2 /* $KAME: if_stf.c,v 1.62 2001/06/07 22:32:16 itojun Exp $ */
5 * Copyright (C) 2000 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
34 * 6to4 interface, based on RFC3056.
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.
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).
44 * Here are interesting symptoms due to the lack of link-local address:
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.
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?).
61 * - Redirects cannot be used due to the lack of link-local address.
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
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.
78 #include "opt_inet6.h"
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/socket.h>
83 #include <sys/sockio.h>
85 #include <sys/errno.h>
86 #include <sys/kernel.h>
87 #include <sys/protosw.h>
88 #include <sys/queue.h>
89 #include <machine/bus.h> /* XXX: Shouldn't really be required! */
91 #include <machine/cpu.h>
93 #include <sys/malloc.h>
96 #include <net/route.h>
97 #include <net/netisr.h>
98 #include <net/if_types.h>
99 #include <net/if_stf.h>
101 #include <netinet/in.h>
102 #include <netinet/in_systm.h>
103 #include <netinet/ip.h>
104 #include <netinet/ip_var.h>
105 #include <netinet/in_var.h>
107 #include <netinet/ip6.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet6/in6_var.h>
110 #include <netinet/ip_ecn.h>
112 #include <netinet/ip_encap.h>
114 #include <machine/stdarg.h>
116 #include <net/net_osdep.h>
120 #define STFNAME "stf"
121 #define STF_MAXUNIT 0 /* only one is currently allowed */
123 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002)
124 #define GET_V4(x) ((struct in_addr *)(&(x)->s6_addr16[1]))
127 struct ifnet sc_if; /* common area */
129 struct route __sc_ro4;
130 struct route_in6 __sc_ro6; /* just for safety */
132 #define sc_ro __sc_ro46.__sc_ro4
133 const struct encaptab *encap_cookie;
134 struct resource *r_unit; /* resource allocated for this unit */
135 LIST_ENTRY(stf_softc) sc_list; /* all stf's are linked */
138 LIST_HEAD(, stf_softc) stf_softc_list;
140 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
141 static struct rman stfunits[1];
142 static int ip_stf_ttl = 40;
144 extern struct domain inetdomain;
145 struct protosw in_stf_protosw =
146 { SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR,
147 in_stf_input, rip_output, 0, rip_ctloutput,
153 static int stfmodevent __P((module_t, int, void *));
154 static int stf_encapcheck __P((const struct mbuf *, int, int, void *));
155 static struct in6_ifaddr *stf_getsrcifa6 __P((struct ifnet *));
156 static int stf_output __P((struct ifnet *, struct mbuf *, struct sockaddr *,
158 static int stf_checkaddr4 __P((struct stf_softc *, struct in_addr *,
160 static int stf_checkaddr6 __P((struct stf_softc *, struct in6_addr *,
162 static void stf_rtrequest __P((int, struct rtentry *, struct rt_addrinfo *));
163 static int stf_ioctl __P((struct ifnet *, u_long, caddr_t));
165 int stf_clone_create __P((struct if_clone *, int *));
166 void stf_clone_destroy __P((struct ifnet *));
168 struct if_clone stf_cloner =
169 IF_CLONE_INITIALIZER(STFNAME, stf_clone_create, stf_clone_destroy);
172 stf_clone_create(ifc, unit)
173 struct if_clone *ifc;
177 struct stf_softc *sc;
179 if (*unit > STF_MAXUNIT)
183 r = rman_reserve_resource(stfunits, 0, STF_MAXUNIT, 1,
184 RF_ALLOCATED | RF_ACTIVE, NULL);
187 *unit = rman_get_start(r);
189 r = rman_reserve_resource(stfunits, *unit, *unit, 1,
190 RF_ALLOCATED | RF_ACTIVE, NULL);
195 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
196 sc->sc_if.if_name = STFNAME;
197 sc->sc_if.if_unit = *unit;
200 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
201 stf_encapcheck, &in_stf_protosw, sc);
202 if (sc->encap_cookie == NULL) {
203 printf("%s: attach failed\n", if_name(&sc->sc_if));
208 sc->sc_if.if_mtu = IPV6_MMTU;
209 sc->sc_if.if_ioctl = stf_ioctl;
210 sc->sc_if.if_output = stf_output;
211 sc->sc_if.if_type = IFT_STF;
212 sc->sc_if.if_snd.ifq_maxlen = IFQ_MAXLEN;
213 if_attach(&sc->sc_if);
214 bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
215 LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
220 stf_clone_destroy(ifp)
224 struct stf_softc *sc = (void *) ifp;
226 LIST_REMOVE(sc, sc_list);
227 err = encap_detach(sc->encap_cookie);
228 KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
232 err = rman_release_resource(sc->r_unit);
233 KASSERT(err == 0, ("Unexpected error freeing resource"));
239 stfmodevent(mod, type, data)
248 stfunits->rm_type = RMAN_ARRAY;
249 stfunits->rm_descr = "configurable if_stf units";
250 err = rman_init(stfunits);
253 err = rman_manage_region(stfunits, 0, STF_MAXUNIT);
255 printf("%s: stfunits: rman_manage_region: Failed %d\n",
260 LIST_INIT(&stf_softc_list);
261 if_clone_attach(&stf_cloner);
265 if_clone_detach(&stf_cloner);
267 while (!LIST_EMPTY(&stf_softc_list))
268 stf_clone_destroy(&LIST_FIRST(&stf_softc_list)->sc_if);
270 err = rman_fini(stfunits);
271 KASSERT(err == 0, ("Unexpected error freeing resource"));
279 static moduledata_t stf_mod = {
285 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
288 stf_encapcheck(m, off, proto, arg)
289 const struct mbuf *m;
295 struct in6_ifaddr *ia6;
296 struct stf_softc *sc;
299 sc = (struct stf_softc *)arg;
303 if ((sc->sc_if.if_flags & IFF_UP) == 0)
306 /* IFF_LINK0 means "no decapsulation" */
307 if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
310 if (proto != IPPROTO_IPV6)
313 /* LINTED const cast */
314 m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip);
319 ia6 = stf_getsrcifa6(&sc->sc_if);
324 * check if IPv4 dst matches the IPv4 address derived from the
325 * local 6to4 address.
326 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
328 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
329 sizeof(ip.ip_dst)) != 0)
333 * check if IPv4 src matches the IPv4 address derived from the
334 * local 6to4 address masked by prefixmask.
335 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
336 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
338 bzero(&a, sizeof(a));
339 a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
340 a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
342 b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
343 if (a.s_addr != b.s_addr)
346 /* stf interface makes single side match only */
350 static struct in6_ifaddr *
355 struct in_ifaddr *ia4;
356 struct sockaddr_in6 *sin6;
359 for (ia = TAILQ_FIRST(&ifp->if_addrlist);
361 ia = TAILQ_NEXT(ia, ifa_list))
363 if (ia->ifa_addr == NULL)
365 if (ia->ifa_addr->sa_family != AF_INET6)
367 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
368 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
371 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
372 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
373 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
378 return (struct in6_ifaddr *)ia;
385 stf_output(ifp, m, dst, rt)
388 struct sockaddr *dst;
391 struct stf_softc *sc;
392 struct sockaddr_in6 *dst6;
394 struct sockaddr_in *dst4;
398 struct in6_ifaddr *ia6;
400 sc = (struct stf_softc*)ifp;
401 dst6 = (struct sockaddr_in6 *)dst;
404 if ((ifp->if_flags & IFF_UP) == 0) {
410 * If we don't have an ip4 address that match my inner ip6 address,
411 * we shouldn't generate output. Without this check, we'll end up
412 * using wrong IPv4 source.
414 ia6 = stf_getsrcifa6(ifp);
420 if (m->m_len < sizeof(*ip6)) {
421 m = m_pullup(m, sizeof(*ip6));
425 ip6 = mtod(m, struct ip6_hdr *);
426 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
429 * Pickup the right outer dst addr from the list of candidates.
430 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
432 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
433 in4 = GET_V4(&ip6->ip6_dst);
434 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
435 in4 = GET_V4(&dst6->sin6_addr);
444 * We need to prepend the address family as
445 * a four byte field. Cons up a dummy header
446 * to pacify bpf. This is safe because bpf
447 * will only read from the mbuf (i.e., it won't
448 * try to free it or keep a pointer a to it).
451 u_int32_t af = AF_INET6;
455 m0.m_data = (char *)⁡
460 bpf_mtap(ifp->if_bpf, &m0);
463 #endif /*NBPFILTER > 0*/
465 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
466 if (m && m->m_len < sizeof(struct ip))
467 m = m_pullup(m, sizeof(struct ip));
470 ip = mtod(m, struct ip *);
472 bzero(ip, sizeof(*ip));
474 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
475 &ip->ip_src, sizeof(ip->ip_src));
476 bcopy(in4, &ip->ip_dst, sizeof(ip->ip_dst));
477 ip->ip_p = IPPROTO_IPV6;
478 ip->ip_ttl = ip_stf_ttl;
479 ip->ip_len = m->m_pkthdr.len; /*host order*/
480 if (ifp->if_flags & IFF_LINK1)
481 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
483 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
485 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
486 if (dst4->sin_family != AF_INET ||
487 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
488 /* cache route doesn't match */
489 dst4->sin_family = AF_INET;
490 dst4->sin_len = sizeof(struct sockaddr_in);
491 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
492 if (sc->sc_ro.ro_rt) {
493 RTFREE(sc->sc_ro.ro_rt);
494 sc->sc_ro.ro_rt = NULL;
498 if (sc->sc_ro.ro_rt == NULL) {
500 if (sc->sc_ro.ro_rt == NULL) {
506 return ip_output(m, NULL, &sc->sc_ro, 0, NULL);
510 stf_checkaddr4(sc, in, inifp)
511 struct stf_softc *sc;
513 struct ifnet *inifp; /* incoming interface */
515 struct in_ifaddr *ia4;
518 * reject packets with the following address:
519 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
521 if (IN_MULTICAST(ntohl(in->s_addr)))
523 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
524 case 0: case 127: case 255:
529 * reject packets with broadcast
531 for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
533 ia4 = TAILQ_NEXT(ia4, ia_link))
535 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
537 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
542 * perform ingress filter
544 if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
545 struct sockaddr_in sin;
548 bzero(&sin, sizeof(sin));
549 sin.sin_family = AF_INET;
550 sin.sin_len = sizeof(struct sockaddr_in);
552 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
553 if (!rt || rt->rt_ifp != inifp) {
555 log(LOG_WARNING, "%s: packet from 0x%x dropped "
556 "due to ingress filter\n", if_name(&sc->sc_if),
557 (u_int32_t)ntohl(sin.sin_addr.s_addr));
570 stf_checkaddr6(sc, in6, inifp)
571 struct stf_softc *sc;
572 struct in6_addr *in6;
573 struct ifnet *inifp; /* incoming interface */
576 * check 6to4 addresses
578 if (IN6_IS_ADDR_6TO4(in6))
579 return stf_checkaddr4(sc, GET_V4(in6), inifp);
582 * reject anything that look suspicious. the test is implemented
583 * in ip6_input too, but we check here as well to
584 * (1) reject bad packets earlier, and
585 * (2) to be safe against future ip6_input change.
587 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
599 struct stf_softc *sc;
604 struct ifqueue *ifq = NULL;
607 proto = mtod(m, struct ip *)->ip_p;
609 if (proto != IPPROTO_IPV6) {
614 ip = mtod(m, struct ip *);
616 sc = (struct stf_softc *)encap_getarg(m);
618 if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
626 * perform sanity check against outer src/dst.
627 * for source, perform ingress filter as well.
629 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
630 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
638 if (m->m_len < sizeof(*ip6)) {
639 m = m_pullup(m, sizeof(*ip6));
643 ip6 = mtod(m, struct ip6_hdr *);
646 * perform sanity check against inner src/dst.
647 * for source, perform ingress filter as well.
649 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
650 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
655 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
656 if ((ifp->if_flags & IFF_LINK1) != 0)
657 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
659 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
660 ip6->ip6_flow &= ~htonl(0xff << 20);
661 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
663 m->m_pkthdr.rcvif = ifp;
667 * We need to prepend the address family as
668 * a four byte field. Cons up a dummy header
669 * to pacify bpf. This is safe because bpf
670 * will only read from the mbuf (i.e., it won't
671 * try to free it or keep a pointer a to it).
674 u_int32_t af = AF_INET6;
678 m0.m_data = (char *)⁡
683 bpf_mtap(ifp->if_bpf, &m0);
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.
696 len = m->m_pkthdr.len;
697 if (! IF_HANDOFF(ifq, m, NULL))
701 ifp->if_ibytes += len;
706 stf_rtrequest(cmd, rt, info)
709 struct rt_addrinfo *info;
713 rt->rt_rmx.rmx_mtu = IPV6_MMTU;
717 stf_ioctl(ifp, cmd, data)
724 struct sockaddr_in6 *sin6;
730 ifa = (struct ifaddr *)data;
731 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
732 error = EAFNOSUPPORT;
735 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
736 if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
737 ifa->ifa_rtrequest = stf_rtrequest;
738 ifp->if_flags |= IFF_UP;
745 ifr = (struct ifreq *)data;
746 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
749 error = EAFNOSUPPORT;