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, *origifp; /* maybe unnecessary */ u_int32_t inzone, outzone; struct in6_addr src_in6, dst_in6; #if IPSEC struct secpolicy *sp = NULL; #endif struct timeval timenow; unsigned int ifscope = IFSCOPE_NONE; #if PF struct pf_mtag *pf_mtag; #endif /* PF */ getmicrotime(&timenow); #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); rtfree(rt); rt = ip6forward_rt->ro_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 (should be handled * by ip6_mforward(). * 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 < timenow.tv_sec) { ip6_log_time = timenow.tv_sec; 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) return (NULL); /* packet is gone - sent over IPv4 */ m = state.m; if (state.ro.ro_rt) { rtfree(state.ro.ro_rt); state.ro.ro_rt = NULL; } 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); } } skip_ipsec: #endif /* 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); } if (!srcrt) { /* * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst */ if (rt == NULL || !(rt->rt_flags & RTF_UP) || rt->generation_id != route_generation) { if (rt != NULL) { /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); rtfree(rt); ip6forward_rt->ro_rt = NULL; } /* 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 (rt == NULL || !(rt->rt_flags & RTF_UP) || !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr) || rt->generation_id != route_generation) { if (rt != NULL) { /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); rtfree(rt); ip6forward_rt->ro_rt = NULL; } 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 < timenow.tv_sec) { ip6_log_time = timenow.tv_sec; 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 = 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); 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.aux_flags & MAUXF_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 */ error = nd6_output(ifp, origifp, m, dst, rt, NULL); if (error) { in6_ifstat_inc(ifp, ifs6_out_discard); ip6stat.ip6s_cantforward++; } else { 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); }
__private_extern__ errno_t arp_route_to_gateway_route(const struct sockaddr *net_dest, route_t hint0, route_t *out_route) { struct timeval timenow; route_t rt = hint0, hint = hint0; errno_t error = 0; *out_route = NULL; /* * Next hop determination. Because we may involve the gateway route * in addition to the original route, locking is rather complicated. * The general concept is that regardless of whether the route points * to the original route or to the gateway route, this routine takes * an extra reference on such a route. This extra reference will be * released at the end. * * Care must be taken to ensure that the "hint0" route never gets freed * via rtfree(), since the caller may have stored it inside a struct * route with a reference held for that placeholder. */ if (rt != NULL) { unsigned int ifindex; RT_LOCK_SPIN(rt); ifindex = rt->rt_ifp->if_index; RT_ADDREF_LOCKED(rt); if (!(rt->rt_flags & RTF_UP)) { RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); /* route is down, find a new one */ hint = rt = rtalloc1_scoped((struct sockaddr *) (size_t)net_dest, 1, 0, ifindex); if (hint != NULL) { RT_LOCK_SPIN(rt); ifindex = rt->rt_ifp->if_index; } else { senderr(EHOSTUNREACH); } } /* * We have a reference to "rt" by now; it will either * be released or freed at the end of this routine. */ RT_LOCK_ASSERT_HELD(rt); if (rt->rt_flags & RTF_GATEWAY) { struct rtentry *gwrt = rt->rt_gwroute; struct sockaddr_in gw; /* If there's no gateway rt, look it up */ if (gwrt == NULL) { gw = *((struct sockaddr_in *)rt->rt_gateway); RT_UNLOCK(rt); goto lookup; } /* Become a regular mutex */ RT_CONVERT_LOCK(rt); /* * Take gwrt's lock while holding route's lock; * this is okay since gwrt never points back * to "rt", so no lock ordering issues. */ RT_LOCK_SPIN(gwrt); if (!(gwrt->rt_flags & RTF_UP)) { struct rtentry *ogwrt; rt->rt_gwroute = NULL; RT_UNLOCK(gwrt); gw = *((struct sockaddr_in *)rt->rt_gateway); RT_UNLOCK(rt); rtfree(gwrt); lookup: gwrt = rtalloc1_scoped( (struct sockaddr *)&gw, 1, 0, ifindex); RT_LOCK(rt); /* * Bail out if the route is down, no route * to gateway, circular route, or if the * gateway portion of "rt" has changed. */ if (!(rt->rt_flags & RTF_UP) || gwrt == NULL || gwrt == rt || !equal(SA(&gw), rt->rt_gateway)) { if (gwrt == rt) { RT_REMREF_LOCKED(gwrt); gwrt = NULL; } RT_UNLOCK(rt); if (gwrt != NULL) rtfree(gwrt); senderr(EHOSTUNREACH); } /* Remove any existing gwrt */ ogwrt = rt->rt_gwroute; if ((rt->rt_gwroute = gwrt) != NULL) RT_ADDREF(gwrt); /* Clean up "rt" now while we can */ if (rt == hint0) { RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); } else { RT_UNLOCK(rt); rtfree(rt); } rt = gwrt; /* Now free the replaced gwrt */ if (ogwrt != NULL) rtfree(ogwrt); /* If still no route to gateway, bail out */ if (rt == NULL) senderr(EHOSTUNREACH); } else { RT_ADDREF_LOCKED(gwrt); RT_UNLOCK(gwrt); /* Clean up "rt" now while we can */ if (rt == hint0) { RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); } else { RT_UNLOCK(rt); rtfree(rt); } rt = gwrt; } /* rt == gwrt; if it is now down, give up */ RT_LOCK_SPIN(rt); if (!(rt->rt_flags & RTF_UP)) { RT_UNLOCK(rt); senderr(EHOSTUNREACH); } } if (rt->rt_flags & RTF_REJECT) { getmicrotime(&timenow); if (rt->rt_rmx.rmx_expire == 0 || timenow.tv_sec < rt->rt_rmx.rmx_expire) { RT_UNLOCK(rt); senderr(rt == hint ? EHOSTDOWN : EHOSTUNREACH); } } /* Become a regular mutex */ RT_CONVERT_LOCK(rt); /* Caller is responsible for cleaning up "rt" */ *out_route = rt; } return (0); bad: /* Clean up route (either it is "rt" or "gwrt") */ if (rt != NULL) { RT_LOCK_SPIN(rt); if (rt == hint0) { RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); } else { RT_UNLOCK(rt); rtfree(rt); } } return (error); }