xref: /freebsd/sys/net/if_stf.c (revision e453e498cbb88570a3ff7b3679de65c88707da95)
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/stdarg.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 <net/bpf.h>
128 
129 #include <security/mac/mac_framework.h>
130 
131 SDT_PROVIDER_DEFINE(if_stf);
132 SDT_PROBE_DEFINE3(if_stf, , encapcheck, in, "struct mbuf *", "int", "int");
133 SDT_PROBE_DEFINE0(if_stf, , encapcheck, accept);
134 SDT_PROBE_DEFINE3(if_stf, , getsrcifa6, in, "struct ifnet *",
135     "struct in6_addr *", "struct in6_addr *");
136 SDT_PROBE_DEFINE2(if_stf, , getsrcifa6, found, "struct in6_addr *",
137     "struct in6_addr *");
138 SDT_PROBE_DEFINE0(if_stf, , getsrcifa6, notfound);
139 
140 SDT_PROBE_DEFINE4(if_stf, , stf_output, in, "struct ifnet *", "struct mbuf *",
141     "struct sockaddr *", "struct route *");
142 SDT_PROBE_DEFINE2(if_stf, , stf_output, error, "int", "int");
143 SDT_PROBE_DEFINE1(if_stf, , stf_output, out, "int");
144 
145 SDT_PROBE_DEFINE3(if_stf, , checkaddr6, in, "struct stf_softc *",
146     "struct in6_addr *", "struct ifnet *");
147 SDT_PROBE_DEFINE2(if_stf, , checkaddr6, out, "int", "int");
148 
149 SDT_PROBE_DEFINE3(if_stf, , stf_input, in, "struct mbuf *", "int", "int");
150 SDT_PROBE_DEFINE2(if_stf, , stf_input, out, "int", "int");
151 
152 SDT_PROBE_DEFINE3(if_stf, , ioctl, sv4net, "struct in_addr *",
153     "struct in_addr *", "int");
154 SDT_PROBE_DEFINE1(if_stf, , ioctl, sdstv4, "struct in_addr *");
155 SDT_PROBE_DEFINE1(if_stf, , ioctl, ifaddr, "struct ifaddr *");
156 
157 SDT_PROBE_DEFINE4(if_stf, , getin4addr_in6, out, "struct in6_addr *",
158     "struct in6_addr *", "struct in6_addr *", "struct sockaddr_in *");
159 
160 SDT_PROBE_DEFINE2(if_stf, , getin4addr, in, "struct in6_addr *", "struct in6_addr *");
161 SDT_PROBE_DEFINE1(if_stf, , getin4addr, out, "struct sockaddr_in *");
162 
163 SYSCTL_DECL(_net_link);
164 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
165     "6to4 Interface");
166 
167 static int stf_permit_rfc1918 = 0;
168 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
169     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
170 
171 #define STFUNIT		0
172 
173 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
174 
175 /*
176  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
177  * struct in_addr *; use bcopy() instead.
178  */
179 #define GET_V4(x)	(&(x)->s6_addr16[1])
180 
181 struct stf_softc {
182 	struct ifnet	*sc_ifp;
183 	in_addr_t	braddr;		/* Border relay IPv4 address */
184 	in_addr_t	srcv4_addr;	/* Our IPv4 WAN address */
185 	u_int		v4prefixlen;	/* How much of the v4 address to include in our address. */
186 	u_int		sc_fibnum;
187 	const struct encaptab *encap_cookie;
188 };
189 #define STF2IFP(sc)	((sc)->sc_ifp)
190 
191 static const char stfname[] = "stf";
192 
193 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
194 static const int ip_stf_ttl = 40;
195 
196 static int in_stf_input(struct mbuf *, int, int, void *);
197 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
198 
199 static int stfmodevent(module_t, int, void *);
200 static int stf_encapcheck(const struct mbuf *, int, int, void *);
201 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
202 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
203 	struct route *);
204 static int isrfc1918addr(struct in_addr *);
205 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
206 	struct ifnet *);
207 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
208 	struct ifnet *);
209 static struct sockaddr_in *stf_getin4addr_in6(struct stf_softc *,
210 	struct sockaddr_in *, struct in6_addr, struct in6_addr,
211 	struct in6_addr);
212 static struct sockaddr_in *stf_getin4addr(struct stf_softc *,
213 	struct sockaddr_in *, struct in6_addr, struct in6_addr);
214 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
215 
216 VNET_DEFINE_STATIC(struct if_clone *, stf_cloner);
217 #define V_stf_cloner	VNET(stf_cloner)
218 
219 static const struct encap_config ipv4_encap_cfg = {
220 	.proto = IPPROTO_IPV6,
221 	.min_length = sizeof(struct ip),
222 	.exact_match = (sizeof(in_addr_t) << 3) + 8,
223 	.check = stf_encapcheck,
224 	.input = in_stf_input
225 };
226 
227 static int
stf_clone_match(struct if_clone * ifc,const char * name)228 stf_clone_match(struct if_clone *ifc, const char *name)
229 {
230 	int i;
231 
232 	for(i = 0; stfnames[i] != NULL; i++) {
233 		if (strcmp(stfnames[i], name) == 0)
234 			return (1);
235 	}
236 
237 	return (0);
238 }
239 
240 static int
stf_clone_create(struct if_clone * ifc,char * name,size_t len,struct ifc_data * ifd,struct ifnet ** ifpp)241 stf_clone_create(struct if_clone *ifc, char *name, size_t len,
242     struct ifc_data *ifd, struct ifnet **ifpp)
243 {
244 	char *dp;
245 	int err, unit, wildcard;
246 	struct stf_softc *sc;
247 	struct ifnet *ifp;
248 
249 	err = ifc_name2unit(name, &unit);
250 	if (err != 0)
251 		return (err);
252 	wildcard = (unit < 0);
253 
254 	/*
255 	 * We can only have one unit, but since unit allocation is
256 	 * already locked, we use it to keep from allocating extra
257 	 * interfaces.
258 	 */
259 	unit = STFUNIT;
260 	err = ifc_alloc_unit(ifc, &unit);
261 	if (err != 0)
262 		return (err);
263 
264 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
265 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
266 	ifp->if_softc = sc;
267 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
268 
269 	/*
270 	 * Set the name manually rather then using if_initname because
271 	 * we don't conform to the default naming convention for interfaces.
272 	 * In the wildcard case, we need to update the name.
273 	 */
274 	if (wildcard) {
275 		for (dp = name; *dp != '\0'; dp++);
276 		if (snprintf(dp, len - (dp-name), "%d", unit) >
277 		    len - (dp-name) - 1) {
278 			/*
279 			 * This can only be a programmer error and
280 			 * there's no straightforward way to recover if
281 			 * it happens.
282 			 */
283 			panic("if_clone_create(): interface name too long");
284 		}
285 	}
286 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
287 	ifp->if_dname = stfname;
288 	ifp->if_dunit = IF_DUNIT_NONE;
289 
290 	sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK);
291 
292 	ifp->if_mtu    = IPV6_MMTU;
293 	ifp->if_ioctl  = stf_ioctl;
294 	ifp->if_output = stf_output;
295 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
296 	if_attach(ifp);
297 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
298 	*ifpp = ifp;
299 
300 	return (0);
301 }
302 
303 static int
stf_clone_destroy(struct if_clone * ifc,struct ifnet * ifp,uint32_t flags)304 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
305 {
306 	struct stf_softc *sc = ifp->if_softc;
307 	int err __unused;
308 
309 	err = ip_encap_detach(sc->encap_cookie);
310 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
311 	bpfdetach(ifp);
312 	if_detach(ifp);
313 	if_free(ifp);
314 
315 	free(sc, M_STF);
316 	ifc_free_unit(ifc, STFUNIT);
317 
318 	return (0);
319 }
320 
321 static void
vnet_stf_init(const void * unused __unused)322 vnet_stf_init(const void *unused __unused)
323 {
324 	struct if_clone_addreq req = {
325 		.match_f = stf_clone_match,
326 		.create_f = stf_clone_create,
327 		.destroy_f = stf_clone_destroy,
328 	};
329 	V_stf_cloner = ifc_attach_cloner(stfname, &req);
330 }
331 VNET_SYSINIT(vnet_stf_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_init, NULL);
332 
333 static void
vnet_stf_uninit(const void * unused __unused)334 vnet_stf_uninit(const void *unused __unused)
335 {
336 	if_clone_detach(V_stf_cloner);
337 	V_stf_cloner = NULL;
338 }
339 VNET_SYSUNINIT(vnet_stf_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_uninit,
340     NULL);
341 
342 static int
stfmodevent(module_t mod,int type,void * data)343 stfmodevent(module_t mod, int type, void *data)
344 {
345 
346 	switch (type) {
347 	case MOD_LOAD:
348 		/* Done in vnet_stf_init() */
349 		break;
350 	case MOD_UNLOAD:
351 		/* Done in vnet_stf_uninit() */
352 		break;
353 	default:
354 		return (EOPNOTSUPP);
355 	}
356 
357 	return (0);
358 }
359 
360 static moduledata_t stf_mod = {
361 	"if_stf",
362 	stfmodevent,
363 	0
364 };
365 
366 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
367 MODULE_VERSION(if_stf, 2);
368 
369 static int
stf_encapcheck(const struct mbuf * m,int off,int proto,void * arg)370 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
371 {
372 	struct ip ip;
373 	struct stf_softc *sc;
374 	struct in6_addr addr6, mask6;
375 	struct sockaddr_in sin4addr, sin4mask;
376 
377 	SDT_PROBE3(if_stf, , encapcheck, in, m, off, proto);
378 
379 	sc = (struct stf_softc *)arg;
380 	if (sc == NULL)
381 		return (0);
382 
383 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
384 		return (0);
385 
386 	/* IFF_LINK0 means "no decapsulation" */
387 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
388 		return (0);
389 
390 	if (proto != IPPROTO_IPV6)
391 		return (0);
392 
393 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
394 
395 	if (ip.ip_v != 4)
396 		return (0);
397 
398 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
399 		return (0);
400 
401 	if (sc->srcv4_addr != INADDR_ANY) {
402 		sin4addr.sin_addr.s_addr = sc->srcv4_addr;
403 		sin4addr.sin_family = AF_INET;
404 	} else
405 		if (stf_getin4addr(sc, &sin4addr, addr6, mask6) == NULL)
406 			return (0);
407 
408 	if (sin4addr.sin_addr.s_addr != ip.ip_dst.s_addr)
409 		return (0);
410 
411 	if (IN6_IS_ADDR_6TO4(&addr6)) {
412 		/*
413 		 * 6to4 (RFC 3056).
414 		 * Check if IPv4 src matches the IPv4 address derived
415 		 * from the local 6to4 address masked by prefixmask.
416 		 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
417 		 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
418 		 */
419 		memcpy(&sin4mask.sin_addr, GET_V4(&mask6),
420 		    sizeof(sin4mask.sin_addr));
421 		if ((sin4addr.sin_addr.s_addr & sin4mask.sin_addr.s_addr) !=
422 		    (ip.ip_src.s_addr & sin4mask.sin_addr.s_addr))
423 			return (0);
424 	} else {
425 		/* 6rd (RFC 5569) */
426 		/*
427 		 * No restriction on the src address in the case of
428 		 * 6rd because the stf(4) interface always has a
429 		 * prefix which covers whole of IPv4 src address
430 		 * range.  So, stf_output() will catch all of
431 		 * 6rd-capsuled IPv4 traffic with suspicious inner dst
432 		 * IPv4 address (i.e. the IPv6 destination address is
433 		 * one the admin does not like to route to outside),
434 		 * and then it discard them silently.
435 		 */
436 	}
437 
438 	SDT_PROBE0(if_stf, , encapcheck, accept);
439 
440 	/* stf interface makes single side match only */
441 	return (32);
442 }
443 
444 static int
stf_getsrcifa6(struct ifnet * ifp,struct in6_addr * addr,struct in6_addr * mask)445 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
446 {
447 	struct ifaddr *ia;
448 	struct in_ifaddr *ia4;
449 	struct in6_addr addr6, mask6;
450 	struct sockaddr_in sin4;
451 	struct stf_softc *sc;
452 	struct in_addr in;
453 
454 	NET_EPOCH_ASSERT();
455 
456 	sc = ifp->if_softc;
457 
458 	SDT_PROBE3(if_stf, , getsrcifa6, in, ifp, addr, mask);
459 
460 	CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
461 		if (ia->ifa_addr->sa_family != AF_INET6)
462 			continue;
463 
464 		addr6 = *IFA_IN6(ia);
465 		mask6 = *IFA_MASKIN6(ia);
466 		if (sc->srcv4_addr != INADDR_ANY)
467 			bcopy(&sc->srcv4_addr, &in, sizeof(in));
468 		else {
469 			if (stf_getin4addr(sc, &sin4, addr6, mask6) == NULL)
470 				continue;
471 			bcopy(&sin4.sin_addr, &in, sizeof(in));
472 		}
473 
474 		CK_LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
475 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
476 				break;
477 		if (ia4 == NULL)
478 			continue;
479 
480 		*addr = addr6;
481 		*mask = mask6;
482 
483 		SDT_PROBE2(if_stf, , getsrcifa6, found, addr, mask);
484 
485 		return (0);
486 	}
487 
488 	SDT_PROBE0(if_stf, , getsrcifa6, notfound);
489 
490 	return (ENOENT);
491 }
492 
493 static int
stf_output(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,struct route * ro)494 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
495     struct route *ro)
496 {
497 	struct stf_softc *sc;
498 	const struct sockaddr_in6 *dst6;
499 	struct sockaddr_in dst4, src4;
500 	u_int8_t tos;
501 	struct ip *ip;
502 	struct ip6_hdr *ip6;
503 	struct in6_addr addr6, mask6;
504 	int error;
505 
506 	SDT_PROBE4(if_stf, , stf_output, in, ifp, m, dst, ro);
507 
508 #ifdef MAC
509 	error = mac_ifnet_check_transmit(ifp, m);
510 	if (error) {
511 		m_freem(m);
512 		SDT_PROBE2(if_stf, , stf_output, error, error, __LINE__);
513 		return (error);
514 	}
515 #endif
516 
517 	sc = ifp->if_softc;
518 	dst6 = (const struct sockaddr_in6 *)dst;
519 
520 	/* just in case */
521 	if ((ifp->if_flags & IFF_UP) == 0) {
522 		m_freem(m);
523 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
524 		SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__);
525 		return (ENETDOWN);
526 	}
527 
528 	/*
529 	 * If we don't have an ip4 address that match my inner ip6 address,
530 	 * we shouldn't generate output.  Without this check, we'll end up
531 	 * using wrong IPv4 source.
532 	 */
533 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
534 		m_freem(m);
535 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
536 		SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__);
537 		return (ENETDOWN);
538 	}
539 
540 	if (m->m_len < sizeof(*ip6)) {
541 		m = m_pullup(m, sizeof(*ip6));
542 		if (!m) {
543 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
544 			SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS,
545 			    __LINE__);
546 			return (ENOBUFS);
547 		}
548 	}
549 	ip6 = mtod(m, struct ip6_hdr *);
550 	tos = IPV6_TRAFFIC_CLASS(ip6);
551 
552 	/*
553 	 * Pickup the right outer dst addr from the list of candidates.
554 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
555 	 */
556 	if (stf_getin4addr_in6(sc, &dst4, addr6, mask6,
557 	    ip6->ip6_dst) == NULL) {
558 		if (sc->braddr != INADDR_ANY)
559 			dst4.sin_addr.s_addr = sc->braddr;
560 		else if (stf_getin4addr_in6(sc, &dst4, addr6, mask6,
561 		    dst6->sin6_addr) == NULL) {
562 			m_freem(m);
563 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
564 			SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH,
565 			    __LINE__);
566 			return (ENETUNREACH);
567 		}
568 	}
569 
570 	if (bpf_peers_present(ifp->if_bpf)) {
571 		/*
572 		 * We need to prepend the address family as
573 		 * a four byte field.  Cons up a dummy header
574 		 * to pacify bpf.  This is safe because bpf
575 		 * will only read from the mbuf (i.e., it won't
576 		 * try to free it or keep a pointer a to it).
577 		 */
578 		u_int af = AF_INET6;
579 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
580 	}
581 
582 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
583 	if (m == NULL) {
584 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
585 		SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS, __LINE__);
586 		return (ENOBUFS);
587 	}
588 	ip = mtod(m, struct ip *);
589 
590 	bzero(ip, sizeof(*ip));
591 
592 	if (sc->srcv4_addr != INADDR_ANY)
593 		src4.sin_addr.s_addr = sc->srcv4_addr;
594 	else if (stf_getin4addr(sc, &src4, addr6, mask6) == NULL) {
595 		m_freem(m);
596 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
597 		SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH, __LINE__);
598 		return (ENETUNREACH);
599 	}
600 	bcopy(&src4.sin_addr, &ip->ip_src, sizeof(ip->ip_src));
601 	bcopy(&dst4.sin_addr, &ip->ip_dst, sizeof(ip->ip_dst));
602 
603 	ip->ip_p = IPPROTO_IPV6;
604 	ip->ip_ttl = ip_stf_ttl;
605 	ip->ip_len = htons(m->m_pkthdr.len);
606 	if (ifp->if_flags & IFF_LINK1)
607 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
608 	else
609 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
610 
611 	M_SETFIB(m, sc->sc_fibnum);
612 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
613 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
614 
615 	SDT_PROBE1(if_stf, , stf_output, out, error);
616 	return (error);
617 }
618 
619 static int
isrfc1918addr(struct in_addr * in)620 isrfc1918addr(struct in_addr *in)
621 {
622 	/*
623 	 * returns 1 if private address range:
624 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
625 	 */
626 	if (stf_permit_rfc1918 == 0 && (
627 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
628 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
629 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
630 		return (1);
631 
632 	return (0);
633 }
634 
635 static int
stf_checkaddr4(struct stf_softc * sc,struct in_addr * in,struct ifnet * inifp)636 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
637 {
638 	struct in_ifaddr *ia4;
639 
640 	/*
641 	 * reject packets with the following address:
642 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
643 	 */
644 	if (IN_MULTICAST(ntohl(in->s_addr)))
645 		return (-1);
646 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
647 	case 0: case 127: case 255:
648 		return (-1);
649 	}
650 
651 	/*
652 	 * reject packets with broadcast
653 	 */
654 	CK_STAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
655 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
656 			continue;
657 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
658 			return (-1);
659 		}
660 	}
661 
662 	/*
663 	 * perform ingress filter
664 	 */
665 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
666 		struct nhop_object *nh;
667 
668 		NET_EPOCH_ASSERT();
669 		nh = fib4_lookup(sc->sc_fibnum, *in, 0, 0, 0);
670 		if (nh == NULL)
671 			return (-1);
672 
673 		if (nh->nh_ifp != inifp)
674 			return (-1);
675 	}
676 
677 	return (0);
678 }
679 
680 static int
stf_checkaddr6(struct stf_softc * sc,struct in6_addr * in6,struct ifnet * inifp)681 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
682 {
683 	SDT_PROBE3(if_stf, , checkaddr6, in, sc, in6, inifp);
684 
685 	/*
686 	 * check 6to4 addresses
687 	 */
688 	if (IN6_IS_ADDR_6TO4(in6)) {
689 		struct in_addr in4;
690 		int ret;
691 
692 		bcopy(GET_V4(in6), &in4, sizeof(in4));
693 		ret = stf_checkaddr4(sc, &in4, inifp);
694 		SDT_PROBE2(if_stf, , checkaddr6, out, ret, __LINE__);
695 		return (ret);
696 	}
697 
698 	/*
699 	 * reject anything that look suspicious.  the test is implemented
700 	 * in ip6_input too, but we check here as well to
701 	 * (1) reject bad packets earlier, and
702 	 * (2) to be safe against future ip6_input change.
703 	 */
704 	if (IN6_IS_ADDR_V4COMPAT(in6)) {
705 		SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__);
706 		return (-1);
707 	}
708 
709 	if (IN6_IS_ADDR_V4MAPPED(in6)) {
710 		SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__);
711 		return (-1);
712 	}
713 
714 	SDT_PROBE2(if_stf, , checkaddr6, out, 0, __LINE__);
715 	return (0);
716 }
717 
718 static int
in_stf_input(struct mbuf * m,int off,int proto,void * arg)719 in_stf_input(struct mbuf *m, int off, int proto, void *arg)
720 {
721 	struct stf_softc *sc = arg;
722 	struct ip ip;
723 	struct ip6_hdr *ip6;
724 	u_int8_t otos, itos;
725 	struct ifnet *ifp;
726 	struct nhop_object *nh;
727 
728 	NET_EPOCH_ASSERT();
729 
730 	SDT_PROBE3(if_stf, , stf_input, in, m, off, proto);
731 
732 	if (proto != IPPROTO_IPV6) {
733 		m_freem(m);
734 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
735 		return (IPPROTO_DONE);
736 	}
737 
738 	m_copydata(m, 0, sizeof(struct ip), (caddr_t)&ip);
739 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
740 		m_freem(m);
741 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
742 		return (IPPROTO_DONE);
743 	}
744 
745 	ifp = STF2IFP(sc);
746 
747 #ifdef MAC
748 	mac_ifnet_create_mbuf(ifp, m);
749 #endif
750 
751 	/*
752 	 * perform sanity check against outer src/dst.
753 	 * for source, perform ingress filter as well.
754 	 */
755 	if (stf_checkaddr4(sc, &ip.ip_dst, NULL) < 0 ||
756 	    stf_checkaddr4(sc, &ip.ip_src, m->m_pkthdr.rcvif) < 0) {
757 		m_freem(m);
758 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
759 		return (IPPROTO_DONE);
760 	}
761 
762 	otos = ip.ip_tos;
763 	m_adj(m, off);
764 
765 	if (m->m_len < sizeof(*ip6)) {
766 		m = m_pullup(m, sizeof(*ip6));
767 		if (!m) {
768 			SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE,
769 			    __LINE__);
770 			return (IPPROTO_DONE);
771 		}
772 	}
773 	ip6 = mtod(m, struct ip6_hdr *);
774 
775 	/*
776 	 * perform sanity check against inner src/dst.
777 	 * for source, perform ingress filter as well.
778 	 */
779 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
780 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
781 		m_freem(m);
782 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
783 		return (IPPROTO_DONE);
784 	}
785 
786 	/*
787 	 * reject packets with private address range.
788 	 * (requirement from RFC3056 section 2 1st paragraph)
789 	 */
790 	if ((IN6_IS_ADDR_6TO4(&ip6->ip6_src) && isrfc1918addr(&ip.ip_src)) ||
791 	    (IN6_IS_ADDR_6TO4(&ip6->ip6_dst) && isrfc1918addr(&ip.ip_dst))) {
792 		m_freem(m);
793 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
794 		return (IPPROTO_DONE);
795 	}
796 
797 	/*
798 	 * Ignore if the destination is the same stf interface because
799 	 * all of valid IPv6 outgoing traffic should go interfaces
800 	 * except for it.
801 	 */
802 	nh = fib6_lookup(sc->sc_fibnum, &ip6->ip6_dst, 0, 0, 0);
803 	if (nh == NULL) {
804 		m_free(m);
805 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
806 		return (IPPROTO_DONE);
807 	}
808 	if ((nh->nh_ifp == ifp) &&
809 	    (!IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &nh->gw6_sa.sin6_addr))) {
810 		m_free(m);
811 		SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
812 		return (IPPROTO_DONE);
813 	}
814 
815 	itos = IPV6_TRAFFIC_CLASS(ip6);
816 	if ((ifp->if_flags & IFF_LINK1) != 0)
817 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
818 	else
819 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
820 	ip6->ip6_flow &= ~htonl(0xff << 20);
821 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
822 
823 	m->m_pkthdr.rcvif = ifp;
824 
825 	if (bpf_peers_present(ifp->if_bpf)) {
826 		/*
827 		 * We need to prepend the address family as
828 		 * a four byte field.  Cons up a dummy header
829 		 * to pacify bpf.  This is safe because bpf
830 		 * will only read from the mbuf (i.e., it won't
831 		 * try to free it or keep a pointer a to it).
832 		 */
833 		u_int32_t af = AF_INET6;
834 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
835 	}
836 
837 	/*
838 	 * Put the packet to the network layer input queue according to the
839 	 * specified address family.
840 	 * See net/if_gif.c for possible issues with packet processing
841 	 * reorder due to extra queueing.
842 	 */
843 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
844 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
845 	M_SETFIB(m, ifp->if_fib);
846 	netisr_dispatch(NETISR_IPV6, m);
847 	SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__);
848 	return (IPPROTO_DONE);
849 }
850 
851 static struct sockaddr_in *
stf_getin4addr_in6(struct stf_softc * sc,struct sockaddr_in * sin,struct in6_addr addr6,struct in6_addr mask6,struct in6_addr in6)852 stf_getin4addr_in6(struct stf_softc *sc, struct sockaddr_in *sin,
853     struct in6_addr addr6, struct in6_addr mask6, struct in6_addr in6)
854 {
855        int i;
856        struct sockaddr_in *out;
857 
858 	/*
859 	* When (src addr & src mask) != (in6 & src mask),
860 	* the dst is not in the 6rd domain.  The IPv4 address must
861 	* not be used.
862 	*/
863 	for (i = 0; i < sizeof(addr6); i++) {
864 		if ((((u_char *)&addr6)[i] & ((u_char *)&mask6)[i]) !=
865 		    (((u_char *)&in6)[i] & ((u_char *)&mask6)[i])) {
866 			SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6,
867 			    &mask6, &in6, NULL);
868 			return (NULL);
869 		}
870 	}
871 
872 	/* After the mask check, use in6 instead of addr6. */
873 	out = stf_getin4addr(sc, sin, in6, mask6);
874 	SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6, &mask6, &in6, out);
875 	return (out);
876 }
877 
878 static struct sockaddr_in *
stf_getin4addr(struct stf_softc * sc,struct sockaddr_in * sin,struct in6_addr addr6,struct in6_addr mask6)879 stf_getin4addr(struct stf_softc *sc, struct sockaddr_in *sin,
880     struct in6_addr addr6, struct in6_addr mask6)
881 {
882 	struct in_addr *in;
883 
884 	SDT_PROBE2(if_stf, , getin4addr, in, &addr6, &mask6);
885 
886 	memset(sin, 0, sizeof(*sin));
887 	in = &sin->sin_addr;
888 	if (IN6_IS_ADDR_6TO4(&addr6)) {
889 		/* 6to4 (RFC 3056) */
890 		bcopy(GET_V4(&addr6), in, sizeof(*in));
891 		if (isrfc1918addr(in))
892 			return (NULL);
893 	} else {
894 		/* 6rd (RFC 5569) */
895 		in_addr_t v4prefix;
896 		uint8_t *v6 = (uint8_t*)&addr6;
897 		uint64_t v6prefix;
898 		u_int plen;
899 		u_int v4suffixlen;
900 
901 		v4prefix = 0;
902 		if (sc->v4prefixlen < 32) {
903 			v4suffixlen = 32 - sc->v4prefixlen;
904 			v4prefix = ntohl(sc->srcv4_addr) &
905 			    (0xffffffffU << v4suffixlen);
906 		} else {
907 			MPASS(sc->v4prefixlen == 32);
908 			v4suffixlen = 32;
909 		}
910 
911 		plen = in6_mask2len(&mask6, NULL);
912 		if (plen > 64)
913 			return (NULL);
914 
915 		/* To make this simple we do not support prefixes longer than
916 		 * 64 bits. RFC5969 says "a 6rd delegated prefix SHOULD be /64
917 		 * or shorter." so this is a moderately safe assumption. */
918 		v6prefix = be64toh(*(uint64_t *)v6);
919 
920 		/* Shift away the v6 prefix itself. */
921 		v6prefix <<= plen;
922 		v6prefix >>= plen;
923 
924 		/* Now shift away everything after the v4 address. */
925 		v6prefix >>= 64 - plen - v4suffixlen;
926 
927 		sin->sin_addr.s_addr = htonl(v4prefix | (uint32_t)v6prefix);
928 	}
929 
930 	SDT_PROBE1(if_stf, , getin4addr, out, sin);
931 
932 	return (sin);
933 }
934 
935 static int
stf_ioctl(struct ifnet * ifp,u_long cmd,caddr_t data)936 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
937 {
938 	struct ifaddr *ifa;
939 	struct ifdrv *ifd;
940 	struct ifreq *ifr;
941 	struct sockaddr_in sin4;
942 	struct stf_softc *sc_cur;
943 	struct stfv4args args;
944 	int error, mtu;
945 
946 	error = 0;
947 	sc_cur = ifp->if_softc;
948 
949 	switch (cmd) {
950 	case SIOCSDRVSPEC:
951 		ifd = (struct ifdrv *)data;
952 		error = priv_check(curthread, PRIV_NET_ADDIFADDR);
953 		if (error)
954 			break;
955 		if (ifd->ifd_cmd == STF6RD_SV4NET) {
956 			if (ifd->ifd_len != sizeof(args)) {
957 				error = EINVAL;
958 				break;
959 			}
960 			bzero(&args, sizeof(args));
961 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
962 			if (error)
963 				break;
964 
965 			if (args.v4_prefixlen < 1 || args.v4_prefixlen > 32) {
966 				error = EINVAL;
967 				break;
968 			}
969 
970 			bcopy(&args.srcv4_addr, &sc_cur->srcv4_addr,
971 			    sizeof(sc_cur->srcv4_addr));
972 			sc_cur->v4prefixlen = args.v4_prefixlen;
973 			SDT_PROBE3(if_stf, , ioctl, sv4net, sc_cur->srcv4_addr,
974 			    sc_cur->srcv4_addr, sc_cur->v4prefixlen);
975 		} else if (ifd->ifd_cmd == STF6RD_SBR) {
976 			if (ifd->ifd_len != sizeof(args)) {
977 				error = EINVAL;
978 				break;
979 			}
980 			bzero(&args, sizeof(args));
981 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
982 			if (error)
983 				break;
984 			sc_cur->braddr = args.braddr.s_addr;
985 			SDT_PROBE1(if_stf, , ioctl, sdstv4,
986 			    sc_cur->braddr);
987 		} else
988 			error = EINVAL;
989 		break;
990 	case SIOCGDRVSPEC:
991 		ifd = (struct ifdrv *)data;
992 		if (ifd->ifd_cmd != STF6RD_GV4NET) {
993 			error = EINVAL;
994 			break;
995 		}
996 		if (ifd->ifd_len != sizeof(args)) {
997 			error = EINVAL;
998 			break;
999 		}
1000 		bzero(&args, sizeof(args));
1001 		args.srcv4_addr.s_addr = sc_cur->srcv4_addr;
1002 		args.braddr.s_addr = sc_cur->braddr;
1003 		args.v4_prefixlen = sc_cur->v4prefixlen;
1004 		error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
1005 		break;
1006 	case SIOCSIFADDR:
1007 		ifa = (struct ifaddr *)data;
1008 		SDT_PROBE1(if_stf, , ioctl, ifaddr, ifa);
1009 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
1010 			error = EAFNOSUPPORT;
1011 			break;
1012 		}
1013 		if (stf_getin4addr(sc_cur, &sin4,
1014 		    satosin6(ifa->ifa_addr)->sin6_addr,
1015 		    satosin6(ifa->ifa_netmask)->sin6_addr) == NULL) {
1016 			error = EINVAL;
1017 			break;
1018 		}
1019 		ifp->if_flags |= IFF_UP;
1020 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1021 		break;
1022 
1023 	case SIOCADDMULTI:
1024 	case SIOCDELMULTI:
1025 		ifr = (struct ifreq *)data;
1026 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
1027 			;
1028 		else
1029 			error = EAFNOSUPPORT;
1030 		break;
1031 
1032 	case SIOCGIFMTU:
1033 		break;
1034 
1035 	case SIOCSIFMTU:
1036 		ifr = (struct ifreq *)data;
1037 		mtu = ifr->ifr_mtu;
1038 		/* RFC 4213 3.2 ideal world MTU */
1039 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
1040 			return (EINVAL);
1041 		ifp->if_mtu = mtu;
1042 		break;
1043 
1044 	default:
1045 		error = EINVAL;
1046 		break;
1047 	}
1048 
1049 	return (error);
1050 }
1051