xref: /freebsd/sys/net/if_stf.c (revision 7899f917b1c0ea178f1d2be0cfb452086d079d23)
1 /*	$KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: BSD-3-Clause
5  *
6  * Copyright (C) 2000 WIDE Project.
7  * Copyright (c) 2010 Hiroki Sato <hrs@FreeBSD.org>
8  * Copyright (c) 2013 Ermal Luci <eri@FreeBSD.org>
9  * Copyright (c) 2017-2021 Rubicon Communications, LLC (Netgate)
10  * All rights reserved.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the project nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 /*
38  * 6to4 interface, based on RFC3056.
39  *
40  * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
41  * There is no address mapping defined from IPv6 multicast address to IPv4
42  * address.  Therefore, we do not have IFF_MULTICAST on the interface.
43  *
44  * Due to the lack of address mapping for link-local addresses, we cannot
45  * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
46  * packets to link-local multicast addresses (ff02::x).
47  *
48  * Here are interesting symptoms due to the lack of link-local address:
49  *
50  * Unicast routing exchange:
51  * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
52  *   and link-local addresses as nexthop.
53  * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
54  *   assigned to the link, and makes use of them.  Also, HELLO packets use
55  *   link-local multicast addresses (ff02::5 and ff02::6).
56  * - BGP4+: Maybe.  You can only use global address as nexthop, and global
57  *   address as TCP endpoint address.
58  *
59  * Multicast routing protocols:
60  * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
61  *   Adjacent PIM routers must be configured manually (is it really spec-wise
62  *   correct thing to do?).
63  *
64  * ICMPv6:
65  * - Redirects cannot be used due to the lack of link-local address.
66  *
67  * stf interface does not have, and will not need, a link-local address.
68  * It seems to have no real benefit and does not help the above symptoms much.
69  * Even if we assign link-locals to interface, we cannot really
70  * use link-local unicast/multicast on top of 6to4 cloud (since there's no
71  * encapsulation defined for link-local address), and the above analysis does
72  * not change.  RFC3056 does not mandate the assignment of link-local address
73  * either.
74  *
75  * 6to4 interface has security issues.  Refer to
76  * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
77  * for details.  The code tries to filter out some of malicious packets.
78  * Note that there is no way to be 100% secure.
79  */
80 
81 #include <sys/param.h>
82 #include <sys/systm.h>
83 #include <sys/socket.h>
84 #include <sys/sockio.h>
85 #include <sys/mbuf.h>
86 #include <sys/endian.h>
87 #include <sys/errno.h>
88 #include <sys/kernel.h>
89 #include <sys/lock.h>
90 #include <sys/module.h>
91 #include <sys/priv.h>
92 #include <sys/proc.h>
93 #include <sys/queue.h>
94 #include <sys/sdt.h>
95 #include <sys/sysctl.h>
96 #include <machine/cpu.h>
97 
98 #include <sys/malloc.h>
99 
100 #include <net/if.h>
101 #include <net/if_var.h>
102 #include <net/if_private.h>
103 #include <net/if_clone.h>
104 #include <net/route.h>
105 #include <net/route/nhop.h>
106 #include <net/netisr.h>
107 #include <net/if_stf.h>
108 #include <net/if_types.h>
109 #include <net/vnet.h>
110 
111 #include <netinet/in.h>
112 #include <netinet/in_fib.h>
113 #include <netinet/in_systm.h>
114 #include <netinet/ip.h>
115 #include <netinet/ip_var.h>
116 #include <netinet/in_var.h>
117 
118 #include <netinet/ip6.h>
119 #include <netinet6/in6_fib.h>
120 #include <netinet6/ip6_var.h>
121 #include <netinet6/in6_var.h>
122 #include <netinet/ip_ecn.h>
123 
124 #include <netinet/ip_encap.h>
125 
126 #include <machine/stdarg.h>
127 
128 #include <net/bpf.h>
129 
130 #include <security/mac/mac_framework.h>
131 
132 SDT_PROVIDER_DEFINE(if_stf);
133 SDT_PROBE_DEFINE3(if_stf, , encapcheck, in, "struct mbuf *", "int", "int");
134 SDT_PROBE_DEFINE0(if_stf, , encapcheck, accept);
135 SDT_PROBE_DEFINE3(if_stf, , getsrcifa6, in, "struct ifnet *",
136     "struct in6_addr *", "struct in6_addr *");
137 SDT_PROBE_DEFINE2(if_stf, , getsrcifa6, found, "struct in6_addr *",
138     "struct in6_addr *");
139 SDT_PROBE_DEFINE0(if_stf, , getsrcifa6, notfound);
140 
141 SDT_PROBE_DEFINE4(if_stf, , stf_output, in, "struct ifnet *", "struct mbuf *",
142     "struct sockaddr *", "struct route *");
143 SDT_PROBE_DEFINE2(if_stf, , stf_output, error, "int", "int");
144 SDT_PROBE_DEFINE1(if_stf, , stf_output, out, "int");
145 
146 SDT_PROBE_DEFINE3(if_stf, , checkaddr6, in, "struct stf_softc *",
147     "struct in6_addr *", "struct ifnet *");
148 SDT_PROBE_DEFINE2(if_stf, , checkaddr6, out, "int", "int");
149 
150 SDT_PROBE_DEFINE3(if_stf, , stf_input, in, "struct mbuf *", "int", "int");
151 SDT_PROBE_DEFINE2(if_stf, , stf_input, out, "int", "int");
152 
153 SDT_PROBE_DEFINE3(if_stf, , ioctl, sv4net, "struct in_addr *",
154     "struct in_addr *", "int");
155 SDT_PROBE_DEFINE1(if_stf, , ioctl, sdstv4, "struct in_addr *");
156 SDT_PROBE_DEFINE1(if_stf, , ioctl, ifaddr, "struct ifaddr *");
157 
158 SDT_PROBE_DEFINE4(if_stf, , getin4addr_in6, out, "struct in6_addr *",
159     "struct in6_addr *", "struct in6_addr *", "struct sockaddr_in *");
160 
161 SDT_PROBE_DEFINE2(if_stf, , getin4addr, in, "struct in6_addr *", "struct in6_addr *");
162 SDT_PROBE_DEFINE1(if_stf, , getin4addr, out, "struct sockaddr_in *");
163 
164 SYSCTL_DECL(_net_link);
165 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
166     "6to4 Interface");
167 
168 static int stf_permit_rfc1918 = 0;
169 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
170     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
171 
172 #define STFUNIT		0
173 
174 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
175 
176 /*
177  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
178  * struct in_addr *; use bcopy() instead.
179  */
180 #define GET_V4(x)	(&(x)->s6_addr16[1])
181 
182 struct stf_softc {
183 	struct ifnet	*sc_ifp;
184 	in_addr_t	braddr;		/* Border relay IPv4 address */
185 	in_addr_t	srcv4_addr;	/* Our IPv4 WAN address */
186 	u_int		v4prefixlen;	/* How much of the v4 address to include in our address. */
187 	u_int		sc_fibnum;
188 	const struct encaptab *encap_cookie;
189 };
190 #define STF2IFP(sc)	((sc)->sc_ifp)
191 
192 static const char stfname[] = "stf";
193 
194 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
195 static const int ip_stf_ttl = 40;
196 
197 static int in_stf_input(struct mbuf *, int, int, void *);
198 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
199 
200 static int stfmodevent(module_t, int, void *);
201 static int stf_encapcheck(const struct mbuf *, int, int, void *);
202 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
203 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
204 	struct route *);
205 static int isrfc1918addr(struct in_addr *);
206 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
207 	struct ifnet *);
208 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
209 	struct ifnet *);
210 static struct sockaddr_in *stf_getin4addr_in6(struct stf_softc *,
211 	struct sockaddr_in *, struct in6_addr, struct in6_addr,
212 	struct in6_addr);
213 static struct sockaddr_in *stf_getin4addr(struct stf_softc *,
214 	struct sockaddr_in *, struct in6_addr, struct in6_addr);
215 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
216 
217 VNET_DEFINE_STATIC(struct if_clone *, stf_cloner);
218 #define V_stf_cloner	VNET(stf_cloner)
219 
220 static const struct encap_config ipv4_encap_cfg = {
221 	.proto = IPPROTO_IPV6,
222 	.min_length = sizeof(struct ip),
223 	.exact_match = (sizeof(in_addr_t) << 3) + 8,
224 	.check = stf_encapcheck,
225 	.input = in_stf_input
226 };
227 
228 static int
229 stf_clone_match(struct if_clone *ifc, const char *name)
230 {
231 	int i;
232 
233 	for(i = 0; stfnames[i] != NULL; i++) {
234 		if (strcmp(stfnames[i], name) == 0)
235 			return (1);
236 	}
237 
238 	return (0);
239 }
240 
241 static int
242 stf_clone_create(struct if_clone *ifc, char *name, size_t len,
243     struct ifc_data *ifd, struct ifnet **ifpp)
244 {
245 	char *dp;
246 	int err, unit, wildcard;
247 	struct stf_softc *sc;
248 	struct ifnet *ifp;
249 
250 	err = ifc_name2unit(name, &unit);
251 	if (err != 0)
252 		return (err);
253 	wildcard = (unit < 0);
254 
255 	/*
256 	 * We can only have one unit, but since unit allocation is
257 	 * already locked, we use it to keep from allocating extra
258 	 * interfaces.
259 	 */
260 	unit = STFUNIT;
261 	err = ifc_alloc_unit(ifc, &unit);
262 	if (err != 0)
263 		return (err);
264 
265 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
266 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
267 	ifp->if_softc = sc;
268 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
269 
270 	/*
271 	 * Set the name manually rather then using if_initname because
272 	 * we don't conform to the default naming convention for interfaces.
273 	 * In the wildcard case, we need to update the name.
274 	 */
275 	if (wildcard) {
276 		for (dp = name; *dp != '\0'; dp++);
277 		if (snprintf(dp, len - (dp-name), "%d", unit) >
278 		    len - (dp-name) - 1) {
279 			/*
280 			 * This can only be a programmer error and
281 			 * there's no straightforward way to recover if
282 			 * it happens.
283 			 */
284 			panic("if_clone_create(): interface name too long");
285 		}
286 	}
287 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
288 	ifp->if_dname = stfname;
289 	ifp->if_dunit = IF_DUNIT_NONE;
290 
291 	sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK);
292 
293 	ifp->if_mtu    = IPV6_MMTU;
294 	ifp->if_ioctl  = stf_ioctl;
295 	ifp->if_output = stf_output;
296 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
297 	if_attach(ifp);
298 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
299 	*ifpp = ifp;
300 
301 	return (0);
302 }
303 
304 static int
305 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
306 {
307 	struct stf_softc *sc = ifp->if_softc;
308 	int err __unused;
309 
310 	err = ip_encap_detach(sc->encap_cookie);
311 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
312 	bpfdetach(ifp);
313 	if_detach(ifp);
314 	if_free(ifp);
315 
316 	free(sc, M_STF);
317 	ifc_free_unit(ifc, STFUNIT);
318 
319 	return (0);
320 }
321 
322 static void
323 vnet_stf_init(const void *unused __unused)
324 {
325 	struct if_clone_addreq req = {
326 		.match_f = stf_clone_match,
327 		.create_f = stf_clone_create,
328 		.destroy_f = stf_clone_destroy,
329 	};
330 	V_stf_cloner = ifc_attach_cloner(stfname, &req);
331 }
332 VNET_SYSINIT(vnet_stf_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_init, NULL);
333 
334 static void
335 vnet_stf_uninit(const void *unused __unused)
336 {
337 	if_clone_detach(V_stf_cloner);
338 	V_stf_cloner = NULL;
339 }
340 VNET_SYSUNINIT(vnet_stf_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_uninit,
341     NULL);
342 
343 static int
344 stfmodevent(module_t mod, int type, void *data)
345 {
346 
347 	switch (type) {
348 	case MOD_LOAD:
349 		/* Done in vnet_stf_init() */
350 		break;
351 	case MOD_UNLOAD:
352 		/* Done in vnet_stf_uninit() */
353 		break;
354 	default:
355 		return (EOPNOTSUPP);
356 	}
357 
358 	return (0);
359 }
360 
361 static moduledata_t stf_mod = {
362 	"if_stf",
363 	stfmodevent,
364 	0
365 };
366 
367 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
368 MODULE_VERSION(if_stf, 2);
369 
370 static int
371 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
372 {
373 	struct ip ip;
374 	struct stf_softc *sc;
375 	struct in6_addr addr6, mask6;
376 	struct sockaddr_in sin4addr, sin4mask;
377 
378 	SDT_PROBE3(if_stf, , encapcheck, in, m, off, proto);
379 
380 	sc = (struct stf_softc *)arg;
381 	if (sc == NULL)
382 		return (0);
383 
384 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
385 		return (0);
386 
387 	/* IFF_LINK0 means "no decapsulation" */
388 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
389 		return (0);
390 
391 	if (proto != IPPROTO_IPV6)
392 		return (0);
393 
394 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
395 
396 	if (ip.ip_v != 4)
397 		return (0);
398 
399 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
400 		return (0);
401 
402 	if (sc->srcv4_addr != INADDR_ANY) {
403 		sin4addr.sin_addr.s_addr = sc->srcv4_addr;
404 		sin4addr.sin_family = AF_INET;
405 	} else
406 		if (stf_getin4addr(sc, &sin4addr, addr6, mask6) == NULL)
407 			return (0);
408 
409 	if (sin4addr.sin_addr.s_addr != ip.ip_dst.s_addr)
410 		return (0);
411 
412 	if (IN6_IS_ADDR_6TO4(&addr6)) {
413 		/*
414 		 * 6to4 (RFC 3056).
415 		 * Check if IPv4 src matches the IPv4 address derived
416 		 * from the local 6to4 address masked by prefixmask.
417 		 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
418 		 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
419 		 */
420 		memcpy(&sin4mask.sin_addr, GET_V4(&mask6),
421 		    sizeof(sin4mask.sin_addr));
422 		if ((sin4addr.sin_addr.s_addr & sin4mask.sin_addr.s_addr) !=
423 		    (ip.ip_src.s_addr & sin4mask.sin_addr.s_addr))
424 			return (0);
425 	} else {
426 		/* 6rd (RFC 5569) */
427 		/*
428 		 * No restriction on the src address in the case of
429 		 * 6rd because the stf(4) interface always has a
430 		 * prefix which covers whole of IPv4 src address
431 		 * range.  So, stf_output() will catch all of
432 		 * 6rd-capsuled IPv4 traffic with suspicious inner dst
433 		 * IPv4 address (i.e. the IPv6 destination address is
434 		 * one the admin does not like to route to outside),
435 		 * and then it discard them silently.
436 		 */
437 	}
438 
439 	SDT_PROBE0(if_stf, , encapcheck, accept);
440 
441 	/* stf interface makes single side match only */
442 	return (32);
443 }
444 
445 static int
446 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
447 {
448 	struct ifaddr *ia;
449 	struct in_ifaddr *ia4;
450 	struct in6_addr addr6, mask6;
451 	struct sockaddr_in sin4;
452 	struct stf_softc *sc;
453 	struct in_addr in;
454 
455 	NET_EPOCH_ASSERT();
456 
457 	sc = ifp->if_softc;
458 
459 	SDT_PROBE3(if_stf, , getsrcifa6, in, ifp, addr, mask);
460 
461 	CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
462 		if (ia->ifa_addr->sa_family != AF_INET6)
463 			continue;
464 
465 		addr6 = *IFA_IN6(ia);
466 		mask6 = *IFA_MASKIN6(ia);
467 		if (sc->srcv4_addr != INADDR_ANY)
468 			bcopy(&sc->srcv4_addr, &in, sizeof(in));
469 		else {
470 			if (stf_getin4addr(sc, &sin4, addr6, mask6) == NULL)
471 				continue;
472 			bcopy(&sin4.sin_addr, &in, sizeof(in));
473 		}
474 
475 		CK_LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
476 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
477 				break;
478 		if (ia4 == NULL)
479 			continue;
480 
481 		*addr = addr6;
482 		*mask = mask6;
483 
484 		SDT_PROBE2(if_stf, , getsrcifa6, found, addr, mask);
485 
486 		return (0);
487 	}
488 
489 	SDT_PROBE0(if_stf, , getsrcifa6, notfound);
490 
491 	return (ENOENT);
492 }
493 
494 static int
495 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
496     struct route *ro)
497 {
498 	struct stf_softc *sc;
499 	const struct sockaddr_in6 *dst6;
500 	struct sockaddr_in dst4, src4;
501 	u_int8_t tos;
502 	struct ip *ip;
503 	struct ip6_hdr *ip6;
504 	struct in6_addr addr6, mask6;
505 	int error;
506 
507 	SDT_PROBE4(if_stf, , stf_output, in, ifp, m, dst, ro);
508 
509 #ifdef MAC
510 	error = mac_ifnet_check_transmit(ifp, m);
511 	if (error) {
512 		m_freem(m);
513 		SDT_PROBE2(if_stf, , stf_output, error, error, __LINE__);
514 		return (error);
515 	}
516 #endif
517 
518 	sc = ifp->if_softc;
519 	dst6 = (const struct sockaddr_in6 *)dst;
520 
521 	/* just in case */
522 	if ((ifp->if_flags & IFF_UP) == 0) {
523 		m_freem(m);
524 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
525 		SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__);
526 		return (ENETDOWN);
527 	}
528 
529 	/*
530 	 * If we don't have an ip4 address that match my inner ip6 address,
531 	 * we shouldn't generate output.  Without this check, we'll end up
532 	 * using wrong IPv4 source.
533 	 */
534 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
535 		m_freem(m);
536 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
537 		SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__);
538 		return (ENETDOWN);
539 	}
540 
541 	if (m->m_len < sizeof(*ip6)) {
542 		m = m_pullup(m, sizeof(*ip6));
543 		if (!m) {
544 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
545 			SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS,
546 			    __LINE__);
547 			return (ENOBUFS);
548 		}
549 	}
550 	ip6 = mtod(m, struct ip6_hdr *);
551 	tos = IPV6_TRAFFIC_CLASS(ip6);
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 (stf_getin4addr_in6(sc, &dst4, addr6, mask6,
558 	    ip6->ip6_dst) == NULL) {
559 		if (sc->braddr != INADDR_ANY)
560 			dst4.sin_addr.s_addr = sc->braddr;
561 		else if (stf_getin4addr_in6(sc, &dst4, addr6, mask6,
562 		    dst6->sin6_addr) == NULL) {
563 			m_freem(m);
564 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
565 			SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH,
566 			    __LINE__);
567 			return (ENETUNREACH);
568 		}
569 	}
570 
571 	if (bpf_peers_present(ifp->if_bpf)) {
572 		/*
573 		 * We need to prepend the address family as
574 		 * a four byte field.  Cons up a dummy header
575 		 * to pacify bpf.  This is safe because bpf
576 		 * will only read from the mbuf (i.e., it won't
577 		 * try to free it or keep a pointer a to it).
578 		 */
579 		u_int af = AF_INET6;
580 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
581 	}
582 
583 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
584 	if (m == NULL) {
585 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
586 		SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS, __LINE__);
587 		return (ENOBUFS);
588 	}
589 	ip = mtod(m, struct ip *);
590 
591 	bzero(ip, sizeof(*ip));
592 
593 	if (sc->srcv4_addr != INADDR_ANY)
594 		src4.sin_addr.s_addr = sc->srcv4_addr;
595 	else if (stf_getin4addr(sc, &src4, addr6, mask6) == NULL) {
596 		m_freem(m);
597 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
598 		SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH, __LINE__);
599 		return (ENETUNREACH);
600 	}
601 	bcopy(&src4.sin_addr, &ip->ip_src, sizeof(ip->ip_src));
602 	bcopy(&dst4.sin_addr, &ip->ip_dst, sizeof(ip->ip_dst));
603 
604 	ip->ip_p = IPPROTO_IPV6;
605 	ip->ip_ttl = ip_stf_ttl;
606 	ip->ip_len = htons(m->m_pkthdr.len);
607 	if (ifp->if_flags & IFF_LINK1)
608 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
609 	else
610 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
611 
612 	M_SETFIB(m, sc->sc_fibnum);
613 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
614 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
615 
616 	SDT_PROBE1(if_stf, , stf_output, out, error);
617 	return (error);
618 }
619 
620 static int
621 isrfc1918addr(struct in_addr *in)
622 {
623 	/*
624 	 * returns 1 if private address range:
625 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
626 	 */
627 	if (stf_permit_rfc1918 == 0 && (
628 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
629 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
630 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
631 		return (1);
632 
633 	return (0);
634 }
635 
636 static int
637 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
638 {
639 	struct in_ifaddr *ia4;
640 
641 	/*
642 	 * reject packets with the following address:
643 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
644 	 */
645 	if (IN_MULTICAST(ntohl(in->s_addr)))
646 		return (-1);
647 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
648 	case 0: case 127: case 255:
649 		return (-1);
650 	}
651 
652 	/*
653 	 * reject packets with broadcast
654 	 */
655 	CK_STAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
656 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
657 			continue;
658 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
659 			return (-1);
660 		}
661 	}
662 
663 	/*
664 	 * perform ingress filter
665 	 */
666 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
667 		struct nhop_object *nh;
668 
669 		NET_EPOCH_ASSERT();
670 		nh = fib4_lookup(sc->sc_fibnum, *in, 0, 0, 0);
671 		if (nh == NULL)
672 			return (-1);
673 
674 		if (nh->nh_ifp != inifp)
675 			return (-1);
676 	}
677 
678 	return (0);
679 }
680 
681 static int
682 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
683 {
684 	SDT_PROBE3(if_stf, , checkaddr6, in, sc, in6, inifp);
685 
686 	/*
687 	 * check 6to4 addresses
688 	 */
689 	if (IN6_IS_ADDR_6TO4(in6)) {
690 		struct in_addr in4;
691 		int ret;
692 
693 		bcopy(GET_V4(in6), &in4, sizeof(in4));
694 		ret = stf_checkaddr4(sc, &in4, inifp);
695 		SDT_PROBE2(if_stf, , checkaddr6, out, ret, __LINE__);
696 		return (ret);
697 	}
698 
699 	/*
700 	 * reject anything that look suspicious.  the test is implemented
701 	 * in ip6_input too, but we check here as well to
702 	 * (1) reject bad packets earlier, and
703 	 * (2) to be safe against future ip6_input change.
704 	 */
705 	if (IN6_IS_ADDR_V4COMPAT(in6)) {
706 		SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__);
707 		return (-1);
708 	}
709 
710 	if (IN6_IS_ADDR_V4MAPPED(in6)) {
711 		SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__);
712 		return (-1);
713 	}
714 
715 	SDT_PROBE2(if_stf, , checkaddr6, out, 0, __LINE__);
716 	return (0);
717 }
718 
719 static int
720 in_stf_input(struct mbuf *m, int off, int proto, void *arg)
721 {
722 	struct stf_softc *sc = arg;
723 	struct ip ip;
724 	struct ip6_hdr *ip6;
725 	u_int8_t otos, itos;
726 	struct ifnet *ifp;
727 	struct nhop_object *nh;
728 
729 	NET_EPOCH_ASSERT();
730 
731 	SDT_PROBE3(if_stf, , stf_input, in, m, off, proto);
732 
733 	if (proto != IPPROTO_IPV6) {
734 		m_freem(m);
735 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
736 		return (IPPROTO_DONE);
737 	}
738 
739 	m_copydata(m, 0, sizeof(struct ip), (caddr_t)&ip);
740 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
741 		m_freem(m);
742 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
743 		return (IPPROTO_DONE);
744 	}
745 
746 	ifp = STF2IFP(sc);
747 
748 #ifdef MAC
749 	mac_ifnet_create_mbuf(ifp, m);
750 #endif
751 
752 	/*
753 	 * perform sanity check against outer src/dst.
754 	 * for source, perform ingress filter as well.
755 	 */
756 	if (stf_checkaddr4(sc, &ip.ip_dst, NULL) < 0 ||
757 	    stf_checkaddr4(sc, &ip.ip_src, m->m_pkthdr.rcvif) < 0) {
758 		m_freem(m);
759 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
760 		return (IPPROTO_DONE);
761 	}
762 
763 	otos = ip.ip_tos;
764 	m_adj(m, off);
765 
766 	if (m->m_len < sizeof(*ip6)) {
767 		m = m_pullup(m, sizeof(*ip6));
768 		if (!m) {
769 			SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE,
770 			    __LINE__);
771 			return (IPPROTO_DONE);
772 		}
773 	}
774 	ip6 = mtod(m, struct ip6_hdr *);
775 
776 	/*
777 	 * perform sanity check against inner src/dst.
778 	 * for source, perform ingress filter as well.
779 	 */
780 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
781 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
782 		m_freem(m);
783 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
784 		return (IPPROTO_DONE);
785 	}
786 
787 	/*
788 	 * reject packets with private address range.
789 	 * (requirement from RFC3056 section 2 1st paragraph)
790 	 */
791 	if ((IN6_IS_ADDR_6TO4(&ip6->ip6_src) && isrfc1918addr(&ip.ip_src)) ||
792 	    (IN6_IS_ADDR_6TO4(&ip6->ip6_dst) && isrfc1918addr(&ip.ip_dst))) {
793 		m_freem(m);
794 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
795 		return (IPPROTO_DONE);
796 	}
797 
798 	/*
799 	 * Ignore if the destination is the same stf interface because
800 	 * all of valid IPv6 outgoing traffic should go interfaces
801 	 * except for it.
802 	 */
803 	nh = fib6_lookup(sc->sc_fibnum, &ip6->ip6_dst, 0, 0, 0);
804 	if (nh == NULL) {
805 		m_free(m);
806 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
807 		return (IPPROTO_DONE);
808 	}
809 	if ((nh->nh_ifp == ifp) &&
810 	    (!IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &nh->gw6_sa.sin6_addr))) {
811 		m_free(m);
812 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
813 		return (IPPROTO_DONE);
814 	}
815 
816 	itos = IPV6_TRAFFIC_CLASS(ip6);
817 	if ((ifp->if_flags & IFF_LINK1) != 0)
818 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
819 	else
820 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
821 	ip6->ip6_flow &= ~htonl(0xff << 20);
822 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
823 
824 	m->m_pkthdr.rcvif = ifp;
825 
826 	if (bpf_peers_present(ifp->if_bpf)) {
827 		/*
828 		 * We need to prepend the address family as
829 		 * a four byte field.  Cons up a dummy header
830 		 * to pacify bpf.  This is safe because bpf
831 		 * will only read from the mbuf (i.e., it won't
832 		 * try to free it or keep a pointer a to it).
833 		 */
834 		u_int32_t af = AF_INET6;
835 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
836 	}
837 
838 	/*
839 	 * Put the packet to the network layer input queue according to the
840 	 * specified address family.
841 	 * See net/if_gif.c for possible issues with packet processing
842 	 * reorder due to extra queueing.
843 	 */
844 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
845 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
846 	M_SETFIB(m, ifp->if_fib);
847 	netisr_dispatch(NETISR_IPV6, m);
848 	SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
849 	return (IPPROTO_DONE);
850 }
851 
852 static struct sockaddr_in *
853 stf_getin4addr_in6(struct stf_softc *sc, struct sockaddr_in *sin,
854     struct in6_addr addr6, struct in6_addr mask6, struct in6_addr in6)
855 {
856        int i;
857        struct sockaddr_in *out;
858 
859 	/*
860 	* When (src addr & src mask) != (in6 & src mask),
861 	* the dst is not in the 6rd domain.  The IPv4 address must
862 	* not be used.
863 	*/
864 	for (i = 0; i < sizeof(addr6); i++) {
865 		if ((((u_char *)&addr6)[i] & ((u_char *)&mask6)[i]) !=
866 		    (((u_char *)&in6)[i] & ((u_char *)&mask6)[i])) {
867 			SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6,
868 			    &mask6, &in6, NULL);
869 			return (NULL);
870 		}
871 	}
872 
873 	/* After the mask check, use in6 instead of addr6. */
874 	out = stf_getin4addr(sc, sin, in6, mask6);
875 	SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6, &mask6, &in6, out);
876 	return (out);
877 }
878 
879 static struct sockaddr_in *
880 stf_getin4addr(struct stf_softc *sc, struct sockaddr_in *sin,
881     struct in6_addr addr6, struct in6_addr mask6)
882 {
883 	struct in_addr *in;
884 
885 	SDT_PROBE2(if_stf, , getin4addr, in, &addr6, &mask6);
886 
887 	memset(sin, 0, sizeof(*sin));
888 	in = &sin->sin_addr;
889 	if (IN6_IS_ADDR_6TO4(&addr6)) {
890 		/* 6to4 (RFC 3056) */
891 		bcopy(GET_V4(&addr6), in, sizeof(*in));
892 		if (isrfc1918addr(in))
893 			return (NULL);
894 	} else {
895 		/* 6rd (RFC 5569) */
896 		in_addr_t v4prefix;
897 		uint8_t *v6 = (uint8_t*)&addr6;
898 		uint64_t v6prefix;
899 		u_int plen;
900 		u_int v4suffixlen;
901 
902 		v4prefix = 0;
903 		if (sc->v4prefixlen < 32) {
904 			v4suffixlen = 32 - sc->v4prefixlen;
905 			v4prefix = ntohl(sc->srcv4_addr) &
906 			    (0xffffffffU << v4suffixlen);
907 		} else {
908 			MPASS(sc->v4prefixlen == 32);
909 			v4suffixlen = 32;
910 		}
911 
912 		plen = in6_mask2len(&mask6, NULL);
913 		if (plen > 64)
914 			return (NULL);
915 
916 		/* To make this simple we do not support prefixes longer than
917 		 * 64 bits. RFC5969 says "a 6rd delegated prefix SHOULD be /64
918 		 * or shorter." so this is a moderately safe assumption. */
919 		v6prefix = be64toh(*(uint64_t *)v6);
920 
921 		/* Shift away the v6 prefix itself. */
922 		v6prefix <<= plen;
923 		v6prefix >>= plen;
924 
925 		/* Now shift away everything after the v4 address. */
926 		v6prefix >>= 64 - plen - v4suffixlen;
927 
928 		sin->sin_addr.s_addr = htonl(v4prefix | (uint32_t)v6prefix);
929 	}
930 
931 	SDT_PROBE1(if_stf, , getin4addr, out, sin);
932 
933 	return (sin);
934 }
935 
936 static int
937 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
938 {
939 	struct ifaddr *ifa;
940 	struct ifdrv *ifd;
941 	struct ifreq *ifr;
942 	struct sockaddr_in sin4;
943 	struct stf_softc *sc_cur;
944 	struct stfv4args args;
945 	int error, mtu;
946 
947 	error = 0;
948 	sc_cur = ifp->if_softc;
949 
950 	switch (cmd) {
951 	case SIOCSDRVSPEC:
952 		ifd = (struct ifdrv *)data;
953 		error = priv_check(curthread, PRIV_NET_ADDIFADDR);
954 		if (error)
955 			break;
956 		if (ifd->ifd_cmd == STF6RD_SV4NET) {
957 			if (ifd->ifd_len != sizeof(args)) {
958 				error = EINVAL;
959 				break;
960 			}
961 			bzero(&args, sizeof(args));
962 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
963 			if (error)
964 				break;
965 
966 			if (args.v4_prefixlen < 1 || args.v4_prefixlen > 32) {
967 				error = EINVAL;
968 				break;
969 			}
970 
971 			bcopy(&args.srcv4_addr, &sc_cur->srcv4_addr,
972 			    sizeof(sc_cur->srcv4_addr));
973 			sc_cur->v4prefixlen = args.v4_prefixlen;
974 			SDT_PROBE3(if_stf, , ioctl, sv4net, sc_cur->srcv4_addr,
975 			    sc_cur->srcv4_addr, sc_cur->v4prefixlen);
976 		} else if (ifd->ifd_cmd == STF6RD_SBR) {
977 			if (ifd->ifd_len != sizeof(args)) {
978 				error = EINVAL;
979 				break;
980 			}
981 			bzero(&args, sizeof(args));
982 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
983 			if (error)
984 				break;
985 			sc_cur->braddr = args.braddr.s_addr;
986 			SDT_PROBE1(if_stf, , ioctl, sdstv4,
987 			    sc_cur->braddr);
988 		} else
989 			error = EINVAL;
990 		break;
991 	case SIOCGDRVSPEC:
992 		ifd = (struct ifdrv *)data;
993 		if (ifd->ifd_cmd != STF6RD_GV4NET) {
994 			error = EINVAL;
995 			break;
996 		}
997 		if (ifd->ifd_len != sizeof(args)) {
998 			error = EINVAL;
999 			break;
1000 		}
1001 		bzero(&args, sizeof(args));
1002 		args.srcv4_addr.s_addr = sc_cur->srcv4_addr;
1003 		args.braddr.s_addr = sc_cur->braddr;
1004 		args.v4_prefixlen = sc_cur->v4prefixlen;
1005 		error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
1006 		break;
1007 	case SIOCSIFADDR:
1008 		ifa = (struct ifaddr *)data;
1009 		SDT_PROBE1(if_stf, , ioctl, ifaddr, ifa);
1010 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
1011 			error = EAFNOSUPPORT;
1012 			break;
1013 		}
1014 		if (stf_getin4addr(sc_cur, &sin4,
1015 		    satosin6(ifa->ifa_addr)->sin6_addr,
1016 		    satosin6(ifa->ifa_netmask)->sin6_addr) == NULL) {
1017 			error = EINVAL;
1018 			break;
1019 		}
1020 		ifp->if_flags |= IFF_UP;
1021 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1022 		break;
1023 
1024 	case SIOCADDMULTI:
1025 	case SIOCDELMULTI:
1026 		ifr = (struct ifreq *)data;
1027 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
1028 			;
1029 		else
1030 			error = EAFNOSUPPORT;
1031 		break;
1032 
1033 	case SIOCGIFMTU:
1034 		break;
1035 
1036 	case SIOCSIFMTU:
1037 		ifr = (struct ifreq *)data;
1038 		mtu = ifr->ifr_mtu;
1039 		/* RFC 4213 3.2 ideal world MTU */
1040 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
1041 			return (EINVAL);
1042 		ifp->if_mtu = mtu;
1043 		break;
1044 
1045 	default:
1046 		error = EINVAL;
1047 		break;
1048 	}
1049 
1050 	return (error);
1051 }
1052