2 /* $KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi 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.
77 #include <sys/param.h>
78 #include <sys/systm.h>
79 #include <sys/socket.h>
80 #include <sys/sockio.h>
82 #include <sys/errno.h>
83 #include <sys/kernel.h>
85 #include <sys/module.h>
86 #include <sys/protosw.h>
88 #include <sys/queue.h>
89 #include <sys/rmlock.h>
90 #include <sys/sysctl.h>
91 #include <machine/cpu.h>
93 #include <sys/malloc.h>
96 #include <net/if_var.h>
97 #include <net/if_clone.h>
98 #include <net/route.h>
99 #include <net/netisr.h>
100 #include <net/if_types.h>
101 #include <net/vnet.h>
103 #include <netinet/in.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/ip.h>
106 #include <netinet/ip_var.h>
107 #include <netinet/in_var.h>
109 #include <netinet/ip6.h>
110 #include <netinet6/ip6_var.h>
111 #include <netinet6/in6_var.h>
112 #include <netinet/ip_ecn.h>
114 #include <netinet/ip_encap.h>
116 #include <machine/stdarg.h>
120 #include <security/mac/mac_framework.h>
122 SYSCTL_DECL(_net_link);
123 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
125 static int stf_permit_rfc1918 = 0;
126 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
127 &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
131 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002)
134 * XXX: Return a pointer with 16-bit aligned. Don't cast it to
135 * struct in_addr *; use bcopy() instead.
137 #define GET_V4(x) (&(x)->s6_addr16[1])
140 struct ifnet *sc_ifp;
141 struct mtx sc_ro_mtx;
143 const struct encaptab *encap_cookie;
145 #define STF2IFP(sc) ((sc)->sc_ifp)
147 static const char stfname[] = "stf";
150 * Note that mutable fields in the softc are not currently locked.
151 * We do lock sc_ro in stf_output though.
153 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
154 static const int ip_stf_ttl = 40;
156 extern struct domain inetdomain;
157 static int in_stf_input(struct mbuf **, int *, int);
158 static struct protosw in_stf_protosw = {
160 .pr_domain = &inetdomain,
161 .pr_protocol = IPPROTO_IPV6,
162 .pr_flags = PR_ATOMIC|PR_ADDR,
163 .pr_input = in_stf_input,
164 .pr_output = rip_output,
165 .pr_ctloutput = rip_ctloutput,
166 .pr_usrreqs = &rip_usrreqs
169 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
171 static int stfmodevent(module_t, int, void *);
172 static int stf_encapcheck(const struct mbuf *, int, int, void *);
173 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
174 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
176 static int isrfc1918addr(struct in_addr *);
177 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
179 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
181 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
183 static int stf_clone_match(struct if_clone *, const char *);
184 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
185 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
186 static struct if_clone *stf_cloner;
189 stf_clone_match(struct if_clone *ifc, const char *name)
193 for(i = 0; stfnames[i] != NULL; i++) {
194 if (strcmp(stfnames[i], name) == 0)
202 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
205 struct stf_softc *sc;
209 * We can only have one unit, but since unit allocation is
210 * already locked, we use it to keep from allocating extra
214 err = ifc_alloc_unit(ifc, &unit);
218 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
219 ifp = STF2IFP(sc) = if_alloc(IFT_STF);
222 ifc_free_unit(ifc, unit);
226 sc->sc_fibnum = curthread->td_proc->p_fibnum;
229 * Set the name manually rather then using if_initname because
230 * we don't conform to the default naming convention for interfaces.
232 strlcpy(ifp->if_xname, name, IFNAMSIZ);
233 ifp->if_dname = stfname;
234 ifp->if_dunit = IF_DUNIT_NONE;
236 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
237 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
238 stf_encapcheck, &in_stf_protosw, sc);
239 if (sc->encap_cookie == NULL) {
240 if_printf(ifp, "attach failed\n");
242 ifc_free_unit(ifc, unit);
246 ifp->if_mtu = IPV6_MMTU;
247 ifp->if_ioctl = stf_ioctl;
248 ifp->if_output = stf_output;
249 ifp->if_snd.ifq_maxlen = ifqmaxlen;
251 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
256 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
258 struct stf_softc *sc = ifp->if_softc;
261 err = encap_detach(sc->encap_cookie);
262 KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
263 mtx_destroy(&(sc)->sc_ro_mtx);
269 ifc_free_unit(ifc, STFUNIT);
275 stfmodevent(module_t mod, int type, void *data)
280 stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
281 stf_clone_create, stf_clone_destroy);
284 if_clone_detach(stf_cloner);
293 static moduledata_t stf_mod = {
299 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
302 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
305 struct stf_softc *sc;
306 struct in_addr a, b, mask;
307 struct in6_addr addr6, mask6;
309 sc = (struct stf_softc *)arg;
313 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
316 /* IFF_LINK0 means "no decapsulation" */
317 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
320 if (proto != IPPROTO_IPV6)
323 /* LINTED const cast */
324 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
329 if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
333 * check if IPv4 dst matches the IPv4 address derived from the
334 * local 6to4 address.
335 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
337 if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
341 * check if IPv4 src matches the IPv4 address derived from the
342 * local 6to4 address masked by prefixmask.
343 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
344 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
346 bzero(&a, sizeof(a));
347 bcopy(GET_V4(&addr6), &a, sizeof(a));
348 bcopy(GET_V4(&mask6), &mask, sizeof(mask));
349 a.s_addr &= mask.s_addr;
351 b.s_addr &= mask.s_addr;
352 if (a.s_addr != b.s_addr)
355 /* stf interface makes single side match only */
360 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
363 struct in_ifaddr *ia4;
364 struct in6_ifaddr *ia6;
365 struct sockaddr_in6 *sin6;
369 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
370 if (ia->ifa_addr->sa_family != AF_INET6)
372 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
373 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
376 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
377 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
378 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
383 ia6 = (struct in6_ifaddr *)ia;
385 *addr = sin6->sin6_addr;
386 *mask = ia6->ia_prefixmask.sin6_addr;
387 if_addr_runlock(ifp);
390 if_addr_runlock(ifp);
396 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
399 struct stf_softc *sc;
400 const struct sockaddr_in6 *dst6;
406 struct in6_addr addr6, mask6;
410 error = mac_ifnet_check_transmit(ifp, m);
418 dst6 = (const struct sockaddr_in6 *)dst;
421 if ((ifp->if_flags & IFF_UP) == 0) {
423 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
428 * If we don't have an ip4 address that match my inner ip6 address,
429 * we shouldn't generate output. Without this check, we'll end up
430 * using wrong IPv4 source.
432 if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
434 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
438 if (m->m_len < sizeof(*ip6)) {
439 m = m_pullup(m, sizeof(*ip6));
441 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
445 ip6 = mtod(m, struct ip6_hdr *);
446 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
449 * Pickup the right outer dst addr from the list of candidates.
450 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
453 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
454 ptr = GET_V4(&ip6->ip6_dst);
455 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
456 ptr = GET_V4(&dst6->sin6_addr);
459 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
462 bcopy(ptr, &in4, sizeof(in4));
464 if (bpf_peers_present(ifp->if_bpf)) {
466 * We need to prepend the address family as
467 * a four byte field. Cons up a dummy header
468 * to pacify bpf. This is safe because bpf
469 * will only read from the mbuf (i.e., it won't
470 * try to free it or keep a pointer a to it).
473 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
476 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
478 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
481 ip = mtod(m, struct ip *);
483 bzero(ip, sizeof(*ip));
485 bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
486 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
487 ip->ip_p = IPPROTO_IPV6;
488 ip->ip_ttl = ip_stf_ttl;
489 ip->ip_len = htons(m->m_pkthdr.len);
490 if (ifp->if_flags & IFF_LINK1)
491 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
493 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
495 M_SETFIB(m, sc->sc_fibnum);
496 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
497 error = ip_output(m, NULL, NULL, 0, NULL, NULL);
503 isrfc1918addr(struct in_addr *in)
506 * returns 1 if private address range:
507 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
509 if (stf_permit_rfc1918 == 0 && (
510 (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
511 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
512 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
519 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
521 struct rm_priotracker in_ifa_tracker;
522 struct in_ifaddr *ia4;
525 * reject packets with the following address:
526 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
528 if (IN_MULTICAST(ntohl(in->s_addr)))
530 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
531 case 0: case 127: case 255:
536 * reject packets with private address range.
537 * (requirement from RFC3056 section 2 1st paragraph)
539 if (isrfc1918addr(in))
543 * reject packets with broadcast
545 IN_IFADDR_RLOCK(&in_ifa_tracker);
546 TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
547 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
549 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
550 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
554 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
557 * perform ingress filter
559 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
560 struct sockaddr_in sin;
563 bzero(&sin, sizeof(sin));
564 sin.sin_family = AF_INET;
565 sin.sin_len = sizeof(struct sockaddr_in);
567 rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
569 if (!rt || rt->rt_ifp != inifp) {
571 log(LOG_WARNING, "%s: packet from 0x%x dropped "
572 "due to ingress filter\n", if_name(STF2IFP(sc)),
573 (u_int32_t)ntohl(sin.sin_addr.s_addr));
586 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
589 * check 6to4 addresses
591 if (IN6_IS_ADDR_6TO4(in6)) {
593 bcopy(GET_V4(in6), &in4, sizeof(in4));
594 return stf_checkaddr4(sc, &in4, inifp);
598 * reject anything that look suspicious. the test is implemented
599 * in ip6_input too, but we check here as well to
600 * (1) reject bad packets earlier, and
601 * (2) to be safe against future ip6_input change.
603 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
610 in_stf_input(struct mbuf **mp, int *offp, int proto)
612 struct stf_softc *sc;
623 if (proto != IPPROTO_IPV6) {
625 return (IPPROTO_DONE);
628 ip = mtod(m, struct ip *);
630 sc = (struct stf_softc *)encap_getarg(m);
632 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
634 return (IPPROTO_DONE);
640 mac_ifnet_create_mbuf(ifp, m);
644 * perform sanity check against outer src/dst.
645 * for source, perform ingress filter as well.
647 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
648 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
650 return (IPPROTO_DONE);
656 if (m->m_len < sizeof(*ip6)) {
657 m = m_pullup(m, sizeof(*ip6));
659 return (IPPROTO_DONE);
661 ip6 = mtod(m, struct ip6_hdr *);
664 * perform sanity check against inner src/dst.
665 * for source, perform ingress filter as well.
667 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
668 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
670 return (IPPROTO_DONE);
673 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
674 if ((ifp->if_flags & IFF_LINK1) != 0)
675 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
677 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
678 ip6->ip6_flow &= ~htonl(0xff << 20);
679 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
681 m->m_pkthdr.rcvif = ifp;
683 if (bpf_peers_present(ifp->if_bpf)) {
685 * We need to prepend the address family as
686 * a four byte field. Cons up a dummy header
687 * to pacify bpf. This is safe because bpf
688 * will only read from the mbuf (i.e., it won't
689 * try to free it or keep a pointer a to it).
691 u_int32_t af = AF_INET6;
692 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
696 * Put the packet to the network layer input queue according to the
697 * specified address family.
698 * See net/if_gif.c for possible issues with packet processing
699 * reorder due to extra queueing.
701 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
702 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
703 M_SETFIB(m, ifp->if_fib);
704 netisr_dispatch(NETISR_IPV6, m);
705 return (IPPROTO_DONE);
709 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
713 struct sockaddr_in6 *sin6;
720 ifa = (struct ifaddr *)data;
721 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
722 error = EAFNOSUPPORT;
725 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
726 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
730 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
731 if (isrfc1918addr(&addr)) {
736 ifp->if_flags |= IFF_UP;
741 ifr = (struct ifreq *)data;
742 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
745 error = EAFNOSUPPORT;
752 ifr = (struct ifreq *)data;
754 /* RFC 4213 3.2 ideal world MTU */
755 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)