/* * Forward a packet. If some error occurs return the sender * an icmp packet. Note we can't always generate a meaningful * icmp message because icmp doesn't have a large enough repertoire * of codes and types. * * If not forwarding, just drop the packet. This could be confusing * if ipforwarding was zero but some routing protocol was advancing * us as a gateway to somewhere. However, we must let the routing * protocol deal with that. * */ void ip6_forward(struct mbuf *m, int srcrt) { struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); struct sockaddr_in6 *dst = NULL; struct rtentry *rt = NULL; struct route_in6 rin6; int error, type = 0, code = 0; struct mbuf *mcopy = NULL; struct ifnet *origifp; /* maybe unnecessary */ u_int32_t inzone, outzone; struct in6_addr src_in6, dst_in6, odst; #ifdef IPSEC struct secpolicy *sp = NULL; #endif #ifdef SCTP int sw_csum; #endif struct m_tag *fwd_tag; char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; /* * Do not forward packets to multicast destination (should be handled * by ip6_mforward(). * Do not forward packets with unspecified source. It was discussed * in July 2000, on the ipngwg mailing list. */ if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { IP6STAT_INC(ip6s_cantforward); /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { V_ip6_log_time = time_uptime; log(LOG_DEBUG, "cannot forward " "from %s to %s nxt %d received on %s\n", ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif)); } m_freem(m); return; } #ifdef IPSEC /* * Check if this packet has an active SA and needs to be dropped * instead of forwarded. */ if (ip6_ipsec_fwd(m) != 0) { IP6STAT_INC(ip6s_cantforward); m_freem(m); return; } #endif /* IPSEC */ #ifdef IPSTEALTH if (!V_ip6stealth) { #endif if (ip6->ip6_hlim <= IPV6_HLIMDEC) { /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ icmp6_error(m, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT, 0); return; } ip6->ip6_hlim -= IPV6_HLIMDEC; #ifdef IPSTEALTH } #endif /* * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - * size of IPv6 + ICMPv6 headers) bytes of the packet in case * we need to generate an ICMP6 message to the src. * Thanks to M_EXT, in most cases copy will not occur. * * It is important to save it before IPsec processing as IPsec * processing may modify the mbuf. */ mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); #ifdef IPSEC /* get a security policy for this packet */ sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, &error); if (sp == NULL) { IPSEC6STAT_INC(ips_out_inval); IP6STAT_INC(ip6s_cantforward); if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return; } error = 0; /* check policy */ switch (sp->policy) { case IPSEC_POLICY_DISCARD: /* * This packet is just discarded. */ IPSEC6STAT_INC(ips_out_polvio); IP6STAT_INC(ip6s_cantforward); KEY_FREESP(&sp); if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return; case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: /* no need to do IPsec. */ KEY_FREESP(&sp); goto skip_ipsec; case IPSEC_POLICY_IPSEC: if (sp->req == NULL) { /* XXX should be panic ? */ printf("ip6_forward: No IPsec request specified.\n"); IP6STAT_INC(ip6s_cantforward); KEY_FREESP(&sp); if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return; } /* do IPsec */ break; case IPSEC_POLICY_ENTRUST: default: /* should be panic ?? */ printf("ip6_forward: Invalid policy found. %d\n", sp->policy); KEY_FREESP(&sp); goto skip_ipsec; } { struct ipsecrequest *isr = NULL; /* * when the kernel forwards a packet, it is not proper to apply * IPsec transport mode to the packet. This check avoid from this. * at present, if there is even a transport mode SA request in the * security policy, the kernel does not apply IPsec to the packet. * this check is not enough because the following case is valid. * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; */ for (isr = sp->req; isr; isr = isr->next) { if (isr->saidx.mode == IPSEC_MODE_ANY) goto doipsectunnel; if (isr->saidx.mode == IPSEC_MODE_TUNNEL) goto doipsectunnel; } /* * if there's no need for tunnel mode IPsec, skip. */ if (!isr) goto skip_ipsec; doipsectunnel: /* * All the extension headers will become inaccessible * (since they can be encrypted). * Don't panic, we need no more updates to extension headers * on inner IPv6 packet (since they are now encapsulated). * * IPv6 [ESP|AH] IPv6 [extension headers] payload */ /* * If we need to encapsulate the packet, do it here * ipsec6_proces_packet will send the packet using ip6_output */ error = ipsec6_process_packet(m, sp->req); /* Release SP if an error occurred */ if (error != 0) KEY_FREESP(&sp); if (error == EJUSTRETURN) { /* * We had a SP with a level of 'use' and no SA. We * will just continue to process the packet without * IPsec processing. */ error = 0; goto skip_ipsec; } if (error) { /* mbuf is already reclaimed in ipsec6_process_packet. */ switch (error) { case EHOSTUNREACH: case ENETUNREACH: case EMSGSIZE: case ENOBUFS: case ENOMEM: break; default: printf("ip6_output (ipsec): error code %d\n", error); /* FALLTHROUGH */ case ENOENT: /* don't show these error codes to the user */ break; } IP6STAT_INC(ip6s_cantforward); if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } return; } else { /* * In the FAST IPSec case we have already * re-injected the packet and it has been freed * by the ipsec_done() function. So, just clean * up after ourselves. */ m = NULL; goto freecopy; } } skip_ipsec: #endif again: bzero(&rin6, sizeof(struct route_in6)); dst = (struct sockaddr_in6 *)&rin6.ro_dst; dst->sin6_len = sizeof(struct sockaddr_in6); dst->sin6_family = AF_INET6; dst->sin6_addr = ip6->ip6_dst; again2: rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m)); rt = rin6.ro_rt; if (rin6.ro_rt != NULL) RT_UNLOCK(rin6.ro_rt); else { IP6STAT_INC(ip6s_noroute); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); if (mcopy) { icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE, 0); } goto bad; } /* * Source scope check: if a packet can't be delivered to its * destination for the reason that the destination is beyond the scope * of the source address, discard the packet and return an icmp6 * destination unreachable error with Code 2 (beyond scope of source * address). We use a local copy of ip6_src, since in6_setscope() * will possibly modify its first argument. * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] */ src_in6 = ip6->ip6_src; if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { /* XXX: this should not happen */ IP6STAT_INC(ip6s_cantforward); IP6STAT_INC(ip6s_badscope); goto bad; } if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { IP6STAT_INC(ip6s_cantforward); IP6STAT_INC(ip6s_badscope); goto bad; } if (inzone != outzone) { IP6STAT_INC(ip6s_cantforward); IP6STAT_INC(ip6s_badscope); in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { V_ip6_log_time = time_uptime; log(LOG_DEBUG, "cannot forward " "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); } if (mcopy) icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_BEYONDSCOPE, 0); goto bad; } /* * Destination scope check: if a packet is going to break the scope * zone of packet's destination address, discard it. This case should * usually be prevented by appropriately-configured routing table, but * we need an explicit check because we may mistakenly forward the * packet to a different zone by (e.g.) a default route. */ dst_in6 = ip6->ip6_dst; if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || inzone != outzone) { IP6STAT_INC(ip6s_cantforward); IP6STAT_INC(ip6s_badscope); goto bad; } if (rt->rt_flags & RTF_GATEWAY) dst = (struct sockaddr_in6 *)rt->rt_gateway; /* * If we are to forward the packet using the same interface * as one we got the packet from, perhaps we should send a redirect * to sender to shortcut a hop. * Only send redirect if source is sending directly to us, * and if packet was not source routed (or has any options). * Also, don't send redirect if forwarding using a route * modified by a redirect. */ if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { /* * If the incoming interface is equal to the outgoing * one, and the link attached to the interface is * point-to-point, then it will be highly probable * that a routing loop occurs. Thus, we immediately * drop the packet and send an ICMPv6 error message. * * type/code is based on suggestion by Rich Draves. * not sure if it is the best pick. */ icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, 0); goto bad; } type = ND_REDIRECT; } /* * Fake scoped addresses. Note that even link-local source or * destinaion can appear, if the originating node just sends the * packet to us (without address resolution for the destination). * Since both icmp6_error and icmp6_redirect_output fill the embedded * link identifiers, we can do this stuff after making a copy for * returning an error. */ if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { /* * See corresponding comments in ip6_output. * XXX: but is it possible that ip6_forward() sends a packet * to a loopback interface? I don't think so, and thus * I bark here. ([email protected]) * XXX: it is common to route invalid packets to loopback. * also, the codepath will be visited on use of ::1 in * rthdr. (itojun) */ #if 1 if (0) #else if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) #endif { printf("ip6_forward: outgoing interface is loopback. " "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); } /* we can just use rcvif in forwarding. */ origifp = m->m_pkthdr.rcvif; } else origifp = rt->rt_ifp; /* * clear embedded scope identifiers if necessary. * in6_clearscope will touch the addresses only when necessary. */ in6_clearscope(&ip6->ip6_src); in6_clearscope(&ip6->ip6_dst); /* Jump over all PFIL processing if hooks are not active. */ if (!PFIL_HOOKED(&V_inet6_pfil_hook)) goto pass; odst = ip6->ip6_dst; /* Run through list of hooks for output packets. */ error = pfil_run_hooks(&V_inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); if (error != 0 || m == NULL) goto freecopy; /* consumed by filter */ ip6 = mtod(m, struct ip6_hdr *); /* See if destination IP address was changed by packet filter. */ if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { m->m_flags |= M_SKIP_FIREWALL; /* If destination is now ourself drop to ip6_input(). */ if (in6_localip(&ip6->ip6_dst)) m->m_flags |= M_FASTFWD_OURS; else { RTFREE(rt); goto again; /* Redo the routing table lookup. */ } } /* See if local, if yes, send it to netisr. */ if (m->m_flags & M_FASTFWD_OURS) { if (m->m_pkthdr.rcvif == NULL) m->m_pkthdr.rcvif = V_loif; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { m->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; m->m_pkthdr.csum_data = 0xffff; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; #endif error = netisr_queue(NETISR_IPV6, m); goto out; } /* Or forward to some other address? */ if ((m->m_flags & M_IP6_NEXTHOP) && (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { dst = (struct sockaddr_in6 *)&rin6.ro_dst; bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6)); m->m_flags |= M_SKIP_FIREWALL; m->m_flags &= ~M_IP6_NEXTHOP; m_tag_delete(m, fwd_tag); RTFREE(rt); goto again2; } pass: /* See if the size was changed by the packet filter. */ if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); if (mcopy) { u_long mtu; #ifdef IPSEC size_t ipsechdrsiz; #endif /* IPSEC */ mtu = IN6_LINKMTU(rt->rt_ifp); #ifdef IPSEC /* * When we do IPsec tunnel ingress, we need to play * with the link value (decrement IPsec header size * from mtu value). The code is much simpler than v4 * case, as we have the outgoing interface for * encapsulated packet as "rt->rt_ifp". */ ipsechdrsiz = ipsec_hdrsiz(mcopy, IPSEC_DIR_OUTBOUND, NULL); if (ipsechdrsiz < mtu) mtu -= ipsechdrsiz; /* * if mtu becomes less than minimum MTU, * tell minimum MTU (and I'll need to fragment it). */ if (mtu < IPV6_MMTU) mtu = IPV6_MMTU; #endif /* IPSEC */ icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); } goto bad; } error = nd6_output_ifp(rt->rt_ifp, origifp, m, dst, NULL); if (error) { in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); IP6STAT_INC(ip6s_cantforward); } else { IP6STAT_INC(ip6s_forward); in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); if (type) IP6STAT_INC(ip6s_redirectsent); else { if (mcopy) goto freecopy; } } if (mcopy == NULL) goto out; switch (error) { case 0: if (type == ND_REDIRECT) { icmp6_redirect_output(mcopy, rt); goto out; } goto freecopy; case EMSGSIZE: /* xxx MTU is constant in PPP? */ goto freecopy; case ENOBUFS: /* Tell source to slow down like source quench in IP? */ goto freecopy; case ENETUNREACH: /* shouldn't happen, checked above */ case EHOSTUNREACH: case ENETDOWN: case EHOSTDOWN: default: type = ICMP6_DST_UNREACH; code = ICMP6_DST_UNREACH_ADDR; break; } icmp6_error(mcopy, type, code, 0); goto out; freecopy: m_freem(mcopy); goto out; bad: m_freem(m); out: if (rt != NULL) RTFREE(rt); }
/* * Type0 routing header processing * * RFC2292 backward compatibility warning: no support for strict/loose bitmap, * as it was dropped between RFC1883 and RFC2460. */ static int ip6_rthdr0(struct mbuf *m, struct ip6_hdr *ip6, struct ip6_rthdr0 *rh0) { int addrs, index; struct in6_addr *nextaddr, tmpaddr; const struct ip6aux *ip6a; if (rh0->ip6r0_segleft == 0) return (0); if (rh0->ip6r0_len % 2 #ifdef COMPAT_RFC1883 || rh0->ip6r0_len > 46 #endif ) { /* * Type 0 routing header can't contain more than 23 addresses. * RFC 2462: this limitation was removed since strict/loose * bitmap field was deleted. */ IP6_STATINC(IP6_STAT_BADOPTIONS); icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, (char *)&rh0->ip6r0_len - (char *)ip6); return (-1); } if ((addrs = rh0->ip6r0_len / 2) < rh0->ip6r0_segleft) { IP6_STATINC(IP6_STAT_BADOPTIONS); icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, (char *)&rh0->ip6r0_segleft - (char *)ip6); return (-1); } index = addrs - rh0->ip6r0_segleft; rh0->ip6r0_segleft--; nextaddr = ((struct in6_addr *)(rh0 + 1)) + index; /* * reject invalid addresses. be proactive about malicious use of * IPv4 mapped/compat address. * XXX need more checks? */ if (IN6_IS_ADDR_MULTICAST(nextaddr) || IN6_IS_ADDR_UNSPECIFIED(nextaddr) || IN6_IS_ADDR_V4MAPPED(nextaddr) || IN6_IS_ADDR_V4COMPAT(nextaddr)) { p6stat[IP6_STAT_BADOPTIONS]++; goto bad; } if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst) || IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst) || IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { IP6_STATINC(IP6_STAT_BADOPTIONS); goto bad; } /* * Determine the scope zone of the next hop, based on the interface * of the current hop. [RFC4007, Section 9] * Then disambiguate the scope zone for the next hop (if necessary). */ if ((ip6a = ip6_getdstifaddr(m)) == NULL) goto bad; if (in6_setzoneid(nextaddr, ip6a->ip6a_scope_id) != 0) { IP6_STATINC(IP6_STAT_BADSCOPE); goto bad; } /* * Swap the IPv6 destination address and nextaddr. Forward the packet. */ tmpaddr = *nextaddr; *nextaddr = ip6->ip6_dst; in6_clearscope(nextaddr); /* XXX */ ip6->ip6_dst = tmpaddr; #ifdef COMPAT_RFC1883 if (rh0->ip6r0_slmap[index / 8] & (1 << (7 - (index % 8)))) ip6_forward(m, IPV6_SRCRT_NEIGHBOR); else ip6_forward(m, IPV6_SRCRT_NOTNEIGHBOR); #else ip6_forward(m, 1); #endif return (-1); /* m would be freed in ip6_forward() */ bad: m_freem(m); return (-1); }
/* * Output a Neighbor Solicitation Message. Caller specifies: * - ICMP6 header source IP6 address * - ND6 header target IP6 address * - ND6 header source datalink address * * Based on RFC 2461 * Based on RFC 2462 (duplicate address detection) * * ln - for source address determination * nonce - If non-NULL, NS is used for duplicate address detection and * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce. */ static void nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6, const struct in6_addr *daddr6, const struct in6_addr *taddr6, uint8_t *nonce, u_int fibnum) { struct mbuf *m; struct m_tag *mtag; struct ip6_hdr *ip6; struct nd_neighbor_solicit *nd_ns; struct ip6_moptions im6o; int icmp6len; int maxlen; caddr_t mac; struct route_in6 ro; if (IN6_IS_ADDR_MULTICAST(taddr6)) return; /* estimate the size of message */ maxlen = sizeof(*ip6) + sizeof(*nd_ns); maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", __func__, max_linkhdr, maxlen, MCLBYTES)); if (max_linkhdr + maxlen > MHLEN) m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); else m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return; M_SETFIB(m, fibnum); bzero(&ro, sizeof(ro)); if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) { m->m_flags |= M_MCAST; im6o.im6o_multicast_ifp = ifp; im6o.im6o_multicast_hlim = 255; im6o.im6o_multicast_loop = 0; } icmp6len = sizeof(*nd_ns); m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len; m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ /* fill neighbor solicitation packet */ ip6 = mtod(m, struct ip6_hdr *); ip6->ip6_flow = 0; ip6->ip6_vfc &= ~IPV6_VERSION_MASK; ip6->ip6_vfc |= IPV6_VERSION; /* ip6->ip6_plen will be set later */ ip6->ip6_nxt = IPPROTO_ICMPV6; ip6->ip6_hlim = 255; if (daddr6) ip6->ip6_dst = *daddr6; else { ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL; ip6->ip6_dst.s6_addr16[1] = 0; ip6->ip6_dst.s6_addr32[1] = 0; ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE; ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3]; ip6->ip6_dst.s6_addr8[12] = 0xff; if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) goto bad; } if (nonce == NULL) { struct ifaddr *ifa = NULL; /* * RFC2461 7.2.2: * "If the source address of the packet prompting the * solicitation is the same as one of the addresses assigned * to the outgoing interface, that address SHOULD be placed * in the IP Source Address of the outgoing solicitation. * Otherwise, any one of the addresses assigned to the * interface should be used." * * We use the source address for the prompting packet * (saddr6), if saddr6 belongs to the outgoing interface. * Otherwise, we perform the source address selection as usual. */ if (saddr6 != NULL) ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6); if (ifa != NULL) { /* ip6_src set already. */ ip6->ip6_src = *saddr6; ifa_free(ifa); } else { int error; struct sockaddr_in6 dst_sa; struct in6_addr src_in; struct ifnet *oifp; bzero(&dst_sa, sizeof(dst_sa)); dst_sa.sin6_family = AF_INET6; dst_sa.sin6_len = sizeof(dst_sa); dst_sa.sin6_addr = ip6->ip6_dst; oifp = ifp; error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src_in); if (error) { char ip6buf[INET6_ADDRSTRLEN]; nd6log((LOG_DEBUG, "%s: source can't be " "determined: dst=%s, error=%d\n", __func__, ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error)); goto bad; } ip6->ip6_src = src_in; } } else { /* * Source address for DAD packet must always be IPv6 * unspecified address. (0::0) * We actually don't have to 0-clear the address (we did it * above), but we do so here explicitly to make the intention * clearer. */ bzero(&ip6->ip6_src, sizeof(ip6->ip6_src)); } nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1); nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT; nd_ns->nd_ns_code = 0; nd_ns->nd_ns_reserved = 0; nd_ns->nd_ns_target = *taddr6; in6_clearscope(&nd_ns->nd_ns_target); /* XXX */ /* * Add source link-layer address option. * * spec implementation * --- --- * DAD packet MUST NOT do not add the option * there's no link layer address: * impossible do not add the option * there's link layer address: * Multicast NS MUST add one add the option * Unicast NS SHOULD add one add the option */ if (nonce == NULL && (mac = nd6_ifptomac(ifp))) { int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); /* 8 byte alignments... */ optlen = (optlen + 7) & ~7; m->m_pkthdr.len += optlen; m->m_len += optlen; icmp6len += optlen; bzero((caddr_t)nd_opt, optlen); nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; nd_opt->nd_opt_len = optlen >> 3; bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); }
struct mbuf * ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt, int srcrt) { struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); struct sockaddr_in6 *dst; struct rtentry *rt; int error, type = 0, code = 0; boolean_t proxy = FALSE; struct mbuf *mcopy = NULL; struct ifnet *ifp, *rcvifp, *origifp; /* maybe unnecessary */ u_int32_t inzone, outzone, len; struct in6_addr src_in6, dst_in6; uint64_t curtime = net_uptime(); #if IPSEC struct secpolicy *sp = NULL; #endif unsigned int ifscope = IFSCOPE_NONE; #if PF struct pf_mtag *pf_mtag; #endif /* PF */ /* * In the prefix proxying case, the route to the proxied node normally * gets created by nd6_prproxy_ns_output(), as part of forwarding a * NS (NUD/AR) packet to the proxied node. In the event that such * packet did not arrive in time before the correct route gets created, * ip6_input() would have performed a rtalloc() which most likely will * create the wrong cloned route; this route points back to the same * interface as the inbound interface, since the parent non-scoped * prefix route points there. Therefore we check if that is the case * and perform the necessary fixup to get the correct route installed. */ if (!srcrt && nd6_prproxy && (rt = ip6forward_rt->ro_rt) != NULL && (rt->rt_flags & RTF_PROXY)) { nd6_proxy_find_fwdroute(m->m_pkthdr.rcvif, ip6forward_rt); if ((rt = ip6forward_rt->ro_rt) != NULL) ifscope = rt->rt_ifp->if_index; } #if PF pf_mtag = pf_find_mtag(m); if (pf_mtag != NULL && pf_mtag->pftag_rtableid != IFSCOPE_NONE) ifscope = pf_mtag->pftag_rtableid; /* * If the caller provides a route which is on a different interface * than the one specified for scoped forwarding, discard the route * and do a lookup below. */ if (ifscope != IFSCOPE_NONE && (rt = ip6forward_rt->ro_rt) != NULL) { RT_LOCK(rt); if (rt->rt_ifp->if_index != ifscope) { RT_UNLOCK(rt); ROUTE_RELEASE(ip6forward_rt); rt = NULL; } else { RT_UNLOCK(rt); } } #endif /* PF */ #if IPSEC /* * Check AH/ESP integrity. */ /* * Don't increment ip6s_cantforward because this is the check * before forwarding packet actually. */ if (ipsec_bypass == 0) { if (ipsec6_in_reject(m, NULL)) { IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio); m_freem(m); return (NULL); } } #endif /*IPSEC*/ /* * Do not forward packets to multicast destination. * Do not forward packets with unspecified source. It was discussed * in July 2000, on ipngwg mailing list. */ if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { ip6stat.ip6s_cantforward++; /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ if (ip6_log_time + ip6_log_interval < curtime) { ip6_log_time = curtime; log(LOG_DEBUG, "cannot forward " "from %s to %s nxt %d received on %s\n", ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif)); } m_freem(m); return (NULL); } if (ip6->ip6_hlim <= IPV6_HLIMDEC) { /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ icmp6_error(m, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT, 0); return (NULL); } /* * See if the destination is a proxied address, and if so pretend * that it's for us. This is mostly to handle NUD probes against * the proxied addresses. We filter for ICMPv6 here and will let * icmp6_input handle the rest. */ if (!srcrt && nd6_prproxy) { VERIFY(!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)); proxy = nd6_prproxy_isours(m, ip6, ip6forward_rt, ifscope); /* * Don't update hop limit while proxying; RFC 4389 4.1. * Also skip IPsec forwarding path processing as this * packet is not to be forwarded. */ if (proxy) goto skip_ipsec; } ip6->ip6_hlim -= IPV6_HLIMDEC; /* * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - * size of IPv6 + ICMPv6 headers) bytes of the packet in case * we need to generate an ICMP6 message to the src. * Thanks to M_EXT, in most cases copy will not occur. * * It is important to save it before IPsec processing as IPsec * processing may modify the mbuf. */ mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); #if IPSEC if (ipsec_bypass != 0) goto skip_ipsec; /* get a security policy for this packet */ sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING, &error); if (sp == NULL) { IPSEC_STAT_INCREMENT(ipsec6stat.out_inval); ip6stat.ip6s_cantforward++; if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return (NULL); } error = 0; /* check policy */ switch (sp->policy) { case IPSEC_POLICY_DISCARD: case IPSEC_POLICY_GENERATE: /* * This packet is just discarded. */ IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio); ip6stat.ip6s_cantforward++; key_freesp(sp, KEY_SADB_UNLOCKED); if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return (NULL); case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: /* no need to do IPsec. */ key_freesp(sp, KEY_SADB_UNLOCKED); goto skip_ipsec; case IPSEC_POLICY_IPSEC: if (sp->req == NULL) { /* XXX should be panic ? */ printf("ip6_forward: No IPsec request specified.\n"); ip6stat.ip6s_cantforward++; key_freesp(sp, KEY_SADB_UNLOCKED); if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return (NULL); } /* do IPsec */ break; case IPSEC_POLICY_ENTRUST: default: /* should be panic ?? */ printf("ip6_forward: Invalid policy found. %d\n", sp->policy); key_freesp(sp, KEY_SADB_UNLOCKED); goto skip_ipsec; } { struct ipsec_output_state state; /* * All the extension headers will become inaccessible * (since they can be encrypted). * Don't panic, we need no more updates to extension headers * on inner IPv6 packet (since they are now encapsulated). * * IPv6 [ESP|AH] IPv6 [extension headers] payload */ bzero(&state, sizeof(state)); state.m = m; state.dst = NULL; /* update at ipsec6_output_tunnel() */ error = ipsec6_output_tunnel(&state, sp, 0); key_freesp(sp, KEY_SADB_UNLOCKED); if (state.tunneled == 4) { ROUTE_RELEASE(&state.ro); return (NULL); /* packet is gone - sent over IPv4 */ } m = state.m; ROUTE_RELEASE(&state.ro); if (error) { /* mbuf is already reclaimed in ipsec6_output_tunnel. */ switch (error) { case EHOSTUNREACH: case ENETUNREACH: case EMSGSIZE: case ENOBUFS: case ENOMEM: break; default: printf("ip6_output (ipsec): error code %d\n", error); /* fall through */ case ENOENT: /* don't show these error codes to the user */ break; } ip6stat.ip6s_cantforward++; if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return (NULL); } } #endif /* IPSEC */ skip_ipsec: dst = (struct sockaddr_in6 *)&ip6forward_rt->ro_dst; if ((rt = ip6forward_rt->ro_rt) != NULL) { RT_LOCK(rt); /* Take an extra ref for ourselves */ RT_ADDREF_LOCKED(rt); } VERIFY(rt == NULL || rt == ip6forward_rt->ro_rt); if (!srcrt) { /* * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst */ if (ROUTE_UNUSABLE(ip6forward_rt)) { if (rt != NULL) { /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); } ROUTE_RELEASE(ip6forward_rt); /* this probably fails but give it a try again */ rtalloc_scoped_ign((struct route *)ip6forward_rt, RTF_PRCLONING, ifscope); if ((rt = ip6forward_rt->ro_rt) != NULL) { RT_LOCK(rt); /* Take an extra ref for ourselves */ RT_ADDREF_LOCKED(rt); } } if (rt == NULL) { ip6stat.ip6s_noroute++; in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); if (mcopy) icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE, 0); m_freem(m); return (NULL); } RT_LOCK_ASSERT_HELD(rt); } else if (ROUTE_UNUSABLE(ip6forward_rt) || !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) { if (rt != NULL) { /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); } ROUTE_RELEASE(ip6forward_rt); bzero(dst, sizeof(*dst)); dst->sin6_len = sizeof(struct sockaddr_in6); dst->sin6_family = AF_INET6; dst->sin6_addr = ip6->ip6_dst; rtalloc_scoped_ign((struct route *)ip6forward_rt, RTF_PRCLONING, ifscope); if ((rt = ip6forward_rt->ro_rt) == NULL) { ip6stat.ip6s_noroute++; in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); if (mcopy) icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE, 0); m_freem(m); return (NULL); } RT_LOCK(rt); /* Take an extra ref for ourselves */ RT_ADDREF_LOCKED(rt); } /* * Source scope check: if a packet can't be delivered to its * destination for the reason that the destination is beyond the scope * of the source address, discard the packet and return an icmp6 * destination unreachable error with Code 2 (beyond scope of source * address) unless we are proxying (source address is link local * for NUDs.) We use a local copy of ip6_src, since in6_setscope() * will possibly modify its first argument. * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] */ src_in6 = ip6->ip6_src; if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { /* XXX: this should not happen */ ip6stat.ip6s_cantforward++; ip6stat.ip6s_badscope++; m_freem(m); return (NULL); } if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { ip6stat.ip6s_cantforward++; ip6stat.ip6s_badscope++; m_freem(m); return (NULL); } if (inzone != outzone && !proxy) { ip6stat.ip6s_cantforward++; ip6stat.ip6s_badscope++; in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); if (ip6_log_time + ip6_log_interval < curtime) { ip6_log_time = curtime; log(LOG_DEBUG, "cannot forward " "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); } /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); if (mcopy) { icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_BEYONDSCOPE, 0); } m_freem(m); return (NULL); } /* * Destination scope check: if a packet is going to break the scope * zone of packet's destination address, discard it. This case should * usually be prevented by appropriately-configured routing table, but * we need an explicit check because we may mistakenly forward the * packet to a different zone by (e.g.) a default route. */ dst_in6 = ip6->ip6_dst; if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || inzone != outzone) { ip6stat.ip6s_cantforward++; ip6stat.ip6s_badscope++; m_freem(m); return (NULL); } if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) { in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); if (mcopy) { uint32_t mtu; #if IPSEC struct secpolicy *sp2; int ipsecerror; size_t ipsechdrsiz; #endif mtu = rt->rt_ifp->if_mtu; #if IPSEC /* * When we do IPsec tunnel ingress, we need to play * with the link value (decrement IPsec header size * from mtu value). The code is much simpler than v4 * case, as we have the outgoing interface for * encapsulated packet as "rt->rt_ifp". */ sp2 = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND, IP_FORWARDING, &ipsecerror); if (sp2) { ipsechdrsiz = ipsec6_hdrsiz(mcopy, IPSEC_DIR_OUTBOUND, NULL); if (ipsechdrsiz < mtu) mtu -= ipsechdrsiz; key_freesp(sp2, KEY_SADB_UNLOCKED); } /* * if mtu becomes less than minimum MTU, * tell minimum MTU (and I'll need to fragment it). */ if (mtu < IPV6_MMTU) mtu = IPV6_MMTU; #endif /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); } else { /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); } m_freem(m); return (NULL); } if (rt->rt_flags & RTF_GATEWAY) dst = (struct sockaddr_in6 *)(void *)rt->rt_gateway; /* * If we are to forward the packet using the same interface * as one we got the packet from, perhaps we should send a redirect * to sender to shortcut a hop. * Only send redirect if source is sending directly to us, * and if packet was not source routed (or has any options). * Also, don't send redirect if forwarding using a route * modified by a redirect. */ if (!proxy && ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { /* * If the incoming interface is equal to the outgoing * one, and the link attached to the interface is * point-to-point, then it will be highly probable * that a routing loop occurs. Thus, we immediately * drop the packet and send an ICMPv6 error message. * * type/code is based on suggestion by Rich Draves. * not sure if it is the best pick. */ RT_REMREF_LOCKED(rt); /* Release extra ref */ RT_UNLOCK(rt); icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, 0); m_freem(m); return (NULL); } type = ND_REDIRECT; } #if IPFW2 /* * Check with the firewall... */ if (ip6_fw_enable && ip6_fw_chk_ptr) { u_short port = 0; ifp = rt->rt_ifp; /* Drop the lock but retain the extra ref */ RT_UNLOCK(rt); /* If ipfw says divert, we have to just drop packet */ if (ip6_fw_chk_ptr(&ip6, ifp, &port, &m)) { m_freem(m); goto freecopy; } if (!m) { goto freecopy; } /* We still have the extra ref on rt */ RT_LOCK(rt); } #endif /* * Fake scoped addresses. Note that even link-local source or * destinaion can appear, if the originating node just sends the * packet to us (without address resolution for the destination). * Since both icmp6_error and icmp6_redirect_output fill the embedded * link identifiers, we can do this stuff after making a copy for * returning an error. */ if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { /* * See corresponding comments in ip6_output. * XXX: but is it possible that ip6_forward() sends a packet * to a loopback interface? I don't think so, and thus * I bark here. ([email protected]) * XXX: it is common to route invalid packets to loopback. * also, the codepath will be visited on use of ::1 in * rthdr. (itojun) */ #if 1 if ((0)) #else if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) #endif { printf("ip6_forward: outgoing interface is loopback. " "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); } /* we can just use rcvif in forwarding. */ origifp = rcvifp = m->m_pkthdr.rcvif; } else if (nd6_prproxy) { /* * In the prefix proxying case, we need to inform nd6_output() * about the inbound interface, so that any subsequent NS * packets generated by nd6_prproxy_ns_output() will not be * sent back to that same interface. */ origifp = rcvifp = m->m_pkthdr.rcvif; } else { rcvifp = m->m_pkthdr.rcvif; origifp = rt->rt_ifp; } /* * clear embedded scope identifiers if necessary. * in6_clearscope will touch the addresses only when necessary. */ in6_clearscope(&ip6->ip6_src); in6_clearscope(&ip6->ip6_dst); ifp = rt->rt_ifp; /* Drop the lock but retain the extra ref */ RT_UNLOCK(rt); /* * If this is to be processed locally, let ip6_input have it. */ if (proxy) { VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST); /* Release extra ref */ RT_REMREF(rt); if (mcopy != NULL) m_freem(mcopy); return (m); } #if PF /* Invoke outbound packet filter */ error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL); if (error != 0 || m == NULL) { if (m != NULL) { panic("%s: unexpected packet %p\n", __func__, m); /* NOTREACHED */ } /* Already freed by callee */ goto senderr; } ip6 = mtod(m, struct ip6_hdr *); #endif /* PF */ /* Mark this packet as being forwarded from another interface */ m->m_pkthdr.pkt_flags |= PKTF_FORWARDED; len = m_pktlen(m); error = nd6_output(ifp, origifp, m, dst, rt, NULL); if (error) { in6_ifstat_inc(ifp, ifs6_out_discard); ip6stat.ip6s_cantforward++; } else { /* * Increment stats on the source interface; the ones * for destination interface has been taken care of * during output above by virtue of PKTF_FORWARDED. */ rcvifp->if_fpackets++; rcvifp->if_fbytes += len; ip6stat.ip6s_forward++; in6_ifstat_inc(ifp, ifs6_out_forward); if (type) ip6stat.ip6s_redirectsent++; else { if (mcopy) { goto freecopy; } } } #if PF senderr: #endif /* PF */ if (mcopy == NULL) { /* Release extra ref */ RT_REMREF(rt); return (NULL); } switch (error) { case 0: #if 1 if (type == ND_REDIRECT) { icmp6_redirect_output(mcopy, rt); /* Release extra ref */ RT_REMREF(rt); return (NULL); } #endif goto freecopy; case EMSGSIZE: /* xxx MTU is constant in PPP? */ goto freecopy; case ENOBUFS: /* Tell source to slow down like source quench in IP? */ goto freecopy; case ENETUNREACH: /* shouldn't happen, checked above */ case EHOSTUNREACH: case ENETDOWN: case EHOSTDOWN: default: type = ICMP6_DST_UNREACH; code = ICMP6_DST_UNREACH_ADDR; break; } icmp6_error(mcopy, type, code, 0); /* Release extra ref */ RT_REMREF(rt); return (NULL); freecopy: m_freem(mcopy); /* Release extra ref */ RT_REMREF(rt); return (NULL); }