xref: /freebsd/sys/net/if_stf.c (revision 3c4ba5f55438f7afd4f4b0b56f88f2bb505fd6a6)
1 /*	$FreeBSD$	*/
2 /*	$KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $	*/
3 
4 /*-
5  * SPDX-License-Identifier: BSD-3-Clause
6  *
7  * Copyright (C) 2000 WIDE Project.
8  * Copyright (c) 2010 Hiroki Sato <hrs@FreeBSD.org>
9  * Copyright (c) 2013 Ermal Luci <eri@FreeBSD.org>
10  * Copyright (c) 2017-2021 Rubicon Communications, LLC (Netgate)
11  * All rights reserved.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the project nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  */
37 
38 /*
39  * 6to4 interface, based on RFC3056.
40  *
41  * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
42  * There is no address mapping defined from IPv6 multicast address to IPv4
43  * address.  Therefore, we do not have IFF_MULTICAST on the interface.
44  *
45  * Due to the lack of address mapping for link-local addresses, we cannot
46  * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
47  * packets to link-local multicast addresses (ff02::x).
48  *
49  * Here are interesting symptoms due to the lack of link-local address:
50  *
51  * Unicast routing exchange:
52  * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
53  *   and link-local addresses as nexthop.
54  * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
55  *   assigned to the link, and makes use of them.  Also, HELLO packets use
56  *   link-local multicast addresses (ff02::5 and ff02::6).
57  * - BGP4+: Maybe.  You can only use global address as nexthop, and global
58  *   address as TCP endpoint address.
59  *
60  * Multicast routing protocols:
61  * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
62  *   Adjacent PIM routers must be configured manually (is it really spec-wise
63  *   correct thing to do?).
64  *
65  * ICMPv6:
66  * - Redirects cannot be used due to the lack of link-local address.
67  *
68  * stf interface does not have, and will not need, a link-local address.
69  * It seems to have no real benefit and does not help the above symptoms much.
70  * Even if we assign link-locals to interface, we cannot really
71  * use link-local unicast/multicast on top of 6to4 cloud (since there's no
72  * encapsulation defined for link-local address), and the above analysis does
73  * not change.  RFC3056 does not mandate the assignment of link-local address
74  * either.
75  *
76  * 6to4 interface has security issues.  Refer to
77  * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
78  * for details.  The code tries to filter out some of malicious packets.
79  * Note that there is no way to be 100% secure.
80  */
81 
82 #include <sys/param.h>
83 #include <sys/systm.h>
84 #include <sys/socket.h>
85 #include <sys/sockio.h>
86 #include <sys/mbuf.h>
87 #include <sys/endian.h>
88 #include <sys/errno.h>
89 #include <sys/kernel.h>
90 #include <sys/lock.h>
91 #include <sys/module.h>
92 #include <sys/priv.h>
93 #include <sys/proc.h>
94 #include <sys/queue.h>
95 #include <sys/sdt.h>
96 #include <sys/sysctl.h>
97 #include <machine/cpu.h>
98 
99 #include <sys/malloc.h>
100 
101 #include <net/if.h>
102 #include <net/if_var.h>
103 #include <net/if_private.h>
104 #include <net/if_clone.h>
105 #include <net/route.h>
106 #include <net/route/nhop.h>
107 #include <net/netisr.h>
108 #include <net/if_stf.h>
109 #include <net/if_types.h>
110 #include <net/vnet.h>
111 
112 #include <netinet/in.h>
113 #include <netinet/in_fib.h>
114 #include <netinet/in_systm.h>
115 #include <netinet/ip.h>
116 #include <netinet/ip_var.h>
117 #include <netinet/in_var.h>
118 
119 #include <netinet/ip6.h>
120 #include <netinet6/in6_fib.h>
121 #include <netinet6/ip6_var.h>
122 #include <netinet6/in6_var.h>
123 #include <netinet/ip_ecn.h>
124 
125 #include <netinet/ip_encap.h>
126 
127 #include <machine/stdarg.h>
128 
129 #include <net/bpf.h>
130 
131 #include <security/mac/mac_framework.h>
132 
133 SDT_PROVIDER_DEFINE(if_stf);
134 SDT_PROBE_DEFINE3(if_stf, , encapcheck, in, "struct mbuf *", "int", "int");
135 SDT_PROBE_DEFINE0(if_stf, , encapcheck, accept);
136 SDT_PROBE_DEFINE3(if_stf, , getsrcifa6, in, "struct ifnet *",
137     "struct in6_addr *", "struct in6_addr *");
138 SDT_PROBE_DEFINE2(if_stf, , getsrcifa6, found, "struct in6_addr *",
139     "struct in6_addr *");
140 SDT_PROBE_DEFINE0(if_stf, , getsrcifa6, notfound);
141 
142 SDT_PROBE_DEFINE4(if_stf, , stf_output, in, "struct ifnet *", "struct mbuf *",
143     "struct sockaddr *", "struct route *");
144 SDT_PROBE_DEFINE2(if_stf, , stf_output, error, "int", "int");
145 SDT_PROBE_DEFINE1(if_stf, , stf_output, out, "int");
146 
147 SDT_PROBE_DEFINE3(if_stf, , checkaddr6, in, "struct stf_softc *",
148     "struct in6_addr *", "struct ifnet *");
149 SDT_PROBE_DEFINE2(if_stf, , checkaddr6, out, "int", "int");
150 
151 SDT_PROBE_DEFINE3(if_stf, , stf_input, in, "struct mbuf *", "int", "int");
152 SDT_PROBE_DEFINE2(if_stf, , stf_input, out, "int", "int");
153 
154 SDT_PROBE_DEFINE3(if_stf, , ioctl, sv4net, "struct in_addr *",
155     "struct in_addr *", "int");
156 SDT_PROBE_DEFINE1(if_stf, , ioctl, sdstv4, "struct in_addr *");
157 SDT_PROBE_DEFINE1(if_stf, , ioctl, ifaddr, "struct ifaddr *");
158 
159 SDT_PROBE_DEFINE4(if_stf, , getin4addr_in6, out, "struct in6_addr *",
160     "struct in6_addr *", "struct in6_addr *", "struct sockaddr_in *");
161 
162 SDT_PROBE_DEFINE2(if_stf, , getin4addr, in, "struct in6_addr *", "struct in6_addr *");
163 SDT_PROBE_DEFINE1(if_stf, , getin4addr, out, "struct sockaddr_in *");
164 
165 SYSCTL_DECL(_net_link);
166 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
167     "6to4 Interface");
168 
169 static int stf_permit_rfc1918 = 0;
170 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
171     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
172 
173 #define STFUNIT		0
174 
175 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
176 
177 /*
178  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
179  * struct in_addr *; use bcopy() instead.
180  */
181 #define GET_V4(x)	(&(x)->s6_addr16[1])
182 
183 struct stf_softc {
184 	struct ifnet	*sc_ifp;
185 	in_addr_t	braddr;		/* Border relay IPv4 address */
186 	in_addr_t	srcv4_addr;	/* Our IPv4 WAN address */
187 	u_int		v4prefixlen;	/* How much of the v4 address to include in our address. */
188 	u_int		sc_fibnum;
189 	const struct encaptab *encap_cookie;
190 };
191 #define STF2IFP(sc)	((sc)->sc_ifp)
192 
193 static const char stfname[] = "stf";
194 
195 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
196 static const int ip_stf_ttl = 40;
197 
198 static int in_stf_input(struct mbuf *, int, int, void *);
199 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
200 
201 static int stfmodevent(module_t, int, void *);
202 static int stf_encapcheck(const struct mbuf *, int, int, void *);
203 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
204 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
205 	struct route *);
206 static int isrfc1918addr(struct in_addr *);
207 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
208 	struct ifnet *);
209 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
210 	struct ifnet *);
211 static struct sockaddr_in *stf_getin4addr_in6(struct stf_softc *,
212 	struct sockaddr_in *, struct in6_addr, struct in6_addr,
213 	struct in6_addr);
214 static struct sockaddr_in *stf_getin4addr(struct stf_softc *,
215 	struct sockaddr_in *, struct in6_addr, struct in6_addr);
216 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
217 
218 VNET_DEFINE_STATIC(struct if_clone *, stf_cloner);
219 #define V_stf_cloner	VNET(stf_cloner)
220 
221 static const struct encap_config ipv4_encap_cfg = {
222 	.proto = IPPROTO_IPV6,
223 	.min_length = sizeof(struct ip),
224 	.exact_match = (sizeof(in_addr_t) << 3) + 8,
225 	.check = stf_encapcheck,
226 	.input = in_stf_input
227 };
228 
229 static int
230 stf_clone_match(struct if_clone *ifc, const char *name)
231 {
232 	int i;
233 
234 	for(i = 0; stfnames[i] != NULL; i++) {
235 		if (strcmp(stfnames[i], name) == 0)
236 			return (1);
237 	}
238 
239 	return (0);
240 }
241 
242 static int
243 stf_clone_create(struct if_clone *ifc, char *name, size_t len,
244     struct ifc_data *ifd, struct ifnet **ifpp)
245 {
246 	char *dp;
247 	int err, unit, wildcard;
248 	struct stf_softc *sc;
249 	struct ifnet *ifp;
250 
251 	err = ifc_name2unit(name, &unit);
252 	if (err != 0)
253 		return (err);
254 	wildcard = (unit < 0);
255 
256 	/*
257 	 * We can only have one unit, but since unit allocation is
258 	 * already locked, we use it to keep from allocating extra
259 	 * interfaces.
260 	 */
261 	unit = STFUNIT;
262 	err = ifc_alloc_unit(ifc, &unit);
263 	if (err != 0)
264 		return (err);
265 
266 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
267 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
268 	if (ifp == NULL) {
269 		free(sc, M_STF);
270 		ifc_free_unit(ifc, unit);
271 		return (ENOSPC);
272 	}
273 	ifp->if_softc = sc;
274 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
275 
276 	/*
277 	 * Set the name manually rather then using if_initname because
278 	 * we don't conform to the default naming convention for interfaces.
279 	 * In the wildcard case, we need to update the name.
280 	 */
281 	if (wildcard) {
282 		for (dp = name; *dp != '\0'; dp++);
283 		if (snprintf(dp, len - (dp-name), "%d", unit) >
284 		    len - (dp-name) - 1) {
285 			/*
286 			 * This can only be a programmer error and
287 			 * there's no straightforward way to recover if
288 			 * it happens.
289 			 */
290 			panic("if_clone_create(): interface name too long");
291 		}
292 	}
293 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
294 	ifp->if_dname = stfname;
295 	ifp->if_dunit = IF_DUNIT_NONE;
296 
297 	sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK);
298 
299 	ifp->if_mtu    = IPV6_MMTU;
300 	ifp->if_ioctl  = stf_ioctl;
301 	ifp->if_output = stf_output;
302 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
303 	if_attach(ifp);
304 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
305 	*ifpp = ifp;
306 
307 	return (0);
308 }
309 
310 static int
311 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
312 {
313 	struct stf_softc *sc = ifp->if_softc;
314 	int err __unused;
315 
316 	err = ip_encap_detach(sc->encap_cookie);
317 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
318 	bpfdetach(ifp);
319 	if_detach(ifp);
320 	if_free(ifp);
321 
322 	free(sc, M_STF);
323 	ifc_free_unit(ifc, STFUNIT);
324 
325 	return (0);
326 }
327 
328 static void
329 vnet_stf_init(const void *unused __unused)
330 {
331 	struct if_clone_addreq req = {
332 		.match_f = stf_clone_match,
333 		.create_f = stf_clone_create,
334 		.destroy_f = stf_clone_destroy,
335 	};
336 	V_stf_cloner = ifc_attach_cloner(stfname, &req);
337 }
338 VNET_SYSINIT(vnet_stf_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_init, NULL);
339 
340 static void
341 vnet_stf_uninit(const void *unused __unused)
342 {
343 	if_clone_detach(V_stf_cloner);
344 	V_stf_cloner = NULL;
345 }
346 VNET_SYSUNINIT(vnet_stf_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_uninit,
347     NULL);
348 
349 static int
350 stfmodevent(module_t mod, int type, void *data)
351 {
352 
353 	switch (type) {
354 	case MOD_LOAD:
355 		/* Done in vnet_stf_init() */
356 		break;
357 	case MOD_UNLOAD:
358 		/* Done in vnet_stf_uninit() */
359 		break;
360 	default:
361 		return (EOPNOTSUPP);
362 	}
363 
364 	return (0);
365 }
366 
367 static moduledata_t stf_mod = {
368 	"if_stf",
369 	stfmodevent,
370 	0
371 };
372 
373 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
374 MODULE_VERSION(if_stf, 2);
375 
376 static int
377 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
378 {
379 	struct ip ip;
380 	struct stf_softc *sc;
381 	struct in6_addr addr6, mask6;
382 	struct sockaddr_in sin4addr, sin4mask;
383 
384 	SDT_PROBE3(if_stf, , encapcheck, in, m, off, proto);
385 
386 	sc = (struct stf_softc *)arg;
387 	if (sc == NULL)
388 		return (0);
389 
390 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
391 		return (0);
392 
393 	/* IFF_LINK0 means "no decapsulation" */
394 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
395 		return (0);
396 
397 	if (proto != IPPROTO_IPV6)
398 		return (0);
399 
400 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
401 
402 	if (ip.ip_v != 4)
403 		return (0);
404 
405 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
406 		return (0);
407 
408 	if (sc->srcv4_addr != INADDR_ANY) {
409 		sin4addr.sin_addr.s_addr = sc->srcv4_addr;
410 		sin4addr.sin_family = AF_INET;
411 	} else
412 		if (stf_getin4addr(sc, &sin4addr, addr6, mask6) == NULL)
413 			return (0);
414 
415 	if (sin4addr.sin_addr.s_addr != ip.ip_dst.s_addr)
416 		return (0);
417 
418 	if (IN6_IS_ADDR_6TO4(&addr6)) {
419 		/*
420 		 * 6to4 (RFC 3056).
421 		 * Check if IPv4 src matches the IPv4 address derived
422 		 * from the local 6to4 address masked by prefixmask.
423 		 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
424 		 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
425 		 */
426 		memcpy(&sin4mask.sin_addr, GET_V4(&mask6),
427 		    sizeof(sin4mask.sin_addr));
428 		if ((sin4addr.sin_addr.s_addr & sin4mask.sin_addr.s_addr) !=
429 		    (ip.ip_src.s_addr & sin4mask.sin_addr.s_addr))
430 			return (0);
431 	} else {
432 		/* 6rd (RFC 5569) */
433 		/*
434 		 * No restriction on the src address in the case of
435 		 * 6rd because the stf(4) interface always has a
436 		 * prefix which covers whole of IPv4 src address
437 		 * range.  So, stf_output() will catch all of
438 		 * 6rd-capsuled IPv4 traffic with suspicious inner dst
439 		 * IPv4 address (i.e. the IPv6 destination address is
440 		 * one the admin does not like to route to outside),
441 		 * and then it discard them silently.
442 		 */
443 	}
444 
445 	SDT_PROBE0(if_stf, , encapcheck, accept);
446 
447 	/* stf interface makes single side match only */
448 	return (32);
449 }
450 
451 static int
452 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
453 {
454 	struct ifaddr *ia;
455 	struct in_ifaddr *ia4;
456 	struct in6_addr addr6, mask6;
457 	struct sockaddr_in sin4;
458 	struct stf_softc *sc;
459 	struct in_addr in;
460 
461 	NET_EPOCH_ASSERT();
462 
463 	sc = ifp->if_softc;
464 
465 	SDT_PROBE3(if_stf, , getsrcifa6, in, ifp, addr, mask);
466 
467 	CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
468 		if (ia->ifa_addr->sa_family != AF_INET6)
469 			continue;
470 
471 		addr6 = *IFA_IN6(ia);
472 		mask6 = *IFA_MASKIN6(ia);
473 		if (sc->srcv4_addr != INADDR_ANY)
474 			bcopy(&sc->srcv4_addr, &in, sizeof(in));
475 		else {
476 			if (stf_getin4addr(sc, &sin4, addr6, mask6) == NULL)
477 				continue;
478 			bcopy(&sin4.sin_addr, &in, sizeof(in));
479 		}
480 
481 		CK_LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
482 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
483 				break;
484 		if (ia4 == NULL)
485 			continue;
486 
487 		*addr = addr6;
488 		*mask = mask6;
489 
490 		SDT_PROBE2(if_stf, , getsrcifa6, found, addr, mask);
491 
492 		return (0);
493 	}
494 
495 	SDT_PROBE0(if_stf, , getsrcifa6, notfound);
496 
497 	return (ENOENT);
498 }
499 
500 static int
501 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
502     struct route *ro)
503 {
504 	struct stf_softc *sc;
505 	const struct sockaddr_in6 *dst6;
506 	struct sockaddr_in dst4, src4;
507 	u_int8_t tos;
508 	struct ip *ip;
509 	struct ip6_hdr *ip6;
510 	struct in6_addr addr6, mask6;
511 	int error;
512 
513 	SDT_PROBE4(if_stf, , stf_output, in, ifp, m, dst, ro);
514 
515 #ifdef MAC
516 	error = mac_ifnet_check_transmit(ifp, m);
517 	if (error) {
518 		m_freem(m);
519 		SDT_PROBE2(if_stf, , stf_output, error, error, __LINE__);
520 		return (error);
521 	}
522 #endif
523 
524 	sc = ifp->if_softc;
525 	dst6 = (const struct sockaddr_in6 *)dst;
526 
527 	/* just in case */
528 	if ((ifp->if_flags & IFF_UP) == 0) {
529 		m_freem(m);
530 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
531 		SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__);
532 		return (ENETDOWN);
533 	}
534 
535 	/*
536 	 * If we don't have an ip4 address that match my inner ip6 address,
537 	 * we shouldn't generate output.  Without this check, we'll end up
538 	 * using wrong IPv4 source.
539 	 */
540 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
541 		m_freem(m);
542 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
543 		SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__);
544 		return (ENETDOWN);
545 	}
546 
547 	if (m->m_len < sizeof(*ip6)) {
548 		m = m_pullup(m, sizeof(*ip6));
549 		if (!m) {
550 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
551 			SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS,
552 			    __LINE__);
553 			return (ENOBUFS);
554 		}
555 	}
556 	ip6 = mtod(m, struct ip6_hdr *);
557 	tos = IPV6_TRAFFIC_CLASS(ip6);
558 
559 	/*
560 	 * Pickup the right outer dst addr from the list of candidates.
561 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
562 	 */
563 	if (stf_getin4addr_in6(sc, &dst4, addr6, mask6,
564 	    ip6->ip6_dst) == NULL) {
565 		if (sc->braddr != INADDR_ANY)
566 			dst4.sin_addr.s_addr = sc->braddr;
567 		else if (stf_getin4addr_in6(sc, &dst4, addr6, mask6,
568 		    dst6->sin6_addr) == NULL) {
569 			m_freem(m);
570 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
571 			SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH,
572 			    __LINE__);
573 			return (ENETUNREACH);
574 		}
575 	}
576 
577 	if (bpf_peers_present(ifp->if_bpf)) {
578 		/*
579 		 * We need to prepend the address family as
580 		 * a four byte field.  Cons up a dummy header
581 		 * to pacify bpf.  This is safe because bpf
582 		 * will only read from the mbuf (i.e., it won't
583 		 * try to free it or keep a pointer a to it).
584 		 */
585 		u_int af = AF_INET6;
586 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
587 	}
588 
589 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
590 	if (m == NULL) {
591 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
592 		SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS, __LINE__);
593 		return (ENOBUFS);
594 	}
595 	ip = mtod(m, struct ip *);
596 
597 	bzero(ip, sizeof(*ip));
598 
599 	if (sc->srcv4_addr != INADDR_ANY)
600 		src4.sin_addr.s_addr = sc->srcv4_addr;
601 	else if (stf_getin4addr(sc, &src4, addr6, mask6) == NULL) {
602 		m_freem(m);
603 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
604 		SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH, __LINE__);
605 		return (ENETUNREACH);
606 	}
607 	bcopy(&src4.sin_addr, &ip->ip_src, sizeof(ip->ip_src));
608 	bcopy(&dst4.sin_addr, &ip->ip_dst, sizeof(ip->ip_dst));
609 
610 	ip->ip_p = IPPROTO_IPV6;
611 	ip->ip_ttl = ip_stf_ttl;
612 	ip->ip_len = htons(m->m_pkthdr.len);
613 	if (ifp->if_flags & IFF_LINK1)
614 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
615 	else
616 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
617 
618 	M_SETFIB(m, sc->sc_fibnum);
619 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
620 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
621 
622 	SDT_PROBE1(if_stf, , stf_output, out, error);
623 	return (error);
624 }
625 
626 static int
627 isrfc1918addr(struct in_addr *in)
628 {
629 	/*
630 	 * returns 1 if private address range:
631 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
632 	 */
633 	if (stf_permit_rfc1918 == 0 && (
634 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
635 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
636 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
637 		return (1);
638 
639 	return (0);
640 }
641 
642 static int
643 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
644 {
645 	struct in_ifaddr *ia4;
646 
647 	/*
648 	 * reject packets with the following address:
649 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
650 	 */
651 	if (IN_MULTICAST(ntohl(in->s_addr)))
652 		return (-1);
653 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
654 	case 0: case 127: case 255:
655 		return (-1);
656 	}
657 
658 	/*
659 	 * reject packets with broadcast
660 	 */
661 	CK_STAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
662 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
663 			continue;
664 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
665 			return (-1);
666 		}
667 	}
668 
669 	/*
670 	 * perform ingress filter
671 	 */
672 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
673 		struct nhop_object *nh;
674 
675 		NET_EPOCH_ASSERT();
676 		nh = fib4_lookup(sc->sc_fibnum, *in, 0, 0, 0);
677 		if (nh == NULL)
678 			return (-1);
679 
680 		if (nh->nh_ifp != inifp)
681 			return (-1);
682 	}
683 
684 	return (0);
685 }
686 
687 static int
688 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
689 {
690 	SDT_PROBE3(if_stf, , checkaddr6, in, sc, in6, inifp);
691 
692 	/*
693 	 * check 6to4 addresses
694 	 */
695 	if (IN6_IS_ADDR_6TO4(in6)) {
696 		struct in_addr in4;
697 		int ret;
698 
699 		bcopy(GET_V4(in6), &in4, sizeof(in4));
700 		ret = stf_checkaddr4(sc, &in4, inifp);
701 		SDT_PROBE2(if_stf, , checkaddr6, out, ret, __LINE__);
702 		return (ret);
703 	}
704 
705 	/*
706 	 * reject anything that look suspicious.  the test is implemented
707 	 * in ip6_input too, but we check here as well to
708 	 * (1) reject bad packets earlier, and
709 	 * (2) to be safe against future ip6_input change.
710 	 */
711 	if (IN6_IS_ADDR_V4COMPAT(in6)) {
712 		SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__);
713 		return (-1);
714 	}
715 
716 	if (IN6_IS_ADDR_V4MAPPED(in6)) {
717 		SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__);
718 		return (-1);
719 	}
720 
721 	SDT_PROBE2(if_stf, , checkaddr6, out, 0, __LINE__);
722 	return (0);
723 }
724 
725 static int
726 in_stf_input(struct mbuf *m, int off, int proto, void *arg)
727 {
728 	struct stf_softc *sc = arg;
729 	struct ip ip;
730 	struct ip6_hdr *ip6;
731 	u_int8_t otos, itos;
732 	struct ifnet *ifp;
733 	struct nhop_object *nh;
734 
735 	NET_EPOCH_ASSERT();
736 
737 	SDT_PROBE3(if_stf, , stf_input, in, m, off, proto);
738 
739 	if (proto != IPPROTO_IPV6) {
740 		m_freem(m);
741 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
742 		return (IPPROTO_DONE);
743 	}
744 
745 	m_copydata(m, 0, sizeof(struct ip), (caddr_t)&ip);
746 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
747 		m_freem(m);
748 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
749 		return (IPPROTO_DONE);
750 	}
751 
752 	ifp = STF2IFP(sc);
753 
754 #ifdef MAC
755 	mac_ifnet_create_mbuf(ifp, m);
756 #endif
757 
758 	/*
759 	 * perform sanity check against outer src/dst.
760 	 * for source, perform ingress filter as well.
761 	 */
762 	if (stf_checkaddr4(sc, &ip.ip_dst, NULL) < 0 ||
763 	    stf_checkaddr4(sc, &ip.ip_src, m->m_pkthdr.rcvif) < 0) {
764 		m_freem(m);
765 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
766 		return (IPPROTO_DONE);
767 	}
768 
769 	otos = ip.ip_tos;
770 	m_adj(m, off);
771 
772 	if (m->m_len < sizeof(*ip6)) {
773 		m = m_pullup(m, sizeof(*ip6));
774 		if (!m) {
775 			SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE,
776 			    __LINE__);
777 			return (IPPROTO_DONE);
778 		}
779 	}
780 	ip6 = mtod(m, struct ip6_hdr *);
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 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
790 		return (IPPROTO_DONE);
791 	}
792 
793 	/*
794 	 * reject packets with private address range.
795 	 * (requirement from RFC3056 section 2 1st paragraph)
796 	 */
797 	if ((IN6_IS_ADDR_6TO4(&ip6->ip6_src) && isrfc1918addr(&ip.ip_src)) ||
798 	    (IN6_IS_ADDR_6TO4(&ip6->ip6_dst) && isrfc1918addr(&ip.ip_dst))) {
799 		m_freem(m);
800 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
801 		return (IPPROTO_DONE);
802 	}
803 
804 	/*
805 	 * Ignore if the destination is the same stf interface because
806 	 * all of valid IPv6 outgoing traffic should go interfaces
807 	 * except for it.
808 	 */
809 	nh = fib6_lookup(sc->sc_fibnum, &ip6->ip6_dst, 0, 0, 0);
810 	if (nh == NULL) {
811 		m_free(m);
812 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
813 		return (IPPROTO_DONE);
814 	}
815 	if ((nh->nh_ifp == ifp) &&
816 	    (!IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &nh->gw6_sa.sin6_addr))) {
817 		m_free(m);
818 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
819 		return (IPPROTO_DONE);
820 	}
821 
822 	itos = IPV6_TRAFFIC_CLASS(ip6);
823 	if ((ifp->if_flags & IFF_LINK1) != 0)
824 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
825 	else
826 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
827 	ip6->ip6_flow &= ~htonl(0xff << 20);
828 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
829 
830 	m->m_pkthdr.rcvif = ifp;
831 
832 	if (bpf_peers_present(ifp->if_bpf)) {
833 		/*
834 		 * We need to prepend the address family as
835 		 * a four byte field.  Cons up a dummy header
836 		 * to pacify bpf.  This is safe because bpf
837 		 * will only read from the mbuf (i.e., it won't
838 		 * try to free it or keep a pointer a to it).
839 		 */
840 		u_int32_t af = AF_INET6;
841 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
842 	}
843 
844 	/*
845 	 * Put the packet to the network layer input queue according to the
846 	 * specified address family.
847 	 * See net/if_gif.c for possible issues with packet processing
848 	 * reorder due to extra queueing.
849 	 */
850 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
851 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
852 	M_SETFIB(m, ifp->if_fib);
853 	netisr_dispatch(NETISR_IPV6, m);
854 	SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
855 	return (IPPROTO_DONE);
856 }
857 
858 static struct sockaddr_in *
859 stf_getin4addr_in6(struct stf_softc *sc, struct sockaddr_in *sin,
860     struct in6_addr addr6, struct in6_addr mask6, struct in6_addr in6)
861 {
862        int i;
863        struct sockaddr_in *out;
864 
865 	/*
866 	* When (src addr & src mask) != (in6 & src mask),
867 	* the dst is not in the 6rd domain.  The IPv4 address must
868 	* not be used.
869 	*/
870 	for (i = 0; i < sizeof(addr6); i++) {
871 		if ((((u_char *)&addr6)[i] & ((u_char *)&mask6)[i]) !=
872 		    (((u_char *)&in6)[i] & ((u_char *)&mask6)[i])) {
873 			SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6,
874 			    &mask6, &in6, NULL);
875 			return (NULL);
876 		}
877 	}
878 
879 	/* After the mask check, use in6 instead of addr6. */
880 	out = stf_getin4addr(sc, sin, in6, mask6);
881 	SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6, &mask6, &in6, out);
882 	return (out);
883 }
884 
885 static struct sockaddr_in *
886 stf_getin4addr(struct stf_softc *sc, struct sockaddr_in *sin,
887     struct in6_addr addr6, struct in6_addr mask6)
888 {
889 	struct in_addr *in;
890 
891 	SDT_PROBE2(if_stf, , getin4addr, in, &addr6, &mask6);
892 
893 	memset(sin, 0, sizeof(*sin));
894 	in = &sin->sin_addr;
895 	if (IN6_IS_ADDR_6TO4(&addr6)) {
896 		/* 6to4 (RFC 3056) */
897 		bcopy(GET_V4(&addr6), in, sizeof(*in));
898 		if (isrfc1918addr(in))
899 			return (NULL);
900 	} else {
901 		/* 6rd (RFC 5569) */
902 		in_addr_t v4prefix;
903 		uint8_t *v6 = (uint8_t*)&addr6;
904 		uint64_t v6prefix;
905 		u_int plen;
906 		u_int v4suffixlen;
907 
908 		v4prefix = 0;
909 		if (sc->v4prefixlen < 32) {
910 			v4suffixlen = 32 - sc->v4prefixlen;
911 			v4prefix = ntohl(sc->srcv4_addr) &
912 			    (0xffffffffU << v4suffixlen);
913 		} else {
914 			MPASS(sc->v4prefixlen == 32);
915 			v4suffixlen = 32;
916 		}
917 
918 		plen = in6_mask2len(&mask6, NULL);
919 		if (plen > 64)
920 			return (NULL);
921 
922 		/* To make this simple we do not support prefixes longer than
923 		 * 64 bits. RFC5969 says "a 6rd delegated prefix SHOULD be /64
924 		 * or shorter." so this is a moderately safe assumption. */
925 		v6prefix = be64toh(*(uint64_t *)v6);
926 
927 		/* Shift away the v6 prefix itself. */
928 		v6prefix <<= plen;
929 		v6prefix >>= plen;
930 
931 		/* Now shift away everything after the v4 address. */
932 		v6prefix >>= 64 - plen - v4suffixlen;
933 
934 		sin->sin_addr.s_addr = htonl(v4prefix | (uint32_t)v6prefix);
935 	}
936 
937 	SDT_PROBE1(if_stf, , getin4addr, out, sin);
938 
939 	return (sin);
940 }
941 
942 static int
943 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
944 {
945 	struct ifaddr *ifa;
946 	struct ifdrv *ifd;
947 	struct ifreq *ifr;
948 	struct sockaddr_in sin4;
949 	struct stf_softc *sc_cur;
950 	struct stfv4args args;
951 	int error, mtu;
952 
953 	error = 0;
954 	sc_cur = ifp->if_softc;
955 
956 	switch (cmd) {
957 	case SIOCSDRVSPEC:
958 		ifd = (struct ifdrv *)data;
959 		error = priv_check(curthread, PRIV_NET_ADDIFADDR);
960 		if (error)
961 			break;
962 		if (ifd->ifd_cmd == STF6RD_SV4NET) {
963 			if (ifd->ifd_len != sizeof(args)) {
964 				error = EINVAL;
965 				break;
966 			}
967 			bzero(&args, sizeof(args));
968 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
969 			if (error)
970 				break;
971 
972 			if (args.v4_prefixlen < 1 || args.v4_prefixlen > 32) {
973 				error = EINVAL;
974 				break;
975 			}
976 
977 			bcopy(&args.srcv4_addr, &sc_cur->srcv4_addr,
978 			    sizeof(sc_cur->srcv4_addr));
979 			sc_cur->v4prefixlen = args.v4_prefixlen;
980 			SDT_PROBE3(if_stf, , ioctl, sv4net, sc_cur->srcv4_addr,
981 			    sc_cur->srcv4_addr, sc_cur->v4prefixlen);
982 		} else if (ifd->ifd_cmd == STF6RD_SBR) {
983 			if (ifd->ifd_len != sizeof(args)) {
984 				error = EINVAL;
985 				break;
986 			}
987 			bzero(&args, sizeof(args));
988 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
989 			if (error)
990 				break;
991 			sc_cur->braddr = args.braddr.s_addr;
992 			SDT_PROBE1(if_stf, , ioctl, sdstv4,
993 			    sc_cur->braddr);
994 		} else
995 			error = EINVAL;
996 		break;
997 	case SIOCGDRVSPEC:
998 		ifd = (struct ifdrv *)data;
999 		if (ifd->ifd_cmd != STF6RD_GV4NET) {
1000 			error = EINVAL;
1001 			break;
1002 		}
1003 		if (ifd->ifd_len != sizeof(args)) {
1004 			error = EINVAL;
1005 			break;
1006 		}
1007 		bzero(&args, sizeof(args));
1008 		args.srcv4_addr.s_addr = sc_cur->srcv4_addr;
1009 		args.braddr.s_addr = sc_cur->braddr;
1010 		args.v4_prefixlen = sc_cur->v4prefixlen;
1011 		error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
1012 		break;
1013 	case SIOCSIFADDR:
1014 		ifa = (struct ifaddr *)data;
1015 		SDT_PROBE1(if_stf, , ioctl, ifaddr, ifa);
1016 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
1017 			error = EAFNOSUPPORT;
1018 			break;
1019 		}
1020 		if (stf_getin4addr(sc_cur, &sin4,
1021 		    satosin6(ifa->ifa_addr)->sin6_addr,
1022 		    satosin6(ifa->ifa_netmask)->sin6_addr) == NULL) {
1023 			error = EINVAL;
1024 			break;
1025 		}
1026 		ifp->if_flags |= IFF_UP;
1027 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1028 		break;
1029 
1030 	case SIOCADDMULTI:
1031 	case SIOCDELMULTI:
1032 		ifr = (struct ifreq *)data;
1033 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
1034 			;
1035 		else
1036 			error = EAFNOSUPPORT;
1037 		break;
1038 
1039 	case SIOCGIFMTU:
1040 		break;
1041 
1042 	case SIOCSIFMTU:
1043 		ifr = (struct ifreq *)data;
1044 		mtu = ifr->ifr_mtu;
1045 		/* RFC 4213 3.2 ideal world MTU */
1046 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
1047 			return (EINVAL);
1048 		ifp->if_mtu = mtu;
1049 		break;
1050 
1051 	default:
1052 		error = EINVAL;
1053 		break;
1054 	}
1055 
1056 	return (error);
1057 }
1058