xref: /freebsd-13.1/sys/net/if_stf.c (revision 563b1596)
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  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the project nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 /*
36  * 6to4 interface, based on RFC3056.
37  *
38  * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
39  * There is no address mapping defined from IPv6 multicast address to IPv4
40  * address.  Therefore, we do not have IFF_MULTICAST on the interface.
41  *
42  * Due to the lack of address mapping for link-local addresses, we cannot
43  * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
44  * packets to link-local multicast addresses (ff02::x).
45  *
46  * Here are interesting symptoms due to the lack of link-local address:
47  *
48  * Unicast routing exchange:
49  * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
50  *   and link-local addresses as nexthop.
51  * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
52  *   assigned to the link, and makes use of them.  Also, HELLO packets use
53  *   link-local multicast addresses (ff02::5 and ff02::6).
54  * - BGP4+: Maybe.  You can only use global address as nexthop, and global
55  *   address as TCP endpoint address.
56  *
57  * Multicast routing protocols:
58  * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
59  *   Adjacent PIM routers must be configured manually (is it really spec-wise
60  *   correct thing to do?).
61  *
62  * ICMPv6:
63  * - Redirects cannot be used due to the lack of link-local address.
64  *
65  * stf interface does not have, and will not need, a link-local address.
66  * It seems to have no real benefit and does not help the above symptoms much.
67  * Even if we assign link-locals to interface, we cannot really
68  * use link-local unicast/multicast on top of 6to4 cloud (since there's no
69  * encapsulation defined for link-local address), and the above analysis does
70  * not change.  RFC3056 does not mandate the assignment of link-local address
71  * either.
72  *
73  * 6to4 interface has security issues.  Refer to
74  * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
75  * for details.  The code tries to filter out some of malicious packets.
76  * Note that there is no way to be 100% secure.
77  */
78 
79 #include <sys/param.h>
80 #include <sys/systm.h>
81 #include <sys/socket.h>
82 #include <sys/sockio.h>
83 #include <sys/mbuf.h>
84 #include <sys/errno.h>
85 #include <sys/kernel.h>
86 #include <sys/lock.h>
87 #include <sys/module.h>
88 #include <sys/proc.h>
89 #include <sys/queue.h>
90 #include <sys/rmlock.h>
91 #include <sys/sysctl.h>
92 #include <machine/cpu.h>
93 
94 #include <sys/malloc.h>
95 
96 #include <net/if.h>
97 #include <net/if_var.h>
98 #include <net/if_clone.h>
99 #include <net/route.h>
100 #include <net/route/nhop.h>
101 #include <net/netisr.h>
102 #include <net/if_types.h>
103 #include <net/vnet.h>
104 
105 #include <netinet/in.h>
106 #include <netinet/in_fib.h>
107 #include <netinet/in_systm.h>
108 #include <netinet/ip.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/in_var.h>
111 
112 #include <netinet/ip6.h>
113 #include <netinet6/ip6_var.h>
114 #include <netinet6/in6_var.h>
115 #include <netinet/ip_ecn.h>
116 
117 #include <netinet/ip_encap.h>
118 
119 #include <machine/stdarg.h>
120 
121 #include <net/bpf.h>
122 
123 #include <security/mac/mac_framework.h>
124 
125 SYSCTL_DECL(_net_link);
126 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
127     "6to4 Interface");
128 
129 static int stf_permit_rfc1918 = 0;
130 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
131     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
132 
133 #define STFUNIT		0
134 
135 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
136 
137 /*
138  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
139  * struct in_addr *; use bcopy() instead.
140  */
141 #define GET_V4(x)	(&(x)->s6_addr16[1])
142 
143 struct stf_softc {
144 	struct ifnet	*sc_ifp;
145 	u_int	sc_fibnum;
146 	const struct encaptab *encap_cookie;
147 };
148 #define STF2IFP(sc)	((sc)->sc_ifp)
149 
150 static const char stfname[] = "stf";
151 
152 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
153 static const int ip_stf_ttl = 40;
154 
155 static int in_stf_input(struct mbuf *, int, int, void *);
156 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
157 
158 static int stfmodevent(module_t, int, void *);
159 static int stf_encapcheck(const struct mbuf *, int, int, void *);
160 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
161 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
162 	struct route *);
163 static int isrfc1918addr(struct in_addr *);
164 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
165 	struct ifnet *);
166 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
167 	struct ifnet *);
168 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
169 
170 static int stf_clone_match(struct if_clone *, const char *);
171 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
172 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
173 VNET_DEFINE_STATIC(struct if_clone *, stf_cloner);
174 #define V_stf_cloner	VNET(stf_cloner)
175 
176 static const struct encap_config ipv4_encap_cfg = {
177 	.proto = IPPROTO_IPV6,
178 	.min_length = sizeof(struct ip),
179 	.exact_match = (sizeof(in_addr_t) << 3) + 8,
180 	.check = stf_encapcheck,
181 	.input = in_stf_input
182 };
183 
184 static int
stf_clone_match(struct if_clone * ifc,const char * name)185 stf_clone_match(struct if_clone *ifc, const char *name)
186 {
187 	int i;
188 
189 	for(i = 0; stfnames[i] != NULL; i++) {
190 		if (strcmp(stfnames[i], name) == 0)
191 			return (1);
192 	}
193 
194 	return (0);
195 }
196 
197 static int
stf_clone_create(struct if_clone * ifc,char * name,size_t len,caddr_t params)198 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
199 {
200 	char *dp;
201 	int err, unit, wildcard;
202 	struct stf_softc *sc;
203 	struct ifnet *ifp;
204 
205 	err = ifc_name2unit(name, &unit);
206 	if (err != 0)
207 		return (err);
208 	wildcard = (unit < 0);
209 
210 	/*
211 	 * We can only have one unit, but since unit allocation is
212 	 * already locked, we use it to keep from allocating extra
213 	 * interfaces.
214 	 */
215 	unit = STFUNIT;
216 	err = ifc_alloc_unit(ifc, &unit);
217 	if (err != 0)
218 		return (err);
219 
220 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
221 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
222 	if (ifp == NULL) {
223 		free(sc, M_STF);
224 		ifc_free_unit(ifc, unit);
225 		return (ENOSPC);
226 	}
227 	ifp->if_softc = sc;
228 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
229 
230 	/*
231 	 * Set the name manually rather then using if_initname because
232 	 * we don't conform to the default naming convention for interfaces.
233 	 * In the wildcard case, we need to update the name.
234 	 */
235 	if (wildcard) {
236 		for (dp = name; *dp != '\0'; dp++);
237 		if (snprintf(dp, len - (dp-name), "%d", unit) >
238 		    len - (dp-name) - 1) {
239 			/*
240 			 * This can only be a programmer error and
241 			 * there's no straightforward way to recover if
242 			 * it happens.
243 			 */
244 			panic("if_clone_create(): interface name too long");
245 		}
246 	}
247 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
248 	ifp->if_dname = stfname;
249 	ifp->if_dunit = IF_DUNIT_NONE;
250 
251 	sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK);
252 	if (sc->encap_cookie == NULL) {
253 		if_printf(ifp, "attach failed\n");
254 		free(sc, M_STF);
255 		ifc_free_unit(ifc, unit);
256 		return (ENOMEM);
257 	}
258 
259 	ifp->if_mtu    = IPV6_MMTU;
260 	ifp->if_ioctl  = stf_ioctl;
261 	ifp->if_output = stf_output;
262 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
263 	if_attach(ifp);
264 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
265 	return (0);
266 }
267 
268 static int
stf_clone_destroy(struct if_clone * ifc,struct ifnet * ifp)269 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
270 {
271 	struct stf_softc *sc = ifp->if_softc;
272 	int err __unused;
273 
274 	err = ip_encap_detach(sc->encap_cookie);
275 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
276 	bpfdetach(ifp);
277 	if_detach(ifp);
278 	if_free(ifp);
279 
280 	free(sc, M_STF);
281 	ifc_free_unit(ifc, STFUNIT);
282 
283 	return (0);
284 }
285 
286 static void
vnet_stf_init(const void * unused __unused)287 vnet_stf_init(const void *unused __unused)
288 {
289 	V_stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
290 	    stf_clone_create, stf_clone_destroy);
291 }
292 VNET_SYSINIT(vnet_stf_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_init, NULL);
293 
294 static void
vnet_stf_uninit(const void * unused __unused)295 vnet_stf_uninit(const void *unused __unused)
296 {
297 	if_clone_detach(V_stf_cloner);
298 	V_stf_cloner = NULL;
299 }
300 VNET_SYSUNINIT(vnet_stf_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_uninit,
301     NULL);
302 
303 static int
stfmodevent(module_t mod,int type,void * data)304 stfmodevent(module_t mod, int type, void *data)
305 {
306 
307 	switch (type) {
308 	case MOD_LOAD:
309 		/* Done in vnet_stf_init() */
310 		break;
311 	case MOD_UNLOAD:
312 		/* Done in vnet_stf_uninit() */
313 		break;
314 	default:
315 		return (EOPNOTSUPP);
316 	}
317 
318 	return (0);
319 }
320 
321 static moduledata_t stf_mod = {
322 	"if_stf",
323 	stfmodevent,
324 	0
325 };
326 
327 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
328 
329 static int
stf_encapcheck(const struct mbuf * m,int off,int proto,void * arg)330 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
331 {
332 	struct ip ip;
333 	struct stf_softc *sc;
334 	struct in_addr a, b, mask;
335 	struct in6_addr addr6, mask6;
336 
337 	sc = (struct stf_softc *)arg;
338 	if (sc == NULL)
339 		return (0);
340 
341 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
342 		return (0);
343 
344 	/* IFF_LINK0 means "no decapsulation" */
345 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
346 		return (0);
347 
348 	if (proto != IPPROTO_IPV6)
349 		return (0);
350 
351 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
352 
353 	if (ip.ip_v != 4)
354 		return (0);
355 
356 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
357 		return (0);
358 
359 	/*
360 	 * check if IPv4 dst matches the IPv4 address derived from the
361 	 * local 6to4 address.
362 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
363 	 */
364 	if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
365 		return (0);
366 
367 	/*
368 	 * check if IPv4 src matches the IPv4 address derived from the
369 	 * local 6to4 address masked by prefixmask.
370 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
371 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
372 	 */
373 	bzero(&a, sizeof(a));
374 	bcopy(GET_V4(&addr6), &a, sizeof(a));
375 	bcopy(GET_V4(&mask6), &mask, sizeof(mask));
376 	a.s_addr &= mask.s_addr;
377 	b = ip.ip_src;
378 	b.s_addr &= mask.s_addr;
379 	if (a.s_addr != b.s_addr)
380 		return (0);
381 
382 	/* stf interface makes single side match only */
383 	return (32);
384 }
385 
386 static int
stf_getsrcifa6(struct ifnet * ifp,struct in6_addr * addr,struct in6_addr * mask)387 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
388 {
389 	struct rm_priotracker in_ifa_tracker;
390 	struct ifaddr *ia;
391 	struct in_ifaddr *ia4;
392 	struct in6_ifaddr *ia6;
393 	struct sockaddr_in6 *sin6;
394 	struct in_addr in;
395 
396 	NET_EPOCH_ASSERT();
397 
398 	CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
399 		if (ia->ifa_addr->sa_family != AF_INET6)
400 			continue;
401 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
402 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
403 			continue;
404 
405 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
406 		IN_IFADDR_RLOCK(&in_ifa_tracker);
407 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
408 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
409 				break;
410 		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
411 		if (ia4 == NULL)
412 			continue;
413 
414 		ia6 = (struct in6_ifaddr *)ia;
415 
416 		*addr = sin6->sin6_addr;
417 		*mask = ia6->ia_prefixmask.sin6_addr;
418 		return (0);
419 	}
420 
421 	return (ENOENT);
422 }
423 
424 static int
stf_output(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,struct route * ro)425 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
426     struct route *ro)
427 {
428 	struct stf_softc *sc;
429 	const struct sockaddr_in6 *dst6;
430 	struct in_addr in4;
431 	const void *ptr;
432 	u_int8_t tos;
433 	struct ip *ip;
434 	struct ip6_hdr *ip6;
435 	struct in6_addr addr6, mask6;
436 	int error;
437 
438 #ifdef MAC
439 	error = mac_ifnet_check_transmit(ifp, m);
440 	if (error) {
441 		m_freem(m);
442 		return (error);
443 	}
444 #endif
445 
446 	sc = ifp->if_softc;
447 	dst6 = (const struct sockaddr_in6 *)dst;
448 
449 	/* just in case */
450 	if ((ifp->if_flags & IFF_UP) == 0) {
451 		m_freem(m);
452 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
453 		return (ENETDOWN);
454 	}
455 
456 	/*
457 	 * If we don't have an ip4 address that match my inner ip6 address,
458 	 * we shouldn't generate output.  Without this check, we'll end up
459 	 * using wrong IPv4 source.
460 	 */
461 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
462 		m_freem(m);
463 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
464 		return (ENETDOWN);
465 	}
466 
467 	if (m->m_len < sizeof(*ip6)) {
468 		m = m_pullup(m, sizeof(*ip6));
469 		if (!m) {
470 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
471 			return (ENOBUFS);
472 		}
473 	}
474 	ip6 = mtod(m, struct ip6_hdr *);
475 	tos = IPV6_TRAFFIC_CLASS(ip6);
476 
477 	/*
478 	 * Pickup the right outer dst addr from the list of candidates.
479 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
480 	 */
481 	ptr = NULL;
482 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
483 		ptr = GET_V4(&ip6->ip6_dst);
484 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
485 		ptr = GET_V4(&dst6->sin6_addr);
486 	else {
487 		m_freem(m);
488 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
489 		return (ENETUNREACH);
490 	}
491 	bcopy(ptr, &in4, sizeof(in4));
492 
493 	if (bpf_peers_present(ifp->if_bpf)) {
494 		/*
495 		 * We need to prepend the address family as
496 		 * a four byte field.  Cons up a dummy header
497 		 * to pacify bpf.  This is safe because bpf
498 		 * will only read from the mbuf (i.e., it won't
499 		 * try to free it or keep a pointer a to it).
500 		 */
501 		u_int af = AF_INET6;
502 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
503 	}
504 
505 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
506 	if (m == NULL) {
507 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
508 		return (ENOBUFS);
509 	}
510 	ip = mtod(m, struct ip *);
511 
512 	bzero(ip, sizeof(*ip));
513 
514 	bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
515 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
516 	ip->ip_p = IPPROTO_IPV6;
517 	ip->ip_ttl = ip_stf_ttl;
518 	ip->ip_len = htons(m->m_pkthdr.len);
519 	if (ifp->if_flags & IFF_LINK1)
520 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
521 	else
522 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
523 
524 	M_SETFIB(m, sc->sc_fibnum);
525 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
526 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
527 
528 	return (error);
529 }
530 
531 static int
isrfc1918addr(struct in_addr * in)532 isrfc1918addr(struct in_addr *in)
533 {
534 	/*
535 	 * returns 1 if private address range:
536 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
537 	 */
538 	if (stf_permit_rfc1918 == 0 && (
539 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
540 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
541 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
542 		return (1);
543 
544 	return (0);
545 }
546 
547 static int
stf_checkaddr4(struct stf_softc * sc,struct in_addr * in,struct ifnet * inifp)548 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
549 {
550 	struct rm_priotracker in_ifa_tracker;
551 	struct in_ifaddr *ia4;
552 
553 	/*
554 	 * reject packets with the following address:
555 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
556 	 */
557 	if (IN_MULTICAST(ntohl(in->s_addr)))
558 		return (-1);
559 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
560 	case 0: case 127: case 255:
561 		return (-1);
562 	}
563 
564 	/*
565 	 * reject packets with private address range.
566 	 * (requirement from RFC3056 section 2 1st paragraph)
567 	 */
568 	if (isrfc1918addr(in))
569 		return (-1);
570 
571 	/*
572 	 * reject packets with broadcast
573 	 */
574 	IN_IFADDR_RLOCK(&in_ifa_tracker);
575 	CK_STAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
576 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
577 			continue;
578 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
579 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
580 			return (-1);
581 		}
582 	}
583 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
584 
585 	/*
586 	 * perform ingress filter
587 	 */
588 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
589 		struct nhop_object *nh;
590 
591 		NET_EPOCH_ASSERT();
592 		nh = fib4_lookup(sc->sc_fibnum, *in, 0, 0, 0);
593 		if (nh == NULL)
594 			return (-1);
595 
596 		if (nh->nh_ifp != inifp)
597 			return (-1);
598 	}
599 
600 	return (0);
601 }
602 
603 static int
stf_checkaddr6(struct stf_softc * sc,struct in6_addr * in6,struct ifnet * inifp)604 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
605 {
606 	/*
607 	 * check 6to4 addresses
608 	 */
609 	if (IN6_IS_ADDR_6TO4(in6)) {
610 		struct in_addr in4;
611 		bcopy(GET_V4(in6), &in4, sizeof(in4));
612 		return (stf_checkaddr4(sc, &in4, inifp));
613 	}
614 
615 	/*
616 	 * reject anything that look suspicious.  the test is implemented
617 	 * in ip6_input too, but we check here as well to
618 	 * (1) reject bad packets earlier, and
619 	 * (2) to be safe against future ip6_input change.
620 	 */
621 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
622 		return (-1);
623 
624 	return (0);
625 }
626 
627 static int
in_stf_input(struct mbuf * m,int off,int proto,void * arg)628 in_stf_input(struct mbuf *m, int off, int proto, void *arg)
629 {
630 	struct stf_softc *sc = arg;
631 	struct ip *ip;
632 	struct ip6_hdr *ip6;
633 	u_int8_t otos, itos;
634 	struct ifnet *ifp;
635 
636 	NET_EPOCH_ASSERT();
637 
638 	if (proto != IPPROTO_IPV6) {
639 		m_freem(m);
640 		return (IPPROTO_DONE);
641 	}
642 
643 	ip = mtod(m, struct ip *);
644 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
645 		m_freem(m);
646 		return (IPPROTO_DONE);
647 	}
648 
649 	ifp = STF2IFP(sc);
650 
651 #ifdef MAC
652 	mac_ifnet_create_mbuf(ifp, m);
653 #endif
654 
655 	/*
656 	 * perform sanity check against outer src/dst.
657 	 * for source, perform ingress filter as well.
658 	 */
659 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
660 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
661 		m_freem(m);
662 		return (IPPROTO_DONE);
663 	}
664 
665 	otos = ip->ip_tos;
666 	m_adj(m, off);
667 
668 	if (m->m_len < sizeof(*ip6)) {
669 		m = m_pullup(m, sizeof(*ip6));
670 		if (!m)
671 			return (IPPROTO_DONE);
672 	}
673 	ip6 = mtod(m, struct ip6_hdr *);
674 
675 	/*
676 	 * perform sanity check against inner src/dst.
677 	 * for source, perform ingress filter as well.
678 	 */
679 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
680 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
681 		m_freem(m);
682 		return (IPPROTO_DONE);
683 	}
684 
685 	itos = IPV6_TRAFFIC_CLASS(ip6);
686 	if ((ifp->if_flags & IFF_LINK1) != 0)
687 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
688 	else
689 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
690 	ip6->ip6_flow &= ~htonl(0xff << 20);
691 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
692 
693 	m->m_pkthdr.rcvif = ifp;
694 
695 	if (bpf_peers_present(ifp->if_bpf)) {
696 		/*
697 		 * We need to prepend the address family as
698 		 * a four byte field.  Cons up a dummy header
699 		 * to pacify bpf.  This is safe because bpf
700 		 * will only read from the mbuf (i.e., it won't
701 		 * try to free it or keep a pointer a to it).
702 		 */
703 		u_int32_t af = AF_INET6;
704 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
705 	}
706 
707 	/*
708 	 * Put the packet to the network layer input queue according to the
709 	 * specified address family.
710 	 * See net/if_gif.c for possible issues with packet processing
711 	 * reorder due to extra queueing.
712 	 */
713 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
714 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
715 	M_SETFIB(m, ifp->if_fib);
716 	netisr_dispatch(NETISR_IPV6, m);
717 	return (IPPROTO_DONE);
718 }
719 
720 static int
stf_ioctl(struct ifnet * ifp,u_long cmd,caddr_t data)721 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
722 {
723 	struct ifaddr *ifa;
724 	struct ifreq *ifr;
725 	struct sockaddr_in6 *sin6;
726 	struct in_addr addr;
727 	int error, mtu;
728 
729 	error = 0;
730 	switch (cmd) {
731 	case SIOCSIFADDR:
732 		ifa = (struct ifaddr *)data;
733 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
734 			error = EAFNOSUPPORT;
735 			break;
736 		}
737 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
738 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
739 			error = EINVAL;
740 			break;
741 		}
742 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
743 		if (isrfc1918addr(&addr)) {
744 			error = EINVAL;
745 			break;
746 		}
747 
748 		ifp->if_flags |= IFF_UP;
749 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
750 		break;
751 
752 	case SIOCADDMULTI:
753 	case SIOCDELMULTI:
754 		ifr = (struct ifreq *)data;
755 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
756 			;
757 		else
758 			error = EAFNOSUPPORT;
759 		break;
760 
761 	case SIOCGIFMTU:
762 		break;
763 
764 	case SIOCSIFMTU:
765 		ifr = (struct ifreq *)data;
766 		mtu = ifr->ifr_mtu;
767 		/* RFC 4213 3.2 ideal world MTU */
768 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
769 			return (EINVAL);
770 		ifp->if_mtu = mtu;
771 		break;
772 
773 	default:
774 		error = EINVAL;
775 		break;
776 	}
777 
778 	return (error);
779 }
780