2 /* $KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $ */
5 * SPDX-License-Identifier: BSD-3-Clause
7 * Copyright (C) 2000 WIDE Project.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the project nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * 6to4 interface, based on RFC3056.
38 * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
39 * There is no address mapping defined from IPv6 multicast address to IPv4
40 * address. Therefore, we do not have IFF_MULTICAST on the interface.
42 * Due to the lack of address mapping for link-local addresses, we cannot
43 * throw packets toward link-local addresses (fe80::x). Also, we cannot throw
44 * packets to link-local multicast addresses (ff02::x).
46 * Here are interesting symptoms due to the lack of link-local address:
48 * Unicast routing exchange:
49 * - RIPng: Impossible. Uses link-local multicast packet toward ff02::9,
50 * and link-local addresses as nexthop.
51 * - OSPFv6: Impossible. OSPFv6 assumes that there's link-local address
52 * assigned to the link, and makes use of them. Also, HELLO packets use
53 * link-local multicast addresses (ff02::5 and ff02::6).
54 * - BGP4+: Maybe. You can only use global address as nexthop, and global
55 * address as TCP endpoint address.
57 * Multicast routing protocols:
58 * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
59 * Adjacent PIM routers must be configured manually (is it really spec-wise
60 * correct thing to do?).
63 * - Redirects cannot be used due to the lack of link-local address.
65 * stf interface does not have, and will not need, a link-local address.
66 * It seems to have no real benefit and does not help the above symptoms much.
67 * Even if we assign link-locals to interface, we cannot really
68 * use link-local unicast/multicast on top of 6to4 cloud (since there's no
69 * encapsulation defined for link-local address), and the above analysis does
70 * not change. RFC3056 does not mandate the assignment of link-local address
73 * 6to4 interface has security issues. Refer to
74 * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
75 * for details. The code tries to filter out some of malicious packets.
76 * Note that there is no way to be 100% secure.
79 #include <sys/param.h>
80 #include <sys/systm.h>
81 #include <sys/socket.h>
82 #include <sys/sockio.h>
84 #include <sys/errno.h>
85 #include <sys/kernel.h>
87 #include <sys/module.h>
89 #include <sys/queue.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/route/nhop.h>
100 #include <net/netisr.h>
101 #include <net/if_types.h>
102 #include <net/vnet.h>
104 #include <netinet/in.h>
105 #include <netinet/in_fib.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 | CTLFLAG_MPSAFE, 0,
128 static int stf_permit_rfc1918 = 0;
129 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
130 &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
134 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002)
137 * XXX: Return a pointer with 16-bit aligned. Don't cast it to
138 * struct in_addr *; use bcopy() instead.
140 #define GET_V4(x) (&(x)->s6_addr16[1])
143 struct ifnet *sc_ifp;
145 const struct encaptab *encap_cookie;
147 #define STF2IFP(sc) ((sc)->sc_ifp)
149 static const char stfname[] = "stf";
151 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
152 static const int ip_stf_ttl = 40;
154 static int in_stf_input(struct mbuf *, int, int, void *);
155 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
157 static int stfmodevent(module_t, int, void *);
158 static int stf_encapcheck(const struct mbuf *, int, int, void *);
159 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
160 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
162 static int isrfc1918addr(struct in_addr *);
163 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
165 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
167 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
169 static int stf_clone_match(struct if_clone *, const char *);
170 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
171 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
172 VNET_DEFINE_STATIC(struct if_clone *, stf_cloner);
173 #define V_stf_cloner VNET(stf_cloner)
175 static const struct encap_config ipv4_encap_cfg = {
176 .proto = IPPROTO_IPV6,
177 .min_length = sizeof(struct ip),
178 .exact_match = (sizeof(in_addr_t) << 3) + 8,
179 .check = stf_encapcheck,
180 .input = in_stf_input
184 stf_clone_match(struct if_clone *ifc, const char *name)
188 for(i = 0; stfnames[i] != NULL; i++) {
189 if (strcmp(stfnames[i], name) == 0)
197 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
200 int err, unit, wildcard;
201 struct stf_softc *sc;
204 err = ifc_name2unit(name, &unit);
207 wildcard = (unit < 0);
210 * We can only have one unit, but since unit allocation is
211 * already locked, we use it to keep from allocating extra
215 err = ifc_alloc_unit(ifc, &unit);
219 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
220 ifp = STF2IFP(sc) = if_alloc(IFT_STF);
223 ifc_free_unit(ifc, unit);
227 sc->sc_fibnum = curthread->td_proc->p_fibnum;
230 * Set the name manually rather then using if_initname because
231 * we don't conform to the default naming convention for interfaces.
232 * In the wildcard case, we need to update the name.
235 for (dp = name; *dp != '\0'; dp++);
236 if (snprintf(dp, len - (dp-name), "%d", unit) >
237 len - (dp-name) - 1) {
239 * This can only be a programmer error and
240 * there's no straightforward way to recover if
243 panic("if_clone_create(): interface name too long");
246 strlcpy(ifp->if_xname, name, IFNAMSIZ);
247 ifp->if_dname = stfname;
248 ifp->if_dunit = IF_DUNIT_NONE;
250 sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK);
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 = ip_encap_detach(sc->encap_cookie);
274 KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
280 ifc_free_unit(ifc, STFUNIT);
286 vnet_stf_init(const void *unused __unused)
288 V_stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
289 stf_clone_create, stf_clone_destroy);
291 VNET_SYSINIT(vnet_stf_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_init, NULL);
294 vnet_stf_uninit(const void *unused __unused)
296 if_clone_detach(V_stf_cloner);
299 VNET_SYSUNINIT(vnet_stf_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_uninit,
303 stfmodevent(module_t mod, int type, void *data)
308 /* Done in vnet_stf_init() */
311 /* Done in vnet_stf_uninit() */
320 static moduledata_t stf_mod = {
326 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
329 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
332 struct stf_softc *sc;
333 struct in_addr a, b, mask;
334 struct in6_addr addr6, mask6;
336 sc = (struct stf_softc *)arg;
340 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
343 /* IFF_LINK0 means "no decapsulation" */
344 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
347 if (proto != IPPROTO_IPV6)
350 m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
355 if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
359 * check if IPv4 dst matches the IPv4 address derived from the
360 * local 6to4 address.
361 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
363 if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
367 * check if IPv4 src matches the IPv4 address derived from the
368 * local 6to4 address masked by prefixmask.
369 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
370 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
372 bzero(&a, sizeof(a));
373 bcopy(GET_V4(&addr6), &a, sizeof(a));
374 bcopy(GET_V4(&mask6), &mask, sizeof(mask));
375 a.s_addr &= mask.s_addr;
377 b.s_addr &= mask.s_addr;
378 if (a.s_addr != b.s_addr)
381 /* stf interface makes single side match only */
386 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
389 struct in_ifaddr *ia4;
390 struct in6_ifaddr *ia6;
391 struct sockaddr_in6 *sin6;
396 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
397 if (ia->ifa_addr->sa_family != AF_INET6)
399 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
400 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
403 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
404 CK_LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
405 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
410 ia6 = (struct in6_ifaddr *)ia;
412 *addr = sin6->sin6_addr;
413 *mask = ia6->ia_prefixmask.sin6_addr;
421 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
424 struct stf_softc *sc;
425 const struct sockaddr_in6 *dst6;
431 struct in6_addr addr6, mask6;
435 error = mac_ifnet_check_transmit(ifp, m);
443 dst6 = (const struct sockaddr_in6 *)dst;
446 if ((ifp->if_flags & IFF_UP) == 0) {
448 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
453 * If we don't have an ip4 address that match my inner ip6 address,
454 * we shouldn't generate output. Without this check, we'll end up
455 * using wrong IPv4 source.
457 if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
459 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
463 if (m->m_len < sizeof(*ip6)) {
464 m = m_pullup(m, sizeof(*ip6));
466 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
470 ip6 = mtod(m, struct ip6_hdr *);
471 tos = IPV6_TRAFFIC_CLASS(ip6);
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);
484 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
485 return (ENETUNREACH);
487 bcopy(ptr, &in4, sizeof(in4));
489 if (bpf_peers_present(ifp->if_bpf)) {
491 * We need to prepend the address family as
492 * a four byte field. Cons up a dummy header
493 * to pacify bpf. This is safe because bpf
494 * will only read from the mbuf (i.e., it won't
495 * try to free it or keep a pointer a to it).
498 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
501 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
503 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
506 ip = mtod(m, struct ip *);
508 bzero(ip, sizeof(*ip));
510 bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
511 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
512 ip->ip_p = IPPROTO_IPV6;
513 ip->ip_ttl = ip_stf_ttl;
514 ip->ip_len = htons(m->m_pkthdr.len);
515 if (ifp->if_flags & IFF_LINK1)
516 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
518 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
520 M_SETFIB(m, sc->sc_fibnum);
521 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
522 error = ip_output(m, NULL, NULL, 0, NULL, NULL);
528 isrfc1918addr(struct in_addr *in)
531 * returns 1 if private address range:
532 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
534 if (stf_permit_rfc1918 == 0 && (
535 (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
536 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
537 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
544 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
546 struct in_ifaddr *ia4;
549 * reject packets with the following address:
550 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
552 if (IN_MULTICAST(ntohl(in->s_addr)))
554 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
555 case 0: case 127: case 255:
560 * reject packets with private address range.
561 * (requirement from RFC3056 section 2 1st paragraph)
563 if (isrfc1918addr(in))
567 * reject packets with broadcast
569 CK_STAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
570 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
572 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
578 * perform ingress filter
580 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
581 struct nhop_object *nh;
584 nh = fib4_lookup(sc->sc_fibnum, *in, 0, 0, 0);
588 if (nh->nh_ifp != inifp)
596 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
599 * check 6to4 addresses
601 if (IN6_IS_ADDR_6TO4(in6)) {
603 bcopy(GET_V4(in6), &in4, sizeof(in4));
604 return (stf_checkaddr4(sc, &in4, inifp));
608 * reject anything that look suspicious. the test is implemented
609 * in ip6_input too, but we check here as well to
610 * (1) reject bad packets earlier, and
611 * (2) to be safe against future ip6_input change.
613 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
620 in_stf_input(struct mbuf *m, int off, int proto, void *arg)
622 struct stf_softc *sc = arg;
630 if (proto != IPPROTO_IPV6) {
632 return (IPPROTO_DONE);
635 ip = mtod(m, struct ip *);
636 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
638 return (IPPROTO_DONE);
644 mac_ifnet_create_mbuf(ifp, m);
648 * perform sanity check against outer src/dst.
649 * for source, perform ingress filter as well.
651 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
652 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
654 return (IPPROTO_DONE);
660 if (m->m_len < sizeof(*ip6)) {
661 m = m_pullup(m, sizeof(*ip6));
663 return (IPPROTO_DONE);
665 ip6 = mtod(m, struct ip6_hdr *);
668 * perform sanity check against inner src/dst.
669 * for source, perform ingress filter as well.
671 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
672 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
674 return (IPPROTO_DONE);
677 itos = IPV6_TRAFFIC_CLASS(ip6);
678 if ((ifp->if_flags & IFF_LINK1) != 0)
679 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
681 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
682 ip6->ip6_flow &= ~htonl(0xff << 20);
683 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
685 m->m_pkthdr.rcvif = ifp;
687 if (bpf_peers_present(ifp->if_bpf)) {
689 * We need to prepend the address family as
690 * a four byte field. Cons up a dummy header
691 * to pacify bpf. This is safe because bpf
692 * will only read from the mbuf (i.e., it won't
693 * try to free it or keep a pointer a to it).
695 u_int32_t af = AF_INET6;
696 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
700 * Put the packet to the network layer input queue according to the
701 * specified address family.
702 * See net/if_gif.c for possible issues with packet processing
703 * reorder due to extra queueing.
705 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
706 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
707 M_SETFIB(m, ifp->if_fib);
708 netisr_dispatch(NETISR_IPV6, m);
709 return (IPPROTO_DONE);
713 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
717 struct sockaddr_in6 *sin6;
724 ifa = (struct ifaddr *)data;
725 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
726 error = EAFNOSUPPORT;
729 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
730 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
734 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
735 if (isrfc1918addr(&addr)) {
740 ifp->if_flags |= IFF_UP;
741 ifp->if_drv_flags |= IFF_DRV_RUNNING;
746 ifr = (struct ifreq *)data;
747 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
750 error = EAFNOSUPPORT;
757 ifr = (struct ifreq *)data;
759 /* RFC 4213 3.2 ideal world MTU */
760 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)