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_fib.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_var.h>
108 #include <netinet/in_var.h>
110 #include <netinet/ip6.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet6/in6_var.h>
113 #include <netinet/ip_ecn.h>
115 #include <netinet/ip_encap.h>
117 #include <machine/stdarg.h>
121 #include <security/mac/mac_framework.h>
123 SYSCTL_DECL(_net_link);
124 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
126 static int stf_permit_rfc1918 = 0;
127 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
128 &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
132 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002)
135 * XXX: Return a pointer with 16-bit aligned. Don't cast it to
136 * struct in_addr *; use bcopy() instead.
138 #define GET_V4(x) (&(x)->s6_addr16[1])
141 struct ifnet *sc_ifp;
142 struct mtx sc_ro_mtx;
144 const struct encaptab *encap_cookie;
146 #define STF2IFP(sc) ((sc)->sc_ifp)
148 static const char stfname[] = "stf";
151 * Note that mutable fields in the softc are not currently locked.
152 * We do lock sc_ro in stf_output though.
154 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
155 static const int ip_stf_ttl = 40;
157 extern struct domain inetdomain;
158 static int in_stf_input(struct mbuf **, int *, int);
159 static struct protosw in_stf_protosw = {
161 .pr_domain = &inetdomain,
162 .pr_protocol = IPPROTO_IPV6,
163 .pr_flags = PR_ATOMIC|PR_ADDR,
164 .pr_input = in_stf_input,
165 .pr_output = rip_output,
166 .pr_ctloutput = rip_ctloutput,
167 .pr_usrreqs = &rip_usrreqs
170 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
172 static int stfmodevent(module_t, int, void *);
173 static int stf_encapcheck(const struct mbuf *, int, int, void *);
174 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
175 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
177 static int isrfc1918addr(struct in_addr *);
178 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
180 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
182 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
184 static int stf_clone_match(struct if_clone *, const char *);
185 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
186 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
187 static struct if_clone *stf_cloner;
190 stf_clone_match(struct if_clone *ifc, const char *name)
194 for(i = 0; stfnames[i] != NULL; i++) {
195 if (strcmp(stfnames[i], name) == 0)
203 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
206 int err, unit, wildcard;
207 struct stf_softc *sc;
210 err = ifc_name2unit(name, &unit);
213 wildcard = (unit < 0);
216 * We can only have one unit, but since unit allocation is
217 * already locked, we use it to keep from allocating extra
221 err = ifc_alloc_unit(ifc, &unit);
225 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
226 ifp = STF2IFP(sc) = if_alloc(IFT_STF);
229 ifc_free_unit(ifc, unit);
233 sc->sc_fibnum = curthread->td_proc->p_fibnum;
236 * Set the name manually rather then using if_initname because
237 * we don't conform to the default naming convention for interfaces.
238 * In the wildcard case, we need to update the name.
241 for (dp = name; *dp != '\0'; dp++);
242 if (snprintf(dp, len - (dp-name), "%d", unit) >
243 len - (dp-name) - 1) {
245 * This can only be a programmer error and
246 * there's no straightforward way to recover if
249 panic("if_clone_create(): interface name too long");
252 strlcpy(ifp->if_xname, name, IFNAMSIZ);
253 ifp->if_dname = stfname;
254 ifp->if_dunit = IF_DUNIT_NONE;
256 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
257 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
258 stf_encapcheck, &in_stf_protosw, sc);
259 if (sc->encap_cookie == NULL) {
260 if_printf(ifp, "attach failed\n");
262 ifc_free_unit(ifc, unit);
266 ifp->if_mtu = IPV6_MMTU;
267 ifp->if_ioctl = stf_ioctl;
268 ifp->if_output = stf_output;
269 ifp->if_snd.ifq_maxlen = ifqmaxlen;
271 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
276 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
278 struct stf_softc *sc = ifp->if_softc;
281 err = encap_detach(sc->encap_cookie);
282 KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
283 mtx_destroy(&(sc)->sc_ro_mtx);
289 ifc_free_unit(ifc, STFUNIT);
295 stfmodevent(module_t mod, int type, void *data)
300 stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
301 stf_clone_create, stf_clone_destroy);
304 if_clone_detach(stf_cloner);
313 static moduledata_t stf_mod = {
319 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
322 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
325 struct stf_softc *sc;
326 struct in_addr a, b, mask;
327 struct in6_addr addr6, mask6;
329 sc = (struct stf_softc *)arg;
333 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
336 /* IFF_LINK0 means "no decapsulation" */
337 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
340 if (proto != IPPROTO_IPV6)
343 m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
348 if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
352 * check if IPv4 dst matches the IPv4 address derived from the
353 * local 6to4 address.
354 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
356 if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
360 * check if IPv4 src matches the IPv4 address derived from the
361 * local 6to4 address masked by prefixmask.
362 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
363 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
365 bzero(&a, sizeof(a));
366 bcopy(GET_V4(&addr6), &a, sizeof(a));
367 bcopy(GET_V4(&mask6), &mask, sizeof(mask));
368 a.s_addr &= mask.s_addr;
370 b.s_addr &= mask.s_addr;
371 if (a.s_addr != b.s_addr)
374 /* stf interface makes single side match only */
379 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
382 struct in_ifaddr *ia4;
383 struct in6_ifaddr *ia6;
384 struct sockaddr_in6 *sin6;
388 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
389 if (ia->ifa_addr->sa_family != AF_INET6)
391 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
392 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
395 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
396 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
397 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
402 ia6 = (struct in6_ifaddr *)ia;
404 *addr = sin6->sin6_addr;
405 *mask = ia6->ia_prefixmask.sin6_addr;
406 if_addr_runlock(ifp);
409 if_addr_runlock(ifp);
415 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
418 struct stf_softc *sc;
419 const struct sockaddr_in6 *dst6;
425 struct in6_addr addr6, mask6;
429 error = mac_ifnet_check_transmit(ifp, m);
437 dst6 = (const struct sockaddr_in6 *)dst;
440 if ((ifp->if_flags & IFF_UP) == 0) {
442 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
447 * If we don't have an ip4 address that match my inner ip6 address,
448 * we shouldn't generate output. Without this check, we'll end up
449 * using wrong IPv4 source.
451 if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
453 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
457 if (m->m_len < sizeof(*ip6)) {
458 m = m_pullup(m, sizeof(*ip6));
460 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
464 ip6 = mtod(m, struct ip6_hdr *);
465 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
468 * Pickup the right outer dst addr from the list of candidates.
469 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
472 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
473 ptr = GET_V4(&ip6->ip6_dst);
474 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
475 ptr = GET_V4(&dst6->sin6_addr);
478 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
481 bcopy(ptr, &in4, sizeof(in4));
483 if (bpf_peers_present(ifp->if_bpf)) {
485 * We need to prepend the address family as
486 * a four byte field. Cons up a dummy header
487 * to pacify bpf. This is safe because bpf
488 * will only read from the mbuf (i.e., it won't
489 * try to free it or keep a pointer a to it).
492 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
495 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
497 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
500 ip = mtod(m, struct ip *);
502 bzero(ip, sizeof(*ip));
504 bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
505 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
506 ip->ip_p = IPPROTO_IPV6;
507 ip->ip_ttl = ip_stf_ttl;
508 ip->ip_len = htons(m->m_pkthdr.len);
509 if (ifp->if_flags & IFF_LINK1)
510 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
512 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
514 M_SETFIB(m, sc->sc_fibnum);
515 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
516 error = ip_output(m, NULL, NULL, 0, NULL, NULL);
522 isrfc1918addr(struct in_addr *in)
525 * returns 1 if private address range:
526 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
528 if (stf_permit_rfc1918 == 0 && (
529 (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
530 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
531 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
538 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
540 struct rm_priotracker in_ifa_tracker;
541 struct in_ifaddr *ia4;
544 * reject packets with the following address:
545 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
547 if (IN_MULTICAST(ntohl(in->s_addr)))
549 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
550 case 0: case 127: case 255:
555 * reject packets with private address range.
556 * (requirement from RFC3056 section 2 1st paragraph)
558 if (isrfc1918addr(in))
562 * reject packets with broadcast
564 IN_IFADDR_RLOCK(&in_ifa_tracker);
565 TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
566 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
568 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
569 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
573 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
576 * perform ingress filter
578 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
579 struct nhop4_basic nh4;
581 if (fib4_lookup_nh_basic(sc->sc_fibnum, *in, 0, 0, &nh4) != 0)
584 if (nh4.nh_ifp != inifp)
592 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
595 * check 6to4 addresses
597 if (IN6_IS_ADDR_6TO4(in6)) {
599 bcopy(GET_V4(in6), &in4, sizeof(in4));
600 return stf_checkaddr4(sc, &in4, inifp);
604 * reject anything that look suspicious. the test is implemented
605 * in ip6_input too, but we check here as well to
606 * (1) reject bad packets earlier, and
607 * (2) to be safe against future ip6_input change.
609 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
616 in_stf_input(struct mbuf **mp, int *offp, int proto)
618 struct stf_softc *sc;
629 if (proto != IPPROTO_IPV6) {
631 return (IPPROTO_DONE);
634 ip = mtod(m, struct ip *);
636 sc = (struct stf_softc *)encap_getarg(m);
638 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
640 return (IPPROTO_DONE);
646 mac_ifnet_create_mbuf(ifp, m);
650 * perform sanity check against outer src/dst.
651 * for source, perform ingress filter as well.
653 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
654 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
656 return (IPPROTO_DONE);
662 if (m->m_len < sizeof(*ip6)) {
663 m = m_pullup(m, sizeof(*ip6));
665 return (IPPROTO_DONE);
667 ip6 = mtod(m, struct ip6_hdr *);
670 * perform sanity check against inner src/dst.
671 * for source, perform ingress filter as well.
673 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
674 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
676 return (IPPROTO_DONE);
679 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
680 if ((ifp->if_flags & IFF_LINK1) != 0)
681 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
683 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
684 ip6->ip6_flow &= ~htonl(0xff << 20);
685 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
687 m->m_pkthdr.rcvif = ifp;
689 if (bpf_peers_present(ifp->if_bpf)) {
691 * We need to prepend the address family as
692 * a four byte field. Cons up a dummy header
693 * to pacify bpf. This is safe because bpf
694 * will only read from the mbuf (i.e., it won't
695 * try to free it or keep a pointer a to it).
697 u_int32_t af = AF_INET6;
698 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
702 * Put the packet to the network layer input queue according to the
703 * specified address family.
704 * See net/if_gif.c for possible issues with packet processing
705 * reorder due to extra queueing.
707 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
708 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
709 M_SETFIB(m, ifp->if_fib);
710 netisr_dispatch(NETISR_IPV6, m);
711 return (IPPROTO_DONE);
715 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
719 struct sockaddr_in6 *sin6;
726 ifa = (struct ifaddr *)data;
727 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
728 error = EAFNOSUPPORT;
731 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
732 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
736 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
737 if (isrfc1918addr(&addr)) {
742 ifp->if_flags |= IFF_UP;
747 ifr = (struct ifreq *)data;
748 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
751 error = EAFNOSUPPORT;
758 ifr = (struct ifreq *)data;
760 /* RFC 4213 3.2 ideal world MTU */
761 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)