Fix DecrementTTL option.
[oweals/tinc.git] / src / route.c
index 2da781e5a8adcce7a10eda3089be6abd07e1e42d..a1254d80832491ee947e505c2c2050a9cd33e0fa 100644 (file)
@@ -1,7 +1,7 @@
 /*
     route.c -- routing
     Copyright (C) 2000-2005 Ivo Timmermans,
-                  2000-2009 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;
@@ -49,6 +53,10 @@ 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);
 
+#ifndef MAX
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+#endif
+
 /* RFC 1071 */
 
 static uint16_t inet_checksum(void *data, int len, uint16_t prevsum) {
@@ -74,13 +82,14 @@ 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;
 }
 
@@ -92,6 +101,85 @@ static bool checklength(node_t *source, vpn_packet_t *packet, length_t length) {
                return true;
 }
 
+static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *packet) {
+       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 = ether_size;
+       uint16_t type = packet->data[12] << 8 | packet->data[13];
+
+       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(packet->len <= start + 20)
+               return;
+
+       /* Use data offset field to calculate length of options field */
+       int len = ((packet->data[start + 12] >> 4) - 5) * 4;
+
+       if(packet->len < start + 20 + len)
+               return;
+
+       /* Search for MSS option header */
+       for(int i = 0; i < len;) {
+               if(packet->data[start + 20 + i] == 0)
+                       break;
+
+               if(packet->data[start + 20 + i] == 1) {
+                       i++;
+                       continue;
+               }
+
+               if(i > len - 2 || i > len - packet->data[start + 21 + i])
+                       break;
+
+               if(packet->data[start + 20 + i] != 2) {
+                       if(packet->data[start + 21 + i] < 2)
+                               break;
+                       i += packet->data[start + 21 + i];
+                       continue;
+               }
+
+               if(packet->data[start + 21] != 4)
+                       break;
+
+               /* Found it */
+               uint16_t oldmss = packet->data[start + 22 + i] << 8 | packet->data[start + 23 + i];
+               uint16_t newmss = mtu - start - 20;
+               uint16_t csum = packet->data[start + 16] << 8 | packet->data[start + 17];
+
+               if(oldmss <= newmss)
+                       break;
+               
+               ifdebug(TRAFFIC) logger(LOG_INFO, "Clamping MSS of packet from %s to %s to %d", source->name, via->name, newmss);
+
+               /* Update the MSS value and the checksum */
+               packet->data[start + 22 + i] = newmss >> 8;
+               packet->data[start + 23 + i] = newmss & 0xff;
+               csum ^= 0xffff;
+               csum -= oldmss;
+               csum += newmss;
+               csum ^= 0xffff;
+               packet->data[start + 16] = csum >> 8;
+               packet->data[start + 17] = csum & 0xff;
+               break;
+       }
+}
+
 static void swap_mac_addresses(vpn_packet_t *packet) {
        mac_t tmp;
        memcpy(&tmp, &packet->data[0], sizeof tmp);
@@ -104,12 +192,12 @@ static void learn_mac(mac_t *address) {
        avl_node_t *node;
        connection_t *c;
 
-       subnet = lookup_subnet_mac(address);
+       subnet = lookup_subnet_mac(myself, 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]);
 
@@ -117,7 +205,9 @@ static void learn_mac(mac_t *address) {
                subnet->type = SUBNET_MAC;
                subnet->expires = now + macexpire;
                subnet->net.mac.address = *address;
+               subnet->weight = 10;
                subnet_add(myself, subnet);
+               subnet_update(myself, subnet, true);
 
                /* And tell all other tinc daemons it's our MAC */
 
@@ -140,7 +230,7 @@ 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))
@@ -153,6 +243,7 @@ void age_subnets(void) {
                                        send_del_subnet(c, s);
                        }
 
+                       subnet_update(myself, s, false);
                        subnet_del(myself, s);
                }
        }
@@ -160,7 +251,7 @@ void age_subnets(void) {
 
 /* 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};
        
@@ -233,7 +324,7 @@ 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;
@@ -249,7 +340,7 @@ 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;
        }
 
@@ -297,7 +388,7 @@ 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;
        }
        
@@ -307,25 +398,38 @@ 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(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 > via->mtu && 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 = via->mtu;
-                       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;
        }
 
+       clamp_mss(source, via, packet);
        send_packet(subnet->owner, packet);
 }
 
@@ -333,11 +437,11 @@ 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))
+                       packet->data[33] == 255)))
                broadcast_packet(source, packet);
        else
                route_ipv4_unicast(source, packet);
@@ -345,7 +449,7 @@ static void route_ipv4(node_t *source, vpn_packet_t *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;      
@@ -443,7 +547,7 @@ 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;
        }
 
@@ -453,17 +557,30 @@ 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);
 
        via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
        
-       if(via && packet->len > via->mtu && 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 = via->mtu;
-               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;
        }
 
+       clamp_mss(source, via, packet);
        send_packet(subnet->owner, packet);
 }
 
@@ -617,7 +734,7 @@ static void route_ipv6(node_t *source, vpn_packet_t *packet) {
                return;
        }
 
-       if(packet->data[38] == 255)
+       if(broadcast_mode && packet->data[38] == 255)
                broadcast_packet(source, packet);
        else
                route_ipv6_unicast(source, packet);
@@ -704,7 +821,7 @@ static void route_mac(node_t *source, vpn_packet_t *packet) {
        /* Lookup destination address */
 
        memcpy(&dest, &packet->data[0], sizeof dest);
-       subnet = lookup_subnet_mac(&dest);
+       subnet = lookup_subnet_mac(NULL, &dest);
 
        if(!subnet) {
                broadcast_packet(source, packet);
@@ -716,35 +833,115 @@ static void route_mac(node_t *source, vpn_packet_t *packet) {
                return;
        }
 
+       if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
+               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) {
-                       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) {
+               } 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;
                }
        }
 
+       clamp_mss(source, via, 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;
 
+       if(decrement_ttl && source != myself)
+               if(!do_decrement_ttl(source, packet))
+                       return;
+
        switch (routing_mode) {
                case RMODE_ROUTER:
                        {