Fix DecrementTTL option for packets destined to the local node.
[oweals/tinc.git] / src / route.c
index 664fed89fbd102edb543b637ed7148edaa2009d9..1b539aa1df60cbb6b89838ac3686069f88433428 100644 (file)
@@ -1,7 +1,7 @@
 /*
     route.c -- routing
     Copyright (C) 2000-2005 Ivo Timmermans,
-                  2000-2010 Guus Sliepen <guus@tinc-vpn.org>
+                  2000-2014 Guus Sliepen <guus@tinc-vpn.org>
 
     This program is free software; you can redistribute it and/or modify
     it under the terms of the GNU General Public License as published by
 #include "utils.h"
 
 rmode_t routing_mode = RMODE_ROUTER;
+fmode_t forwarding_mode = FMODE_INTERNAL;
+bmode_t broadcast_mode = BMODE_MST;
+bool decrement_ttl = false;
+bool directonly = false;
 bool priorityinheritance = false;
 int macexpire = 600;
 bool overwrite_mac = false;
@@ -48,7 +52,12 @@ static const size_t ip6_size = sizeof(struct ip6_hdr);
 static const size_t icmp6_size = sizeof(struct icmp6_hdr);
 static const size_t ns_size = sizeof(struct nd_neighbor_solicit);
 static const size_t opt_size = sizeof(struct nd_opt_hdr);
-#define max(a, b) ((a) > (b) ? (a) : (b))
+
+static bool do_decrement_ttl(node_t *source, vpn_packet_t *packet);
+
+#ifndef MAX
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+#endif
 
 /* RFC 1071 */
 
@@ -75,13 +84,14 @@ static bool ratelimit(int frequency) {
        static int count = 0;
        
        if(lasttime == now) {
-               if(++count > frequency)
+               if(count >= frequency)
                        return true;
        } else {
                lasttime = now;
                count = 0;
        }
 
+       count++;
        return false;
 }
 
@@ -94,19 +104,30 @@ static bool checklength(node_t *source, vpn_packet_t *packet, length_t length) {
 }
 
 static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *packet) {
-       if(!via || via == myself || !(via->options & OPTION_CLAMP_MSS))
+       if(!source || !via || !(via->options & OPTION_CLAMP_MSS))
                return;
 
+       uint16_t mtu = source->mtu;
+       if(via != myself && via->mtu < mtu)
+               mtu = via->mtu;
+
        /* Find TCP header */
-       int start = 0;
+       int start = ether_size;
        uint16_t type = packet->data[12] << 8 | packet->data[13];
 
-       if(type == ETH_P_IP && packet->data[23] == 6)
-               start = 14 + (packet->data[14] & 0xf) * 4;
-       else if(type == ETH_P_IPV6 && packet->data[20] == 6)
-               start = 14 + 40;
+       if(type == ETH_P_8021Q) {
+               start += 4;
+               type = packet->data[16] << 8 | packet->data[17];
+       }
+
+       if(type == ETH_P_IP && packet->data[start + 9] == 6)
+               start += (packet->data[start] & 0xf) * 4;
+       else if(type == ETH_P_IPV6 && packet->data[start + 6] == 6)
+               start += 40;
+       else
+               return;
 
-       if(!start || packet->len <= start + 20)
+       if(packet->len <= start + 20)
                return;
 
        /* Use data offset field to calculate length of options field */
@@ -140,8 +161,8 @@ static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *pac
 
                /* Found it */
                uint16_t oldmss = packet->data[start + 22 + i] << 8 | packet->data[start + 23 + i];
-               uint16_t newmss = via->mtu - start - 20;
-               uint16_t csum = packet->data[start + 16] << 8 | packet->data[start + 17];
+               uint16_t newmss = mtu - start - 20;
+               uint32_t csum = packet->data[start + 16] << 8 | packet->data[start + 17];
 
                if(oldmss <= newmss)
                        break;
@@ -152,11 +173,13 @@ static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *pac
                packet->data[start + 22 + i] = newmss >> 8;
                packet->data[start + 23 + i] = newmss & 0xff;
                csum ^= 0xffff;
-               csum -= oldmss;
+               csum += oldmss ^ 0xffff;
                csum += newmss;
+               csum = (csum & 0xffff) + (csum >> 16);
+               csum += csum >> 16;
                csum ^= 0xffff;
                packet->data[start + 16] = csum >> 8;
-               packet->data[start + 17] = csum & 0xff;
+               packet->data[start + 17] = csum;
                break;
        }
 }
@@ -178,7 +201,7 @@ static void learn_mac(mac_t *address) {
        /* If we don't know this MAC address yet, store it */
 
        if(!subnet) {
-               ifdebug(TRAFFIC) logger(LOG_INFO, "Learned new MAC address %hx:%hx:%hx:%hx:%hx:%hx",
+               ifdebug(TRAFFIC) logger(LOG_INFO, "Learned new MAC address %x:%x:%x:%x:%x:%x",
                                   address->x[0], address->x[1], address->x[2], address->x[3],
                                   address->x[4], address->x[5]);
 
@@ -211,7 +234,7 @@ void age_subnets(void) {
        for(node = myself->subnet_tree->head; node; node = next) {
                next = node->next;
                s = node->data;
-               if(s->expires && s->expires < now) {
+               if(s->expires && s->expires <= now) {
                        ifdebug(TRAFFIC) {
                                char netstr[MAXNETSTR];
                                if(net2str(netstr, sizeof netstr, s))
@@ -230,9 +253,17 @@ void age_subnets(void) {
        }
 }
 
+static void broadcast_packet_helper(node_t *source, vpn_packet_t *packet) {
+       if(decrement_ttl && source != myself)
+               if(!do_decrement_ttl(source, packet))
+                       return;
+
+       broadcast_packet(source, packet);
+}
+
 /* RFC 792 */
 
-static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, uint8_t type, uint8_t code) {
+static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
        struct ip ip = {0};
        struct icmp icmp = {0};
        
@@ -256,6 +287,27 @@ static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, uint8_t
        ip_src = ip.ip_src;
        ip_dst = ip.ip_dst;
 
+       /* Try to reply with an IP address assigned to the local machine */
+
+       if (type == ICMP_TIME_EXCEEDED && code == ICMP_EXC_TTL) {
+               int sockfd = socket(AF_INET, SOCK_DGRAM, 0);
+               if (sockfd != -1) {
+                       struct sockaddr_in addr;
+                       memset(&addr, 0, sizeof(addr));
+                       addr.sin_family = AF_INET;
+                       addr.sin_addr = ip.ip_src;
+                       if (!connect(sockfd, (const struct sockaddr*) &addr, sizeof(addr))) {
+                               memset(&addr, 0, sizeof(addr));
+                               addr.sin_family = AF_INET;
+                               socklen_t addrlen = sizeof(addr);
+                               if (!getsockname(sockfd, (struct sockaddr*) &addr, &addrlen) && addrlen <= sizeof(addr)) {
+                                       ip_dst = addr.sin_addr;
+                               }
+                       }
+                       close(sockfd);
+               }
+       }
+
        oldlen = packet->len - ether_size;
 
        if(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
@@ -305,7 +357,7 @@ static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, uint8_t
 
 /* RFC 791 */
 
-static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet) {
+static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet, length_t ether_size) {
        struct ip ip;
        vpn_packet_t fragment;
        int len, maxlen, todo;
@@ -321,7 +373,7 @@ static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet) {
        todo = ntohs(ip.ip_len) - ip_size;
 
        if(ether_size + ip_size + todo != packet->len) {
-               ifdebug(TRAFFIC) logger(LOG_WARNING, "Length of packet (%d) doesn't match length in IPv4 header (%zd)", packet->len, ether_size + ip_size + todo);
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Length of packet (%d) doesn't match length in IPv4 header (%d)", packet->len, (int)(ether_size + ip_size + todo));
                return;
        }
 
@@ -369,7 +421,7 @@ static void route_ipv4_unicast(node_t *source, vpn_packet_t *packet) {
                                dest.x[2],
                                dest.x[3]);
 
-               route_ipv4_unreachable(source, packet, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
+               route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
                return;
        }
        
@@ -379,20 +431,35 @@ static void route_ipv4_unicast(node_t *source, vpn_packet_t *packet) {
        }
 
        if(!subnet->owner->status.reachable)
-               route_ipv4_unreachable(source, packet, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
+               return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
+
+       if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
+               return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
+
+       if(decrement_ttl && source != myself && subnet->owner != myself)
+               if(!do_decrement_ttl(source, packet))
+                       return;
 
        if(priorityinheritance)
                packet->priority = packet->data[15];
 
        via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
+
+       if(via == source) {
+               ifdebug(TRAFFIC) logger(LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
+               return;
+       }
        
-       if(via && packet->len > max(via->mtu, 590) && via != myself) {
+       if(directonly && subnet->owner != via)
+               return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
+
+       if(via && packet->len > MAX(via->mtu, 590) && via != myself) {
                ifdebug(TRAFFIC) logger(LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
                if(packet->data[20] & 0x40) {
-                       packet->len = max(via->mtu, 590);
-                       route_ipv4_unreachable(source, packet, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
+                       packet->len = MAX(via->mtu, 590);
+                       route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
                } else {
-                       fragment_ipv4_packet(via, packet);
+                       fragment_ipv4_packet(via, packet, ether_size);
                }
 
                return;
@@ -407,19 +474,19 @@ static void route_ipv4(node_t *source, vpn_packet_t *packet) {
        if(!checklength(source, packet, ether_size + ip_size))
                return;
 
-       if(((packet->data[30] & 0xf0) == 0xe0) || (
+       if(broadcast_mode && (((packet->data[30] & 0xf0) == 0xe0) || (
                        packet->data[30] == 255 &&
                        packet->data[31] == 255 &&
                        packet->data[32] == 255 &&
-                       packet->data[33] == 255))
-               broadcast_packet(source, packet);
+                       packet->data[33] == 255)))
+               broadcast_packet_helper(source, packet);
        else
                route_ipv4_unicast(source, packet);
 }
 
 /* RFC 2463 */
 
-static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, uint8_t type, uint8_t code) {
+static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
        struct ip6_hdr ip6;
        struct icmp6_hdr icmp6 = {0};
        uint16_t checksum;      
@@ -447,6 +514,27 @@ static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, uint8_t
        pseudo.ip6_src = ip6.ip6_dst;
        pseudo.ip6_dst = ip6.ip6_src;
 
+       /* Try to reply with an IP address assigned to the local machine */
+
+       if (type == ICMP6_TIME_EXCEEDED && code == ICMP6_TIME_EXCEED_TRANSIT) {
+               int sockfd = socket(AF_INET6, SOCK_DGRAM, 0);
+               if (sockfd != -1) {
+                       struct sockaddr_in6 addr;
+                       memset(&addr, 0, sizeof(addr));
+                       addr.sin6_family = AF_INET6;
+                       addr.sin6_addr = ip6.ip6_src;
+                       if (!connect(sockfd, (const struct sockaddr*) &addr, sizeof(addr))) {
+                               memset(&addr, 0, sizeof(addr));
+                               addr.sin6_family = AF_INET6;
+                               socklen_t addrlen = sizeof(addr);
+                               if (!getsockname(sockfd, (struct sockaddr*) &addr, &addrlen) && addrlen <= sizeof(addr)) {
+                                       pseudo.ip6_src = addr.sin6_addr;
+                               }
+                       }
+                       close(sockfd);
+               }
+       }
+
        pseudo.length = packet->len - ether_size;
 
        if(type == ICMP6_PACKET_TOO_BIG)
@@ -517,7 +605,7 @@ static void route_ipv6_unicast(node_t *source, vpn_packet_t *packet) {
                                ntohs(dest.x[6]),
                                ntohs(dest.x[7]));
 
-               route_ipv6_unreachable(source, packet, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
+               route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
                return;
        }
 
@@ -527,14 +615,29 @@ static void route_ipv6_unicast(node_t *source, vpn_packet_t *packet) {
        }
 
        if(!subnet->owner->status.reachable)
-               route_ipv6_unreachable(source, packet, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
+               return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
+
+       if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
+               return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
+
+       if(decrement_ttl && source != myself && subnet->owner != myself)
+               if(!do_decrement_ttl(source, packet))
+                       return;
 
        via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
        
-       if(via && packet->len > max(via->mtu, 1294) && via != myself) {
+       if(via == source) {
+               ifdebug(TRAFFIC) logger(LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
+               return;
+       }
+       
+       if(directonly && subnet->owner != via)
+               return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
+
+       if(via && packet->len > MAX(via->mtu, 1294) && via != myself) {
                ifdebug(TRAFFIC) logger(LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
-               packet->len = max(via->mtu, 1294);
-               route_ipv6_unreachable(source, packet, ICMP6_PACKET_TOO_BIG, 0);
+               packet->len = MAX(via->mtu, 1294);
+               route_ipv6_unreachable(source, packet, ether_size, ICMP6_PACKET_TOO_BIG, 0);
                return;
        }
 
@@ -637,6 +740,10 @@ static void route_neighborsol(node_t *source, vpn_packet_t *packet) {
        if(subnet->owner == myself)
                return;                                 /* silently ignore */
 
+       if(decrement_ttl)
+               if(!do_decrement_ttl(source, packet))
+                       return;
+
        /* Create neighbor advertation reply */
 
        memcpy(packet->data, packet->data + ETH_ALEN, ETH_ALEN);        /* copy destination address */
@@ -693,8 +800,8 @@ static void route_ipv6(node_t *source, vpn_packet_t *packet) {
                return;
        }
 
-       if(packet->data[38] == 255)
-               broadcast_packet(source, packet);
+       if(broadcast_mode && packet->data[38] == 255)
+               broadcast_packet_helper(source, packet);
        else
                route_ipv6_unicast(source, packet);
 }
@@ -747,6 +854,10 @@ static void route_arp(node_t *source, vpn_packet_t *packet) {
        if(subnet->owner == myself)
                return;                                 /* silently ignore */
 
+       if(decrement_ttl)
+               if(!do_decrement_ttl(source, packet))
+                       return;
+
        memcpy(packet->data, packet->data + ETH_ALEN, ETH_ALEN);        /* copy destination address */
        packet->data[ETH_ALEN * 2 - 1] ^= 0xFF; /* mangle source address so it looks like it's not from us */
 
@@ -783,7 +894,7 @@ static void route_mac(node_t *source, vpn_packet_t *packet) {
        subnet = lookup_subnet_mac(NULL, &dest);
 
        if(!subnet) {
-               broadcast_packet(source, packet);
+               broadcast_packet_helper(source, packet);
                return;
        }
 
@@ -792,24 +903,45 @@ static void route_mac(node_t *source, vpn_packet_t *packet) {
                return;
        }
 
+       if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
+               return;
+
+       if(decrement_ttl && source != myself && subnet->owner != myself)
+               if(!do_decrement_ttl(source, packet))
+                       return;
+
+       uint16_t type = packet->data[12] << 8 | packet->data[13];
+
+       if(priorityinheritance && type == ETH_P_IP && packet->len >= ether_size + ip_size)
+               packet->priority = packet->data[15];
+
        // Handle packets larger than PMTU
 
        node_t *via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
+
+       if(directonly && subnet->owner != via)
+               return;
        
        if(via && packet->len > via->mtu && via != myself) {
                ifdebug(TRAFFIC) logger(LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
-               uint16_t type = packet->data[12] << 8 | packet->data[13];
-               if(type == ETH_P_IP && packet->len > 590) {
-                       if(packet->data[20] & 0x40) {
+               length_t ethlen = 14;
+
+               if(type == ETH_P_8021Q) {
+                       type = packet->data[16] << 8 | packet->data[17];
+                       ethlen += 4;
+               }
+
+               if(type == ETH_P_IP && packet->len > 576 + ethlen) {
+                       if(packet->data[6 + ethlen] & 0x40) {
                                packet->len = via->mtu;
-                               route_ipv4_unreachable(source, packet, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
+                               route_ipv4_unreachable(source, packet, ethlen, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
                        } else {
-                               fragment_ipv4_packet(via, packet);
+                               fragment_ipv4_packet(via, packet, ethlen);
                        }
                        return;
-               } else if(type == ETH_P_IPV6 && packet->len > 1294) {
+               } else if(type == ETH_P_IPV6 && packet->len > 1280 + ethlen) {
                        packet->len = via->mtu;
-                       route_ipv6_unreachable(source, packet, ICMP6_PACKET_TOO_BIG, 0);
+                       route_ipv6_unreachable(source, packet, ethlen, ICMP6_PACKET_TOO_BIG, 0);
                        return;
                }
        }
@@ -819,7 +951,64 @@ static void route_mac(node_t *source, vpn_packet_t *packet) {
        send_packet(subnet->owner, packet);
 }
 
+static bool do_decrement_ttl(node_t *source, vpn_packet_t *packet) {
+       uint16_t type = packet->data[12] << 8 | packet->data[13];
+       length_t ethlen = ether_size;
+
+       if(type == ETH_P_8021Q) {
+               type = packet->data[16] << 8 | packet->data[17];
+               ethlen += 4;
+       }
+
+       switch (type) {
+               case ETH_P_IP:
+                       if(!checklength(source, packet, ethlen + ip_size))
+                               return false;
+
+                       if(packet->data[ethlen + 8] <= 1) {
+                               if(packet->data[ethlen + 11] != IPPROTO_ICMP || packet->data[ethlen + 32] != ICMP_TIME_EXCEEDED)
+                                       route_ipv4_unreachable(source, packet, ethlen, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL);
+                               return false;
+                       }
+
+                       uint16_t old = packet->data[ethlen + 8] << 8 | packet->data[ethlen + 9];
+                       packet->data[ethlen + 8]--;
+                       uint16_t new = packet->data[ethlen + 8] << 8 | packet->data[ethlen + 9];
+
+                       uint32_t checksum = packet->data[ethlen + 10] << 8 | packet->data[ethlen + 11];
+                       checksum += old + (~new & 0xFFFF);
+                       while(checksum >> 16)
+                               checksum = (checksum & 0xFFFF) + (checksum >> 16);
+                       packet->data[ethlen + 10] = checksum >> 8;
+                       packet->data[ethlen + 11] = checksum & 0xff;
+
+                       return true;
+
+               case ETH_P_IPV6:
+                       if(!checklength(source, packet, ethlen + ip6_size))
+                               return false;
+
+                       if(packet->data[ethlen + 7] <= 1) {
+                               if(packet->data[ethlen + 6] != IPPROTO_ICMPV6 || packet->data[ethlen + 40] != ICMP6_TIME_EXCEEDED)
+                                       route_ipv6_unreachable(source, packet, ethlen, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT);
+                               return false;
+                       }
+
+                       packet->data[ethlen + 7]--;
+
+                       return true;
+
+               default:
+                       return true;
+       }
+}
+
 void route(node_t *source, vpn_packet_t *packet) {
+       if(forwarding_mode == FMODE_KERNEL && source != myself) {
+               send_packet(myself, packet);
+               return;
+       }
+
        if(!checklength(source, packet, ether_size))
                return;
 
@@ -853,7 +1042,7 @@ void route(node_t *source, vpn_packet_t *packet) {
                        break;
 
                case RMODE_HUB:
-                       broadcast_packet(source, packet);
+                       broadcast_packet_helper(source, packet);
                        break;
        }
 }