xref: /netbsd/sys/net/if_stf.c (revision d10e2ccb)
1 /*	$NetBSD: if_stf.c,v 1.109 2022/09/03 02:47:59 thorpej Exp $	*/
2 /*	$KAME: if_stf.c,v 1.62 2001/06/07 22:32:16 itojun Exp $ */
3 
4 /*
5  * Copyright (C) 2000 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * 6to4 interface, based on RFC3056.
35  *
36  * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
37  * There is no address mapping defined from IPv6 multicast address to IPv4
38  * address.  Therefore, we do not have IFF_MULTICAST on the interface.
39  *
40  * Due to the lack of address mapping for link-local addresses, we cannot
41  * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
42  * packets to link-local multicast addresses (ff02::x).
43  *
44  * Here are interesting symptoms due to the lack of link-local address:
45  *
46  * Unicast routing exchange:
47  * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
48  *   and link-local addresses as nexthop.
49  * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
50  *   assigned to the link, and makes use of them.  Also, HELLO packets use
51  *   link-local multicast addresses (ff02::5 and ff02::6).
52  * - BGP4+: Maybe.  You can only use global address as nexthop, and global
53  *   address as TCP endpoint address.
54  *
55  * Multicast routing protocols:
56  * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
57  *   Adjacent PIM routers must be configured manually (is it really spec-wise
58  *   correct thing to do?).
59  *
60  * ICMPv6:
61  * - Redirects cannot be used due to the lack of link-local address.
62  *
63  * stf interface does not have, and will not need, a link-local address.
64  * It seems to have no real benefit and does not help the above symptoms much.
65  * Even if we assign link-locals to interface, we cannot really
66  * use link-local unicast/multicast on top of 6to4 cloud (since there's no
67  * encapsulation defined for link-local address), and the above analysis does
68  * not change.  RFC3056 does not mandate the assignment of link-local address
69  * either.
70  *
71  * 6to4 interface has security issues.  Refer to
72  * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
73  * for details.  The code tries to filter out some of malicious packets.
74  * Note that there is no way to be 100% secure.
75  */
76 
77 #include <sys/cdefs.h>
78 __KERNEL_RCSID(0, "$NetBSD: if_stf.c,v 1.109 2022/09/03 02:47:59 thorpej Exp $");
79 
80 #ifdef _KERNEL_OPT
81 #include "opt_inet.h"
82 #include "stf.h"
83 #endif
84 
85 #ifndef INET6
86 	#error "pseudo-device stf requires options INET6"
87 #endif
88 
89 #include <sys/param.h>
90 #include <sys/systm.h>
91 #include <sys/socket.h>
92 #include <sys/sockio.h>
93 #include <sys/mbuf.h>
94 #include <sys/errno.h>
95 #include <sys/ioctl.h>
96 #include <sys/proc.h>
97 #include <sys/queue.h>
98 #include <sys/syslog.h>
99 #include <sys/device.h>
100 #include <sys/module.h>
101 
102 #include <sys/cpu.h>
103 
104 #include <net/if.h>
105 #include <net/route.h>
106 #include <net/if_types.h>
107 #include <net/if_stf.h>
108 
109 #include <netinet/in.h>
110 #include <netinet/in_systm.h>
111 #include <netinet/ip.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/in_var.h>
114 
115 #include <netinet/ip6.h>
116 #include <netinet6/ip6_var.h>
117 #include <netinet6/in6_var.h>
118 #include <netinet/ip_ecn.h>
119 
120 #include <netinet/ip_encap.h>
121 
122 #include <net/bpf.h>
123 
124 #include "ioconf.h"
125 
126 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
127 #define GET_V4(x)	((const struct in_addr *)(&(x)->s6_addr16[1]))
128 
129 struct stf_softc {
130 	struct ifnet	sc_if;	   /* common area */
131 	struct route	sc_ro;
132 	const struct encaptab *encap_cookie;
133 	LIST_ENTRY(stf_softc) sc_list;
134 };
135 
136 static LIST_HEAD(, stf_softc) stf_softc_list;
137 
138 static int	stf_clone_create(struct if_clone *, int);
139 static int	stf_clone_destroy(struct ifnet *);
140 
141 struct if_clone stf_cloner =
142     IF_CLONE_INITIALIZER("stf", stf_clone_create, stf_clone_destroy);
143 
144 static int ip_stf_ttl = STF_TTL;
145 
146 extern struct domain inetdomain;
147 
148 static const struct encapsw in_stf_encapsw =
149 {
150 	.encapsw4 = {
151 		.pr_input	= in_stf_input,
152 		.pr_ctlinput	= NULL,
153 	}
154 };
155 
156 static int stf_encapcheck(struct mbuf *, int, int, void *);
157 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
158 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
159 	const struct rtentry *);
160 static int isrfc1918addr(const struct in_addr *);
161 static int stf_checkaddr4(struct stf_softc *, const struct in_addr *,
162 	struct ifnet *);
163 static int stf_checkaddr6(struct stf_softc *, const struct in6_addr *,
164 	struct ifnet *);
165 static void stf_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
166 static int stf_ioctl(struct ifnet *, u_long, void *);
167 
168 /* ARGSUSED */
169 void
stfattach(int count)170 stfattach(int count)
171 {
172 
173 	/*
174 	 * Nothing to do here, initialization is handled by the
175 	 * module initialization code in stfinit() below).
176 	 */
177 }
178 
179 static void
stfinit(void)180 stfinit(void)
181 {
182 
183 	LIST_INIT(&stf_softc_list);
184 	if_clone_attach(&stf_cloner);
185 }
186 
187 static int
stfdetach(void)188 stfdetach(void)
189 {
190 	int error = 0;
191 
192 	if (!LIST_EMPTY(&stf_softc_list))
193 		error = EBUSY;
194 
195 	if (error == 0)
196 		if_clone_detach(&stf_cloner);
197 
198 	return error;
199 }
200 
201 static int
stf_clone_create(struct if_clone * ifc,int unit)202 stf_clone_create(struct if_clone *ifc, int unit)
203 {
204 	struct stf_softc *sc;
205 	int error;
206 
207 	sc = malloc(sizeof(struct stf_softc), M_DEVBUF, M_WAIT|M_ZERO);
208 	if_initname(&sc->sc_if, ifc->ifc_name, unit);
209 
210 	error = encap_lock_enter();
211 	if (error) {
212 		free(sc, M_DEVBUF);
213 		return error;
214 	}
215 
216 	if (LIST_FIRST(&stf_softc_list) != NULL) {
217 		/* Only one stf interface is allowed. */
218 		encap_lock_exit();
219 		free(sc, M_DEVBUF);
220 		return EEXIST;
221 	}
222 
223 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
224 	    stf_encapcheck, &in_stf_encapsw, sc);
225 	encap_lock_exit();
226 	if (sc->encap_cookie == NULL) {
227 		printf("%s: unable to attach encap\n", if_name(&sc->sc_if));
228 		free(sc, M_DEVBUF);
229 		return EIO;	/* XXX */
230 	}
231 
232 	sc->sc_if.if_mtu    = STF_MTU;
233 	sc->sc_if.if_flags  = 0;
234 	sc->sc_if.if_ioctl  = stf_ioctl;
235 	sc->sc_if.if_output = stf_output;
236 	sc->sc_if.if_type   = IFT_STF;
237 	sc->sc_if.if_dlt    = DLT_NULL;
238 	if_attach(&sc->sc_if);
239 	if_alloc_sadl(&sc->sc_if);
240 	bpf_attach(&sc->sc_if, DLT_NULL, sizeof(u_int));
241 	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
242 	return 0;
243 }
244 
245 static int
stf_clone_destroy(struct ifnet * ifp)246 stf_clone_destroy(struct ifnet *ifp)
247 {
248 	struct stf_softc *sc = (void *) ifp;
249 
250 	encap_lock_enter();
251 	LIST_REMOVE(sc, sc_list);
252 	encap_detach(sc->encap_cookie);
253 	encap_lock_exit();
254 	bpf_detach(ifp);
255 	if_detach(ifp);
256 	rtcache_free(&sc->sc_ro);
257 	free(sc, M_DEVBUF);
258 
259 	return 0;
260 }
261 
262 static int
stf_encapcheck(struct mbuf * m,int off,int proto,void * arg)263 stf_encapcheck(struct mbuf *m, int off, int proto, void *arg)
264 {
265 	struct ip ip;
266 	struct in6_ifaddr *ia6;
267 	struct stf_softc *sc;
268 	struct in_addr a, b;
269 
270 	sc = (struct stf_softc *)arg;
271 	if (sc == NULL)
272 		return 0;
273 
274 	if ((sc->sc_if.if_flags & IFF_UP) == 0)
275 		return 0;
276 
277 	/* IFF_LINK0 means "no decapsulation" */
278 	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
279 		return 0;
280 
281 	if (proto != IPPROTO_IPV6)
282 		return 0;
283 
284 	m_copydata(m, 0, sizeof(ip), (void *)&ip);
285 
286 	if (ip.ip_v != 4)
287 		return 0;
288 
289 	ia6 = stf_getsrcifa6(&sc->sc_if);
290 	if (ia6 == NULL)
291 		return 0;
292 
293 	/*
294 	 * check if IPv4 dst matches the IPv4 address derived from the
295 	 * local 6to4 address.
296 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
297 	 */
298 	if (memcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
299 	    sizeof(ip.ip_dst)) != 0)
300 		return 0;
301 
302 	/*
303 	 * check if IPv4 src matches the IPv4 address derived from the
304 	 * local 6to4 address masked by prefixmask.
305 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
306 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
307 	 */
308 	memset(&a, 0, sizeof(a));
309 	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
310 	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
311 	b = ip.ip_src;
312 	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
313 	if (a.s_addr != b.s_addr)
314 		return 0;
315 
316 	/* stf interface makes single side match only */
317 	return 32;
318 }
319 
320 static struct in6_ifaddr *
stf_getsrcifa6(struct ifnet * ifp)321 stf_getsrcifa6(struct ifnet *ifp)
322 {
323 	struct ifaddr *ifa;
324 	struct in_ifaddr *ia4;
325 	struct sockaddr_in6 *sin6;
326 	struct in_addr in;
327 	int s;
328 
329 	s = pserialize_read_enter();
330 	IFADDR_READER_FOREACH(ifa, ifp) {
331 		if (ifa->ifa_addr->sa_family != AF_INET6)
332 			continue;
333 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
334 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
335 			continue;
336 
337 		memcpy(&in, GET_V4(&sin6->sin6_addr), sizeof(in));
338 		ia4 = in_get_ia(in);
339 		if (ia4 == NULL)
340 			continue;
341 
342 		pserialize_read_exit(s);
343 		/* TODO NOMPSAFE */
344 		return (struct in6_ifaddr *)ifa;
345 	}
346 	pserialize_read_exit(s);
347 
348 	return NULL;
349 }
350 
351 static int
stf_output(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,const struct rtentry * rt0)352 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
353     const struct rtentry *rt0)
354 {
355 	struct rtentry *rt;
356 	struct stf_softc *sc;
357 	const struct sockaddr_in6 *dst6;
358 	const struct in_addr *in4;
359 	uint8_t tos;
360 	struct ip *ip;
361 	struct ip6_hdr *ip6;
362 	struct in6_ifaddr *ia6;
363 	union {
364 		struct sockaddr		dst;
365 		struct sockaddr_in	dst4;
366 	} u;
367 
368 	sc = (struct stf_softc*)ifp;
369 	dst6 = (const struct sockaddr_in6 *)dst;
370 
371 	/* just in case */
372 	if ((ifp->if_flags & IFF_UP) == 0) {
373 		m_freem(m);
374 		return ENETDOWN;
375 	}
376 
377 	/*
378 	 * If we don't have an ip4 address that match my inner ip6 address,
379 	 * we shouldn't generate output.  Without this check, we'll end up
380 	 * using wrong IPv4 source.
381 	 */
382 	ia6 = stf_getsrcifa6(ifp);
383 	if (ia6 == NULL) {
384 		m_freem(m);
385 		if_statinc(ifp, if_oerrors);
386 		return ENETDOWN;
387 	}
388 
389 	if (m->m_len < sizeof(*ip6)) {
390 		m = m_pullup(m, sizeof(*ip6));
391 		if (m == NULL) {
392 			if_statinc(ifp, if_oerrors);
393 			return ENOBUFS;
394 		}
395 	}
396 	ip6 = mtod(m, struct ip6_hdr *);
397 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
398 
399 	/*
400 	 * Pickup the right outer dst addr from the list of candidates.
401 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
402 	 */
403 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
404 		in4 = GET_V4(&ip6->ip6_dst);
405 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
406 		in4 = GET_V4(&dst6->sin6_addr);
407 	else {
408 		m_freem(m);
409 		if_statinc(ifp, if_oerrors);
410 		return ENETUNREACH;
411 	}
412 
413 	bpf_mtap_af(ifp, AF_INET6, m, BPF_D_OUT);
414 
415 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
416 	if (m && m->m_len < sizeof(struct ip))
417 		m = m_pullup(m, sizeof(struct ip));
418 	if (m == NULL) {
419 		if_statinc(ifp, if_oerrors);
420 		return ENOBUFS;
421 	}
422 	ip = mtod(m, struct ip *);
423 
424 	memset(ip, 0, sizeof(*ip));
425 
426 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
427 	    &ip->ip_src, sizeof(ip->ip_src));
428 	memcpy(&ip->ip_dst, in4, sizeof(ip->ip_dst));
429 	ip->ip_p = IPPROTO_IPV6;
430 	ip->ip_ttl = ip_stf_ttl;
431 	ip->ip_len = htons(m->m_pkthdr.len);
432 	if (ifp->if_flags & IFF_LINK1)
433 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
434 	else
435 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
436 
437 	sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
438 	if ((rt = rtcache_lookup(&sc->sc_ro, &u.dst)) == NULL) {
439 		m_freem(m);
440 		if_statinc(ifp, if_oerrors);
441 		return ENETUNREACH;
442 	}
443 
444 	/* If the route constitutes infinite encapsulation, punt. */
445 	if (rt->rt_ifp == ifp) {
446 		rtcache_unref(rt, &sc->sc_ro);
447 		rtcache_free(&sc->sc_ro);
448 		m_freem(m);
449 		if_statinc(ifp, if_oerrors);
450 		return ENETUNREACH;
451 	}
452 	rtcache_unref(rt, &sc->sc_ro);
453 
454 	if_statadd2(ifp, if_opackets, 1,
455 	    if_obytes, m->m_pkthdr.len - sizeof(struct ip));
456 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
457 }
458 
459 static int
isrfc1918addr(const struct in_addr * in)460 isrfc1918addr(const struct in_addr *in)
461 {
462 	/*
463 	 * returns 1 if private address range:
464 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
465 	 */
466 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
467 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
468 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
469 		return 1;
470 
471 	return 0;
472 }
473 
474 static int
stf_checkaddr4(struct stf_softc * sc,const struct in_addr * in,struct ifnet * inifp)475 stf_checkaddr4(struct stf_softc *sc, const struct in_addr *in,
476     struct ifnet *inifp /*incoming interface*/)
477 {
478 	struct in_ifaddr *ia4;
479 
480 	/*
481 	 * reject packets with the following address:
482 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
483 	 */
484 	if (IN_MULTICAST(in->s_addr))
485 		return -1;
486 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
487 	case 0: case 127: case 255:
488 		return -1;
489 	}
490 
491 	/*
492 	 * reject packets with private address range.
493 	 * (requirement from RFC3056 section 2 1st paragraph)
494 	 */
495 	if (isrfc1918addr(in))
496 		return -1;
497 
498 	/*
499 	 * reject packet with IPv4 link-local (169.254.0.0/16),
500 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
501 	 */
502 	if (((ntohl(in->s_addr) & 0xff000000) >> 24) == 169 &&
503 	    ((ntohl(in->s_addr) & 0x00ff0000) >> 16) == 254)
504 		return -1;
505 
506 	/*
507 	 * reject packets with broadcast
508 	 */
509 	IN_ADDRLIST_READER_FOREACH(ia4) {
510 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
511 			continue;
512 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
513 			return -1;
514 	}
515 
516 	/*
517 	 * perform ingress filter
518 	 */
519 	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
520 		struct sockaddr_in sin;
521 		struct rtentry *rt;
522 
523 		memset(&sin, 0, sizeof(sin));
524 		sin.sin_family = AF_INET;
525 		sin.sin_len = sizeof(struct sockaddr_in);
526 		sin.sin_addr = *in;
527 		rt = rtalloc1((struct sockaddr *)&sin, 0);
528 		if (!rt || rt->rt_ifp != inifp) {
529 #if 0
530 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
531 			    "due to ingress filter\n", if_name(&sc->sc_if),
532 			    (uint32_t)ntohl(sin.sin_addr.s_addr));
533 #endif
534 			if (rt)
535 				rt_unref(rt);
536 			return -1;
537 		}
538 		rt_unref(rt);
539 	}
540 
541 	return 0;
542 }
543 
544 static int
stf_checkaddr6(struct stf_softc * sc,const struct in6_addr * in6,struct ifnet * inifp)545 stf_checkaddr6(struct stf_softc *sc, const struct in6_addr *in6,
546     struct ifnet *inifp /*incoming interface*/)
547 {
548 
549 	/*
550 	 * check 6to4 addresses
551 	 */
552 	if (IN6_IS_ADDR_6TO4(in6))
553 		return stf_checkaddr4(sc, GET_V4(in6), inifp);
554 
555 	/*
556 	 * reject anything that look suspicious.  the test is implemented
557 	 * in ip6_input too, but we check here as well to
558 	 * (1) reject bad packets earlier, and
559 	 * (2) to be safe against future ip6_input change.
560 	 */
561 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
562 		return -1;
563 
564 	/*
565 	 * reject link-local and site-local unicast
566 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
567 	 */
568 	if (IN6_IS_ADDR_LINKLOCAL(in6) || IN6_IS_ADDR_SITELOCAL(in6))
569 		return -1;
570 
571 	/*
572 	 * reject node-local and link-local multicast
573 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
574 	 */
575 	if (IN6_IS_ADDR_MC_NODELOCAL(in6) || IN6_IS_ADDR_MC_LINKLOCAL(in6))
576 		return -1;
577 
578 	return 0;
579 }
580 
581 void
in_stf_input(struct mbuf * m,int off,int proto,void * eparg)582 in_stf_input(struct mbuf *m, int off, int proto, void *eparg)
583 {
584 	int s;
585 	struct stf_softc *sc = eparg;
586 	struct ip *ip;
587 	struct ip6_hdr *ip6;
588 	uint8_t otos, itos;
589 	struct ifnet *ifp;
590 	size_t pktlen;
591 
592 	KASSERT(sc != NULL);
593 
594 	if (proto != IPPROTO_IPV6) {
595 		m_freem(m);
596 		return;
597 	}
598 
599 	ip = mtod(m, struct ip *);
600 
601 	if ((sc->sc_if.if_flags & IFF_UP) == 0) {
602 		m_freem(m);
603 		return;
604 	}
605 
606 	ifp = &sc->sc_if;
607 
608 	/*
609 	 * perform sanity check against outer src/dst.
610 	 * for source, perform ingress filter as well.
611 	 */
612 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
613 	    stf_checkaddr4(sc, &ip->ip_src, m_get_rcvif_NOMPSAFE(m)) < 0) {
614 		m_freem(m);
615 		return;
616 	}
617 
618 	otos = ip->ip_tos;
619 	m_adj(m, off);
620 
621 	if (m->m_len < sizeof(*ip6)) {
622 		m = m_pullup(m, sizeof(*ip6));
623 		if (!m)
624 			return;
625 	}
626 	ip6 = mtod(m, struct ip6_hdr *);
627 
628 	/*
629 	 * perform sanity check against inner src/dst.
630 	 * for source, perform ingress filter as well.
631 	 */
632 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
633 	    stf_checkaddr6(sc, &ip6->ip6_src, m_get_rcvif_NOMPSAFE(m)) < 0) {
634 		m_freem(m);
635 		return;
636 	}
637 
638 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
639 	if ((ifp->if_flags & IFF_LINK1) != 0)
640 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
641 	else
642 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
643 	ip6->ip6_flow &= ~htonl(0xff << 20);
644 	ip6->ip6_flow |= htonl((uint32_t)itos << 20);
645 
646 	pktlen = m->m_pkthdr.len;
647 	m_set_rcvif(m, ifp);
648 
649 	bpf_mtap_af(ifp, AF_INET6, m, BPF_D_IN);
650 
651 	/*
652 	 * Put the packet to the network layer input queue according to the
653 	 * specified address family.
654 	 * See net/if_gif.c for possible issues with packet processing
655 	 * reorder due to extra queueing.
656 	 */
657 
658 	s = splnet();
659 	if (__predict_true(pktq_enqueue(ip6_pktq, m, 0))) {
660 		if_statadd2(ifp, if_ipackets, 1, if_ibytes, pktlen);
661 	} else {
662 		m_freem(m);
663 	}
664 	splx(s);
665 
666 	return;
667 }
668 
669 /* ARGSUSED */
670 static void
stf_rtrequest(int cmd,struct rtentry * rt,const struct rt_addrinfo * info)671 stf_rtrequest(int cmd, struct rtentry *rt,
672     const struct rt_addrinfo *info)
673 {
674 	if (rt != NULL) {
675 		struct stf_softc *sc;
676 
677 		sc = LIST_FIRST(&stf_softc_list);
678 		rt->rt_rmx.rmx_mtu = (sc != NULL) ? sc->sc_if.if_mtu : STF_MTU;
679 	}
680 }
681 
682 static int
stf_ioctl(struct ifnet * ifp,u_long cmd,void * data)683 stf_ioctl(struct ifnet *ifp, u_long cmd, void *data)
684 {
685 	struct ifaddr		*ifa;
686 	struct ifreq		*ifr = data;
687 	struct sockaddr_in6	*sin6;
688 	int			error;
689 
690 	error = 0;
691 	switch (cmd) {
692 	case SIOCINITIFADDR:
693 		ifa = (struct ifaddr *)data;
694 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
695 			error = EAFNOSUPPORT;
696 			break;
697 		}
698 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
699 		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr) &&
700 		    !isrfc1918addr(GET_V4(&sin6->sin6_addr))) {
701 			ifa->ifa_rtrequest = stf_rtrequest;
702 			ifp->if_flags |= IFF_UP;
703 		} else
704 			error = EINVAL;
705 		break;
706 
707 	case SIOCADDMULTI:
708 	case SIOCDELMULTI:
709 		if (ifr != NULL &&
710 		    ifreq_getaddr(cmd, ifr)->sa_family == AF_INET6)
711 			;
712 		else
713 			error = EAFNOSUPPORT;
714 		break;
715 
716 	case SIOCSIFMTU:
717 		if (ifr->ifr_mtu < STF_MTU_MIN || ifr->ifr_mtu > STF_MTU_MAX)
718 			return EINVAL;
719 		else if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
720 			error = 0;
721 		break;
722 
723 	default:
724 		error = ifioctl_common(ifp, cmd, data);
725 		break;
726 	}
727 
728 	return error;
729 }
730 
731 /*
732  * Module infrastructure
733  */
734 #include "if_module.h"
735 
736 IF_MODULE(MODULE_CLASS_DRIVER, stf, NULL)
737