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_clone.h>
98 #include <net/route.h>
99 #include <net/netisr.h>
100 #include <net/if_types.h>
101 #include <net/if_stf.h>
102 #include <net/vnet.h>
104 #include <netinet/in.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 SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
126 static int stf_route_cache = 1;
127 SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW,
128 &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output");
130 #define STFNAME "stf"
133 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002)
136 * XXX: Return a pointer with 16-bit aligned. Don't cast it to
137 * struct in_addr *; use bcopy() instead.
139 #define GET_V4(x) ((caddr_t)(&(x)->s6_addr16[1]))
142 struct ifnet *sc_ifp;
144 struct route __sc_ro4;
145 struct route_in6 __sc_ro6; /* just for safety */
147 #define sc_ro __sc_ro46.__sc_ro4
148 struct mtx sc_ro_mtx;
150 const struct encaptab *encap_cookie;
152 #define STF2IFP(sc) ((sc)->sc_ifp)
155 * Note that mutable fields in the softc are not currently locked.
156 * We do lock sc_ro in stf_output though.
158 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
159 static const int ip_stf_ttl = 40;
161 extern struct domain inetdomain;
162 struct protosw in_stf_protosw = {
164 .pr_domain = &inetdomain,
165 .pr_protocol = IPPROTO_IPV6,
166 .pr_flags = PR_ATOMIC|PR_ADDR,
167 .pr_input = in_stf_input,
168 .pr_output = (pr_output_t *)rip_output,
169 .pr_ctloutput = rip_ctloutput,
170 .pr_usrreqs = &rip_usrreqs
173 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
175 static int stfmodevent(module_t, int, void *);
176 static int stf_encapcheck(const struct mbuf *, int, int, void *);
177 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
178 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
180 static int isrfc1918addr(struct in_addr *);
181 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
183 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
185 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
186 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
188 static int stf_clone_match(struct if_clone *, const char *);
189 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
190 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
191 struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0,
192 NULL, stf_clone_match, stf_clone_create, stf_clone_destroy);
195 stf_clone_match(struct if_clone *ifc, const char *name)
199 for(i = 0; stfnames[i] != NULL; i++) {
200 if (strcmp(stfnames[i], name) == 0)
208 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
211 struct stf_softc *sc;
215 * We can only have one unit, but since unit allocation is
216 * already locked, we use it to keep from allocating extra
220 err = ifc_alloc_unit(ifc, &unit);
224 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
225 ifp = STF2IFP(sc) = if_alloc(IFT_STF);
228 ifc_free_unit(ifc, unit);
232 sc->sc_fibnum = curthread->td_proc->p_fibnum;
235 * Set the name manually rather then using if_initname because
236 * we don't conform to the default naming convention for interfaces.
238 strlcpy(ifp->if_xname, name, IFNAMSIZ);
239 ifp->if_dname = ifc->ifc_name;
240 ifp->if_dunit = IF_DUNIT_NONE;
242 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
243 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
244 stf_encapcheck, &in_stf_protosw, sc);
245 if (sc->encap_cookie == NULL) {
246 if_printf(ifp, "attach failed\n");
248 ifc_free_unit(ifc, unit);
252 ifp->if_mtu = IPV6_MMTU;
253 ifp->if_ioctl = stf_ioctl;
254 ifp->if_output = stf_output;
255 ifp->if_snd.ifq_maxlen = ifqmaxlen;
257 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
262 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
264 struct stf_softc *sc = ifp->if_softc;
267 err = encap_detach(sc->encap_cookie);
268 KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
269 mtx_destroy(&(sc)->sc_ro_mtx);
275 ifc_free_unit(ifc, STFUNIT);
281 stfmodevent(mod, type, data)
289 if_clone_attach(&stf_cloner);
292 if_clone_detach(&stf_cloner);
301 static moduledata_t stf_mod = {
307 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
310 stf_encapcheck(m, off, proto, arg)
311 const struct mbuf *m;
317 struct in6_ifaddr *ia6;
318 struct stf_softc *sc;
319 struct in_addr a, b, mask;
321 sc = (struct stf_softc *)arg;
325 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
328 /* IFF_LINK0 means "no decapsulation" */
329 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
332 if (proto != IPPROTO_IPV6)
335 /* LINTED const cast */
336 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
341 ia6 = stf_getsrcifa6(STF2IFP(sc));
346 * check if IPv4 dst matches the IPv4 address derived from the
347 * local 6to4 address.
348 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
350 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
351 sizeof(ip.ip_dst)) != 0) {
352 ifa_free(&ia6->ia_ifa);
357 * check if IPv4 src matches the IPv4 address derived from the
358 * local 6to4 address masked by prefixmask.
359 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
360 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
362 bzero(&a, sizeof(a));
363 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
364 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
365 ifa_free(&ia6->ia_ifa);
366 a.s_addr &= mask.s_addr;
368 b.s_addr &= mask.s_addr;
369 if (a.s_addr != b.s_addr)
372 /* stf interface makes single side match only */
376 static struct in6_ifaddr *
381 struct in_ifaddr *ia4;
382 struct sockaddr_in6 *sin6;
386 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
387 if (ia->ifa_addr->sa_family != AF_INET6)
389 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
390 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
393 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
394 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
395 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
401 if_addr_runlock(ifp);
402 return (struct in6_ifaddr *)ia;
404 if_addr_runlock(ifp);
410 stf_output(ifp, m, dst, ro)
413 struct sockaddr *dst;
416 struct stf_softc *sc;
417 struct sockaddr_in6 *dst6;
418 struct route *cached_route;
421 struct sockaddr_in *dst4;
425 struct in6_ifaddr *ia6;
430 error = mac_ifnet_check_transmit(ifp, m);
438 dst6 = (struct sockaddr_in6 *)dst;
441 if ((ifp->if_flags & IFF_UP) == 0) {
448 * If we don't have an ip4 address that match my inner ip6 address,
449 * we shouldn't generate output. Without this check, we'll end up
450 * using wrong IPv4 source.
452 ia6 = stf_getsrcifa6(ifp);
459 if (m->m_len < sizeof(*ip6)) {
460 m = m_pullup(m, sizeof(*ip6));
462 ifa_free(&ia6->ia_ifa);
467 ip6 = mtod(m, struct ip6_hdr *);
468 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
471 * BPF writes need to be handled specially.
472 * This is a null operation, nothing here checks dst->sa_family.
474 if (dst->sa_family == AF_UNSPEC) {
475 bcopy(dst->sa_data, &af, sizeof(af));
480 * Pickup the right outer dst addr from the list of candidates.
481 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
484 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
485 ptr = GET_V4(&ip6->ip6_dst);
486 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
487 ptr = GET_V4(&dst6->sin6_addr);
489 ifa_free(&ia6->ia_ifa);
494 bcopy(ptr, &in4, sizeof(in4));
496 if (bpf_peers_present(ifp->if_bpf)) {
498 * We need to prepend the address family as
499 * a four byte field. Cons up a dummy header
500 * to pacify bpf. This is safe because bpf
501 * will only read from the mbuf (i.e., it won't
502 * try to free it or keep a pointer a to it).
505 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
508 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
509 if (m && m->m_len < sizeof(struct ip))
510 m = m_pullup(m, sizeof(struct ip));
512 ifa_free(&ia6->ia_ifa);
516 ip = mtod(m, struct ip *);
518 bzero(ip, sizeof(*ip));
520 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
521 &ip->ip_src, sizeof(ip->ip_src));
522 ifa_free(&ia6->ia_ifa);
523 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
524 ip->ip_p = IPPROTO_IPV6;
525 ip->ip_ttl = ip_stf_ttl;
526 ip->ip_len = m->m_pkthdr.len; /*host order*/
527 if (ifp->if_flags & IFF_LINK1)
528 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
530 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
532 if (!stf_route_cache) {
538 * Do we have a cached route?
540 mtx_lock(&(sc)->sc_ro_mtx);
541 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
542 if (dst4->sin_family != AF_INET ||
543 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
544 /* cache route doesn't match */
545 dst4->sin_family = AF_INET;
546 dst4->sin_len = sizeof(struct sockaddr_in);
547 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
548 if (sc->sc_ro.ro_rt) {
549 RTFREE(sc->sc_ro.ro_rt);
550 sc->sc_ro.ro_rt = NULL;
554 if (sc->sc_ro.ro_rt == NULL) {
555 rtalloc_fib(&sc->sc_ro, sc->sc_fibnum);
556 if (sc->sc_ro.ro_rt == NULL) {
558 mtx_unlock(&(sc)->sc_ro_mtx);
563 cached_route = &sc->sc_ro;
566 M_SETFIB(m, sc->sc_fibnum);
568 error = ip_output(m, NULL, cached_route, 0, NULL, NULL);
570 if (cached_route != NULL)
571 mtx_unlock(&(sc)->sc_ro_mtx);
580 * returns 1 if private address range:
581 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
583 if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
584 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
585 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
592 stf_checkaddr4(sc, in, inifp)
593 struct stf_softc *sc;
595 struct ifnet *inifp; /* incoming interface */
597 struct in_ifaddr *ia4;
600 * reject packets with the following address:
601 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
603 if (IN_MULTICAST(ntohl(in->s_addr)))
605 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
606 case 0: case 127: case 255:
611 * reject packets with private address range.
612 * (requirement from RFC3056 section 2 1st paragraph)
614 if (isrfc1918addr(in))
618 * reject packets with broadcast
621 for (ia4 = TAILQ_FIRST(&V_in_ifaddrhead);
623 ia4 = TAILQ_NEXT(ia4, ia_link))
625 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
627 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
635 * perform ingress filter
637 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
638 struct sockaddr_in sin;
641 bzero(&sin, sizeof(sin));
642 sin.sin_family = AF_INET;
643 sin.sin_len = sizeof(struct sockaddr_in);
645 rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
647 if (!rt || rt->rt_ifp != inifp) {
649 log(LOG_WARNING, "%s: packet from 0x%x dropped "
650 "due to ingress filter\n", if_name(STF2IFP(sc)),
651 (u_int32_t)ntohl(sin.sin_addr.s_addr));
664 stf_checkaddr6(sc, in6, inifp)
665 struct stf_softc *sc;
666 struct in6_addr *in6;
667 struct ifnet *inifp; /* incoming interface */
670 * check 6to4 addresses
672 if (IN6_IS_ADDR_6TO4(in6)) {
674 bcopy(GET_V4(in6), &in4, sizeof(in4));
675 return stf_checkaddr4(sc, &in4, inifp);
679 * reject anything that look suspicious. the test is implemented
680 * in ip6_input too, but we check here as well to
681 * (1) reject bad packets earlier, and
682 * (2) to be safe against future ip6_input change.
684 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
696 struct stf_softc *sc;
702 proto = mtod(m, struct ip *)->ip_p;
704 if (proto != IPPROTO_IPV6) {
709 ip = mtod(m, struct ip *);
711 sc = (struct stf_softc *)encap_getarg(m);
713 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
721 mac_ifnet_create_mbuf(ifp, m);
725 * perform sanity check against outer src/dst.
726 * for source, perform ingress filter as well.
728 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
729 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
737 if (m->m_len < sizeof(*ip6)) {
738 m = m_pullup(m, sizeof(*ip6));
742 ip6 = mtod(m, struct ip6_hdr *);
745 * perform sanity check against inner src/dst.
746 * for source, perform ingress filter as well.
748 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
749 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
754 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
755 if ((ifp->if_flags & IFF_LINK1) != 0)
756 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
758 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
759 ip6->ip6_flow &= ~htonl(0xff << 20);
760 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
762 m->m_pkthdr.rcvif = ifp;
764 if (bpf_peers_present(ifp->if_bpf)) {
766 * We need to prepend the address family as
767 * a four byte field. Cons up a dummy header
768 * to pacify bpf. This is safe because bpf
769 * will only read from the mbuf (i.e., it won't
770 * try to free it or keep a pointer a to it).
772 u_int32_t af = AF_INET6;
773 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
777 * Put the packet to the network layer input queue according to the
778 * specified address family.
779 * See net/if_gif.c for possible issues with packet processing
780 * reorder due to extra queueing.
783 ifp->if_ibytes += m->m_pkthdr.len;
784 M_SETFIB(m, ifp->if_fib);
785 netisr_dispatch(NETISR_IPV6, m);
790 stf_rtrequest(cmd, rt, info)
793 struct rt_addrinfo *info;
796 rt->rt_rmx.rmx_mtu = IPV6_MMTU;
800 stf_ioctl(ifp, cmd, data)
807 struct sockaddr_in6 *sin6;
814 ifa = (struct ifaddr *)data;
815 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
816 error = EAFNOSUPPORT;
819 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
820 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
824 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
825 if (isrfc1918addr(&addr)) {
830 ifa->ifa_rtrequest = stf_rtrequest;
831 ifp->if_flags |= IFF_UP;
836 ifr = (struct ifreq *)data;
837 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
840 error = EAFNOSUPPORT;