Also clamp MSS of TCP over IPv6 packets.
[tinc] / src / route.c
index 7dd7635..77bec04 100644 (file)
@@ -1,7 +1,7 @@
 /*
     route.c -- routing
-    Copyright (C) 2000-2003 Ivo Timmermans <ivo@o2w.nl>,
-                  2000-2003 Guus Sliepen <guus@sliepen.eu.org>
+    Copyright (C) 2000-2005 Ivo Timmermans,
+                  2000-2009 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
     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
     GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License
-    along with this program; if not, write to the Free Software
-    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
-
-    $Id: route.c,v 1.1.2.62 2003/07/22 20:55:20 guus Exp $
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 */
 
 #include "system.h"
 
-#ifdef HAVE_NET_ETHERNET_H
-#include <net/ethernet.h>
-#endif
-#ifdef HAVE_NET_IF_ARP_H
-#include <net/if_arp.h>
-#endif
-#ifdef HAVE_NETINET_IP_ICMP_H
-#include <netinet/ip_icmp.h>
-#endif
-#ifdef HAVE_NETINET_IP6_H
-#include <netinet/ip6.h>
-#endif
-#ifdef HAVE_NETINET_ICMP6_H
-#include <netinet/icmp6.h>
-#endif
-#ifdef HAVE_NETINET_IF_ETHER_H
-#include <netinet/if_ether.h>
-#endif
-
 #include "avl_tree.h"
 #include "connection.h"
-#include "device.h"
 #include "ethernet.h"
 #include "ipv4.h"
 #include "ipv6.h"
@@ -60,10 +38,21 @@ int macexpire = 600;
 bool overwrite_mac = false;
 mac_t mymac = {{0xFE, 0xFD, 0, 0, 0, 0}};
 
+/* Sizes of various headers */
+
+static const size_t ether_size = sizeof(struct ether_header);
+static const size_t arp_size = sizeof(struct ether_arp);
+static const size_t ip_size = sizeof(struct ip);
+static const size_t icmp_size = sizeof(struct icmp) - sizeof(struct ip);
+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))
+
 /* RFC 1071 */
 
-static uint16_t inet_checksum(void *data, int len, uint16_t prevsum)
-{
+static uint16_t inet_checksum(void *data, int len, uint16_t prevsum) {
        uint16_t *p = data;
        uint32_t checksum = prevsum ^ 0xFFFF;
 
@@ -73,7 +62,7 @@ static uint16_t inet_checksum(void *data, int len, uint16_t prevsum)
        }
        
        if(len)
-               checksum += *(unsigned char *)p;
+               checksum += *(uint8_t *)p;
 
        while(checksum >> 16)
                checksum = (checksum & 0xFFFF) + (checksum >> 16);
@@ -81,69 +70,152 @@ static uint16_t inet_checksum(void *data, int len, uint16_t prevsum)
        return ~checksum;
 }
 
-static bool ratelimit(void) {
+static bool ratelimit(int frequency) {
        static time_t lasttime = 0;
+       static int count = 0;
        
-       if(lasttime == now)
-               return true;
+       if(lasttime == now) {
+               if(++count > frequency)
+                       return true;
+       } else {
+               lasttime = now;
+               count = 0;
+       }
 
-       lasttime = now;
        return false;
 }
+
+static bool checklength(node_t *source, vpn_packet_t *packet, length_t length) {
+       if(packet->len < length) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Got too short packet from %s (%s)", source->name, source->hostname);
+               return false;
+       } else
+               return true;
+}
+
+static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *packet) {
+       if(!via || via == myself)
+               return;
+
+       /* Find TCP header */
+       int start = 0;
+       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(!start || packet->len <= start + 20)
+               return;
+
+       /* Use data offset field to calculate length of options field */
+       int len = ((packet->data[start + 12] >> 4) - 5) * 4;
+
+       /* 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 = via->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);
+       memcpy(&packet->data[0], &packet->data[6], sizeof tmp);
+       memcpy(&packet->data[6], &tmp, sizeof tmp);
+}
        
-static void learn_mac(mac_t *address)
-{
+static void learn_mac(mac_t *address) {
        subnet_t *subnet;
        avl_node_t *node;
        connection_t *c;
 
-       cp();
-
-       subnet = lookup_subnet_mac(address);
+       subnet = lookup_subnet_mac(myself, address);
 
        /* If we don't know this MAC address yet, store it */
 
-       if(!subnet || subnet->owner != myself) {
-               ifdebug(TRAFFIC) logger(LOG_INFO, _("Learned new MAC address %hx:%hx:%hx:%hx:%hx:%hx"),
+       if(!subnet) {
+               ifdebug(TRAFFIC) logger(LOG_INFO, "Learned new MAC address %hx:%hx:%hx:%hx:%hx:%hx",
                                   address->x[0], address->x[1], address->x[2], address->x[3],
                                   address->x[4], address->x[5]);
 
                subnet = new_subnet();
                subnet->type = SUBNET_MAC;
-               memcpy(&subnet->net.mac.address, address, sizeof(mac_t));
+               subnet->expires = now + macexpire;
+               subnet->net.mac.address = *address;
+               subnet->weight = 10;
                subnet_add(myself, subnet);
 
                /* And tell all other tinc daemons it's our MAC */
 
                for(node = connection_tree->head; node; node = node->next) {
-                       c = (connection_t *) node->data;
+                       c = node->data;
                        if(c->status.active)
                                send_add_subnet(c, subnet);
                }
        }
 
-       subnet->net.mac.lastseen = now;
+       if(subnet->expires)
+               subnet->expires = now + macexpire;
 }
 
-void age_mac(void)
-{
+void age_subnets(void) {
        subnet_t *s;
        connection_t *c;
        avl_node_t *node, *next, *node2;
 
-       cp();
-
        for(node = myself->subnet_tree->head; node; node = next) {
                next = node->next;
-               s = (subnet_t *) node->data;
-               if(s->type == SUBNET_MAC && s->net.mac.lastseen && s->net.mac.lastseen + macexpire < now) {
-                       ifdebug(TRAFFIC) logger(LOG_INFO, _("MAC address %hx:%hx:%hx:%hx:%hx:%hx expired"),
-                                          s->net.mac.address.x[0], s->net.mac.address.x[1],
-                                          s->net.mac.address.x[2], s->net.mac.address.x[3],
-                                          s->net.mac.address.x[4], s->net.mac.address.x[5]);
+               s = node->data;
+               if(s->expires && s->expires < now) {
+                       ifdebug(TRAFFIC) {
+                               char netstr[MAXNETSTR];
+                               if(net2str(netstr, sizeof netstr, s))
+                                       logger(LOG_INFO, "Subnet %s expired", netstr);
+                       }
 
                        for(node2 = connection_tree->head; node2; node2 = node2->next) {
-                               c = (connection_t *) node2->data;
+                               c = node2->data;
                                if(c->status.active)
                                        send_del_subnet(c, s);
                        }
@@ -153,119 +225,198 @@ void age_mac(void)
        }
 }
 
-static node_t *route_mac(vpn_packet_t *packet)
-{
-       subnet_t *subnet;
-
-       cp();
-
-       /* Learn source address */
-
-       learn_mac((mac_t *)(&packet->data[6]));
-
-       /* Lookup destination address */
-
-       subnet = lookup_subnet_mac((mac_t *)(&packet->data[0]));
-
-       if(subnet)
-               return subnet->owner;
-       else
-               return NULL;
-}
-
 /* RFC 792 */
 
-static void route_ipv4_unreachable(vpn_packet_t *packet, uint8_t code)
-{
-       struct ip *hdr;
-       struct icmp *icmp;
+static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, uint8_t type, uint8_t code) {
+       struct ip ip = {0};
+       struct icmp icmp = {0};
        
        struct in_addr ip_src;
        struct in_addr ip_dst;
        uint32_t oldlen;
 
-       if(ratelimit())
+       if(ratelimit(3))
                return;
        
-       cp();
+       /* Swap Ethernet source and destination addresses */
+
+       swap_mac_addresses(packet);
+
+       /* Copy headers from packet into properly aligned structs on the stack */
 
-       hdr = (struct ip *)(packet->data + 14);
-       icmp = (struct icmp *)(packet->data + 14 + 20);
+       memcpy(&ip, packet->data + ether_size, ip_size);
 
        /* Remember original source and destination */
-               
-       memcpy(&ip_src, &hdr->ip_src, 4);
-       memcpy(&ip_dst, &hdr->ip_dst, 4);
-       oldlen = packet->len - 14;
        
-       if(oldlen >= IP_MSS - sizeof(*hdr) - sizeof(*icmp))
-               oldlen = IP_MSS - sizeof(*hdr) - sizeof(*icmp);
+       ip_src = ip.ip_src;
+       ip_dst = ip.ip_dst;
+
+       oldlen = packet->len - ether_size;
+
+       if(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
+               icmp.icmp_nextmtu = htons(packet->len - ether_size);
+
+       if(oldlen >= IP_MSS - ip_size - icmp_size)
+               oldlen = IP_MSS - ip_size - icmp_size;
        
        /* Copy first part of original contents to ICMP message */
        
-       memmove(&icmp->icmp_ip, hdr, oldlen);
+       memmove(packet->data + ether_size + ip_size + icmp_size, packet->data + ether_size, oldlen);
 
        /* Fill in IPv4 header */
        
-       hdr->ip_v = 4;
-       hdr->ip_hl = sizeof(*hdr) / 4;
-       hdr->ip_tos = 0;
-       hdr->ip_len = htons(20 + 8 + oldlen);
-       hdr->ip_id = 0;
-       hdr->ip_off = 0;
-       hdr->ip_ttl = 255;
-       hdr->ip_p = IPPROTO_ICMP;
-       hdr->ip_sum = 0;
-       memcpy(&hdr->ip_src, &ip_dst, 4);
-       memcpy(&hdr->ip_dst, &ip_src, 4);
-
-       hdr->ip_sum = inet_checksum(hdr, 20, ~0);
+       ip.ip_v = 4;
+       ip.ip_hl = ip_size / 4;
+       ip.ip_tos = 0;
+       ip.ip_len = htons(ip_size + icmp_size + oldlen);
+       ip.ip_id = 0;
+       ip.ip_off = 0;
+       ip.ip_ttl = 255;
+       ip.ip_p = IPPROTO_ICMP;
+       ip.ip_sum = 0;
+       ip.ip_src = ip_dst;
+       ip.ip_dst = ip_src;
+
+       ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
        
        /* Fill in ICMP header */
        
-       icmp->icmp_type = ICMP_DEST_UNREACH;
-       icmp->icmp_code = code;
-       icmp->icmp_cksum = 0;
+       icmp.icmp_type = type;
+       icmp.icmp_code = code;
+       icmp.icmp_cksum = 0;
+       
+       icmp.icmp_cksum = inet_checksum(&icmp, icmp_size, ~0);
+       icmp.icmp_cksum = inet_checksum(packet->data + ether_size + ip_size + icmp_size, oldlen, icmp.icmp_cksum);
+
+       /* Copy structs on stack back to packet */
+
+       memcpy(packet->data + ether_size, &ip, ip_size);
+       memcpy(packet->data + ether_size + ip_size, &icmp, icmp_size);
+       
+       packet->len = ether_size + ip_size + icmp_size + oldlen;
+
+       send_packet(source, packet);
+}
+
+/* RFC 791 */
+
+static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet) {
+       struct ip ip;
+       vpn_packet_t fragment;
+       int len, maxlen, todo;
+       uint8_t *offset;
+       uint16_t ip_off, origf;
        
-       icmp->icmp_cksum = inet_checksum(icmp, 8 + oldlen, ~0);
+       memcpy(&ip, packet->data + ether_size, ip_size);
+       fragment.priority = packet->priority;
+
+       if(ip.ip_hl != ip_size / 4)
+               return;
        
-       packet->len = 14 + 20 + 8 + oldlen;
+       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);
+               return;
+       }
+
+       ifdebug(TRAFFIC) logger(LOG_INFO, "Fragmenting packet of %d bytes to %s (%s)", packet->len, dest->name, dest->hostname);
+
+       offset = packet->data + ether_size + ip_size;
+       maxlen = (dest->mtu - ether_size - ip_size) & ~0x7;
+       ip_off = ntohs(ip.ip_off);
+       origf = ip_off & ~IP_OFFMASK;
+       ip_off &= IP_OFFMASK;
        
-       write_packet(packet);
+       while(todo) {
+               len = todo > maxlen ? maxlen : todo;
+               memcpy(fragment.data + ether_size + ip_size, offset, len);
+               todo -= len;
+               offset += len;
+
+               ip.ip_len = htons(ip_size + len);
+               ip.ip_off = htons(ip_off | origf | (todo ? IP_MF : 0));
+               ip.ip_sum = 0;
+               ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
+               memcpy(fragment.data, packet->data, ether_size);
+               memcpy(fragment.data + ether_size, &ip, ip_size);
+               fragment.len = ether_size + ip_size + len;
+
+               send_packet(dest, &fragment);
+
+               ip_off += len / 8;
+       }       
 }
 
-static node_t *route_ipv4(vpn_packet_t *packet)
-{
+static void route_ipv4_unicast(node_t *source, vpn_packet_t *packet) {
        subnet_t *subnet;
+       node_t *via;
+       ipv4_t dest;
 
-       cp();
+       memcpy(&dest, &packet->data[30], sizeof dest);
+       subnet = lookup_subnet_ipv4(&dest);
+
+       if(!subnet) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv4 destination address %d.%d.%d.%d",
+                               source->name, source->hostname,
+                               dest.x[0],
+                               dest.x[1],
+                               dest.x[2],
+                               dest.x[3]);
+
+               route_ipv4_unreachable(source, packet, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
+               return;
+       }
+       
+       if(subnet->owner == source) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
+               return;
+       }
+
+       if(!subnet->owner->status.reachable)
+               route_ipv4_unreachable(source, packet, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
 
        if(priorityinheritance)
                packet->priority = packet->data[15];
 
-       subnet = lookup_subnet_ipv4((ipv4_t *) & packet->data[30]);
-
-       if(!subnet) {
-               ifdebug(TRAFFIC) logger(LOG_WARNING, _("Cannot route packet: unknown IPv4 destination address %d.%d.%d.%d"),
-                                  packet->data[30], packet->data[31], packet->data[32],
-                                  packet->data[33]);
+       via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
+       
+       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);
+               } else {
+                       fragment_ipv4_packet(via, packet);
+               }
 
-               route_ipv4_unreachable(packet, ICMP_NET_UNKNOWN);
-               return NULL;
+               return;
        }
-       
-       if(!subnet->owner->status.reachable)
-               route_ipv4_unreachable(packet, ICMP_NET_UNREACH);
 
-       return subnet->owner;
+       clamp_mss(source, via, packet);
+       send_packet(subnet->owner, packet);
+}
+
+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) || (
+                       packet->data[30] == 255 &&
+                       packet->data[31] == 255 &&
+                       packet->data[32] == 255 &&
+                       packet->data[33] == 255))
+               broadcast_packet(source, packet);
+       else
+               route_ipv4_unicast(source, packet);
 }
 
 /* RFC 2463 */
 
-static void route_ipv6_unreachable(vpn_packet_t *packet, uint8_t code)
-{
-       struct ip6_hdr *hdr;
-       struct icmp6_hdr *icmp;
+static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, uint8_t type, uint8_t code) {
+       struct ip6_hdr ip6;
+       struct icmp6_hdr icmp6 = {0};
        uint16_t checksum;      
 
        struct {
@@ -275,97 +426,127 @@ static void route_ipv6_unreachable(vpn_packet_t *packet, uint8_t code)
                uint32_t next;
        } pseudo;
 
-       if(ratelimit())
+       if(ratelimit(3))
                return;
        
-       cp();
+       /* Swap Ethernet source and destination addresses */
+
+       swap_mac_addresses(packet);
 
-       hdr = (struct ip6_hdr *)(packet->data + 14);
-       icmp = (struct icmp6_hdr *)(packet->data + 14 + sizeof(*hdr));
+       /* Copy headers from packet to structs on the stack */
+
+       memcpy(&ip6, packet->data + ether_size, ip6_size);
 
        /* Remember original source and destination */
-               
-       memcpy(&pseudo.ip6_src, &hdr->ip6_dst, 16);
-       memcpy(&pseudo.ip6_dst, &hdr->ip6_src, 16);
-       pseudo.length = ntohs(hdr->ip6_plen) + sizeof(*hdr);
        
-       if(pseudo.length >= IP_MSS - sizeof(*hdr) - sizeof(*icmp))
-               pseudo.length = IP_MSS - sizeof(*hdr) - sizeof(*icmp);
+       pseudo.ip6_src = ip6.ip6_dst;
+       pseudo.ip6_dst = ip6.ip6_src;
+
+       pseudo.length = packet->len - ether_size;
+
+       if(type == ICMP6_PACKET_TOO_BIG)
+               icmp6.icmp6_mtu = htonl(pseudo.length);
+       
+       if(pseudo.length >= IP_MSS - ip6_size - icmp6_size)
+               pseudo.length = IP_MSS - ip6_size - icmp6_size;
        
        /* Copy first part of original contents to ICMP message */
        
-       memmove(((char *)icmp) + sizeof(*icmp), hdr, pseudo.length);
+       memmove(packet->data + ether_size + ip6_size + icmp6_size, packet->data + ether_size, pseudo.length);
 
        /* Fill in IPv6 header */
        
-       hdr->ip6_flow = htonl(0x60000000UL);
-       hdr->ip6_plen = htons(sizeof(*icmp) + pseudo.length);
-       hdr->ip6_nxt = IPPROTO_ICMPV6;
-       hdr->ip6_hlim = 255;
-       memcpy(&hdr->ip6_dst, &pseudo.ip6_dst, 16);
-       memcpy(&hdr->ip6_src, &pseudo.ip6_src, 16);
+       ip6.ip6_flow = htonl(0x60000000UL);
+       ip6.ip6_plen = htons(icmp6_size + pseudo.length);
+       ip6.ip6_nxt = IPPROTO_ICMPV6;
+       ip6.ip6_hlim = 255;
+       ip6.ip6_src = pseudo.ip6_src;
+       ip6.ip6_dst = pseudo.ip6_dst;
 
        /* Fill in ICMP header */
        
-       icmp->icmp6_type = ICMP6_DST_UNREACH;
-       icmp->icmp6_code = code;
-       icmp->icmp6_cksum = 0;
+       icmp6.icmp6_type = type;
+       icmp6.icmp6_code = code;
+       icmp6.icmp6_cksum = 0;
 
        /* Create pseudo header */
                
-       pseudo.length = htonl(sizeof(*icmp) + pseudo.length);
+       pseudo.length = htonl(icmp6_size + pseudo.length);
        pseudo.next = htonl(IPPROTO_ICMPV6);
 
        /* Generate checksum */
        
        checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
-       checksum = inet_checksum(icmp, ntohl(pseudo.length), checksum);
+       checksum = inet_checksum(&icmp6, icmp6_size, checksum);
+       checksum = inet_checksum(packet->data + ether_size + ip6_size + icmp6_size, ntohl(pseudo.length) - icmp6_size, checksum);
+
+       icmp6.icmp6_cksum = checksum;
+
+       /* Copy structs on stack back to packet */
 
-       icmp->icmp6_cksum = checksum;
+       memcpy(packet->data + ether_size, &ip6, ip6_size);
+       memcpy(packet->data + ether_size + ip6_size, &icmp6, icmp6_size);
        
-       packet->len = 14 + sizeof(*hdr) + ntohl(pseudo.length);
+       packet->len = ether_size + ip6_size + ntohl(pseudo.length);
        
-       write_packet(packet);
+       send_packet(source, packet);
 }
 
-static node_t *route_ipv6(vpn_packet_t *packet)
-{
+static void route_ipv6_unicast(node_t *source, vpn_packet_t *packet) {
        subnet_t *subnet;
+       node_t *via;
+       ipv6_t dest;
 
-       cp();
-
-       subnet = lookup_subnet_ipv6((ipv6_t *) & packet->data[38]);
+       memcpy(&dest, &packet->data[38], sizeof dest);
+       subnet = lookup_subnet_ipv6(&dest);
 
        if(!subnet) {
-               ifdebug(TRAFFIC) logger(LOG_WARNING, _("Cannot route packet: unknown IPv6 destination address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx"),
-                                  ntohs(*(uint16_t *) & packet->data[38]),
-                                  ntohs(*(uint16_t *) & packet->data[40]),
-                                  ntohs(*(uint16_t *) & packet->data[42]),
-                                  ntohs(*(uint16_t *) & packet->data[44]),
-                                  ntohs(*(uint16_t *) & packet->data[46]),
-                                  ntohs(*(uint16_t *) & packet->data[48]),
-                                  ntohs(*(uint16_t *) & packet->data[50]),
-                                  ntohs(*(uint16_t *) & packet->data[52]));
-               route_ipv6_unreachable(packet, ICMP6_DST_UNREACH_ADDR);
-
-               return NULL;
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv6 destination address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
+                               source->name, source->hostname,
+                               ntohs(dest.x[0]),
+                               ntohs(dest.x[1]),
+                               ntohs(dest.x[2]),
+                               ntohs(dest.x[3]),
+                               ntohs(dest.x[4]),
+                               ntohs(dest.x[5]),
+                               ntohs(dest.x[6]),
+                               ntohs(dest.x[7]));
+
+               route_ipv6_unreachable(source, packet, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
+               return;
+       }
+
+       if(subnet->owner == source) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
+               return;
        }
 
        if(!subnet->owner->status.reachable)
-               route_ipv6_unreachable(packet, ICMP6_DST_UNREACH_NOROUTE);
+               route_ipv6_unreachable(source, packet, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
+
+       via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
        
-       return subnet->owner;
+       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);
+               return;
+       }
+
+       clamp_mss(source, via, packet);
+       send_packet(subnet->owner, packet);
 }
 
 /* RFC 2461 */
 
-static void route_neighborsol(vpn_packet_t *packet)
-{
-       struct ip6_hdr *hdr;
-       struct nd_neighbor_solicit *ns;
-       struct nd_opt_hdr *opt;
+static void route_neighborsol(node_t *source, vpn_packet_t *packet) {
+       struct ip6_hdr ip6;
+       struct nd_neighbor_solicit ns;
+       struct nd_opt_hdr opt;
        subnet_t *subnet;
        uint16_t checksum;
+       bool has_opt;
 
        struct {
                struct in6_addr ip6_src;        /* source address */
@@ -374,56 +555,74 @@ static void route_neighborsol(vpn_packet_t *packet)
                uint32_t next;
        } pseudo;
 
-       cp();
+       if(!checklength(source, packet, ether_size + ip6_size + ns_size))
+               return;
+       
+       has_opt = packet->len >= ether_size + ip6_size + ns_size + opt_size + ETH_ALEN;
+       
+       if(source != myself) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Got neighbor solicitation request from %s (%s) while in router mode!", source->name, source->hostname);
+               return;
+       }
 
-       hdr = (struct ip6_hdr *)(packet->data + 14);
-       ns = (struct nd_neighbor_solicit *)(packet->data + 14 + sizeof(*hdr));
-       opt = (struct nd_opt_hdr *)(packet->data + 14 + sizeof(*hdr) + sizeof(*ns));
+       /* Copy headers from packet to structs on the stack */
+
+       memcpy(&ip6, packet->data + ether_size, ip6_size);
+       memcpy(&ns, packet->data + ether_size + ip6_size, ns_size);
+       if(has_opt)
+               memcpy(&opt, packet->data + ether_size + ip6_size + ns_size, opt_size);
 
        /* First, snatch the source address from the neighbor solicitation packet */
 
        if(overwrite_mac)
-               memcpy(mymac.x, packet->data + 6, 6);
+               memcpy(mymac.x, packet->data + ETH_ALEN, ETH_ALEN);
 
        /* Check if this is a valid neighbor solicitation request */
 
-       if(ns->nd_ns_hdr.icmp6_type != ND_NEIGHBOR_SOLICIT ||
-          opt->nd_opt_type != ND_OPT_SOURCE_LINKADDR) {
-               ifdebug(TRAFFIC) logger(LOG_WARNING, _("Cannot route packet: received unknown type neighbor solicitation request"));
+       if(ns.nd_ns_hdr.icmp6_type != ND_NEIGHBOR_SOLICIT ||
+          (has_opt && opt.nd_opt_type != ND_OPT_SOURCE_LINKADDR)) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: received unknown type neighbor solicitation request");
                return;
        }
 
        /* Create pseudo header */
 
-       memcpy(&pseudo.ip6_src, &hdr->ip6_src, 16);
-       memcpy(&pseudo.ip6_dst, &hdr->ip6_dst, 16);
-       pseudo.length = htonl(sizeof(*ns) + sizeof(*opt) + 6);
+       pseudo.ip6_src = ip6.ip6_src;
+       pseudo.ip6_dst = ip6.ip6_dst;
+       if(has_opt)
+               pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
+       else
+               pseudo.length = htonl(ns_size);
        pseudo.next = htonl(IPPROTO_ICMPV6);
 
        /* Generate checksum */
 
        checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
-       checksum = inet_checksum(ns, sizeof(*ns) + 8, checksum);
+       checksum = inet_checksum(&ns, ns_size, checksum);
+       if(has_opt) {
+               checksum = inet_checksum(&opt, opt_size, checksum);
+               checksum = inet_checksum(packet->data + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
+       }
 
        if(checksum) {
-               ifdebug(TRAFFIC) logger(LOG_WARNING, _("Cannot route packet: checksum error for neighbor solicitation request"));
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: checksum error for neighbor solicitation request");
                return;
        }
 
        /* Check if the IPv6 address exists on the VPN */
 
-       subnet = lookup_subnet_ipv6((ipv6_t *) & ns->nd_ns_target);
+       subnet = lookup_subnet_ipv6((ipv6_t *) &ns.nd_ns_target);
 
        if(!subnet) {
-               ifdebug(TRAFFIC) logger(LOG_WARNING, _("Cannot route packet: neighbor solicitation request for unknown address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx"),
-                                  ntohs(((uint16_t *) & ns->nd_ns_target)[0]),
-                                  ntohs(((uint16_t *) & ns->nd_ns_target)[1]),
-                                  ntohs(((uint16_t *) & ns->nd_ns_target)[2]),
-                                  ntohs(((uint16_t *) & ns->nd_ns_target)[3]),
-                                  ntohs(((uint16_t *) & ns->nd_ns_target)[4]),
-                                  ntohs(((uint16_t *) & ns->nd_ns_target)[5]),
-                                  ntohs(((uint16_t *) & ns->nd_ns_target)[6]),
-                                  ntohs(((uint16_t *) & ns->nd_ns_target)[7]));
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: neighbor solicitation request for unknown address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
+                                  ntohs(((uint16_t *) &ns.nd_ns_target)[0]),
+                                  ntohs(((uint16_t *) &ns.nd_ns_target)[1]),
+                                  ntohs(((uint16_t *) &ns.nd_ns_target)[2]),
+                                  ntohs(((uint16_t *) &ns.nd_ns_target)[3]),
+                                  ntohs(((uint16_t *) &ns.nd_ns_target)[4]),
+                                  ntohs(((uint16_t *) &ns.nd_ns_target)[5]),
+                                  ntohs(((uint16_t *) &ns.nd_ns_target)[6]),
+                                  ntohs(((uint16_t *) &ns.nd_ns_target)[7]));
 
                return;
        }
@@ -435,77 +634,106 @@ static void route_neighborsol(vpn_packet_t *packet)
 
        /* Create neighbor advertation reply */
 
-       memcpy(packet->data, packet->data + ETHER_ADDR_LEN, ETHER_ADDR_LEN);    /* copy destination address */
-       packet->data[ETHER_ADDR_LEN * 2 - 1] ^= 0xFF;   /* mangle source address so it looks like it's not from us */
+       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 */
 
-       memcpy(&hdr->ip6_dst, &hdr->ip6_src, 16);       /* swap destination and source protocol address */
-       memcpy(&hdr->ip6_src, &ns->nd_ns_target, 16);   /* ... */
+       ip6.ip6_dst = ip6.ip6_src;                      /* swap destination and source protocoll address */
+       ip6.ip6_src = ns.nd_ns_target;
 
-       memcpy((char *) opt + sizeof(*opt), packet->data + ETHER_ADDR_LEN, 6);  /* add fake source hard addr */
+       if(has_opt)
+               memcpy(packet->data + ether_size + ip6_size + ns_size + opt_size, packet->data + ETH_ALEN, ETH_ALEN);   /* add fake source hard addr */
 
-       ns->nd_ns_hdr.icmp6_cksum = 0;
-       ns->nd_ns_hdr.icmp6_type = ND_NEIGHBOR_ADVERT;
-       ns->nd_ns_hdr.icmp6_dataun.icmp6_un_data8[0] = 0x40;    /* Set solicited flag */
-       ns->nd_ns_hdr.icmp6_dataun.icmp6_un_data8[1] =
-               ns->nd_ns_hdr.icmp6_dataun.icmp6_un_data8[2] =
-               ns->nd_ns_hdr.icmp6_dataun.icmp6_un_data8[3] = 0;
-       opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
+       ns.nd_ns_cksum = 0;
+       ns.nd_ns_type = ND_NEIGHBOR_ADVERT;
+       ns.nd_ns_reserved = htonl(0x40000000UL);        /* Set solicited flag */
+       opt.nd_opt_type = ND_OPT_TARGET_LINKADDR;
 
        /* Create pseudo header */
 
-       memcpy(&pseudo.ip6_src, &hdr->ip6_src, 16);
-       memcpy(&pseudo.ip6_dst, &hdr->ip6_dst, 16);
-       pseudo.length = htonl(sizeof(*ns) + sizeof(*opt) + 6);
+       pseudo.ip6_src = ip6.ip6_src;
+       pseudo.ip6_dst = ip6.ip6_dst;
+       if(has_opt)
+               pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
+       else
+               pseudo.length = htonl(ns_size);
        pseudo.next = htonl(IPPROTO_ICMPV6);
 
        /* Generate checksum */
 
        checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
-       checksum = inet_checksum(ns, sizeof(*ns) + 8, checksum);
+       checksum = inet_checksum(&ns, ns_size, checksum);
+       if(has_opt) {
+               checksum = inet_checksum(&opt, opt_size, checksum);
+               checksum = inet_checksum(packet->data + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
+       }
+
+       ns.nd_ns_hdr.icmp6_cksum = checksum;
+
+       /* Copy structs on stack back to packet */
+
+       memcpy(packet->data + ether_size, &ip6, ip6_size);
+       memcpy(packet->data + ether_size + ip6_size, &ns, ns_size);
+       if(has_opt)
+               memcpy(packet->data + ether_size + ip6_size + ns_size, &opt, opt_size);
+
+       send_packet(source, packet);
+}
+
+static void route_ipv6(node_t *source, vpn_packet_t *packet) {
+       if(!checklength(source, packet, ether_size + ip6_size))
+               return;
 
-       ns->nd_ns_hdr.icmp6_cksum = checksum;
+       if(packet->data[20] == IPPROTO_ICMPV6 && checklength(source, packet, ether_size + ip6_size + icmp6_size) && packet->data[54] == ND_NEIGHBOR_SOLICIT) {
+               route_neighborsol(source, packet);
+               return;
+       }
 
-       write_packet(packet);
+       if(packet->data[38] == 255)
+               broadcast_packet(source, packet);
+       else
+               route_ipv6_unicast(source, packet);
 }
 
 /* RFC 826 */
 
-static void route_arp(vpn_packet_t *packet)
-{
-       struct ether_arp *arp;
+static void route_arp(node_t *source, vpn_packet_t *packet) {
+       struct ether_arp arp;
        subnet_t *subnet;
-       uint8_t ipbuf[4];
+       struct in_addr addr;
 
-       cp();
+       if(!checklength(source, packet, ether_size + arp_size))
+               return;
+
+       if(source != myself) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Got ARP request from %s (%s) while in router mode!", source->name, source->hostname);
+               return;
+       }
 
        /* First, snatch the source address from the ARP packet */
 
        if(overwrite_mac)
-               memcpy(mymac.x, packet->data + 6, 6);
+               memcpy(mymac.x, packet->data + ETH_ALEN, ETH_ALEN);
 
-       /* This routine generates replies to ARP requests.
-          You don't need to set NOARP flag on the interface anymore (which is broken on FreeBSD).
-          Most of the code here is taken from choparp.c by Takamichi Tateoka (tree@mma.club.uec.ac.jp)
-        */
+       /* Copy headers from packet to structs on the stack */
 
-       arp = (struct ether_arp *)(packet->data + 14);
+       memcpy(&arp, packet->data + ether_size, arp_size);
 
        /* Check if this is a valid ARP request */
 
-       if(ntohs(arp->arp_hrd) != ARPHRD_ETHER || ntohs(arp->arp_pro) != ETHERTYPE_IP ||
-          arp->arp_hln != ETHER_ADDR_LEN || arp->arp_pln != 4 || ntohs(arp->arp_op) != ARPOP_REQUEST) {
-               ifdebug(TRAFFIC) logger(LOG_WARNING, _("Cannot route packet: received unknown type ARP request"));
+       if(ntohs(arp.arp_hrd) != ARPHRD_ETHER || ntohs(arp.arp_pro) != ETH_P_IP ||
+          arp.arp_hln != ETH_ALEN || arp.arp_pln != sizeof(addr) || ntohs(arp.arp_op) != ARPOP_REQUEST) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: received unknown type ARP request");
                return;
        }
 
        /* Check if the IPv4 address exists on the VPN */
 
-       subnet = lookup_subnet_ipv4((ipv4_t *) arp->arp_tpa);
+       subnet = lookup_subnet_ipv4((ipv4_t *) &arp.arp_tpa);
 
        if(!subnet) {
-               ifdebug(TRAFFIC) logger(LOG_WARNING, _("Cannot route packet: ARP request for unknown address %d.%d.%d.%d"),
-                                  arp->arp_tpa[0], arp->arp_tpa[1], arp->arp_tpa[2],
-                                  arp->arp_tpa[3]);
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: ARP request for unknown address %d.%d.%d.%d",
+                                  arp.arp_tpa[0], arp.arp_tpa[1], arp.arp_tpa[2],
+                                  arp.arp_tpa[3]);
                return;
        }
 
@@ -514,126 +742,113 @@ static void route_arp(vpn_packet_t *packet)
        if(subnet->owner == myself)
                return;                                 /* silently ignore */
 
-       memcpy(packet->data, packet->data + ETHER_ADDR_LEN, ETHER_ADDR_LEN);    /* copy destination address */
-       packet->data[ETHER_ADDR_LEN * 2 - 1] ^= 0xFF;   /* mangle source address so it looks like it's not from us */
+       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 */
+
+       memcpy(&addr, arp.arp_tpa, sizeof(addr));       /* save protocol addr */
+       memcpy(arp.arp_tpa, arp.arp_spa, sizeof(addr)); /* swap destination and source protocol address */
+       memcpy(arp.arp_spa, &addr, sizeof(addr));       /* ... */
+
+       memcpy(arp.arp_tha, arp.arp_sha, ETH_ALEN);     /* set target hard/proto addr */
+       memcpy(arp.arp_sha, packet->data + ETH_ALEN, ETH_ALEN); /* add fake source hard addr */
+       arp.arp_op = htons(ARPOP_REPLY);
 
-       memcpy(ipbuf, arp->arp_tpa, 4); /* save protocol addr */
-       memcpy(arp->arp_tpa, arp->arp_spa, 4);  /* swap destination and source protocol address */
-       memcpy(arp->arp_spa, ipbuf, 4); /* ... */
+       /* Copy structs on stack back to packet */
 
-       memcpy(arp->arp_tha, arp->arp_sha, 10); /* set target hard/proto addr */
-       memcpy(arp->arp_sha, packet->data + ETHER_ADDR_LEN, ETHER_ADDR_LEN);    /* add fake source hard addr */
-       arp->arp_op = htons(ARPOP_REPLY);
+       memcpy(packet->data + ether_size, &arp, arp_size);
 
-       write_packet(packet);
+       send_packet(source, packet);
 }
 
-void route_outgoing(vpn_packet_t *packet)
-{
-       uint16_t type;
-       node_t *n = NULL;
+static void route_mac(node_t *source, vpn_packet_t *packet) {
+       subnet_t *subnet;
+       mac_t dest;
 
-       cp();
+       /* Learn source address */
 
-       /* FIXME: multicast? */
+       if(source == myself) {
+               mac_t src;
+               memcpy(&src, &packet->data[6], sizeof src);
+               learn_mac(&src);
+       }
 
-       switch (routing_mode) {
-               case RMODE_ROUTER:
-                       type = ntohs(*((uint16_t *)(&packet->data[12])));
-                       switch (type) {
-                               case 0x0800:
-                                       n = route_ipv4(packet);
-                                       break;
-
-                               case 0x86DD:
-                                       if(packet->data[20] == IPPROTO_ICMPV6 && packet->data[54] == ND_NEIGHBOR_SOLICIT) {
-                                               route_neighborsol(packet);
-                                               return;
-                                       }
-                                       n = route_ipv6(packet);
-                                       break;
-
-                               case 0x0806:
-                                       route_arp(packet);
-                                       return;
-
-                               default:
-                                       ifdebug(TRAFFIC) logger(LOG_WARNING, _("Cannot route packet: unknown type %hx"), type);
-                                       return;
-                       }
-                       if(n)
-                               send_packet(n, packet);
-                       break;
+       /* Lookup destination address */
 
-               case RMODE_SWITCH:
-                       n = route_mac(packet);
-                       if(n)
-                               send_packet(n, packet);
-                       else
-                               broadcast_packet(myself, packet);
-                       break;
+       memcpy(&dest, &packet->data[0], sizeof dest);
+       subnet = lookup_subnet_mac(NULL, &dest);
 
-               case RMODE_HUB:
-                       broadcast_packet(myself, packet);
-                       break;
+       if(!subnet) {
+               broadcast_packet(source, packet);
+               return;
+       }
+
+       if(subnet->owner == source) {
+               ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
+               return;
+       }
+
+       // Handle packets larger than PMTU
+
+       node_t *via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
+       
+       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) {
+                               packet->len = via->mtu;
+                               route_ipv4_unreachable(source, packet, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
+                       } else {
+                               fragment_ipv4_packet(via, packet);
+                       }
+                       return;
+               } else if(type == ETH_P_IPV6 && packet->len > 1294) {
+                       packet->len = via->mtu;
+                       route_ipv6_unreachable(source, packet, ICMP6_PACKET_TOO_BIG, 0);
+                       return;
+               }
        }
+
+       clamp_mss(source, via, packet);
+       send_packet(subnet->owner, packet);
 }
 
-void route_incoming(node_t *source, vpn_packet_t *packet)
-{
+void route(node_t *source, vpn_packet_t *packet) {
+       if(!checklength(source, packet, ether_size))
+               return;
+
        switch (routing_mode) {
                case RMODE_ROUTER:
                        {
-                               node_t *n = NULL;
-                               uint16_t type;
+                               uint16_t type = packet->data[12] << 8 | packet->data[13];
 
-                               type = ntohs(*((uint16_t *)(&packet->data[12])));
                                switch (type) {
-                                       case 0x0800:
-                                               n = route_ipv4(packet);
+                                       case ETH_P_ARP:
+                                               route_arp(source, packet);
                                                break;
 
-                                       case 0x86DD:
-                                               n = route_ipv6(packet);
+                                       case ETH_P_IP:
+                                               route_ipv4(source, packet);
                                                break;
 
-                                       default:
-                                               n = myself;
+                                       case ETH_P_IPV6:
+                                               route_ipv6(source, packet);
                                                break;
-                               }
 
-                               if(n) {
-                                       if(n == myself) {
-                                               if(overwrite_mac)
-                                                       memcpy(packet->data, mymac.x, 6);
-                                               write_packet(packet);
-                                       } else
-                                               send_packet(n, packet);
+                                       default:
+                                               ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown type %hx", source->name, source->hostname, type);
+                                               break;
                                }
                        }
                        break;
 
                case RMODE_SWITCH:
-                       {
-                               subnet_t *subnet;
-
-                               subnet = lookup_subnet_mac((mac_t *)(&packet->data[0]));
-
-                               if(subnet) {
-                                       if(subnet->owner == myself)
-                                               write_packet(packet);
-                                       else
-                                               send_packet(subnet->owner, packet);
-                               } else {
-                                       broadcast_packet(source, packet);
-                                       write_packet(packet);
-                               }
-                       }
+                       route_mac(source, packet);
                        break;
 
                case RMODE_HUB:
-                       broadcast_packet(source, packet);       /* Spread it on */
-                       write_packet(packet);
+                       broadcast_packet(source, packet);
                        break;
        }
 }