xref: /freebsd/sys/net/if_stf.c (revision 344d411c676dc92e648be86de369c4d1dd070209)
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/netisr.h>
101 #include <net/if_types.h>
102 #include <net/vnet.h>
103 
104 #include <netinet/in.h>
105 #include <netinet/in_fib.h>
106 #include <netinet/in_systm.h>
107 #include <netinet/ip.h>
108 #include <netinet/ip_var.h>
109 #include <netinet/in_var.h>
110 
111 #include <netinet/ip6.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet6/in6_var.h>
114 #include <netinet/ip_ecn.h>
115 
116 #include <netinet/ip_encap.h>
117 
118 #include <machine/stdarg.h>
119 
120 #include <net/bpf.h>
121 
122 #include <security/mac/mac_framework.h>
123 
124 SYSCTL_DECL(_net_link);
125 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
126     "6to4 Interface");
127 
128 static int stf_permit_rfc1918 = 0;
129 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
130     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
131 
132 #define STFUNIT		0
133 
134 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
135 
136 /*
137  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
138  * struct in_addr *; use bcopy() instead.
139  */
140 #define GET_V4(x)	(&(x)->s6_addr16[1])
141 
142 struct stf_softc {
143 	struct ifnet	*sc_ifp;
144 	u_int	sc_fibnum;
145 	const struct encaptab *encap_cookie;
146 };
147 #define STF2IFP(sc)	((sc)->sc_ifp)
148 
149 static const char stfname[] = "stf";
150 
151 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
152 static const int ip_stf_ttl = 40;
153 
154 static int in_stf_input(struct mbuf *, int, int, void *);
155 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
156 
157 static int stfmodevent(module_t, int, void *);
158 static int stf_encapcheck(const struct mbuf *, int, int, void *);
159 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
160 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
161 	struct route *);
162 static int isrfc1918addr(struct in_addr *);
163 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
164 	struct ifnet *);
165 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
166 	struct ifnet *);
167 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
168 
169 static int stf_clone_match(struct if_clone *, const char *);
170 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
171 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
172 static struct if_clone *stf_cloner;
173 
174 static const struct encap_config ipv4_encap_cfg = {
175 	.proto = IPPROTO_IPV6,
176 	.min_length = sizeof(struct ip),
177 	.exact_match = (sizeof(in_addr_t) << 3) + 8,
178 	.check = stf_encapcheck,
179 	.input = in_stf_input
180 };
181 
182 static int
183 stf_clone_match(struct if_clone *ifc, const char *name)
184 {
185 	int i;
186 
187 	for(i = 0; stfnames[i] != NULL; i++) {
188 		if (strcmp(stfnames[i], name) == 0)
189 			return (1);
190 	}
191 
192 	return (0);
193 }
194 
195 static int
196 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
197 {
198 	char *dp;
199 	int err, unit, wildcard;
200 	struct stf_softc *sc;
201 	struct ifnet *ifp;
202 
203 	err = ifc_name2unit(name, &unit);
204 	if (err != 0)
205 		return (err);
206 	wildcard = (unit < 0);
207 
208 	/*
209 	 * We can only have one unit, but since unit allocation is
210 	 * already locked, we use it to keep from allocating extra
211 	 * interfaces.
212 	 */
213 	unit = STFUNIT;
214 	err = ifc_alloc_unit(ifc, &unit);
215 	if (err != 0)
216 		return (err);
217 
218 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
219 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
220 	if (ifp == NULL) {
221 		free(sc, M_STF);
222 		ifc_free_unit(ifc, unit);
223 		return (ENOSPC);
224 	}
225 	ifp->if_softc = sc;
226 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
227 
228 	/*
229 	 * Set the name manually rather then using if_initname because
230 	 * we don't conform to the default naming convention for interfaces.
231 	 * In the wildcard case, we need to update the name.
232 	 */
233 	if (wildcard) {
234 		for (dp = name; *dp != '\0'; dp++);
235 		if (snprintf(dp, len - (dp-name), "%d", unit) >
236 		    len - (dp-name) - 1) {
237 			/*
238 			 * This can only be a programmer error and
239 			 * there's no straightforward way to recover if
240 			 * it happens.
241 			 */
242 			panic("if_clone_create(): interface name too long");
243 		}
244 	}
245 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
246 	ifp->if_dname = stfname;
247 	ifp->if_dunit = IF_DUNIT_NONE;
248 
249 	sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK);
250 	if (sc->encap_cookie == NULL) {
251 		if_printf(ifp, "attach failed\n");
252 		free(sc, M_STF);
253 		ifc_free_unit(ifc, unit);
254 		return (ENOMEM);
255 	}
256 
257 	ifp->if_mtu    = IPV6_MMTU;
258 	ifp->if_ioctl  = stf_ioctl;
259 	ifp->if_output = stf_output;
260 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
261 	if_attach(ifp);
262 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
263 	return (0);
264 }
265 
266 static int
267 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
268 {
269 	struct stf_softc *sc = ifp->if_softc;
270 	int err __unused;
271 
272 	err = ip_encap_detach(sc->encap_cookie);
273 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
274 	bpfdetach(ifp);
275 	if_detach(ifp);
276 	if_free(ifp);
277 
278 	free(sc, M_STF);
279 	ifc_free_unit(ifc, STFUNIT);
280 
281 	return (0);
282 }
283 
284 static int
285 stfmodevent(module_t mod, int type, void *data)
286 {
287 
288 	switch (type) {
289 	case MOD_LOAD:
290 		stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
291 		    stf_clone_create, stf_clone_destroy);
292 		break;
293 	case MOD_UNLOAD:
294 		if_clone_detach(stf_cloner);
295 		break;
296 	default:
297 		return (EOPNOTSUPP);
298 	}
299 
300 	return (0);
301 }
302 
303 static moduledata_t stf_mod = {
304 	"if_stf",
305 	stfmodevent,
306 	0
307 };
308 
309 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
310 
311 static int
312 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
313 {
314 	struct ip ip;
315 	struct stf_softc *sc;
316 	struct in_addr a, b, mask;
317 	struct in6_addr addr6, mask6;
318 
319 	sc = (struct stf_softc *)arg;
320 	if (sc == NULL)
321 		return 0;
322 
323 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
324 		return 0;
325 
326 	/* IFF_LINK0 means "no decapsulation" */
327 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
328 		return 0;
329 
330 	if (proto != IPPROTO_IPV6)
331 		return 0;
332 
333 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
334 
335 	if (ip.ip_v != 4)
336 		return 0;
337 
338 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
339 		return (0);
340 
341 	/*
342 	 * check if IPv4 dst matches the IPv4 address derived from the
343 	 * local 6to4 address.
344 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
345 	 */
346 	if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
347 		return 0;
348 
349 	/*
350 	 * check if IPv4 src matches the IPv4 address derived from the
351 	 * local 6to4 address masked by prefixmask.
352 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
353 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
354 	 */
355 	bzero(&a, sizeof(a));
356 	bcopy(GET_V4(&addr6), &a, sizeof(a));
357 	bcopy(GET_V4(&mask6), &mask, sizeof(mask));
358 	a.s_addr &= mask.s_addr;
359 	b = ip.ip_src;
360 	b.s_addr &= mask.s_addr;
361 	if (a.s_addr != b.s_addr)
362 		return 0;
363 
364 	/* stf interface makes single side match only */
365 	return 32;
366 }
367 
368 static int
369 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
370 {
371 	struct rm_priotracker in_ifa_tracker;
372 	struct ifaddr *ia;
373 	struct in_ifaddr *ia4;
374 	struct in6_ifaddr *ia6;
375 	struct sockaddr_in6 *sin6;
376 	struct in_addr in;
377 
378 	NET_EPOCH_ASSERT();
379 
380 	CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
381 		if (ia->ifa_addr->sa_family != AF_INET6)
382 			continue;
383 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
384 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
385 			continue;
386 
387 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
388 		IN_IFADDR_RLOCK(&in_ifa_tracker);
389 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
390 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
391 				break;
392 		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
393 		if (ia4 == NULL)
394 			continue;
395 
396 		ia6 = (struct in6_ifaddr *)ia;
397 
398 		*addr = sin6->sin6_addr;
399 		*mask = ia6->ia_prefixmask.sin6_addr;
400 		return (0);
401 	}
402 
403 	return (ENOENT);
404 }
405 
406 static int
407 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
408     struct route *ro)
409 {
410 	struct stf_softc *sc;
411 	const struct sockaddr_in6 *dst6;
412 	struct in_addr in4;
413 	const void *ptr;
414 	u_int8_t tos;
415 	struct ip *ip;
416 	struct ip6_hdr *ip6;
417 	struct in6_addr addr6, mask6;
418 	int error;
419 
420 #ifdef MAC
421 	error = mac_ifnet_check_transmit(ifp, m);
422 	if (error) {
423 		m_freem(m);
424 		return (error);
425 	}
426 #endif
427 
428 	sc = ifp->if_softc;
429 	dst6 = (const struct sockaddr_in6 *)dst;
430 
431 	/* just in case */
432 	if ((ifp->if_flags & IFF_UP) == 0) {
433 		m_freem(m);
434 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
435 		return ENETDOWN;
436 	}
437 
438 	/*
439 	 * If we don't have an ip4 address that match my inner ip6 address,
440 	 * we shouldn't generate output.  Without this check, we'll end up
441 	 * using wrong IPv4 source.
442 	 */
443 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
444 		m_freem(m);
445 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
446 		return ENETDOWN;
447 	}
448 
449 	if (m->m_len < sizeof(*ip6)) {
450 		m = m_pullup(m, sizeof(*ip6));
451 		if (!m) {
452 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
453 			return ENOBUFS;
454 		}
455 	}
456 	ip6 = mtod(m, struct ip6_hdr *);
457 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
458 
459 	/*
460 	 * Pickup the right outer dst addr from the list of candidates.
461 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
462 	 */
463 	ptr = NULL;
464 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
465 		ptr = GET_V4(&ip6->ip6_dst);
466 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
467 		ptr = GET_V4(&dst6->sin6_addr);
468 	else {
469 		m_freem(m);
470 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
471 		return ENETUNREACH;
472 	}
473 	bcopy(ptr, &in4, sizeof(in4));
474 
475 	if (bpf_peers_present(ifp->if_bpf)) {
476 		/*
477 		 * We need to prepend the address family as
478 		 * a four byte field.  Cons up a dummy header
479 		 * to pacify bpf.  This is safe because bpf
480 		 * will only read from the mbuf (i.e., it won't
481 		 * try to free it or keep a pointer a to it).
482 		 */
483 		u_int af = AF_INET6;
484 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
485 	}
486 
487 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
488 	if (m == NULL) {
489 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
490 		return ENOBUFS;
491 	}
492 	ip = mtod(m, struct ip *);
493 
494 	bzero(ip, sizeof(*ip));
495 
496 	bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
497 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
498 	ip->ip_p = IPPROTO_IPV6;
499 	ip->ip_ttl = ip_stf_ttl;
500 	ip->ip_len = htons(m->m_pkthdr.len);
501 	if (ifp->if_flags & IFF_LINK1)
502 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
503 	else
504 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
505 
506 	M_SETFIB(m, sc->sc_fibnum);
507 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
508 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
509 
510 	return error;
511 }
512 
513 static int
514 isrfc1918addr(struct in_addr *in)
515 {
516 	/*
517 	 * returns 1 if private address range:
518 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
519 	 */
520 	if (stf_permit_rfc1918 == 0 && (
521 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
522 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
523 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
524 		return 1;
525 
526 	return 0;
527 }
528 
529 static int
530 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
531 {
532 	struct rm_priotracker in_ifa_tracker;
533 	struct in_ifaddr *ia4;
534 
535 	/*
536 	 * reject packets with the following address:
537 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
538 	 */
539 	if (IN_MULTICAST(ntohl(in->s_addr)))
540 		return -1;
541 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
542 	case 0: case 127: case 255:
543 		return -1;
544 	}
545 
546 	/*
547 	 * reject packets with private address range.
548 	 * (requirement from RFC3056 section 2 1st paragraph)
549 	 */
550 	if (isrfc1918addr(in))
551 		return -1;
552 
553 	/*
554 	 * reject packets with broadcast
555 	 */
556 	IN_IFADDR_RLOCK(&in_ifa_tracker);
557 	CK_STAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
558 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
559 			continue;
560 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
561 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
562 			return -1;
563 		}
564 	}
565 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
566 
567 	/*
568 	 * perform ingress filter
569 	 */
570 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
571 		struct nhop4_basic nh4;
572 
573 		if (fib4_lookup_nh_basic(sc->sc_fibnum, *in, 0, 0, &nh4) != 0)
574 			return (-1);
575 
576 		if (nh4.nh_ifp != inifp)
577 			return (-1);
578 	}
579 
580 	return 0;
581 }
582 
583 static int
584 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
585 {
586 	/*
587 	 * check 6to4 addresses
588 	 */
589 	if (IN6_IS_ADDR_6TO4(in6)) {
590 		struct in_addr in4;
591 		bcopy(GET_V4(in6), &in4, sizeof(in4));
592 		return stf_checkaddr4(sc, &in4, inifp);
593 	}
594 
595 	/*
596 	 * reject anything that look suspicious.  the test is implemented
597 	 * in ip6_input too, but we check here as well to
598 	 * (1) reject bad packets earlier, and
599 	 * (2) to be safe against future ip6_input change.
600 	 */
601 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
602 		return -1;
603 
604 	return 0;
605 }
606 
607 static int
608 in_stf_input(struct mbuf *m, int off, int proto, void *arg)
609 {
610 	struct stf_softc *sc = arg;
611 	struct ip *ip;
612 	struct ip6_hdr *ip6;
613 	u_int8_t otos, itos;
614 	struct ifnet *ifp;
615 
616 	NET_EPOCH_ASSERT();
617 
618 	if (proto != IPPROTO_IPV6) {
619 		m_freem(m);
620 		return (IPPROTO_DONE);
621 	}
622 
623 	ip = mtod(m, struct ip *);
624 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
625 		m_freem(m);
626 		return (IPPROTO_DONE);
627 	}
628 
629 	ifp = STF2IFP(sc);
630 
631 #ifdef MAC
632 	mac_ifnet_create_mbuf(ifp, m);
633 #endif
634 
635 	/*
636 	 * perform sanity check against outer src/dst.
637 	 * for source, perform ingress filter as well.
638 	 */
639 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
640 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
641 		m_freem(m);
642 		return (IPPROTO_DONE);
643 	}
644 
645 	otos = ip->ip_tos;
646 	m_adj(m, off);
647 
648 	if (m->m_len < sizeof(*ip6)) {
649 		m = m_pullup(m, sizeof(*ip6));
650 		if (!m)
651 			return (IPPROTO_DONE);
652 	}
653 	ip6 = mtod(m, struct ip6_hdr *);
654 
655 	/*
656 	 * perform sanity check against inner src/dst.
657 	 * for source, perform ingress filter as well.
658 	 */
659 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
660 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
661 		m_freem(m);
662 		return (IPPROTO_DONE);
663 	}
664 
665 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
666 	if ((ifp->if_flags & IFF_LINK1) != 0)
667 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
668 	else
669 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
670 	ip6->ip6_flow &= ~htonl(0xff << 20);
671 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
672 
673 	m->m_pkthdr.rcvif = ifp;
674 
675 	if (bpf_peers_present(ifp->if_bpf)) {
676 		/*
677 		 * We need to prepend the address family as
678 		 * a four byte field.  Cons up a dummy header
679 		 * to pacify bpf.  This is safe because bpf
680 		 * will only read from the mbuf (i.e., it won't
681 		 * try to free it or keep a pointer a to it).
682 		 */
683 		u_int32_t af = AF_INET6;
684 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
685 	}
686 
687 	/*
688 	 * Put the packet to the network layer input queue according to the
689 	 * specified address family.
690 	 * See net/if_gif.c for possible issues with packet processing
691 	 * reorder due to extra queueing.
692 	 */
693 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
694 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
695 	M_SETFIB(m, ifp->if_fib);
696 	netisr_dispatch(NETISR_IPV6, m);
697 	return (IPPROTO_DONE);
698 }
699 
700 static int
701 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
702 {
703 	struct ifaddr *ifa;
704 	struct ifreq *ifr;
705 	struct sockaddr_in6 *sin6;
706 	struct in_addr addr;
707 	int error, mtu;
708 
709 	error = 0;
710 	switch (cmd) {
711 	case SIOCSIFADDR:
712 		ifa = (struct ifaddr *)data;
713 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
714 			error = EAFNOSUPPORT;
715 			break;
716 		}
717 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
718 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
719 			error = EINVAL;
720 			break;
721 		}
722 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
723 		if (isrfc1918addr(&addr)) {
724 			error = EINVAL;
725 			break;
726 		}
727 
728 		ifp->if_flags |= IFF_UP;
729 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
730 		break;
731 
732 	case SIOCADDMULTI:
733 	case SIOCDELMULTI:
734 		ifr = (struct ifreq *)data;
735 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
736 			;
737 		else
738 			error = EAFNOSUPPORT;
739 		break;
740 
741 	case SIOCGIFMTU:
742 		break;
743 
744 	case SIOCSIFMTU:
745 		ifr = (struct ifreq *)data;
746 		mtu = ifr->ifr_mtu;
747 		/* RFC 4213 3.2 ideal world MTU */
748 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
749 			return (EINVAL);
750 		ifp->if_mtu = mtu;
751 		break;
752 
753 	default:
754 		error = EINVAL;
755 		break;
756 	}
757 
758 	return error;
759 }
760