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.
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/module.h>
88 #include <sys/protosw.h>
90 #include <sys/queue.h>
91 #include <sys/sysctl.h>
92 #include <machine/cpu.h>
94 #include <sys/malloc.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/if_stf.h>
103 #include <net/vnet.h>
105 #include <netinet/in.h>
106 #include <netinet/in_systm.h>
107 #include <netinet/ip.h>
108 #include <netinet/ip_var.h>
109 #include <netinet/in_var.h>
111 #include <netinet/ip6.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet6/in6_var.h>
114 #include <netinet/ip_ecn.h>
116 #include <netinet/ip_encap.h>
118 #include <machine/stdarg.h>
122 #include <security/mac/mac_framework.h>
124 SYSCTL_DECL(_net_link);
125 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
127 static int stf_route_cache = 1;
128 SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW,
129 &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output");
131 static int stf_permit_rfc1918 = 0;
132 TUNABLE_INT("net.link.stf.permit_rfc1918", &stf_permit_rfc1918);
133 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RW | CTLFLAG_TUN,
134 &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
138 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002)
141 * XXX: Return a pointer with 16-bit aligned. Don't cast it to
142 * struct in_addr *; use bcopy() instead.
144 #define GET_V4(x) (&(x)->s6_addr16[1])
147 struct ifnet *sc_ifp;
149 struct route __sc_ro4;
150 struct route_in6 __sc_ro6; /* just for safety */
152 #define sc_ro __sc_ro46.__sc_ro4
153 struct mtx sc_ro_mtx;
155 const struct encaptab *encap_cookie;
157 #define STF2IFP(sc) ((sc)->sc_ifp)
159 static const char stfname[] = "stf";
162 * Note that mutable fields in the softc are not currently locked.
163 * We do lock sc_ro in stf_output though.
165 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
166 static const int ip_stf_ttl = 40;
168 extern struct domain inetdomain;
169 struct protosw in_stf_protosw = {
171 .pr_domain = &inetdomain,
172 .pr_protocol = IPPROTO_IPV6,
173 .pr_flags = PR_ATOMIC|PR_ADDR,
174 .pr_input = in_stf_input,
175 .pr_output = (pr_output_t *)rip_output,
176 .pr_ctloutput = rip_ctloutput,
177 .pr_usrreqs = &rip_usrreqs
180 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
182 static int stfmodevent(module_t, int, void *);
183 static int stf_encapcheck(const struct mbuf *, int, int, void *);
184 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
185 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
187 static int isrfc1918addr(struct in_addr *);
188 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
190 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
192 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
193 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
195 static int stf_clone_match(struct if_clone *, const char *);
196 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
197 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
198 static struct if_clone *stf_cloner;
201 stf_clone_match(struct if_clone *ifc, const char *name)
205 for(i = 0; stfnames[i] != NULL; i++) {
206 if (strcmp(stfnames[i], name) == 0)
214 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
217 struct stf_softc *sc;
221 * We can only have one unit, but since unit allocation is
222 * already locked, we use it to keep from allocating extra
226 err = ifc_alloc_unit(ifc, &unit);
230 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
231 ifp = STF2IFP(sc) = if_alloc(IFT_STF);
234 ifc_free_unit(ifc, unit);
238 sc->sc_fibnum = curthread->td_proc->p_fibnum;
241 * Set the name manually rather then using if_initname because
242 * we don't conform to the default naming convention for interfaces.
244 strlcpy(ifp->if_xname, name, IFNAMSIZ);
245 ifp->if_dname = stfname;
246 ifp->if_dunit = IF_DUNIT_NONE;
248 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
249 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
250 stf_encapcheck, &in_stf_protosw, sc);
251 if (sc->encap_cookie == NULL) {
252 if_printf(ifp, "attach failed\n");
254 ifc_free_unit(ifc, unit);
258 ifp->if_mtu = IPV6_MMTU;
259 ifp->if_ioctl = stf_ioctl;
260 ifp->if_output = stf_output;
261 ifp->if_snd.ifq_maxlen = ifqmaxlen;
263 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
268 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
270 struct stf_softc *sc = ifp->if_softc;
273 err = encap_detach(sc->encap_cookie);
274 KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
275 mtx_destroy(&(sc)->sc_ro_mtx);
281 ifc_free_unit(ifc, STFUNIT);
287 stfmodevent(mod, type, data)
295 stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
296 stf_clone_create, stf_clone_destroy);
299 if_clone_detach(stf_cloner);
308 static moduledata_t stf_mod = {
314 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
317 stf_encapcheck(m, off, proto, arg)
318 const struct mbuf *m;
324 struct in6_ifaddr *ia6;
325 struct stf_softc *sc;
326 struct in_addr a, b, mask;
328 sc = (struct stf_softc *)arg;
332 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
335 /* IFF_LINK0 means "no decapsulation" */
336 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
339 if (proto != IPPROTO_IPV6)
342 /* LINTED const cast */
343 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
348 ia6 = stf_getsrcifa6(STF2IFP(sc));
353 * check if IPv4 dst matches the IPv4 address derived from the
354 * local 6to4 address.
355 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
357 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
358 sizeof(ip.ip_dst)) != 0) {
359 ifa_free(&ia6->ia_ifa);
364 * check if IPv4 src matches the IPv4 address derived from the
365 * local 6to4 address masked by prefixmask.
366 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
367 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
369 bzero(&a, sizeof(a));
370 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
371 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
372 ifa_free(&ia6->ia_ifa);
373 a.s_addr &= mask.s_addr;
375 b.s_addr &= mask.s_addr;
376 if (a.s_addr != b.s_addr)
379 /* stf interface makes single side match only */
383 static struct in6_ifaddr *
388 struct in_ifaddr *ia4;
389 struct sockaddr_in6 *sin6;
393 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
394 if (ia->ifa_addr->sa_family != AF_INET6)
396 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
397 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
400 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
401 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
402 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
408 if_addr_runlock(ifp);
409 return (struct in6_ifaddr *)ia;
411 if_addr_runlock(ifp);
417 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
420 struct stf_softc *sc;
421 const struct sockaddr_in6 *dst6;
422 struct route *cached_route;
425 struct sockaddr_in *dst4;
429 struct in6_ifaddr *ia6;
433 error = mac_ifnet_check_transmit(ifp, m);
441 dst6 = (const struct sockaddr_in6 *)dst;
444 if ((ifp->if_flags & IFF_UP) == 0) {
451 * If we don't have an ip4 address that match my inner ip6 address,
452 * we shouldn't generate output. Without this check, we'll end up
453 * using wrong IPv4 source.
455 ia6 = stf_getsrcifa6(ifp);
462 if (m->m_len < sizeof(*ip6)) {
463 m = m_pullup(m, sizeof(*ip6));
465 ifa_free(&ia6->ia_ifa);
470 ip6 = mtod(m, struct ip6_hdr *);
471 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
474 * Pickup the right outer dst addr from the list of candidates.
475 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
478 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
479 ptr = GET_V4(&ip6->ip6_dst);
480 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
481 ptr = GET_V4(&dst6->sin6_addr);
483 ifa_free(&ia6->ia_ifa);
488 bcopy(ptr, &in4, sizeof(in4));
490 if (bpf_peers_present(ifp->if_bpf)) {
492 * We need to prepend the address family as
493 * a four byte field. Cons up a dummy header
494 * to pacify bpf. This is safe because bpf
495 * will only read from the mbuf (i.e., it won't
496 * try to free it or keep a pointer a to it).
499 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
502 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
503 if (m && m->m_len < sizeof(struct ip))
504 m = m_pullup(m, sizeof(struct ip));
506 ifa_free(&ia6->ia_ifa);
510 ip = mtod(m, struct ip *);
512 bzero(ip, sizeof(*ip));
514 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
515 &ip->ip_src, sizeof(ip->ip_src));
516 ifa_free(&ia6->ia_ifa);
517 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
518 ip->ip_p = IPPROTO_IPV6;
519 ip->ip_ttl = ip_stf_ttl;
520 ip->ip_len = htons(m->m_pkthdr.len);
521 if (ifp->if_flags & IFF_LINK1)
522 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
524 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
526 if (!stf_route_cache) {
532 * Do we have a cached route?
534 mtx_lock(&(sc)->sc_ro_mtx);
535 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
536 if (dst4->sin_family != AF_INET ||
537 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
538 /* cache route doesn't match */
539 dst4->sin_family = AF_INET;
540 dst4->sin_len = sizeof(struct sockaddr_in);
541 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
542 if (sc->sc_ro.ro_rt) {
543 RTFREE(sc->sc_ro.ro_rt);
544 sc->sc_ro.ro_rt = NULL;
548 if (sc->sc_ro.ro_rt == NULL) {
549 rtalloc_fib(&sc->sc_ro, sc->sc_fibnum);
550 if (sc->sc_ro.ro_rt == NULL) {
552 mtx_unlock(&(sc)->sc_ro_mtx);
557 cached_route = &sc->sc_ro;
560 M_SETFIB(m, sc->sc_fibnum);
562 error = ip_output(m, NULL, cached_route, 0, NULL, NULL);
564 if (cached_route != NULL)
565 mtx_unlock(&(sc)->sc_ro_mtx);
574 * returns 1 if private address range:
575 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
577 if (stf_permit_rfc1918 == 0 && (
578 (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
579 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
580 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
587 stf_checkaddr4(sc, in, inifp)
588 struct stf_softc *sc;
590 struct ifnet *inifp; /* incoming interface */
592 struct in_ifaddr *ia4;
595 * reject packets with the following address:
596 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
598 if (IN_MULTICAST(ntohl(in->s_addr)))
600 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
601 case 0: case 127: case 255:
606 * reject packets with private address range.
607 * (requirement from RFC3056 section 2 1st paragraph)
609 if (isrfc1918addr(in))
613 * reject packets with broadcast
616 TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
617 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
619 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
627 * perform ingress filter
629 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
630 struct sockaddr_in sin;
633 bzero(&sin, sizeof(sin));
634 sin.sin_family = AF_INET;
635 sin.sin_len = sizeof(struct sockaddr_in);
637 rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
639 if (!rt || rt->rt_ifp != inifp) {
641 log(LOG_WARNING, "%s: packet from 0x%x dropped "
642 "due to ingress filter\n", if_name(STF2IFP(sc)),
643 (u_int32_t)ntohl(sin.sin_addr.s_addr));
656 stf_checkaddr6(sc, in6, inifp)
657 struct stf_softc *sc;
658 struct in6_addr *in6;
659 struct ifnet *inifp; /* incoming interface */
662 * check 6to4 addresses
664 if (IN6_IS_ADDR_6TO4(in6)) {
666 bcopy(GET_V4(in6), &in4, sizeof(in4));
667 return stf_checkaddr4(sc, &in4, inifp);
671 * reject anything that look suspicious. the test is implemented
672 * in ip6_input too, but we check here as well to
673 * (1) reject bad packets earlier, and
674 * (2) to be safe against future ip6_input change.
676 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
688 struct stf_softc *sc;
694 proto = mtod(m, struct ip *)->ip_p;
696 if (proto != IPPROTO_IPV6) {
701 ip = mtod(m, struct ip *);
703 sc = (struct stf_softc *)encap_getarg(m);
705 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
713 mac_ifnet_create_mbuf(ifp, m);
717 * perform sanity check against outer src/dst.
718 * for source, perform ingress filter as well.
720 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
721 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
729 if (m->m_len < sizeof(*ip6)) {
730 m = m_pullup(m, sizeof(*ip6));
734 ip6 = mtod(m, struct ip6_hdr *);
737 * perform sanity check against inner src/dst.
738 * for source, perform ingress filter as well.
740 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
741 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
746 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
747 if ((ifp->if_flags & IFF_LINK1) != 0)
748 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
750 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
751 ip6->ip6_flow &= ~htonl(0xff << 20);
752 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
754 m->m_pkthdr.rcvif = ifp;
756 if (bpf_peers_present(ifp->if_bpf)) {
758 * We need to prepend the address family as
759 * a four byte field. Cons up a dummy header
760 * to pacify bpf. This is safe because bpf
761 * will only read from the mbuf (i.e., it won't
762 * try to free it or keep a pointer a to it).
764 u_int32_t af = AF_INET6;
765 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
769 * Put the packet to the network layer input queue according to the
770 * specified address family.
771 * See net/if_gif.c for possible issues with packet processing
772 * reorder due to extra queueing.
775 ifp->if_ibytes += m->m_pkthdr.len;
776 M_SETFIB(m, ifp->if_fib);
777 netisr_dispatch(NETISR_IPV6, m);
782 stf_rtrequest(cmd, rt, info)
785 struct rt_addrinfo *info;
788 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
792 stf_ioctl(ifp, cmd, data)
799 struct sockaddr_in6 *sin6;
806 ifa = (struct ifaddr *)data;
807 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
808 error = EAFNOSUPPORT;
811 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
812 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
816 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
817 if (isrfc1918addr(&addr)) {
822 ifa->ifa_rtrequest = stf_rtrequest;
823 ifp->if_flags |= IFF_UP;
828 ifr = (struct ifreq *)data;
829 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
832 error = EAFNOSUPPORT;
839 ifr = (struct ifreq *)data;
841 /* RFC 4213 3.2 ideal world MTU */
842 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)