interface scope */
1196 			sa = rtm_scrub(type, i, hint, sa, &ssbuf,
1197 			    sizeof(ssbuf), NULL);
1198 			break;
1199 
1200 		default:
1201 			break;
1202 		}
1203 
1204 		rtinfo->rti_addrs |= (1 << i);
1205 		dlen = sa->sa_len;
1206 		m_copyback(m, off, dlen, (caddr_t)sa);
1207 		len = off + dlen;
1208 		off += ROUNDUP32(dlen);
1209 	}
1210 	if (m->m_pkthdr.len != len) {
1211 		m_freem(m);
1212 		return NULL;
1213 	}
1214 	rtm->rtm_msglen = (u_short)len;
1215 	rtm->rtm_version = RTM_VERSION;
1216 	rtm->rtm_type = type;
1217 	return m;
1218 }
1219 
1220 static int
rt_msg2(u_char type,struct rt_addrinfo * rtinfo,caddr_t cp,struct walkarg * w,kauth_cred_t * credp)1221 rt_msg2(u_char type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w,
1222     kauth_cred_t* credp)
1223 {
1224 	int i;
1225 	int len, dlen, rlen, second_time = 0;
1226 	caddr_t cp0;
1227 
1228 	rtinfo->rti_addrs = 0;
1229 again:
1230 	switch (type) {
1231 	case RTM_DELADDR:
1232 	case RTM_NEWADDR:
1233 		len = sizeof(struct ifa_msghdr);
1234 		break;
1235 
1236 	case RTM_DELMADDR:
1237 	case RTM_NEWMADDR:
1238 		len = sizeof(struct ifma_msghdr);
1239 		break;
1240 
1241 	case RTM_IFINFO:
1242 		len = sizeof(struct if_msghdr);
1243 		break;
1244 
1245 	case RTM_IFINFO2:
1246 		len = sizeof(struct if_msghdr2);
1247 		break;
1248 
1249 	case RTM_NEWMADDR2:
1250 		len = sizeof(struct ifma_msghdr2);
1251 		break;
1252 
1253 	case RTM_GET_EXT:
1254 		len = sizeof(struct rt_msghdr_ext);
1255 		break;
1256 
1257 	case RTM_GET2:
1258 		len = sizeof(struct rt_msghdr2);
1259 		break;
1260 
1261 	default:
1262 		len = sizeof(struct rt_msghdr);
1263 	}
1264 	cp0 = cp;
1265 	if (cp0) {
1266 		cp += len;
1267 	}
1268 	for (i = 0; i < RTAX_MAX; i++) {
1269 		struct sockaddr *sa, *hint;
1270 		uint8_t ssbuf[SOCK_MAXADDRLEN + 1];
1271 
1272 		/*
1273 		 * Make sure to accomodate the largest possible size of sa_len.
1274 		 */
1275 		_CASSERT(sizeof(ssbuf) == (SOCK_MAXADDRLEN + 1));
1276 
1277 		if ((sa = rtinfo->rti_info[i]) == NULL) {
1278 			continue;
1279 		}
1280 
1281 		switch (i) {
1282 		case RTAX_DST:
1283 		case RTAX_NETMASK:
1284 			if ((hint = rtinfo->rti_info[RTAX_DST]) == NULL) {
1285 				hint = rtinfo->rti_info[RTAX_IFA];
1286 			}
1287 
1288 			/* Scrub away any trace of embedded interface scope */
1289 			sa = rtm_scrub(type, i, hint, sa, &ssbuf,
1290 			    sizeof(ssbuf), NULL);
1291 			break;
1292 		case RTAX_GATEWAY:
1293 		case RTAX_IFP:
1294 			sa = rtm_scrub(type, i, NULL, sa, &ssbuf,
1295 			    sizeof(ssbuf), credp);
1296 			break;
1297 
1298 		default:
1299 			break;
1300 		}
1301 
1302 		rtinfo->rti_addrs |= (1 << i);
1303 		dlen = sa->sa_len;
1304 		rlen = ROUNDUP32(dlen);
1305 		if (cp) {
1306 			bcopy((caddr_t)sa, cp, (size_t)dlen);
1307 			if (dlen != rlen) {
1308 				bzero(cp + dlen, rlen - dlen);
1309 			}
1310 			cp += rlen;
1311 		}
1312 		len += rlen;
1313 	}
1314 	if (cp == NULL && w != NULL && !second_time) {
1315 		struct walkarg *rw = w;
1316 
1317 		if (rw->w_req != NULL) {
1318 			if (rw->w_tmemsize < len) {
1319 				if (rw->w_tmem != NULL) {
1320 					kfree_data(rw->w_tmem, rw->w_tmemsize);
1321 				}
1322 				rw->w_tmem = (caddr_t) kalloc_data(len, Z_ZERO | Z_WAITOK);
1323 				if (rw->w_tmem != NULL) {
1324 					rw->w_tmemsize = len;
1325 				}
1326 			}
1327 			if (rw->w_tmem != NULL) {
1328 				cp = rw->w_tmem;
1329 				second_time = 1;
1330 				goto again;
1331 			}
1332 		}
1333 	}
1334 	if (cp) {
1335 		struct rt_msghdr *rtm = (struct rt_msghdr *)(void *)cp0;
1336 
1337 		rtm->rtm_version = RTM_VERSION;
1338 		rtm->rtm_type = type;
1339 		rtm->rtm_msglen = (u_short)len;
1340 	}
1341 	return len;
1342 }
1343 
1344 /*
1345  * This routine is called to generate a message from the routing
1346  * socket indicating that a redirect has occurred, a routing lookup
1347  * has failed, or that a protocol has detected timeouts to a particular
1348  * destination.
1349  */
1350 void
rt_missmsg(u_char type,struct rt_addrinfo * rtinfo,int flags,int error)1351 rt_missmsg(u_char type, struct rt_addrinfo *rtinfo, int flags, int error)
1352 {
1353 	struct rt_msghdr *rtm;
1354 	struct mbuf *m;
1355 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1356 	struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1357 
1358 	if (route_cb.any_count == 0) {
1359 		return;
1360 	}
1361 	m = rt_msg1(type, rtinfo);
1362 	if (m == NULL) {
1363 		return;
1364 	}
1365 	rtm = mtod(m, struct rt_msghdr *);
1366 	rtm->rtm_flags = RTF_DONE | flags;
1367 	rtm->rtm_errno = error;
1368 	rtm->rtm_addrs = rtinfo->rti_addrs;
1369 	route_proto.sp_family = sa ? sa->sa_family : 0;
1370 	raw_input(m, &route_proto, &route_src, &route_dst);
1371 }
1372 
1373 /*
1374  * This routine is called to generate a message from the routing
1375  * socket indicating that the status of a network interface has changed.
1376  */
1377 void
rt_ifmsg(struct ifnet * ifp)1378 rt_ifmsg(struct ifnet *ifp)
1379 {
1380 	struct if_msghdr *ifm;
1381 	struct mbuf *m;
1382 	struct rt_addrinfo info;
1383 	struct  sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1384 
1385 	if (route_cb.any_count == 0) {
1386 		return;
1387 	}
1388 	bzero((caddr_t)&info, sizeof(info));
1389 	m = rt_msg1(RTM_IFINFO, &info);
1390 	if (m == NULL) {
1391 		return;
1392 	}
1393 	ifm = mtod(m, struct if_msghdr *);
1394 	ifm->ifm_index = ifp->if_index;
1395 	ifm->ifm_flags = (u_short)ifp->if_flags;
1396 	if_data_internal_to_if_data(ifp, &ifp->if_data, &ifm->ifm_data);
1397 	ifm->ifm_addrs = 0;
1398 	raw_input(m, &route_proto, &route_src, &route_dst);
1399 }
1400 
1401 /*
1402  * This is called to generate messages from the routing socket
1403  * indicating a network interface has had addresses associated with it.
1404  * if we ever reverse the logic and replace messages TO the routing
1405  * socket indicate a request to configure interfaces, then it will
1406  * be unnecessary as the routing socket will automatically generate
1407  * copies of it.
1408  *
1409  * Since this is coming from the interface, it is expected that the
1410  * interface will be locked.  Caller must hold rnh_lock and rt_lock.
1411  */
1412 void
rt_newaddrmsg(u_char cmd,struct ifaddr * ifa,int error,struct rtentry * rt)1413 rt_newaddrmsg(u_char cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1414 {
1415 	struct rt_addrinfo info;
1416 	struct sockaddr *sa = 0;
1417 	int pass;
1418 	struct mbuf *m = 0;
1419 	struct ifnet *ifp = ifa->ifa_ifp;
1420 	struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1421 
1422 	LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_OWNED);
1423 	RT_LOCK_ASSERT_HELD(rt);
1424 
1425 	if (route_cb.any_count == 0) {
1426 		return;
1427 	}
1428 
1429 	/* Become a regular mutex, just in case */
1430 	RT_CONVERT_LOCK(rt);
1431 	for (pass = 1; pass < 3; pass++) {
1432 		bzero((caddr_t)&info, sizeof(info));
1433 		if ((cmd == RTM_ADD && pass == 1) ||
1434 		    (cmd == RTM_DELETE && pass == 2)) {
1435 			struct ifa_msghdr *ifam;
1436 			u_char ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1437 
1438 			/* Lock ifp for if_lladdr */
1439 			ifnet_lock_shared(ifp);
1440 			IFA_LOCK(ifa);
1441 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1442 			/*
1443 			 * Holding ifnet lock here prevents the link address
1444 			 * from changing contents, so no need to hold its
1445 			 * lock.  The link address is always present; it's
1446 			 * never freed.
1447 			 */
1448 			info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr;
1449 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1450 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1451 			if ((m = rt_msg1(ncmd, &info)) == NULL) {
1452 				IFA_UNLOCK(ifa);
1453 				ifnet_lock_done(ifp);
1454 				continue;
1455 			}
1456 			IFA_UNLOCK(ifa);
1457 			ifnet_lock_done(ifp);
1458 			ifam = mtod(m, struct ifa_msghdr *);
1459 			ifam->ifam_index = ifp->if_index;
1460 			IFA_LOCK_SPIN(ifa);
1461 			ifam->ifam_metric = ifa->ifa_metric;
1462 			ifam->ifam_flags = ifa->ifa_flags;
1463 			IFA_UNLOCK(ifa);
1464 			ifam->ifam_addrs = info.rti_addrs;
1465 		}
1466 		if ((cmd == RTM_ADD && pass == 2) ||
1467 		    (cmd == RTM_DELETE && pass == 1)) {
1468 			struct rt_msghdr *rtm;
1469 
1470 			if (rt == NULL) {
1471 				continue;
1472 			}
1473 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1474 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1475 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1476 			if ((m = rt_msg1(cmd, &info)) == NULL) {
1477 				continue;
1478 			}
1479 			rtm = mtod(m, struct rt_msghdr *);
1480 			rtm->rtm_index = ifp->if_index;
1481 			rtm->rtm_flags |= rt->rt_flags;
1482 			rtm->rtm_errno = error;
1483 			rtm->rtm_addrs = info.rti_addrs;
1484 		}
1485 		route_proto.sp_protocol = sa ? sa->sa_family : 0;
1486 		raw_input(m, &route_proto, &route_src, &route_dst);
1487 	}
1488 }
1489 
1490 /*
1491  * This is the analogue to the rt_newaddrmsg which performs the same
1492  * function but for multicast group memberhips.  This is easier since
1493  * there is no route state to worry about.
1494  */
1495 void
rt_newmaddrmsg(u_char cmd,struct ifmultiaddr * ifma)1496 rt_newmaddrmsg(u_char cmd, struct ifmultiaddr *ifma)
1497 {
1498 	struct rt_addrinfo info;
1499 	struct mbuf *m = 0;
1500 	struct ifnet *ifp = ifma->ifma_ifp;
1501 	struct ifma_msghdr *ifmam;
1502 	struct sockproto route_proto = { .sp_family = PF_ROUTE, .sp_protocol = 0 };
1503 
1504 	if (route_cb.any_count == 0) {
1505 		return;
1506 	}
1507 
1508 	/* Lock ifp for if_lladdr */
1509 	ifnet_lock_shared(ifp);
1510 	bzero((caddr_t)&info, sizeof(info));
1511 	IFMA_LOCK(ifma);
1512 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1513 	/* lladdr doesn't need lock */
1514 	info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr;
1515 
1516 	/*
1517 	 * If a link-layer address is present, present it as a ``gateway''
1518 	 * (similarly to how ARP entries, e.g., are presented).
1519 	 */
1520 	info.rti_info[RTAX_GATEWAY] = (ifma->ifma_ll != NULL) ?
1521 	    ifma->ifma_ll->ifma_addr : NULL;
1522 	if ((m = rt_msg1(cmd, &info)) == NULL) {
1523 		IFMA_UNLOCK(ifma);
1524 		ifnet_lock_done(ifp);
1525 		return;
1526 	}
1527 	ifmam = mtod(m, struct ifma_msghdr *);
1528 	ifmam->ifmam_index = ifp->if_index;
1529 	ifmam->ifmam_addrs = info.rti_addrs;
1530 	route_proto.sp_protocol = ifma->ifma_addr->sa_family;
1531 	IFMA_UNLOCK(ifma);
1532 	ifnet_lock_done(ifp);
1533 	raw_input(m, &route_proto, &route_src, &route_dst);
1534 }
1535 
1536 const char *
rtm2str(int cmd)1537 rtm2str(int cmd)
1538 {
1539 	const char *c = "RTM_?";
1540 
1541 	switch (cmd) {
1542 	case RTM_ADD:
1543 		c = "RTM_ADD";
1544 		break;
1545 	case RTM_DELETE:
1546 		c = "RTM_DELETE";
1547 		break;
1548 	case RTM_CHANGE:
1549 		c = "RTM_CHANGE";
1550 		break;
1551 	case RTM_GET:
1552 		c = "RTM_GET";
1553 		break;
1554 	case RTM_LOSING:
1555 		c = "RTM_LOSING";
1556 		break;
1557 	case RTM_REDIRECT:
1558 		c = "RTM_REDIRECT";
1559 		break;
1560 	case RTM_MISS:
1561 		c = "RTM_MISS";
1562 		break;
1563 	case RTM_LOCK:
1564 		c = "RTM_LOCK";
1565 		break;
1566 	case RTM_OLDADD:
1567 		c = "RTM_OLDADD";
1568 		break;
1569 	case RTM_OLDDEL:
1570 		c = "RTM_OLDDEL";
1571 		break;
1572 	case RTM_RESOLVE:
1573 		c = "RTM_RESOLVE";
1574 		break;
1575 	case RTM_NEWADDR:
1576 		c = "RTM_NEWADDR";
1577 		break;
1578 	case RTM_DELADDR:
1579 		c = "RTM_DELADDR";
1580 		break;
1581 	case RTM_IFINFO:
1582 		c = "RTM_IFINFO";
1583 		break;
1584 	case RTM_NEWMADDR:
1585 		c = "RTM_NEWMADDR";
1586 		break;
1587 	case RTM_DELMADDR:
1588 		c = "RTM_DELMADDR";
1589 		break;
1590 	case RTM_GET_SILENT:
1591 		c = "RTM_GET_SILENT";
1592 		break;
1593 	case RTM_IFINFO2:
1594 		c = "RTM_IFINFO2";
1595 		break;
1596 	case RTM_NEWMADDR2:
1597 		c = "RTM_NEWMADDR2";
1598 		break;
1599 	case RTM_GET2:
1600 		c = "RTM_GET2";
1601 		break;
1602 	case RTM_GET_EXT:
1603 		c = "RTM_GET_EXT";
1604 		break;
1605 	}
1606 
1607 	return c;
1608 }
1609 
1610 /*
1611  * This is used in dumping the kernel table via sysctl().
1612  */
1613 static int
sysctl_dumpentry(struct radix_node * rn,void * vw)1614 sysctl_dumpentry(struct radix_node *rn, void *vw)
1615 {
1616 	struct walkarg *w = vw;
1617 	struct rtentry *rt = (struct rtentry *)rn;
1618 	int error = 0, size;
1619 	struct rt_addrinfo info;
1620 	kauth_cred_t cred;
1621 	kauth_cred_t *credp;
1622 
1623 	cred = kauth_cred_proc_ref(current_proc());
1624 	credp = &cred;
1625 
1626 	RT_LOCK(rt);
1627 	if ((w->w_op == NET_RT_FLAGS || w->w_op == NET_RT_FLAGS_PRIV) &&
1628 	    !(rt->rt_flags & w->w_arg)) {
1629 		goto done;
1630 	}
1631 
1632 	/*
1633 	 * If the matching route has RTF_LLINFO set, then we can skip scrubbing the MAC
1634 	 * only if the outgoing interface is not loopback and the process has entitlement
1635 	 * for neighbor cache read.
1636 	 */
1637 	if (w->w_op == NET_RT_FLAGS_PRIV && (rt->rt_flags & RTF_LLINFO)) {
1638 		if (rt->rt_ifp != lo_ifp &&
1639 		    (route_op_entitlement_check(NULL, cred, ROUTE_OP_READ, TRUE) == 0)) {
1640 			credp = NULL;
1641 		}
1642 	}
1643 
1644 	bzero((caddr_t)&info, sizeof(info));
1645 	info.rti_info[RTAX_DST] = rt_key(rt);
1646 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1647 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1648 	info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1649 	if (RT_HAS_IFADDR(rt)) {
1650 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1651 	}
1652 
1653 	if (w->w_op != NET_RT_DUMP2) {
1654 		size = rt_msg2(RTM_GET, &info, NULL, w, credp);
1655 		if (w->w_req != NULL && w->w_tmem != NULL) {
1656 			struct rt_msghdr *rtm =
1657 			    (struct rt_msghdr *)(void *)w->w_tmem;
1658 
1659 			rtm->rtm_flags = rt->rt_flags;
1660 			rtm->rtm_use = rt->rt_use;
1661 			rt_getmetrics(rt, &rtm->rtm_rmx);
1662 			rtm->rtm_index = rt->rt_ifp->if_index;
1663 			rtm->rtm_pid = 0;
1664 			rtm->rtm_seq = 0;
1665 			rtm->rtm_errno = 0;
1666 			rtm->rtm_addrs = info.rti_addrs;
1667 			error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1668 		}
1669 	} else {
1670 		size = rt_msg2(RTM_GET2, &info, NULL, w, credp);
1671 		if (w->w_req != NULL && w->w_tmem != NULL) {
1672 			struct rt_msghdr2 *rtm =
1673 			    (struct rt_msghdr2 *)(void *)w->w_tmem;
1674 
1675 			rtm->rtm_flags = rt->rt_flags;
1676 			rtm->rtm_use = rt->rt_use;
1677 			rt_getmetrics(rt, &rtm->rtm_rmx);
1678 			rtm->rtm_index = rt->rt_ifp->if_index;
1679 			rtm->rtm_refcnt = rt->rt_refcnt;
1680 			if (rt->rt_parent) {
1681 				rtm->rtm_parentflags = rt->rt_parent->rt_flags;
1682 			} else {
1683 				rtm->rtm_parentflags = 0;
1684 			}
1685 			rtm->rtm_reserved = 0;
1686 			rtm->rtm_addrs = info.rti_addrs;
1687 			error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1688 		}
1689 	}
1690 
1691 done:
1692 	RT_UNLOCK(rt);
1693 	kauth_cred_unref(&cred);
1694 	return error;
1695 }
1696 
1697 /*
1698  * This is used for dumping extended information from route entries.
1699  */
1700 static int
sysctl_dumpentry_ext(struct radix_node * rn,void * vw)1701 sysctl_dumpentry_ext(struct radix_node *rn, void *vw)
1702 {
1703 	struct walkarg *w = vw;
1704 	struct rtentry *rt = (struct rtentry *)rn;
1705 	int error = 0, size;
1706 	struct rt_addrinfo info;
1707 	kauth_cred_t cred;
1708 
1709 	cred = kauth_cred_proc_ref(current_proc());
1710 
1711 	RT_LOCK(rt);
1712 	if (w->w_op == NET_RT_DUMPX_FLAGS && !(rt->rt_flags & w->w_arg)) {
1713 		goto done;
1714 	}
1715 	bzero(&info, sizeof(info));
1716 	info.rti_info[RTAX_DST] = rt_key(rt);
1717 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1718 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1719 	info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1720 
1721 	size = rt_msg2(RTM_GET_EXT, &info, NULL, w, &cred);
1722 	if (w->w_req != NULL && w->w_tmem != NULL) {
1723 		struct rt_msghdr_ext *ertm =
1724 		    (struct rt_msghdr_ext *)(void *)w->w_tmem;
1725 
1726 		ertm->rtm_flags = rt->rt_flags;
1727 		ertm->rtm_use = rt->rt_use;
1728 		rt_getmetrics(rt, &ertm->rtm_rmx);
1729 		ertm->rtm_index = rt->rt_ifp->if_index;
1730 		ertm->rtm_pid = 0;
1731 		ertm->rtm_seq = 0;
1732 		ertm->rtm_errno = 0;
1733 		ertm->rtm_addrs = info.rti_addrs;
1734 		if (rt->rt_llinfo_get_ri == NULL) {
1735 			bzero(&ertm->rtm_ri, sizeof(ertm->rtm_ri));
1736 			ertm->rtm_ri.ri_rssi = IFNET_RSSI_UNKNOWN;
1737 			ertm->rtm_ri.ri_lqm = IFNET_LQM_THRESH_OFF;
1738 			ertm->rtm_ri.ri_npm = IFNET_NPM_THRESH_UNKNOWN;
1739 		} else {
1740 			rt->rt_llinfo_get_ri(rt, &ertm->rtm_ri);
1741 		}
1742 		error = SYSCTL_OUT(w->w_req, (caddr_t)ertm, size);
1743 	}
1744 
1745 done:
1746 	RT_UNLOCK(rt);
1747 	kauth_cred_unref(&cred);
1748 	return error;
1749 }
1750 
1751 /*
1752  * rdar://9307819
1753  * To avoid to call copyout() while holding locks and to cause problems
1754  * in the paging path, sysctl_iflist() and sysctl_iflist2() contstruct
1755  * the list in two passes. In the first pass we compute the total
1756  * length of the data we are going to copyout, then we release
1757  * all locks to allocate a temporary buffer that gets filled
1758  * in the second pass.
1759  *
1760  * Note that we are verifying the assumption that kalloc() returns a buffer
1761  * that is at least 32 bits aligned and that the messages and addresses are
1762  * 32 bits aligned.
1763  */
1764 static int
sysctl_iflist(int af,struct walkarg * w)1765 sysctl_iflist(int af, struct walkarg *w)
1766 {
1767 	struct ifnet *ifp;
1768 	struct ifaddr *ifa;
1769 	struct  rt_addrinfo info;
1770 	int     error = 0;
1771 	int     pass = 0;
1772 	size_t  len = 0, total_len = 0, total_buffer_len = 0, current_len = 0;
1773 	char    *total_buffer = NULL, *cp = NULL;
1774 	kauth_cred_t cred;
1775 
1776 	cred = kauth_cred_proc_ref(current_proc());
1777 
1778 	bzero((caddr_t)&info, sizeof(info));
1779 
1780 	for (pass = 0; pass < 2; pass++) {
1781 		ifnet_head_lock_shared();
1782 
1783 		TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1784 			if (error) {
1785 				break;
1786 			}
1787 			if (w->w_arg && w->w_arg != ifp->if_index) {
1788 				continue;
1789 			}
1790 			ifnet_lock_shared(ifp);
1791 			/*
1792 			 * Holding ifnet lock here prevents the link address
1793 			 * from changing contents, so no need to hold the ifa
1794 			 * lock.  The link address is always present; it's
1795 			 * never freed.
1796 			 */
1797 			ifa = ifp->if_lladdr;
1798 			info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1799 			len = rt_msg2(RTM_IFINFO, &info, NULL, NULL, &cred);
1800 			if (pass == 0) {
1801 				if (os_add_overflow(total_len, len, &total_len)) {
1802 					ifnet_lock_done(ifp);
1803 					error = ENOBUFS;
1804 					break;
1805 				}
1806 			} else {
1807 				struct if_msghdr *ifm;
1808 
1809 				if (current_len + len > total_len) {
1810 					ifnet_lock_done(ifp);
1811 					error = ENOBUFS;
1812 					break;
1813 				}
1814 				info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1815 				len = rt_msg2(RTM_IFINFO, &info,
1816 				    (caddr_t)cp, NULL, &cred);
1817 				info.rti_info[RTAX_IFP] = NULL;
1818 
1819 				ifm = (struct if_msghdr *)(void *)cp;
1820 				ifm->ifm_index = ifp->if_index;
1821 				ifm->ifm_flags = (u_short)ifp->if_flags;
1822 				if_data_internal_to_if_data(ifp, &ifp->if_data,
1823 				    &ifm->ifm_data);
1824 				ifm->ifm_addrs = info.rti_addrs;
1825 				/*
1826 				 * <rdar://problem/32940901>
1827 				 * Round bytes only for non-platform
1828 				 */
1829 				if (!csproc_get_platform_binary(w->w_req->p)) {
1830 					ALIGN_BYTES(ifm->ifm_data.ifi_ibytes);
1831 					ALIGN_BYTES(ifm->ifm_data.ifi_obytes);
1832 				}
1833 
1834 				cp += len;
1835 				VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1836 				current_len += len;
1837 				VERIFY(current_len <= total_len);
1838 			}
1839 			while ((ifa = ifa->ifa_link.tqe_next) != NULL) {
1840 				IFA_LOCK(ifa);
1841 				if (af && af != ifa->ifa_addr->sa_family) {
1842 					IFA_UNLOCK(ifa);
1843 					continue;
1844 				}
1845 				if (ifa->ifa_addr->sa_family == AF_INET6 &&
1846 				    (((struct in6_ifaddr *)ifa)->ia6_flags &
1847 				    IN6_IFF_CLAT46) != 0) {
1848 					IFA_UNLOCK(ifa);
1849 					continue;
1850 				}
1851 				info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1852 				info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1853 				info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1854 				len = rt_msg2(RTM_NEWADDR, &info, NULL, NULL,
1855 				    &cred);
1856 				if (pass == 0) {
1857 					if (os_add_overflow(total_len, len, &total_len)) {
1858 						IFA_UNLOCK(ifa);
1859 						error = ENOBUFS;
1860 						break;
1861 					}
1862 				} else {
1863 					struct ifa_msghdr *ifam;
1864 
1865 					if (current_len + len > total_len) {
1866 						IFA_UNLOCK(ifa);
1867 						error = ENOBUFS;
1868 						break;
1869 					}
1870 					len = rt_msg2(RTM_NEWADDR, &info,
1871 					    (caddr_t)cp, NULL, &cred);
1872 
1873 					ifam = (struct ifa_msghdr *)(void *)cp;
1874 					ifam->ifam_index =
1875 					    ifa->ifa_ifp->if_index;
1876 					ifam->ifam_flags = ifa->ifa_flags;
1877 					ifam->ifam_metric = ifa->ifa_metric;
1878 					ifam->ifam_addrs = info.rti_addrs;
1879 
1880 					cp += len;
1881 					VERIFY(IS_P2ALIGNED(cp,
1882 					    sizeof(u_int32_t)));
1883 					current_len += len;
1884 					VERIFY(current_len <= total_len);
1885 				}
1886 				IFA_UNLOCK(ifa);
1887 			}
1888 			ifnet_lock_done(ifp);
1889 			info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1890 			    info.rti_info[RTAX_BRD] = NULL;
1891 		}
1892 
1893 		ifnet_head_done();
1894 
1895 		if (error != 0) {
1896 			if (error == ENOBUFS) {
1897 				printf("%s: current_len (%lu) + len (%lu) > "
1898 				    "total_len (%lu)\n", __func__, current_len,
1899 				    len, total_len);
1900 			}
1901 			break;
1902 		}
1903 
1904 		if (pass == 0) {
1905 			/* Better to return zero length buffer than ENOBUFS */
1906 			if (total_len == 0) {
1907 				total_len = 1;
1908 			}
1909 			total_len += total_len >> 3;
1910 			total_buffer_len = total_len;
1911 			total_buffer = (char *) kalloc_data(total_len, Z_ZERO | Z_WAITOK);
1912 			if (total_buffer == NULL) {
1913 				printf("%s: kalloc_data(%lu) failed\n", __func__,
1914 				    total_len);
1915 				error = ENOBUFS;
1916 				break;
1917 			}
1918 			cp = total_buffer;
1919 			VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1920 		} else {
1921 			error = SYSCTL_OUT(w->w_req, total_buffer, current_len);
1922 			if (error) {
1923 				break;
1924 			}
1925 		}
1926 	}
1927 
1928 	if (total_buffer != NULL) {
1929 		kfree_data(total_buffer, total_buffer_len);
1930 	}
1931 
1932 	kauth_cred_unref(&cred);
1933 	return error;
1934 }
1935 
1936 static int
sysctl_iflist2(int af,struct walkarg * w)1937 sysctl_iflist2(int af, struct walkarg *w)
1938 {
1939 	struct ifnet *ifp;
1940 	struct ifaddr *ifa;
1941 	struct  rt_addrinfo info;
1942 	int     error = 0;
1943 	int     pass = 0;
1944 	size_t  len = 0, total_len = 0, total_buffer_len = 0, current_len = 0;
1945 	char    *total_buffer = NULL, *cp = NULL;
1946 	kauth_cred_t cred;
1947 
1948 	cred = kauth_cred_proc_ref(current_proc());
1949 
1950 	bzero((caddr_t)&info, sizeof(info));
1951 
1952 	for (pass = 0; pass < 2; pass++) {
1953 		struct ifmultiaddr *ifma;
1954 
1955 		ifnet_head_lock_shared();
1956 
1957 		TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1958 			if (error) {
1959 				break;
1960 			}
1961 			if (w->w_arg && w->w_arg != ifp->if_index) {
1962 				continue;
1963 			}
1964 			ifnet_lock_shared(ifp);
1965 			/*
1966 			 * Holding ifnet lock here prevents the link address
1967 			 * from changing contents, so no need to hold the ifa
1968 			 * lock.  The link address is always present; it's
1969 			 * never freed.
1970 			 */
1971 			ifa = ifp->if_lladdr;
1972 			info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1973 			len = rt_msg2(RTM_IFINFO2, &info, NULL, NULL, &cred);
1974 			if (pass == 0) {
1975 				if (os_add_overflow(total_len, len, &total_len)) {
1976 					ifnet_lock_done(ifp);
1977 					error = ENOBUFS;
1978 					break;
1979 				}
1980 			} else {
1981 				struct if_msghdr2 *ifm;
1982 
1983 				if (current_len + len > total_len) {
1984 					ifnet_lock_done(ifp);
1985 					error = ENOBUFS;
1986 					break;
1987 				}
1988 				info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1989 				len = rt_msg2(RTM_IFINFO2, &info,
1990 				    (caddr_t)cp, NULL, &cred);
1991 				info.rti_info[RTAX_IFP] = NULL;
1992 
1993 				ifm = (struct if_msghdr2 *)(void *)cp;
1994 				ifm->ifm_addrs = info.rti_addrs;
1995 				ifm->ifm_flags = (u_short)ifp->if_flags;
1996 				ifm->ifm_index = ifp->if_index;
1997 				ifm->ifm_snd_len = IFCQ_LEN(ifp->if_snd);
1998 				ifm->ifm_snd_maxlen = IFCQ_MAXLEN(ifp->if_snd);
1999 				ifm->ifm_snd_drops =
2000 				    (int)ifp->if_snd->ifcq_dropcnt.packets;
2001 				ifm->ifm_timer = ifp->if_timer;
2002 				if_data_internal_to_if_data64(ifp,
2003 				    &ifp->if_data, &ifm->ifm_data);
2004 				/*
2005 				 * <rdar://problem/32940901>
2006 				 * Round bytes only for non-platform
2007 				 */
2008 				if (!csproc_get_platform_binary(w->w_req->p)) {
2009 					ALIGN_BYTES(ifm->ifm_data.ifi_ibytes);
2010 					ALIGN_BYTES(ifm->ifm_data.ifi_obytes);
2011 				}
2012 
2013 				cp += len;
2014 				VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
2015 				current_len += len;
2016 				VERIFY(current_len <= total_len);
2017 			}
2018 			while ((ifa = ifa->ifa_link.tqe_next) != NULL) {
2019 				IFA_LOCK(ifa);
2020 				if (af && af != ifa->ifa_addr->sa_family) {
2021 					IFA_UNLOCK(ifa);
2022 					continue;
2023 				}
2024 				if (ifa->ifa_addr->sa_family == AF_INET6 &&
2025 				    (((struct in6_ifaddr *)ifa)->ia6_flags &
2026 				    IN6_IFF_CLAT46) != 0) {
2027 					IFA_UNLOCK(ifa);
2028 					continue;
2029 				}
2030 
2031 				info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2032 				info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
2033 				info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2034 				len = rt_msg2(RTM_NEWADDR, &info, NULL, NULL,
2035 				    &cred);
2036 				if (pass == 0) {
2037 					if (os_add_overflow(total_len, len, &total_len)) {
2038 						IFA_UNLOCK(ifa);
2039 						error = ENOBUFS;
2040 						break;
2041 					}
2042 				} else {
2043 					struct ifa_msghdr *ifam;
2044 
2045 					if (current_len + len > total_len) {
2046 						IFA_UNLOCK(ifa);
2047 						error = ENOBUFS;
2048 						break;
2049 					}
2050 					len = rt_msg2(RTM_NEWADDR, &info,
2051 					    (caddr_t)cp, NULL, &cred);
2052 
2053 					ifam = (struct ifa_msghdr *)(void *)cp;
2054 					ifam->ifam_index =
2055 					    ifa->ifa_ifp->if_index;
2056 					ifam->ifam_flags = ifa->ifa_flags;
2057 					ifam->ifam_metric = ifa->ifa_metric;
2058 					ifam->ifam_addrs = info.rti_addrs;
2059 
2060 					cp += len;
2061 					VERIFY(IS_P2ALIGNED(cp,
2062 					    sizeof(u_int32_t)));
2063 					current_len += len;
2064 					VERIFY(current_len <= total_len);
2065 				}
2066 				IFA_UNLOCK(ifa);
2067 			}
2068 			if (error) {
2069 				ifnet_lock_done(ifp);
2070 				break;
2071 			}
2072 
2073 			for (ifma = LIST_FIRST(&ifp->if_multiaddrs);
2074 			    ifma != NULL; ifma = LIST_NEXT(ifma, ifma_link)) {
2075 				struct ifaddr *ifa0;
2076 
2077 				IFMA_LOCK(ifma);
2078 				if (af && af != ifma->ifma_addr->sa_family) {
2079 					IFMA_UNLOCK(ifma);
2080 					continue;
2081 				}
2082 				bzero((caddr_t)&info, sizeof(info));
2083 				info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2084 				/*
2085 				 * Holding ifnet lock here prevents the link
2086 				 * address from changing contents, so no need
2087 				 * to hold the ifa0 lock.  The link address is
2088 				 * always present; it's never freed.
2089 				 */
2090 				ifa0 = ifp->if_lladdr;
2091 				info.rti_info[RTAX_IFP] = ifa0->ifa_addr;
2092 				if (ifma->ifma_ll != NULL) {
2093 					info.rti_info[RTAX_GATEWAY] =
2094 					    ifma->ifma_ll->ifma_addr;
2095 				}
2096 				len = rt_msg2(RTM_NEWMADDR2, &info, NULL, NULL,
2097 				    &cred);
2098 				if (pass == 0) {
2099 					total_len += len;
2100 				} else {
2101 					struct ifma_msghdr2 *ifmam;
2102 
2103 					if (current_len + len > total_len) {
2104 						IFMA_UNLOCK(ifma);
2105 						error = ENOBUFS;
2106 						break;
2107 					}
2108 					len = rt_msg2(RTM_NEWMADDR2, &info,
2109 					    (caddr_t)cp, NULL, &cred);
2110 
2111 					ifmam =
2112 					    (struct ifma_msghdr2 *)(void *)cp;
2113 					ifmam->ifmam_addrs = info.rti_addrs;
2114 					ifmam->ifmam_flags = 0;
2115 					ifmam->ifmam_index =
2116 					    ifma->ifma_ifp->if_index;
2117 					ifmam->ifmam_refcount =
2118 					    ifma->ifma_reqcnt;
2119 
2120 					cp += len;
2121 					VERIFY(IS_P2ALIGNED(cp,
2122 					    sizeof(u_int32_t)));
2123 					current_len += len;
2124 				}
2125 				IFMA_UNLOCK(ifma);
2126 			}
2127 			ifnet_lock_done(ifp);
2128 			info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
2129 			    info.rti_info[RTAX_BRD] = NULL;
2130 		}
2131 		ifnet_head_done();
2132 
2133 		if (error) {
2134 			if (error == ENOBUFS) {
2135 				printf("%s: current_len (%lu) + len (%lu) > "
2136 				    "total_len (%lu)\n", __func__, current_len,
2137 				    len, total_len);
2138 			}
2139 			break;
2140 		}
2141 
2142 		if (pass == 0) {
2143 			/* Better to return zero length buffer than ENOBUFS */
2144 			if (total_len == 0) {
2145 				total_len = 1;
2146 			}
2147 			total_len += total_len >> 3;
2148 			total_buffer_len = total_len;
2149 			total_buffer = (char *) kalloc_data(total_len, Z_ZERO | Z_WAITOK);
2150 			if (total_buffer == NULL) {
2151 				printf("%s: kalloc_data(%lu) failed\n", __func__,
2152 				    total_len);
2153 				error = ENOBUFS;
2154 				break;
2155 			}
2156 			cp = total_buffer;
2157 			VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
2158 		} else {
2159 			error = SYSCTL_OUT(w->w_req, total_buffer, current_len);
2160 			if (error) {
2161 				break;
2162 			}
2163 		}
2164 	}
2165 
2166 	if (total_buffer != NULL) {
2167 		kfree_data(total_buffer, total_buffer_len);
2168 	}
2169 
2170 	kauth_cred_unref(&cred);
2171 	return error;
2172 }
2173 
2174 
2175 static int
sysctl_rtstat(struct sysctl_req * req)2176 sysctl_rtstat(struct sysctl_req *req)
2177 {
2178 	return SYSCTL_OUT(req, &rtstat, sizeof(struct rtstat));
2179 }
2180 
2181 static int
sysctl_rttrash(struct sysctl_req * req)2182 sysctl_rttrash(struct sysctl_req *req)
2183 {
2184 	return SYSCTL_OUT(req, &rttrash, sizeof(rttrash));
2185 }
2186 
2187 static int
2188 sysctl_rtsock SYSCTL_HANDLER_ARGS
2189 {
2190 #pragma unused(oidp)
2191 	int     *name = (int *)arg1;
2192 	u_int   namelen = arg2;
2193 	struct radix_node_head *rnh;
2194 	int     i, error = EINVAL;
2195 	u_char  af;
2196 	struct  walkarg w;
2197 
2198 	name++;
2199 	namelen--;
2200 	if (req->newptr) {
2201 		return EPERM;
2202 	}
2203 	if (namelen != 3) {
2204 		return EINVAL;
2205 	}
2206 	af = (u_char)name[0];
2207 	Bzero(&w, sizeof(w));
2208 	w.w_op = name[1];
2209 	w.w_arg = name[2];
2210 	w.w_req = req;
2211 
2212 	switch (w.w_op) {
2213 	case NET_RT_DUMP:
2214 	case NET_RT_DUMP2:
2215 	case NET_RT_FLAGS:
2216 	case NET_RT_FLAGS_PRIV:
2217 		lck_mtx_lock(rnh_lock);
2218 		for (i = 1; i <= AF_MAX; i++) {
2219 			if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
2220 			    (error = rnh->rnh_walktree(rnh,
2221 			    sysctl_dumpentry, &w))) {
2222 				break;
2223 			}
2224 		}
2225 		lck_mtx_unlock(rnh_lock);
2226 		break;
2227 	case NET_RT_DUMPX:
2228 	case NET_RT_DUMPX_FLAGS:
2229 		lck_mtx_lock(rnh_lock);
2230 		for (i = 1; i <= AF_MAX; i++) {
2231 			if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
2232 			    (error = rnh->rnh_walktree(rnh,
2233 			    sysctl_dumpentry_ext, &w))) {
2234 				break;
2235 			}
2236 		}
2237 		lck_mtx_unlock(rnh_lock);
2238 		break;
2239 	case NET_RT_IFLIST:
2240 		error = sysctl_iflist(af, &w);
2241 		break;
2242 	case NET_RT_IFLIST2:
2243 		error = sysctl_iflist2(af, &w);
2244 		break;
2245 	case NET_RT_STAT:
2246 		error = sysctl_rtstat(req);
2247 		break;
2248 	case NET_RT_TRASH:
2249 		error = sysctl_rttrash(req);
2250 		break;
2251 	}
2252 	if (w.w_tmem != NULL) {
2253 		kfree_data(w.w_tmem, w.w_tmemsize);
2254 	}
2255 	return error;
2256 }
2257 
2258 /*
2259  * Definitions of protocols supported in the ROUTE domain.
2260  */
2261 static struct protosw routesw[] = {
2262 	{
2263 		.pr_type =              SOCK_RAW,
2264 		.pr_protocol =          0,
2265 		.pr_flags =             PR_ATOMIC | PR_ADDR,
2266 		.pr_output =            route_output,
2267 		.pr_ctlinput =          raw_ctlinput,
2268 		.pr_usrreqs =           &route_usrreqs,
2269 	}
2270 };
2271 
2272 static int route_proto_count = (sizeof(routesw) / sizeof(struct protosw));
2273 
2274 struct domain routedomain_s = {
2275 	.dom_family =           PF_ROUTE,
2276 	.dom_name =             "route",
2277 	.dom_init =             route_dinit,
2278 };
2279 
2280 static void
route_dinit(struct domain * dp)2281 route_dinit(struct domain *dp)
2282 {
2283 	struct protosw *pr;
2284 	int i;
2285 
2286 	VERIFY(!(dp->dom_flags & DOM_INITIALIZED));
2287 	VERIFY(routedomain == NULL);
2288 
2289 	routedomain = dp;
2290 
2291 	for (i = 0, pr = &routesw[0]; i < route_proto_count; i++, pr++) {
2292 		net_add_proto(pr, dp, 1);
2293 	}
2294 
2295 	route_init();
2296 }
2297