xref: /freebsd/sys/net/if_stf.c (revision 5f4c09dd85bff675e0ca63c55ea3c517e0fddfcc)
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 	if (ifp == NULL) {
268 		free(sc, M_STF);
269 		ifc_free_unit(ifc, unit);
270 		return (ENOSPC);
271 	}
272 	ifp->if_softc = sc;
273 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
274 
275 	/*
276 	 * Set the name manually rather then using if_initname because
277 	 * we don't conform to the default naming convention for interfaces.
278 	 * In the wildcard case, we need to update the name.
279 	 */
280 	if (wildcard) {
281 		for (dp = name; *dp != '\0'; dp++);
282 		if (snprintf(dp, len - (dp-name), "%d", unit) >
283 		    len - (dp-name) - 1) {
284 			/*
285 			 * This can only be a programmer error and
286 			 * there's no straightforward way to recover if
287 			 * it happens.
288 			 */
289 			panic("if_clone_create(): interface name too long");
290 		}
291 	}
292 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
293 	ifp->if_dname = stfname;
294 	ifp->if_dunit = IF_DUNIT_NONE;
295 
296 	sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK);
297 
298 	ifp->if_mtu    = IPV6_MMTU;
299 	ifp->if_ioctl  = stf_ioctl;
300 	ifp->if_output = stf_output;
301 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
302 	if_attach(ifp);
303 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
304 	*ifpp = ifp;
305 
306 	return (0);
307 }
308 
309 static int
310 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
311 {
312 	struct stf_softc *sc = ifp->if_softc;
313 	int err __unused;
314 
315 	err = ip_encap_detach(sc->encap_cookie);
316 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
317 	bpfdetach(ifp);
318 	if_detach(ifp);
319 	if_free(ifp);
320 
321 	free(sc, M_STF);
322 	ifc_free_unit(ifc, STFUNIT);
323 
324 	return (0);
325 }
326 
327 static void
328 vnet_stf_init(const void *unused __unused)
329 {
330 	struct if_clone_addreq req = {
331 		.match_f = stf_clone_match,
332 		.create_f = stf_clone_create,
333 		.destroy_f = stf_clone_destroy,
334 	};
335 	V_stf_cloner = ifc_attach_cloner(stfname, &req);
336 }
337 VNET_SYSINIT(vnet_stf_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_init, NULL);
338 
339 static void
340 vnet_stf_uninit(const void *unused __unused)
341 {
342 	if_clone_detach(V_stf_cloner);
343 	V_stf_cloner = NULL;
344 }
345 VNET_SYSUNINIT(vnet_stf_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_uninit,
346     NULL);
347 
348 static int
349 stfmodevent(module_t mod, int type, void *data)
350 {
351 
352 	switch (type) {
353 	case MOD_LOAD:
354 		/* Done in vnet_stf_init() */
355 		break;
356 	case MOD_UNLOAD:
357 		/* Done in vnet_stf_uninit() */
358 		break;
359 	default:
360 		return (EOPNOTSUPP);
361 	}
362 
363 	return (0);
364 }
365 
366 static moduledata_t stf_mod = {
367 	"if_stf",
368 	stfmodevent,
369 	0
370 };
371 
372 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
373 MODULE_VERSION(if_stf, 2);
374 
375 static int
376 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
377 {
378 	struct ip ip;
379 	struct stf_softc *sc;
380 	struct in6_addr addr6, mask6;
381 	struct sockaddr_in sin4addr, sin4mask;
382 
383 	SDT_PROBE3(if_stf, , encapcheck, in, m, off, proto);
384 
385 	sc = (struct stf_softc *)arg;
386 	if (sc == NULL)
387 		return (0);
388 
389 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
390 		return (0);
391 
392 	/* IFF_LINK0 means "no decapsulation" */
393 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
394 		return (0);
395 
396 	if (proto != IPPROTO_IPV6)
397 		return (0);
398 
399 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
400 
401 	if (ip.ip_v != 4)
402 		return (0);
403 
404 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
405 		return (0);
406 
407 	if (sc->srcv4_addr != INADDR_ANY) {
408 		sin4addr.sin_addr.s_addr = sc->srcv4_addr;
409 		sin4addr.sin_family = AF_INET;
410 	} else
411 		if (stf_getin4addr(sc, &sin4addr, addr6, mask6) == NULL)
412 			return (0);
413 
414 	if (sin4addr.sin_addr.s_addr != ip.ip_dst.s_addr)
415 		return (0);
416 
417 	if (IN6_IS_ADDR_6TO4(&addr6)) {
418 		/*
419 		 * 6to4 (RFC 3056).
420 		 * Check if IPv4 src matches the IPv4 address derived
421 		 * from the local 6to4 address masked by prefixmask.
422 		 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
423 		 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
424 		 */
425 		memcpy(&sin4mask.sin_addr, GET_V4(&mask6),
426 		    sizeof(sin4mask.sin_addr));
427 		if ((sin4addr.sin_addr.s_addr & sin4mask.sin_addr.s_addr) !=
428 		    (ip.ip_src.s_addr & sin4mask.sin_addr.s_addr))
429 			return (0);
430 	} else {
431 		/* 6rd (RFC 5569) */
432 		/*
433 		 * No restriction on the src address in the case of
434 		 * 6rd because the stf(4) interface always has a
435 		 * prefix which covers whole of IPv4 src address
436 		 * range.  So, stf_output() will catch all of
437 		 * 6rd-capsuled IPv4 traffic with suspicious inner dst
438 		 * IPv4 address (i.e. the IPv6 destination address is
439 		 * one the admin does not like to route to outside),
440 		 * and then it discard them silently.
441 		 */
442 	}
443 
444 	SDT_PROBE0(if_stf, , encapcheck, accept);
445 
446 	/* stf interface makes single side match only */
447 	return (32);
448 }
449 
450 static int
451 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
452 {
453 	struct ifaddr *ia;
454 	struct in_ifaddr *ia4;
455 	struct in6_addr addr6, mask6;
456 	struct sockaddr_in sin4;
457 	struct stf_softc *sc;
458 	struct in_addr in;
459 
460 	NET_EPOCH_ASSERT();
461 
462 	sc = ifp->if_softc;
463 
464 	SDT_PROBE3(if_stf, , getsrcifa6, in, ifp, addr, mask);
465 
466 	CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
467 		if (ia->ifa_addr->sa_family != AF_INET6)
468 			continue;
469 
470 		addr6 = *IFA_IN6(ia);
471 		mask6 = *IFA_MASKIN6(ia);
472 		if (sc->srcv4_addr != INADDR_ANY)
473 			bcopy(&sc->srcv4_addr, &in, sizeof(in));
474 		else {
475 			if (stf_getin4addr(sc, &sin4, addr6, mask6) == NULL)
476 				continue;
477 			bcopy(&sin4.sin_addr, &in, sizeof(in));
478 		}
479 
480 		CK_LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
481 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
482 				break;
483 		if (ia4 == NULL)
484 			continue;
485 
486 		*addr = addr6;
487 		*mask = mask6;
488 
489 		SDT_PROBE2(if_stf, , getsrcifa6, found, addr, mask);
490 
491 		return (0);
492 	}
493 
494 	SDT_PROBE0(if_stf, , getsrcifa6, notfound);
495 
496 	return (ENOENT);
497 }
498 
499 static int
500 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
501     struct route *ro)
502 {
503 	struct stf_softc *sc;
504 	const struct sockaddr_in6 *dst6;
505 	struct sockaddr_in dst4, src4;
506 	u_int8_t tos;
507 	struct ip *ip;
508 	struct ip6_hdr *ip6;
509 	struct in6_addr addr6, mask6;
510 	int error;
511 
512 	SDT_PROBE4(if_stf, , stf_output, in, ifp, m, dst, ro);
513 
514 #ifdef MAC
515 	error = mac_ifnet_check_transmit(ifp, m);
516 	if (error) {
517 		m_freem(m);
518 		SDT_PROBE2(if_stf, , stf_output, error, error, __LINE__);
519 		return (error);
520 	}
521 #endif
522 
523 	sc = ifp->if_softc;
524 	dst6 = (const struct sockaddr_in6 *)dst;
525 
526 	/* just in case */
527 	if ((ifp->if_flags & IFF_UP) == 0) {
528 		m_freem(m);
529 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
530 		SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__);
531 		return (ENETDOWN);
532 	}
533 
534 	/*
535 	 * If we don't have an ip4 address that match my inner ip6 address,
536 	 * we shouldn't generate output.  Without this check, we'll end up
537 	 * using wrong IPv4 source.
538 	 */
539 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
540 		m_freem(m);
541 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
542 		SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__);
543 		return (ENETDOWN);
544 	}
545 
546 	if (m->m_len < sizeof(*ip6)) {
547 		m = m_pullup(m, sizeof(*ip6));
548 		if (!m) {
549 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
550 			SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS,
551 			    __LINE__);
552 			return (ENOBUFS);
553 		}
554 	}
555 	ip6 = mtod(m, struct ip6_hdr *);
556 	tos = IPV6_TRAFFIC_CLASS(ip6);
557 
558 	/*
559 	 * Pickup the right outer dst addr from the list of candidates.
560 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
561 	 */
562 	if (stf_getin4addr_in6(sc, &dst4, addr6, mask6,
563 	    ip6->ip6_dst) == NULL) {
564 		if (sc->braddr != INADDR_ANY)
565 			dst4.sin_addr.s_addr = sc->braddr;
566 		else if (stf_getin4addr_in6(sc, &dst4, addr6, mask6,
567 		    dst6->sin6_addr) == NULL) {
568 			m_freem(m);
569 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
570 			SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH,
571 			    __LINE__);
572 			return (ENETUNREACH);
573 		}
574 	}
575 
576 	if (bpf_peers_present(ifp->if_bpf)) {
577 		/*
578 		 * We need to prepend the address family as
579 		 * a four byte field.  Cons up a dummy header
580 		 * to pacify bpf.  This is safe because bpf
581 		 * will only read from the mbuf (i.e., it won't
582 		 * try to free it or keep a pointer a to it).
583 		 */
584 		u_int af = AF_INET6;
585 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
586 	}
587 
588 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
589 	if (m == NULL) {
590 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
591 		SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS, __LINE__);
592 		return (ENOBUFS);
593 	}
594 	ip = mtod(m, struct ip *);
595 
596 	bzero(ip, sizeof(*ip));
597 
598 	if (sc->srcv4_addr != INADDR_ANY)
599 		src4.sin_addr.s_addr = sc->srcv4_addr;
600 	else if (stf_getin4addr(sc, &src4, addr6, mask6) == NULL) {
601 		m_freem(m);
602 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
603 		SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH, __LINE__);
604 		return (ENETUNREACH);
605 	}
606 	bcopy(&src4.sin_addr, &ip->ip_src, sizeof(ip->ip_src));
607 	bcopy(&dst4.sin_addr, &ip->ip_dst, sizeof(ip->ip_dst));
608 
609 	ip->ip_p = IPPROTO_IPV6;
610 	ip->ip_ttl = ip_stf_ttl;
611 	ip->ip_len = htons(m->m_pkthdr.len);
612 	if (ifp->if_flags & IFF_LINK1)
613 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
614 	else
615 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
616 
617 	M_SETFIB(m, sc->sc_fibnum);
618 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
619 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
620 
621 	SDT_PROBE1(if_stf, , stf_output, out, error);
622 	return (error);
623 }
624 
625 static int
626 isrfc1918addr(struct in_addr *in)
627 {
628 	/*
629 	 * returns 1 if private address range:
630 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
631 	 */
632 	if (stf_permit_rfc1918 == 0 && (
633 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
634 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
635 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
636 		return (1);
637 
638 	return (0);
639 }
640 
641 static int
642 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
643 {
644 	struct in_ifaddr *ia4;
645 
646 	/*
647 	 * reject packets with the following address:
648 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
649 	 */
650 	if (IN_MULTICAST(ntohl(in->s_addr)))
651 		return (-1);
652 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
653 	case 0: case 127: case 255:
654 		return (-1);
655 	}
656 
657 	/*
658 	 * reject packets with broadcast
659 	 */
660 	CK_STAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
661 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
662 			continue;
663 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
664 			return (-1);
665 		}
666 	}
667 
668 	/*
669 	 * perform ingress filter
670 	 */
671 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
672 		struct nhop_object *nh;
673 
674 		NET_EPOCH_ASSERT();
675 		nh = fib4_lookup(sc->sc_fibnum, *in, 0, 0, 0);
676 		if (nh == NULL)
677 			return (-1);
678 
679 		if (nh->nh_ifp != inifp)
680 			return (-1);
681 	}
682 
683 	return (0);
684 }
685 
686 static int
687 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
688 {
689 	SDT_PROBE3(if_stf, , checkaddr6, in, sc, in6, inifp);
690 
691 	/*
692 	 * check 6to4 addresses
693 	 */
694 	if (IN6_IS_ADDR_6TO4(in6)) {
695 		struct in_addr in4;
696 		int ret;
697 
698 		bcopy(GET_V4(in6), &in4, sizeof(in4));
699 		ret = stf_checkaddr4(sc, &in4, inifp);
700 		SDT_PROBE2(if_stf, , checkaddr6, out, ret, __LINE__);
701 		return (ret);
702 	}
703 
704 	/*
705 	 * reject anything that look suspicious.  the test is implemented
706 	 * in ip6_input too, but we check here as well to
707 	 * (1) reject bad packets earlier, and
708 	 * (2) to be safe against future ip6_input change.
709 	 */
710 	if (IN6_IS_ADDR_V4COMPAT(in6)) {
711 		SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__);
712 		return (-1);
713 	}
714 
715 	if (IN6_IS_ADDR_V4MAPPED(in6)) {
716 		SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__);
717 		return (-1);
718 	}
719 
720 	SDT_PROBE2(if_stf, , checkaddr6, out, 0, __LINE__);
721 	return (0);
722 }
723 
724 static int
725 in_stf_input(struct mbuf *m, int off, int proto, void *arg)
726 {
727 	struct stf_softc *sc = arg;
728 	struct ip ip;
729 	struct ip6_hdr *ip6;
730 	u_int8_t otos, itos;
731 	struct ifnet *ifp;
732 	struct nhop_object *nh;
733 
734 	NET_EPOCH_ASSERT();
735 
736 	SDT_PROBE3(if_stf, , stf_input, in, m, off, proto);
737 
738 	if (proto != IPPROTO_IPV6) {
739 		m_freem(m);
740 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
741 		return (IPPROTO_DONE);
742 	}
743 
744 	m_copydata(m, 0, sizeof(struct ip), (caddr_t)&ip);
745 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
746 		m_freem(m);
747 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
748 		return (IPPROTO_DONE);
749 	}
750 
751 	ifp = STF2IFP(sc);
752 
753 #ifdef MAC
754 	mac_ifnet_create_mbuf(ifp, m);
755 #endif
756 
757 	/*
758 	 * perform sanity check against outer src/dst.
759 	 * for source, perform ingress filter as well.
760 	 */
761 	if (stf_checkaddr4(sc, &ip.ip_dst, NULL) < 0 ||
762 	    stf_checkaddr4(sc, &ip.ip_src, m->m_pkthdr.rcvif) < 0) {
763 		m_freem(m);
764 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
765 		return (IPPROTO_DONE);
766 	}
767 
768 	otos = ip.ip_tos;
769 	m_adj(m, off);
770 
771 	if (m->m_len < sizeof(*ip6)) {
772 		m = m_pullup(m, sizeof(*ip6));
773 		if (!m) {
774 			SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE,
775 			    __LINE__);
776 			return (IPPROTO_DONE);
777 		}
778 	}
779 	ip6 = mtod(m, struct ip6_hdr *);
780 
781 	/*
782 	 * perform sanity check against inner src/dst.
783 	 * for source, perform ingress filter as well.
784 	 */
785 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
786 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
787 		m_freem(m);
788 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
789 		return (IPPROTO_DONE);
790 	}
791 
792 	/*
793 	 * reject packets with private address range.
794 	 * (requirement from RFC3056 section 2 1st paragraph)
795 	 */
796 	if ((IN6_IS_ADDR_6TO4(&ip6->ip6_src) && isrfc1918addr(&ip.ip_src)) ||
797 	    (IN6_IS_ADDR_6TO4(&ip6->ip6_dst) && isrfc1918addr(&ip.ip_dst))) {
798 		m_freem(m);
799 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
800 		return (IPPROTO_DONE);
801 	}
802 
803 	/*
804 	 * Ignore if the destination is the same stf interface because
805 	 * all of valid IPv6 outgoing traffic should go interfaces
806 	 * except for it.
807 	 */
808 	nh = fib6_lookup(sc->sc_fibnum, &ip6->ip6_dst, 0, 0, 0);
809 	if (nh == NULL) {
810 		m_free(m);
811 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
812 		return (IPPROTO_DONE);
813 	}
814 	if ((nh->nh_ifp == ifp) &&
815 	    (!IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &nh->gw6_sa.sin6_addr))) {
816 		m_free(m);
817 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
818 		return (IPPROTO_DONE);
819 	}
820 
821 	itos = IPV6_TRAFFIC_CLASS(ip6);
822 	if ((ifp->if_flags & IFF_LINK1) != 0)
823 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
824 	else
825 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
826 	ip6->ip6_flow &= ~htonl(0xff << 20);
827 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
828 
829 	m->m_pkthdr.rcvif = ifp;
830 
831 	if (bpf_peers_present(ifp->if_bpf)) {
832 		/*
833 		 * We need to prepend the address family as
834 		 * a four byte field.  Cons up a dummy header
835 		 * to pacify bpf.  This is safe because bpf
836 		 * will only read from the mbuf (i.e., it won't
837 		 * try to free it or keep a pointer a to it).
838 		 */
839 		u_int32_t af = AF_INET6;
840 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
841 	}
842 
843 	/*
844 	 * Put the packet to the network layer input queue according to the
845 	 * specified address family.
846 	 * See net/if_gif.c for possible issues with packet processing
847 	 * reorder due to extra queueing.
848 	 */
849 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
850 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
851 	M_SETFIB(m, ifp->if_fib);
852 	netisr_dispatch(NETISR_IPV6, m);
853 	SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
854 	return (IPPROTO_DONE);
855 }
856 
857 static struct sockaddr_in *
858 stf_getin4addr_in6(struct stf_softc *sc, struct sockaddr_in *sin,
859     struct in6_addr addr6, struct in6_addr mask6, struct in6_addr in6)
860 {
861        int i;
862        struct sockaddr_in *out;
863 
864 	/*
865 	* When (src addr & src mask) != (in6 & src mask),
866 	* the dst is not in the 6rd domain.  The IPv4 address must
867 	* not be used.
868 	*/
869 	for (i = 0; i < sizeof(addr6); i++) {
870 		if ((((u_char *)&addr6)[i] & ((u_char *)&mask6)[i]) !=
871 		    (((u_char *)&in6)[i] & ((u_char *)&mask6)[i])) {
872 			SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6,
873 			    &mask6, &in6, NULL);
874 			return (NULL);
875 		}
876 	}
877 
878 	/* After the mask check, use in6 instead of addr6. */
879 	out = stf_getin4addr(sc, sin, in6, mask6);
880 	SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6, &mask6, &in6, out);
881 	return (out);
882 }
883 
884 static struct sockaddr_in *
885 stf_getin4addr(struct stf_softc *sc, struct sockaddr_in *sin,
886     struct in6_addr addr6, struct in6_addr mask6)
887 {
888 	struct in_addr *in;
889 
890 	SDT_PROBE2(if_stf, , getin4addr, in, &addr6, &mask6);
891 
892 	memset(sin, 0, sizeof(*sin));
893 	in = &sin->sin_addr;
894 	if (IN6_IS_ADDR_6TO4(&addr6)) {
895 		/* 6to4 (RFC 3056) */
896 		bcopy(GET_V4(&addr6), in, sizeof(*in));
897 		if (isrfc1918addr(in))
898 			return (NULL);
899 	} else {
900 		/* 6rd (RFC 5569) */
901 		in_addr_t v4prefix;
902 		uint8_t *v6 = (uint8_t*)&addr6;
903 		uint64_t v6prefix;
904 		u_int plen;
905 		u_int v4suffixlen;
906 
907 		v4prefix = 0;
908 		if (sc->v4prefixlen < 32) {
909 			v4suffixlen = 32 - sc->v4prefixlen;
910 			v4prefix = ntohl(sc->srcv4_addr) &
911 			    (0xffffffffU << v4suffixlen);
912 		} else {
913 			MPASS(sc->v4prefixlen == 32);
914 			v4suffixlen = 32;
915 		}
916 
917 		plen = in6_mask2len(&mask6, NULL);
918 		if (plen > 64)
919 			return (NULL);
920 
921 		/* To make this simple we do not support prefixes longer than
922 		 * 64 bits. RFC5969 says "a 6rd delegated prefix SHOULD be /64
923 		 * or shorter." so this is a moderately safe assumption. */
924 		v6prefix = be64toh(*(uint64_t *)v6);
925 
926 		/* Shift away the v6 prefix itself. */
927 		v6prefix <<= plen;
928 		v6prefix >>= plen;
929 
930 		/* Now shift away everything after the v4 address. */
931 		v6prefix >>= 64 - plen - v4suffixlen;
932 
933 		sin->sin_addr.s_addr = htonl(v4prefix | (uint32_t)v6prefix);
934 	}
935 
936 	SDT_PROBE1(if_stf, , getin4addr, out, sin);
937 
938 	return (sin);
939 }
940 
941 static int
942 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
943 {
944 	struct ifaddr *ifa;
945 	struct ifdrv *ifd;
946 	struct ifreq *ifr;
947 	struct sockaddr_in sin4;
948 	struct stf_softc *sc_cur;
949 	struct stfv4args args;
950 	int error, mtu;
951 
952 	error = 0;
953 	sc_cur = ifp->if_softc;
954 
955 	switch (cmd) {
956 	case SIOCSDRVSPEC:
957 		ifd = (struct ifdrv *)data;
958 		error = priv_check(curthread, PRIV_NET_ADDIFADDR);
959 		if (error)
960 			break;
961 		if (ifd->ifd_cmd == STF6RD_SV4NET) {
962 			if (ifd->ifd_len != sizeof(args)) {
963 				error = EINVAL;
964 				break;
965 			}
966 			bzero(&args, sizeof(args));
967 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
968 			if (error)
969 				break;
970 
971 			if (args.v4_prefixlen < 1 || args.v4_prefixlen > 32) {
972 				error = EINVAL;
973 				break;
974 			}
975 
976 			bcopy(&args.srcv4_addr, &sc_cur->srcv4_addr,
977 			    sizeof(sc_cur->srcv4_addr));
978 			sc_cur->v4prefixlen = args.v4_prefixlen;
979 			SDT_PROBE3(if_stf, , ioctl, sv4net, sc_cur->srcv4_addr,
980 			    sc_cur->srcv4_addr, sc_cur->v4prefixlen);
981 		} else if (ifd->ifd_cmd == STF6RD_SBR) {
982 			if (ifd->ifd_len != sizeof(args)) {
983 				error = EINVAL;
984 				break;
985 			}
986 			bzero(&args, sizeof(args));
987 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
988 			if (error)
989 				break;
990 			sc_cur->braddr = args.braddr.s_addr;
991 			SDT_PROBE1(if_stf, , ioctl, sdstv4,
992 			    sc_cur->braddr);
993 		} else
994 			error = EINVAL;
995 		break;
996 	case SIOCGDRVSPEC:
997 		ifd = (struct ifdrv *)data;
998 		if (ifd->ifd_cmd != STF6RD_GV4NET) {
999 			error = EINVAL;
1000 			break;
1001 		}
1002 		if (ifd->ifd_len != sizeof(args)) {
1003 			error = EINVAL;
1004 			break;
1005 		}
1006 		bzero(&args, sizeof(args));
1007 		args.srcv4_addr.s_addr = sc_cur->srcv4_addr;
1008 		args.braddr.s_addr = sc_cur->braddr;
1009 		args.v4_prefixlen = sc_cur->v4prefixlen;
1010 		error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
1011 		break;
1012 	case SIOCSIFADDR:
1013 		ifa = (struct ifaddr *)data;
1014 		SDT_PROBE1(if_stf, , ioctl, ifaddr, ifa);
1015 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
1016 			error = EAFNOSUPPORT;
1017 			break;
1018 		}
1019 		if (stf_getin4addr(sc_cur, &sin4,
1020 		    satosin6(ifa->ifa_addr)->sin6_addr,
1021 		    satosin6(ifa->ifa_netmask)->sin6_addr) == NULL) {
1022 			error = EINVAL;
1023 			break;
1024 		}
1025 		ifp->if_flags |= IFF_UP;
1026 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1027 		break;
1028 
1029 	case SIOCADDMULTI:
1030 	case SIOCDELMULTI:
1031 		ifr = (struct ifreq *)data;
1032 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
1033 			;
1034 		else
1035 			error = EAFNOSUPPORT;
1036 		break;
1037 
1038 	case SIOCGIFMTU:
1039 		break;
1040 
1041 	case SIOCSIFMTU:
1042 		ifr = (struct ifreq *)data;
1043 		mtu = ifr->ifr_mtu;
1044 		/* RFC 4213 3.2 ideal world MTU */
1045 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
1046 			return (EINVAL);
1047 		ifp->if_mtu = mtu;
1048 		break;
1049 
1050 	default:
1051 		error = EINVAL;
1052 		break;
1053 	}
1054 
1055 	return (error);
1056 }
1057