xref: /xnu-8020.140.41/bsd/net/if_stf.c (revision 27b03b360a988dfd3dfdf34262bb0042026747cc)
1 /*
2  * Copyright (c) 2000-2020 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 /*	$FreeBSD: src/sys/net/if_stf.c,v 1.1.2.6 2001/07/24 19:10:18 brooks Exp $	*/
30 /*	$KAME: if_stf.c,v 1.62 2001/06/07 22:32:16 itojun Exp $	*/
31 
32 /*
33  * Copyright (C) 2000 WIDE Project.
34  * All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the project 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 PROJECT 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 PROJECT 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 /*
61  * NOTICE: This file was modified by SPARTA, Inc. in 2006 to introduce
62  * support for mandatory and extensible security protections.  This notice
63  * is included in support of clause 2.2 (b) of the Apple Public License,
64  * Version 2.0.
65  */
66 
67 /*
68  * 6to4 interface, based on RFC3056.
69  *
70  * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
71  * There is no address mapping defined from IPv6 multicast address to IPv4
72  * address.  Therefore, we do not have IFF_MULTICAST on the interface.
73  *
74  * Due to the lack of address mapping for link-local addresses, we cannot
75  * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
76  * packets to link-local multicast addresses (ff02::x).
77  *
78  * Here are interesting symptoms due to the lack of link-local address:
79  *
80  * Unicast routing exchange:
81  * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
82  *   and link-local addresses as nexthop.
83  * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
84  *   assigned to the link, and makes use of them.  Also, HELLO packets use
85  *   link-local multicast addresses (ff02::5 and ff02::6).
86  * - BGP4+: Maybe.  You can only use global address as nexthop, and global
87  *   address as TCP endpoint address.
88  *
89  * Multicast routing protocols:
90  * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
91  *   Adjacent PIM routers must be configured manually (is it really spec-wise
92  *   correct thing to do?).
93  *
94  * ICMPv6:
95  * - Redirects cannot be used due to the lack of link-local address.
96  *
97  * stf interface does not have, and will not need, a link-local address.
98  * It seems to have no real benefit and does not help the above symptoms much.
99  * Even if we assign link-locals to interface, we cannot really
100  * use link-local unicast/multicast on top of 6to4 cloud (since there's no
101  * encapsulation defined for link-local address), and the above analysis does
102  * not change.  RFC3056 does not mandate the assignment of link-local address
103  * either.
104  *
105  * 6to4 interface has security issues.  Refer to
106  * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
107  * for details.  The code tries to filter out some of malicious packets.
108  * Note that there is no way to be 100% secure.
109  */
110 
111 #include <sys/param.h>
112 #include <sys/systm.h>
113 #include <sys/socket.h>
114 #include <sys/sockio.h>
115 #include <sys/mbuf.h>
116 #include <sys/errno.h>
117 #include <sys/protosw.h>
118 #include <sys/kernel.h>
119 #include <sys/syslog.h>
120 
121 #include <sys/malloc.h>
122 
123 #include <kern/locks.h>
124 
125 #include <net/if.h>
126 #include <net/route.h>
127 #include <net/if_types.h>
128 
129 #include <netinet/in.h>
130 #include <netinet/in_systm.h>
131 #include <netinet/ip.h>
132 #include <netinet/ip_var.h>
133 #include <netinet/in_var.h>
134 
135 #include <netinet/ip6.h>
136 #include <netinet6/ip6_var.h>
137 #include <netinet6/in6_var.h>
138 #include <netinet/ip_ecn.h>
139 
140 #include <netinet/ip_encap.h>
141 #include <net/kpi_interface.h>
142 #include <net/kpi_protocol.h>
143 
144 
145 #include <net/net_osdep.h>
146 
147 #include <net/bpf.h>
148 
149 #define GET_V4(x) ((const struct in_addr *)(const void *)(&(x)->s6_addr16[1]))
150 
151 static LCK_GRP_DECLARE(stf_mtx_grp, "stf");
152 
153 struct stf_softc {
154 	ifnet_t                         sc_if;     /* common area */
155 	u_int32_t                               sc_protocol_family; /* dlil protocol attached */
156 	union {
157 		struct route  __sc_ro4;
158 		struct route_in6 __sc_ro6; /* just for safety */
159 	} __sc_ro46;
160 #define sc_ro   __sc_ro46.__sc_ro4
161 	decl_lck_mtx_data(, sc_ro_mtx);
162 	const struct encaptab *encap_cookie;
163 	bpf_tap_mode            tap_mode;
164 	bpf_packet_func         tap_callback;
165 };
166 
167 void stfattach(void);
168 
169 static int ip_stf_ttl = 40;
170 
171 static void in_stf_input(struct mbuf *, int);
172 
173 static struct protosw in_stf_protosw =
174 {
175 	.pr_type =              SOCK_RAW,
176 	.pr_protocol =          IPPROTO_IPV6,
177 	.pr_flags =             PR_ATOMIC | PR_ADDR,
178 	.pr_input =             in_stf_input,
179 	.pr_ctloutput =         rip_ctloutput,
180 	.pr_usrreqs =           &rip_usrreqs,
181 	.pr_unlock =            rip_unlock,
182 };
183 
184 static int stf_encapcheck(const struct mbuf *, int, int, void *);
185 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
186 int stf_pre_output(struct ifnet *, protocol_family_t, struct mbuf **,
187     const struct sockaddr *, void *, char *, char *);
188 static int stf_checkaddr4(struct stf_softc *, const struct in_addr *,
189     struct ifnet *);
190 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
191     struct ifnet *);
192 static void stf_rtrequest(int, struct rtentry *, struct sockaddr *);
193 static errno_t stf_ioctl(ifnet_t ifp, u_long cmd, void *data);
194 static errno_t stf_output(ifnet_t ifp, mbuf_t m);
195 
196 /*
197  * gif_input is the input handler for IP and IPv6 attached to gif
198  */
199 static errno_t
stf_media_input(__unused ifnet_t ifp,protocol_family_t protocol_family,mbuf_t m,__unused char * frame_header)200 stf_media_input(
201 	__unused ifnet_t        ifp,
202 	protocol_family_t       protocol_family,
203 	mbuf_t                          m,
204 	__unused char           *frame_header)
205 {
206 	if (proto_input(protocol_family, m) != 0) {
207 		m_freem(m);
208 	}
209 
210 	return 0;
211 }
212 
213 
214 
215 static errno_t
stf_add_proto(ifnet_t ifp,protocol_family_t protocol_family,__unused const struct ifnet_demux_desc * demux_array,__unused u_int32_t demux_count)216 stf_add_proto(
217 	ifnet_t                                                                 ifp,
218 	protocol_family_t                                               protocol_family,
219 	__unused const struct ifnet_demux_desc  *demux_array,
220 	__unused u_int32_t                                              demux_count)
221 {
222 	/* Only one protocol may be attached at a time */
223 	struct stf_softc* stf = ifnet_softc(ifp);
224 	if (stf->sc_protocol_family == 0) {
225 		stf->sc_protocol_family = protocol_family;
226 	} else {
227 		printf("stf_add_proto: stf already has a proto\n");
228 		return EBUSY;
229 	}
230 
231 	return 0;
232 }
233 
234 static errno_t
stf_del_proto(ifnet_t ifp,protocol_family_t protocol_family)235 stf_del_proto(
236 	ifnet_t                         ifp,
237 	protocol_family_t       protocol_family)
238 {
239 	if (((struct stf_softc*)ifnet_softc(ifp))->sc_protocol_family == protocol_family) {
240 		((struct stf_softc*)ifnet_softc(ifp))->sc_protocol_family = 0;
241 	}
242 
243 	return 0;
244 }
245 
246 static errno_t
stf_attach_inet6(ifnet_t ifp,protocol_family_t protocol_family)247 stf_attach_inet6(
248 	ifnet_t                         ifp,
249 	protocol_family_t       protocol_family)
250 {
251 	struct ifnet_attach_proto_param     reg;
252 	errno_t                                                     stat;
253 
254 	if (protocol_family != PF_INET6) {
255 		return EPROTONOSUPPORT;
256 	}
257 
258 	bzero(&reg, sizeof(reg));
259 	reg.input = stf_media_input;
260 	reg.pre_output = stf_pre_output;
261 
262 	stat = ifnet_attach_protocol(ifp, protocol_family, &reg);
263 	if (stat && stat != EEXIST) {
264 		printf("stf_attach_proto_family can't attach interface fam=%d\n",
265 		    protocol_family);
266 	}
267 
268 	return stat;
269 }
270 
271 static errno_t
stf_demux(ifnet_t ifp,__unused mbuf_t m,__unused char * frame_ptr,protocol_family_t * protocol_family)272 stf_demux(
273 	ifnet_t                                 ifp,
274 	__unused mbuf_t                 m,
275 	__unused char                   *frame_ptr,
276 	protocol_family_t               *protocol_family)
277 {
278 	struct stf_softc* stf = ifnet_softc(ifp);
279 	*protocol_family = stf->sc_protocol_family;
280 	return 0;
281 }
282 
283 static errno_t
stf_set_bpf_tap(ifnet_t ifp,bpf_tap_mode mode,bpf_packet_func callback)284 stf_set_bpf_tap(
285 	ifnet_t                 ifp,
286 	bpf_tap_mode    mode,
287 	bpf_packet_func callback)
288 {
289 	struct stf_softc        *sc = ifnet_softc(ifp);
290 
291 	sc->tap_mode = mode;
292 	sc->tap_callback = callback;
293 
294 	return 0;
295 }
296 
297 void
stfattach(void)298 stfattach(void)
299 {
300 	struct stf_softc *sc;
301 	int error;
302 	const struct encaptab *p;
303 	struct ifnet_init_eparams       stf_init;
304 
305 	error = proto_register_plumber(PF_INET6, APPLE_IF_FAM_STF,
306 	    stf_attach_inet6, NULL);
307 	if (error != 0) {
308 		printf("proto_register_plumber failed for AF_INET6 error=%d\n", error);
309 	}
310 
311 	sc = kalloc_type(struct stf_softc, Z_WAITOK_ZERO_NOFAIL);
312 	lck_mtx_init(&sc->sc_ro_mtx, &stf_mtx_grp, LCK_ATTR_NULL);
313 
314 	p = encap_attach_func(AF_INET, IPPROTO_IPV6, stf_encapcheck,
315 	    &in_stf_protosw, sc);
316 	if (p == NULL) {
317 		printf("sftattach encap_attach_func failed\n");
318 		goto free_sc;
319 	}
320 	sc->encap_cookie = p;
321 
322 	bzero(&stf_init, sizeof(stf_init));
323 	stf_init.ver = IFNET_INIT_CURRENT_VERSION;
324 	stf_init.len = sizeof(stf_init);
325 	stf_init.flags = IFNET_INIT_LEGACY;
326 	stf_init.name = "stf";
327 	stf_init.unit = 0;
328 	stf_init.type = IFT_STF;
329 	stf_init.family = IFNET_FAMILY_STF;
330 	stf_init.output = stf_output;
331 	stf_init.demux = stf_demux;
332 	stf_init.add_proto = stf_add_proto;
333 	stf_init.del_proto = stf_del_proto;
334 	stf_init.softc = sc;
335 	stf_init.ioctl = stf_ioctl;
336 	stf_init.set_bpf_tap = stf_set_bpf_tap;
337 
338 	error = ifnet_allocate_extended(&stf_init, &sc->sc_if);
339 	if (error != 0) {
340 		printf("stfattach, ifnet_allocate failed - %d\n", error);
341 		encap_detach(sc->encap_cookie);
342 		goto free_sc;
343 	}
344 	ifnet_set_mtu(sc->sc_if, IPV6_MMTU);
345 	ifnet_set_flags(sc->sc_if, 0, 0xffff); /* clear all flags */
346 #if 0
347 	/* turn off ingress filter */
348 	ifnet_set_flags(sc->sc_if, IFF_LINK2, IFF_LINK2);
349 #endif
350 
351 	error = ifnet_attach(sc->sc_if, NULL);
352 	if (error != 0) {
353 		printf("stfattach: ifnet_attach returned error=%d\n", error);
354 		encap_detach(sc->encap_cookie);
355 		ifnet_release(sc->sc_if);
356 		goto free_sc;
357 	}
358 
359 	bpfattach(sc->sc_if, DLT_NULL, sizeof(u_int));
360 
361 	return;
362 
363 free_sc:
364 	lck_mtx_destroy(&sc->sc_ro_mtx, &stf_mtx_grp);
365 	kfree_type(struct stf_softc, sc);
366 }
367 
368 static int
stf_encapcheck(const struct mbuf * m,__unused int off,int proto,void * arg)369 stf_encapcheck(
370 	const struct mbuf *m,
371 	__unused int off,
372 	int proto,
373 	void *arg)
374 {
375 	struct ip ip;
376 	struct in6_ifaddr *ia6;
377 	struct stf_softc *sc;
378 	struct in_addr a, b;
379 
380 	sc = (struct stf_softc *)arg;
381 	if (sc == NULL) {
382 		return 0;
383 	}
384 
385 	if ((ifnet_flags(sc->sc_if) & IFF_UP) == 0) {
386 		return 0;
387 	}
388 
389 	/* IFF_LINK0 means "no decapsulation" */
390 	if ((ifnet_flags(sc->sc_if) & IFF_LINK0) != 0) {
391 		return 0;
392 	}
393 
394 	if (proto != IPPROTO_IPV6) {
395 		return 0;
396 	}
397 
398 	mbuf_copydata((struct mbuf *)(size_t)m, 0, sizeof(ip), &ip);
399 
400 	if (ip.ip_v != 4) {
401 		return 0;
402 	}
403 
404 	ia6 = stf_getsrcifa6(sc->sc_if);
405 	if (ia6 == NULL) {
406 		return 0;
407 	}
408 
409 	/*
410 	 * check if IPv4 dst matches the IPv4 address derived from the
411 	 * local 6to4 address.
412 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
413 	 */
414 	IFA_LOCK(&ia6->ia_ifa);
415 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
416 	    sizeof(ip.ip_dst)) != 0) {
417 		IFA_UNLOCK(&ia6->ia_ifa);
418 		IFA_REMREF(&ia6->ia_ifa);
419 		return 0;
420 	}
421 	/*
422 	 * check if IPv4 src matches the IPv4 address derived from the
423 	 * local 6to4 address masked by prefixmask.
424 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
425 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
426 	 */
427 	bzero(&a, sizeof(a));
428 	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
429 	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
430 	b = ip.ip_src;
431 	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
432 	if (a.s_addr != b.s_addr) {
433 		IFA_UNLOCK(&ia6->ia_ifa);
434 		IFA_REMREF(&ia6->ia_ifa);
435 		return 0;
436 	}
437 	/* stf interface makes single side match only */
438 	IFA_UNLOCK(&ia6->ia_ifa);
439 	IFA_REMREF(&ia6->ia_ifa);
440 	return 32;
441 }
442 
443 static struct in6_ifaddr *
stf_getsrcifa6(struct ifnet * ifp)444 stf_getsrcifa6(struct ifnet *ifp)
445 {
446 	struct ifaddr *ia;
447 	struct in_ifaddr *ia4;
448 	struct sockaddr_in6 *sin6;
449 	struct in_addr in;
450 
451 	ifnet_lock_shared(ifp);
452 	for (ia = ifp->if_addrlist.tqh_first; ia; ia = ia->ifa_list.tqe_next) {
453 		IFA_LOCK(ia);
454 		if (ia->ifa_addr == NULL) {
455 			IFA_UNLOCK(ia);
456 			continue;
457 		}
458 		if (ia->ifa_addr->sa_family != AF_INET6) {
459 			IFA_UNLOCK(ia);
460 			continue;
461 		}
462 		sin6 = (struct sockaddr_in6 *)(void *)ia->ifa_addr;
463 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
464 			IFA_UNLOCK(ia);
465 			continue;
466 		}
467 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
468 		IFA_UNLOCK(ia);
469 		lck_rw_lock_shared(&in_ifaddr_rwlock);
470 		for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
471 		    ia4;
472 		    ia4 = TAILQ_NEXT(ia4, ia_link)) {
473 			IFA_LOCK(&ia4->ia_ifa);
474 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr) {
475 				IFA_UNLOCK(&ia4->ia_ifa);
476 				break;
477 			}
478 			IFA_UNLOCK(&ia4->ia_ifa);
479 		}
480 		lck_rw_done(&in_ifaddr_rwlock);
481 		if (ia4 == NULL) {
482 			continue;
483 		}
484 
485 		IFA_ADDREF(ia);         /* for caller */
486 		ifnet_lock_done(ifp);
487 		return (struct in6_ifaddr *)ia;
488 	}
489 	ifnet_lock_done(ifp);
490 
491 	return NULL;
492 }
493 
494 int
stf_pre_output(struct ifnet * ifp,__unused protocol_family_t protocol_family,struct mbuf ** m0,const struct sockaddr * dst,__unused void * route,__unused char * desk_linkaddr,__unused char * frame_type)495 stf_pre_output(
496 	struct ifnet    *ifp,
497 	__unused protocol_family_t  protocol_family,
498 	struct mbuf     **m0,
499 	const struct sockaddr   *dst,
500 	__unused void *route,
501 	__unused char *desk_linkaddr,
502 	__unused char *frame_type)
503 {
504 	struct mbuf *m = *m0;
505 	struct stf_softc *sc;
506 	const struct sockaddr_in6 *dst6;
507 	const struct in_addr *in4;
508 	u_int8_t tos;
509 	struct ip *ip;
510 	struct ip6_hdr *ip6;
511 	struct in6_ifaddr *ia6;
512 	struct sockaddr_in      *dst4;
513 	struct ip_out_args ipoa;
514 	errno_t result = 0;
515 
516 	bzero(&ipoa, sizeof(ipoa));
517 	ipoa.ipoa_boundif = IFSCOPE_NONE;
518 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF;
519 	ipoa.ipoa_sotc = SO_TC_UNSPEC;
520 	ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
521 
522 	sc = ifnet_softc(ifp);
523 	dst6 = (const struct sockaddr_in6 *)(const void *)dst;
524 
525 	/* just in case */
526 	if ((ifnet_flags(ifp) & IFF_UP) == 0) {
527 		printf("stf: IFF_DOWN\n");
528 		return ENETDOWN;
529 	}
530 
531 	/*
532 	 * If we don't have an ip4 address that match my inner ip6 address,
533 	 * we shouldn't generate output.  Without this check, we'll end up
534 	 * using wrong IPv4 source.
535 	 */
536 	ia6 = stf_getsrcifa6(ifp);
537 	if (ia6 == NULL) {
538 		return ENETDOWN;
539 	}
540 
541 	if (mbuf_len(m) < sizeof(*ip6)) {
542 		m = m_pullup(m, sizeof(*ip6));
543 		if (!m) {
544 			*m0 = NULL; /* makes sure this won't be double freed */
545 			IFA_REMREF(&ia6->ia_ifa);
546 			return ENOBUFS;
547 		}
548 		*m0 = m;
549 	}
550 	ip6 = mtod(m, struct ip6_hdr *);
551 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
552 
553 	/*
554 	 * Pickup the right outer dst addr from the list of candidates.
555 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
556 	 */
557 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst)) {
558 		in4 = GET_V4(&ip6->ip6_dst);
559 	} else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr)) {
560 		in4 = GET_V4(&dst6->sin6_addr);
561 	} else {
562 		IFA_REMREF(&ia6->ia_ifa);
563 		return ENETUNREACH;
564 	}
565 
566 	if (ifp->if_bpf) {
567 		/* We need to prepend the address family as a four byte field. */
568 		u_int32_t af = AF_INET6;
569 
570 		bpf_tap_out(ifp, 0, m, &af, sizeof(af));
571 	}
572 
573 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT, 1);
574 	if (m && mbuf_len(m) < sizeof(struct ip)) {
575 		m = m_pullup(m, sizeof(struct ip));
576 	}
577 	if (m == NULL) {
578 		*m0 = NULL;
579 		IFA_REMREF(&ia6->ia_ifa);
580 		return ENOBUFS;
581 	}
582 
583 	*m0 = m;
584 	ip = mtod(m, struct ip *);
585 
586 	bzero(ip, sizeof(*ip));
587 
588 	IFA_LOCK_SPIN(&ia6->ia_ifa);
589 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
590 	    &ip->ip_src, sizeof(ip->ip_src));
591 	IFA_UNLOCK(&ia6->ia_ifa);
592 	bcopy(in4, &ip->ip_dst, sizeof(ip->ip_dst));
593 	ip->ip_p = IPPROTO_IPV6;
594 	ip->ip_ttl = ip_stf_ttl;
595 	ip->ip_len = m->m_pkthdr.len;   /*host order*/
596 	if (ifp->if_flags & IFF_LINK1) {
597 		ip_ecn_ingress(ECN_NORMAL, &ip->ip_tos, &tos);
598 	} else {
599 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
600 	}
601 
602 	lck_mtx_lock(&sc->sc_ro_mtx);
603 	dst4 = (struct sockaddr_in *)(void *)&sc->sc_ro.ro_dst;
604 	if (ROUTE_UNUSABLE(&sc->sc_ro) || dst4->sin_family != AF_INET ||
605 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
606 		ROUTE_RELEASE(&sc->sc_ro);
607 		/* cache route doesn't match: always the case during the first use */
608 		dst4->sin_family = AF_INET;
609 		dst4->sin_len = sizeof(struct sockaddr_in);
610 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
611 	}
612 
613 	result = ip_output(m, NULL, &sc->sc_ro, IP_OUTARGS, NULL, &ipoa);
614 	lck_mtx_unlock(&sc->sc_ro_mtx);
615 
616 	/* Assumption: ip_output will free mbuf on errors */
617 	/* All the output processing is done here, don't let stf_output be called */
618 	if (result == 0) {
619 		result = EJUSTRETURN;
620 	}
621 	*m0 = NULL;
622 	IFA_REMREF(&ia6->ia_ifa);
623 	return result;
624 }
625 static errno_t
stf_output(__unused ifnet_t ifp,__unused mbuf_t m)626 stf_output(
627 	__unused ifnet_t        ifp,
628 	__unused mbuf_t m)
629 {
630 	/* All processing is done in stf_pre_output
631 	 * this shouldn't be called as the pre_output returns "EJUSTRETURN"
632 	 */
633 	return 0;
634 }
635 
636 static int
stf_checkaddr4(struct stf_softc * sc,const struct in_addr * in,struct ifnet * inifp)637 stf_checkaddr4(
638 	struct stf_softc *sc,
639 	const struct in_addr *in,
640 	struct ifnet *inifp)    /* incoming interface */
641 {
642 	struct in_ifaddr *ia4;
643 
644 	/*
645 	 * reject packets with the following address:
646 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
647 	 */
648 	if (IN_MULTICAST(ntohl(in->s_addr))) {
649 		return -1;
650 	}
651 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
652 	case 0: case 127: case 255:
653 		return -1;
654 	}
655 
656 	/*
657 	 * reject packets with broadcast
658 	 */
659 	lck_rw_lock_shared(&in_ifaddr_rwlock);
660 	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
661 	    ia4;
662 	    ia4 = TAILQ_NEXT(ia4, ia_link)) {
663 		IFA_LOCK(&ia4->ia_ifa);
664 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) {
665 			IFA_UNLOCK(&ia4->ia_ifa);
666 			continue;
667 		}
668 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
669 			IFA_UNLOCK(&ia4->ia_ifa);
670 			lck_rw_done(&in_ifaddr_rwlock);
671 			return -1;
672 		}
673 		IFA_UNLOCK(&ia4->ia_ifa);
674 	}
675 	lck_rw_done(&in_ifaddr_rwlock);
676 
677 	/*
678 	 * perform ingress filter
679 	 */
680 	if (sc && (ifnet_flags(sc->sc_if) & IFF_LINK2) == 0 && inifp) {
681 		struct sockaddr_in sin;
682 		struct rtentry *rt;
683 
684 		bzero(&sin, sizeof(sin));
685 		sin.sin_family = AF_INET;
686 		sin.sin_len = sizeof(struct sockaddr_in);
687 		sin.sin_addr = *in;
688 		rt = rtalloc1((struct sockaddr *)&sin, 0, 0);
689 		if (rt != NULL) {
690 			RT_LOCK(rt);
691 		}
692 		if (rt == NULL || rt->rt_ifp != inifp) {
693 #if 1
694 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
695 			    "due to ingress filter\n", if_name(sc->sc_if),
696 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
697 #endif
698 			if (rt != NULL) {
699 				RT_UNLOCK(rt);
700 				rtfree(rt);
701 			}
702 			return -1;
703 		}
704 		RT_UNLOCK(rt);
705 		rtfree(rt);
706 	}
707 
708 	return 0;
709 }
710 
711 static int
stf_checkaddr6(struct stf_softc * sc,struct in6_addr * in6,struct ifnet * inifp)712 stf_checkaddr6(
713 	struct stf_softc *sc,
714 	struct in6_addr *in6,
715 	struct ifnet *inifp)    /* incoming interface */
716 {
717 	/*
718 	 * check 6to4 addresses
719 	 */
720 	if (IN6_IS_ADDR_6TO4(in6)) {
721 		return stf_checkaddr4(sc, GET_V4(in6), inifp);
722 	}
723 
724 	/*
725 	 * reject anything that look suspicious.  the test is implemented
726 	 * in ip6_input too, but we check here as well to
727 	 * (1) reject bad packets earlier, and
728 	 * (2) to be safe against future ip6_input change.
729 	 */
730 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6)) {
731 		return -1;
732 	}
733 
734 	return 0;
735 }
736 
737 static void
in_stf_input(struct mbuf * m,int off)738 in_stf_input(
739 	struct mbuf *m,
740 	int off)
741 {
742 	struct stf_softc *sc;
743 	struct ip *ip;
744 	struct ip6_hdr ip6;
745 	u_int8_t otos, itos;
746 	int proto;
747 	struct ifnet *ifp;
748 	struct ifnet_stat_increment_param       stats;
749 
750 	ip = mtod(m, struct ip *);
751 	proto = ip->ip_p;
752 
753 	if (proto != IPPROTO_IPV6) {
754 		m_freem(m);
755 		return;
756 	}
757 
758 	ip = mtod(m, struct ip *);
759 
760 	sc = (struct stf_softc *)encap_getarg(m);
761 
762 	if (sc == NULL || (ifnet_flags(sc->sc_if) & IFF_UP) == 0) {
763 		m_freem(m);
764 		return;
765 	}
766 
767 	ifp = sc->sc_if;
768 
769 	/*
770 	 * perform sanity check against outer src/dst.
771 	 * for source, perform ingress filter as well.
772 	 */
773 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
774 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
775 		m_freem(m);
776 		return;
777 	}
778 
779 	otos = ip->ip_tos;
780 	mbuf_copydata(m, off, sizeof(ip6), &ip6);
781 
782 	/*
783 	 * perform sanity check against inner src/dst.
784 	 * for source, perform ingress filter as well.
785 	 */
786 	if (stf_checkaddr6(sc, &ip6.ip6_dst, NULL) < 0 ||
787 	    stf_checkaddr6(sc, &ip6.ip6_src, m->m_pkthdr.rcvif) < 0) {
788 		m_freem(m);
789 		return;
790 	}
791 
792 	itos = (ntohl(ip6.ip6_flow) >> 20) & 0xff;
793 	if ((ifnet_flags(ifp) & IFF_LINK1) != 0) {
794 		ip_ecn_egress(ECN_NORMAL, &otos, &itos);
795 	} else {
796 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
797 	}
798 	ip6.ip6_flow &= ~htonl(0xff << 20);
799 	ip6.ip6_flow |= htonl((u_int32_t)itos << 20);
800 
801 	m->m_pkthdr.rcvif = ifp;
802 	mbuf_pkthdr_setheader(m, mbuf_data(m));
803 	mbuf_adj(m, off);
804 
805 	if (ifp->if_bpf) {
806 		/* We need to prepend the address family as a four byte field. */
807 		u_int32_t af = AF_INET6;
808 		bpf_tap_in(ifp, 0, m, &af, sizeof(af));
809 	}
810 
811 	/*
812 	 * Put the packet to the network layer input queue according to the
813 	 * specified address family.
814 	 * See net/if_gif.c for possible issues with packet processing
815 	 * reorder due to extra queueing.
816 	 */
817 	bzero(&stats, sizeof(stats));
818 	stats.packets_in = 1;
819 	stats.bytes_in = mbuf_pkthdr_len(m);
820 	mbuf_pkthdr_setrcvif(m, ifp);
821 	ifnet_input(ifp, m, &stats);
822 
823 	return;
824 }
825 
826 static void
stf_rtrequest(__unused int cmd,struct rtentry * rt,__unused struct sockaddr * sa)827 stf_rtrequest(
828 	__unused int cmd,
829 	struct rtentry *rt,
830 	__unused struct sockaddr *sa)
831 {
832 	if (rt != NULL) {
833 		RT_LOCK_ASSERT_HELD(rt);
834 		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
835 	}
836 }
837 
838 static errno_t
stf_ioctl(ifnet_t ifp,u_long cmd,void * data)839 stf_ioctl(
840 	ifnet_t         ifp,
841 	u_long          cmd,
842 	void            *data)
843 {
844 	struct ifaddr *ifa;
845 	struct ifreq *ifr;
846 	struct sockaddr_in6 *sin6;
847 	int error;
848 
849 	error = 0;
850 	switch (cmd) {
851 	case SIOCSIFADDR:
852 		ifa = (struct ifaddr *)data;
853 		if (ifa == NULL) {
854 			error = EAFNOSUPPORT;
855 			break;
856 		}
857 		IFA_LOCK(ifa);
858 		if (ifa->ifa_addr->sa_family != AF_INET6) {
859 			IFA_UNLOCK(ifa);
860 			error = EAFNOSUPPORT;
861 			break;
862 		}
863 		sin6 = (struct sockaddr_in6 *)(void *)ifa->ifa_addr;
864 		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
865 			if (!(ifnet_flags( ifp ) & IFF_UP)) {
866 				/* do this only if the interface is not already up */
867 				ifa->ifa_rtrequest = stf_rtrequest;
868 				IFA_UNLOCK(ifa);
869 				ifnet_set_flags(ifp, IFF_UP, IFF_UP);
870 			} else {
871 				IFA_UNLOCK(ifa);
872 			}
873 		} else {
874 			IFA_UNLOCK(ifa);
875 			error = EINVAL;
876 		}
877 		IFA_LOCK_ASSERT_NOTHELD(ifa);
878 		break;
879 
880 	case SIOCADDMULTI:
881 	case SIOCDELMULTI:
882 		ifr = (struct ifreq *)data;
883 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6) {
884 			;
885 		} else {
886 			error = EAFNOSUPPORT;
887 		}
888 		break;
889 
890 	default:
891 		error = EOPNOTSUPP;
892 		break;
893 	}
894 
895 	return error;
896 }
897