www.tinc-vpn.org Git - tinc/blob - src/route.c

   1 /*
   2     route.c -- routing
   3     Copyright (C) 2000-2005 Ivo Timmermans,
   4                   2000-2018 Guus Sliepen <guus@tinc-vpn.org>
   5
   6     This program is free software; you can redistribute it and/or modify
   7     it under the terms of the GNU General Public License as published by
   8     the Free Software Foundation; either version 2 of the License, or
   9     (at your option) any later version.
  10
  11     This program is distributed in the hope that it will be useful,
  12     but WITHOUT ANY WARRANTY; without even the implied warranty of
  13     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14     GNU General Public License for more details.
  15
  16     You should have received a copy of the GNU General Public License along
  17     with this program; if not, write to the Free Software Foundation, Inc.,
  18     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  19 */
  20
  21 #include "system.h"
  22
  23 #include "connection.h"
  24 #include "control_common.h"
  25 #include "ethernet.h"
  26 #include "ipv4.h"
  27 #include "ipv6.h"
  28 #include "logger.h"
  29 #include "meta.h"
  30 #include "net.h"
  31 #include "protocol.h"
  32 #include "route.h"
  33 #include "subnet.h"
  34 #include "utils.h"
  35
  36 rmode_t routing_mode = RMODE_ROUTER;
  37 fmode_t forwarding_mode = FMODE_INTERNAL;
  38 bmode_t broadcast_mode = BMODE_MST;
  39 bool decrement_ttl = false;
  40 bool directonly = false;
  41 bool priorityinheritance = false;
  42 int macexpire = 600;
  43 bool overwrite_mac = false;
  44 mac_t mymac = {{0xFE, 0xFD, 0, 0, 0, 0}};
  45 bool pcap = false;
  46
  47 /* Sizes of various headers */
  48
  49 static const size_t ether_size = sizeof(struct ether_header);
  50 static const size_t arp_size = sizeof(struct ether_arp);
  51 static const size_t ip_size = sizeof(struct ip);
  52 static const size_t icmp_size = sizeof(struct icmp) - sizeof(struct ip);
  53 static const size_t ip6_size = sizeof(struct ip6_hdr);
  54 static const size_t icmp6_size = sizeof(struct icmp6_hdr);
  55 static const size_t ns_size = sizeof(struct nd_neighbor_solicit);
  56 static const size_t opt_size = sizeof(struct nd_opt_hdr);
  57
  58 #ifndef MAX
  59 #define MAX(a, b) ((a) > (b) ? (a) : (b))
  60 #endif
  61
  62 static timeout_t age_subnets_timeout;
  63
  64 /* RFC 1071 */
  65
  66 static uint16_t inet_checksum(void *vdata, int len, uint16_t prevsum) {
  67         uint8_t *data = vdata;
  68         uint16_t word;
  69         uint32_t checksum = prevsum ^ 0xFFFF;
  70
  71         while(len >= 2) {
  72                 memcpy(&word, data, sizeof(word));
  73                 checksum += word;
  74                 data += 2;
  75                 len -= 2;
  76         }
  77
  78         if(len) {
  79                 checksum += *data;
  80         }
  81
  82         while(checksum >> 16) {
  83                 checksum = (checksum & 0xFFFF) + (checksum >> 16);
  84         }
  85
  86         return ~checksum;
  87 }
  88
  89 static bool ratelimit(int frequency) {
  90         static time_t lasttime = 0;
  91         static int count = 0;
  92
  93         if(lasttime == now.tv_sec) {
  94                 if(count >= frequency) {
  95                         return true;
  96                 }
  97         } else {
  98                 lasttime = now.tv_sec;
  99                 count = 0;
 100         }
 101
 102         count++;
 103         return false;
 104 }
 105
 106 static bool checklength(node_t *source, vpn_packet_t *packet, length_t length) {
 107         if(packet->len < length) {
 108                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got too short packet from %s (%s)", source->name, source->hostname);
 109                 return false;
 110         } else {
 111                 return true;
 112         }
 113 }
 114
 115 static void swap_mac_addresses(vpn_packet_t *packet) {
 116         mac_t tmp;
 117         memcpy(&tmp, &DATA(packet)[0], sizeof(tmp));
 118         memcpy(&DATA(packet)[0], &DATA(packet)[6], sizeof(tmp));
 119         memcpy(&DATA(packet)[6], &tmp, sizeof(tmp));
 120 }
 121
 122 /* RFC 792 */
 123
 124 static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
 125         struct ip ip = {0};
 126         struct icmp icmp = {0};
 127
 128         struct in_addr ip_src;
 129         struct in_addr ip_dst;
 130         uint32_t oldlen;
 131
 132         if(ratelimit(3)) {
 133                 return;
 134         }
 135
 136         /* Swap Ethernet source and destination addresses */
 137
 138         swap_mac_addresses(packet);
 139
 140         /* Copy headers from packet into properly aligned structs on the stack */
 141
 142         memcpy(&ip, DATA(packet) + ether_size, ip_size);
 143
 144         /* Remember original source and destination */
 145
 146         ip_src = ip.ip_src;
 147         ip_dst = ip.ip_dst;
 148
 149         /* Try to reply with an IP address assigned to the local machine */
 150
 151         if(type == ICMP_TIME_EXCEEDED && code == ICMP_EXC_TTL) {
 152                 int sockfd = socket(AF_INET, SOCK_DGRAM, 0);
 153
 154                 if(sockfd != -1) {
 155                         struct sockaddr_in addr;
 156                         memset(&addr, 0, sizeof(addr));
 157                         addr.sin_family = AF_INET;
 158                         addr.sin_addr = ip.ip_src;
 159
 160                         if(!connect(sockfd, (const struct sockaddr *) &addr, sizeof(addr))) {
 161                                 memset(&addr, 0, sizeof(addr));
 162                                 addr.sin_family = AF_INET;
 163                                 socklen_t addrlen = sizeof(addr);
 164
 165                                 if(!getsockname(sockfd, (struct sockaddr *) &addr, &addrlen) && (size_t)addrlen <= sizeof(addr)) {
 166                                         ip_dst = addr.sin_addr;
 167                                 }
 168                         }
 169
 170                         close(sockfd);
 171                 }
 172         }
 173
 174         oldlen = packet->len - ether_size;
 175
 176         if(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 177                 icmp.icmp_nextmtu = htons(packet->len - ether_size);
 178         }
 179
 180         if(oldlen >= IP_MSS - ip_size - icmp_size) {
 181                 oldlen = IP_MSS - ip_size - icmp_size;
 182         }
 183
 184         /* Copy first part of original contents to ICMP message */
 185
 186         memmove(DATA(packet) + ether_size + ip_size + icmp_size, DATA(packet) + ether_size, oldlen);
 187
 188         /* Fill in IPv4 header */
 189
 190         ip.ip_v = 4;
 191         ip.ip_hl = ip_size / 4;
 192         ip.ip_tos = 0;
 193         ip.ip_len = htons(ip_size + icmp_size + oldlen);
 194         ip.ip_id = 0;
 195         ip.ip_off = 0;
 196         ip.ip_ttl = 255;
 197         ip.ip_p = IPPROTO_ICMP;
 198         ip.ip_sum = 0;
 199         ip.ip_src = ip_dst;
 200         ip.ip_dst = ip_src;
 201
 202         ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
 203
 204         /* Fill in ICMP header */
 205
 206         icmp.icmp_type = type;
 207         icmp.icmp_code = code;
 208         icmp.icmp_cksum = 0;
 209
 210         icmp.icmp_cksum = inet_checksum(&icmp, icmp_size, ~0);
 211         icmp.icmp_cksum = inet_checksum(DATA(packet) + ether_size + ip_size + icmp_size, oldlen, icmp.icmp_cksum);
 212
 213         /* Copy structs on stack back to packet */
 214
 215         memcpy(DATA(packet) + ether_size, &ip, ip_size);
 216         memcpy(DATA(packet) + ether_size + ip_size, &icmp, icmp_size);
 217
 218         packet->len = ether_size + ip_size + icmp_size + oldlen;
 219
 220         send_packet(source, packet);
 221 }
 222
 223 /* RFC 2463 */
 224
 225 static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
 226         struct ip6_hdr ip6;
 227         struct icmp6_hdr icmp6 = {0};
 228         uint16_t checksum;
 229
 230         struct {
 231                 struct in6_addr ip6_src;        /* source address */
 232                 struct in6_addr ip6_dst;        /* destination address */
 233                 uint32_t length;
 234                 uint32_t next;
 235         } pseudo;
 236
 237         if(ratelimit(3)) {
 238                 return;
 239         }
 240
 241         /* Swap Ethernet source and destination addresses */
 242
 243         swap_mac_addresses(packet);
 244
 245         /* Copy headers from packet to structs on the stack */
 246
 247         memcpy(&ip6, DATA(packet) + ether_size, ip6_size);
 248
 249         /* Remember original source and destination */
 250
 251         pseudo.ip6_src = ip6.ip6_dst;
 252         pseudo.ip6_dst = ip6.ip6_src;
 253
 254         /* Try to reply with an IP address assigned to the local machine */
 255
 256         if(type == ICMP6_TIME_EXCEEDED && code == ICMP6_TIME_EXCEED_TRANSIT) {
 257                 int sockfd = socket(AF_INET6, SOCK_DGRAM, 0);
 258
 259                 if(sockfd != -1) {
 260                         struct sockaddr_in6 addr;
 261                         memset(&addr, 0, sizeof(addr));
 262                         addr.sin6_family = AF_INET6;
 263                         addr.sin6_addr = ip6.ip6_src;
 264
 265                         if(!connect(sockfd, (const struct sockaddr *) &addr, sizeof(addr))) {
 266                                 memset(&addr, 0, sizeof(addr));
 267                                 addr.sin6_family = AF_INET6;
 268                                 socklen_t addrlen = sizeof(addr);
 269
 270                                 if(!getsockname(sockfd, (struct sockaddr *) &addr, &addrlen) && (size_t)addrlen <= sizeof(addr)) {
 271                                         pseudo.ip6_src = addr.sin6_addr;
 272                                 }
 273                         }
 274
 275                         close(sockfd);
 276                 }
 277         }
 278
 279         pseudo.length = packet->len - ether_size;
 280
 281         if(type == ICMP6_PACKET_TOO_BIG) {
 282                 icmp6.icmp6_mtu = htonl(pseudo.length);
 283         }
 284
 285         if(pseudo.length >= IP_MSS - ip6_size - icmp6_size) {
 286                 pseudo.length = IP_MSS - ip6_size - icmp6_size;
 287         }
 288
 289         /* Copy first part of original contents to ICMP message */
 290
 291         memmove(DATA(packet) + ether_size + ip6_size + icmp6_size, DATA(packet) + ether_size, pseudo.length);
 292
 293         /* Fill in IPv6 header */
 294
 295         ip6.ip6_flow = htonl(0x60000000UL);
 296         ip6.ip6_plen = htons(icmp6_size + pseudo.length);
 297         ip6.ip6_nxt = IPPROTO_ICMPV6;
 298         ip6.ip6_hlim = 255;
 299         ip6.ip6_src = pseudo.ip6_src;
 300         ip6.ip6_dst = pseudo.ip6_dst;
 301
 302         /* Fill in ICMP header */
 303
 304         icmp6.icmp6_type = type;
 305         icmp6.icmp6_code = code;
 306         icmp6.icmp6_cksum = 0;
 307
 308         /* Create pseudo header */
 309
 310         pseudo.length = htonl(icmp6_size + pseudo.length);
 311         pseudo.next = htonl(IPPROTO_ICMPV6);
 312
 313         /* Generate checksum */
 314
 315         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
 316         checksum = inet_checksum(&icmp6, icmp6_size, checksum);
 317         checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + icmp6_size, ntohl(pseudo.length) - icmp6_size, checksum);
 318
 319         icmp6.icmp6_cksum = checksum;
 320
 321         /* Copy structs on stack back to packet */
 322
 323         memcpy(DATA(packet) + ether_size, &ip6, ip6_size);
 324         memcpy(DATA(packet) + ether_size + ip6_size, &icmp6, icmp6_size);
 325
 326         packet->len = ether_size + ip6_size + ntohl(pseudo.length);
 327
 328         send_packet(source, packet);
 329 }
 330
 331 static bool do_decrement_ttl(node_t *source, vpn_packet_t *packet) {
 332         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
 333         length_t ethlen = ether_size;
 334
 335         if(type == ETH_P_8021Q) {
 336                 type = DATA(packet)[16] << 8 | DATA(packet)[17];
 337                 ethlen += 4;
 338         }
 339
 340         switch(type) {
 341         case ETH_P_IP:
 342                 if(!checklength(source, packet, ethlen + ip_size)) {
 343                         return false;
 344                 }
 345
 346                 if(DATA(packet)[ethlen + 8] <= 1) {
 347                         if(DATA(packet)[ethlen + 11] != IPPROTO_ICMP || DATA(packet)[ethlen + 32] != ICMP_TIME_EXCEEDED) {
 348                                 route_ipv4_unreachable(source, packet, ethlen, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL);
 349                         }
 350
 351                         return false;
 352                 }
 353
 354                 uint16_t old = DATA(packet)[ethlen + 8] << 8 | DATA(packet)[ethlen + 9];
 355                 DATA(packet)[ethlen + 8]--;
 356                 uint16_t new = DATA(packet)[ethlen + 8] << 8 | DATA(packet)[ethlen + 9];
 357
 358                 uint32_t checksum = DATA(packet)[ethlen + 10] << 8 | DATA(packet)[ethlen + 11];
 359                 checksum += old + (~new & 0xFFFF);
 360
 361                 while(checksum >> 16) {
 362                         checksum = (checksum & 0xFFFF) + (checksum >> 16);
 363                 }
 364
 365                 DATA(packet)[ethlen + 10] = checksum >> 8;
 366                 DATA(packet)[ethlen + 11] = checksum & 0xff;
 367
 368                 return true;
 369
 370         case ETH_P_IPV6:
 371                 if(!checklength(source, packet, ethlen + ip6_size)) {
 372                         return false;
 373                 }
 374
 375                 if(DATA(packet)[ethlen + 7] <= 1) {
 376                         if(DATA(packet)[ethlen + 6] != IPPROTO_ICMPV6 || DATA(packet)[ethlen + 40] != ICMP6_TIME_EXCEEDED) {
 377                                 route_ipv6_unreachable(source, packet, ethlen, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT);
 378                         }
 379
 380                         return false;
 381                 }
 382
 383                 DATA(packet)[ethlen + 7]--;
 384
 385                 return true;
 386
 387         default:
 388                 return true;
 389         }
 390 }
 391
 392 static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *packet) {
 393         if(!source || !via || !(via->options & OPTION_CLAMP_MSS)) {
 394                 return;
 395         }
 396
 397         uint16_t mtu = source->mtu;
 398
 399         if(via != myself && via->mtu < mtu) {
 400                 mtu = via->mtu;
 401         }
 402
 403         /* Find TCP header */
 404         int start = ether_size;
 405         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
 406
 407         if(type == ETH_P_8021Q) {
 408                 start += 4;
 409                 type = DATA(packet)[16] << 8 | DATA(packet)[17];
 410         }
 411
 412         /* IP in IP (RFC 2003) packet */
 413         if(type == ETH_P_IP && DATA(packet)[start + 9] == 4) {
 414                 start += 20;
 415         }
 416
 417         if(packet->len <= start + 20) {
 418                 return;
 419         }
 420
 421         if(type == ETH_P_IP && DATA(packet)[start + 9] == 6) {
 422                 start += (DATA(packet)[start] & 0xf) * 4;
 423         } else if(type == ETH_P_IPV6 && DATA(packet)[start + 6] == 6) {
 424                 start += 40;
 425         } else {
 426                 return;
 427         }
 428
 429         if(packet->len <= start + 20) {
 430                 return;
 431         }
 432
 433         /* Use data offset field to calculate length of options field */
 434         int len = ((DATA(packet)[start + 12] >> 4) - 5) * 4;
 435
 436         if(packet->len < start + 20 + len) {
 437                 return;
 438         }
 439
 440         /* Search for MSS option header */
 441         for(int i = 0; i < len;) {
 442                 if(DATA(packet)[start + 20 + i] == 0) {
 443                         break;
 444                 }
 445
 446                 if(DATA(packet)[start + 20 + i] == 1) {
 447                         i++;
 448                         continue;
 449                 }
 450
 451                 if(i > len - 2 || i > len - DATA(packet)[start + 21 + i]) {
 452                         break;
 453                 }
 454
 455                 if(DATA(packet)[start + 20 + i] != 2) {
 456                         if(DATA(packet)[start + 21 + i] < 2) {
 457                                 break;
 458                         }
 459
 460                         i += DATA(packet)[start + 21 + i];
 461                         continue;
 462                 }
 463
 464                 if(DATA(packet)[start + 21] != 4) {
 465                         break;
 466                 }
 467
 468                 /* Found it */
 469                 uint16_t oldmss = DATA(packet)[start + 22 + i] << 8 | DATA(packet)[start + 23 + i];
 470                 uint16_t newmss = mtu - start - 20;
 471                 uint32_t csum = DATA(packet)[start + 16] << 8 | DATA(packet)[start + 17];
 472
 473                 if(oldmss <= newmss) {
 474                         break;
 475                 }
 476
 477                 logger(DEBUG_TRAFFIC, LOG_INFO, "Clamping MSS of packet from %s to %s to %d", source->name, via->name, newmss);
 478
 479                 /* Update the MSS value and the checksum */
 480                 DATA(packet)[start + 22 + i] = newmss >> 8;
 481                 DATA(packet)[start + 23 + i] = newmss & 0xff;
 482                 csum ^= 0xffff;
 483                 csum += oldmss ^ 0xffff;
 484                 csum += newmss;
 485                 csum = (csum & 0xffff) + (csum >> 16);
 486                 csum += csum >> 16;
 487                 csum ^= 0xffff;
 488                 DATA(packet)[start + 16] = csum >> 8;
 489                 DATA(packet)[start + 17] = csum;
 490                 break;
 491         }
 492 }
 493
 494 static void age_subnets(void *data) {
 495         (void)data;
 496         bool left = false;
 497
 498         for splay_each(subnet_t, s, myself->subnet_tree) {
 499                 if(s->expires && s->expires < now.tv_sec) {
 500                         if(debug_level >= DEBUG_TRAFFIC) {
 501                                 char netstr[MAXNETSTR];
 502
 503                                 if(net2str(netstr, sizeof(netstr), s)) {
 504                                         logger(DEBUG_TRAFFIC, LOG_INFO, "Subnet %s expired", netstr);
 505                                 }
 506                         }
 507
 508                         for list_each(connection_t, c, connection_list)
 509                                 if(c->edge) {
 510                                         send_del_subnet(c, s);
 511                                 }
 512
 513                         subnet_del(myself, s);
 514                 } else {
 515                         if(s->expires) {
 516                                 left = true;
 517                         }
 518                 }
 519         }
 520
 521         if(left)
 522                 timeout_set(&age_subnets_timeout, &(struct timeval) {
 523                 10, rand() % 100000
 524         });
 525 }
 526
 527 static void learn_mac(mac_t *address) {
 528         subnet_t *subnet = lookup_subnet_mac(myself, address);
 529
 530         /* If we don't know this MAC address yet, store it */
 531
 532         if(!subnet) {
 533                 logger(DEBUG_TRAFFIC, LOG_INFO, "Learned new MAC address %x:%x:%x:%x:%x:%x",
 534                        address->x[0], address->x[1], address->x[2], address->x[3],
 535                        address->x[4], address->x[5]);
 536
 537                 subnet = new_subnet();
 538                 subnet->type = SUBNET_MAC;
 539                 subnet->expires = now.tv_sec + macexpire;
 540                 subnet->net.mac.address = *address;
 541                 subnet->weight = 10;
 542                 subnet_add(myself, subnet);
 543                 subnet_update(myself, subnet, true);
 544
 545                 /* And tell all other tinc daemons it's our MAC */
 546
 547                 for list_each(connection_t, c, connection_list)
 548                         if(c->edge) {
 549                                 send_add_subnet(c, subnet);
 550                         }
 551
 552                 timeout_add(&age_subnets_timeout, age_subnets, NULL, &(struct timeval) {
 553                         10, rand() % 100000
 554                 });
 555         } else {
 556                 if(subnet->expires) {
 557                         subnet->expires = now.tv_sec + macexpire;
 558                 }
 559         }
 560 }
 561
 562 static void route_broadcast(node_t *source, vpn_packet_t *packet) {
 563         if(decrement_ttl && source != myself)
 564                 if(!do_decrement_ttl(source, packet)) {
 565                         return;
 566                 }
 567
 568         broadcast_packet(source, packet);
 569 }
 570
 571 /* RFC 791 */
 572
 573 static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet, length_t ether_size) {
 574         struct ip ip;
 575         vpn_packet_t fragment;
 576         int maxlen, todo;
 577         uint8_t *offset;
 578         uint16_t ip_off, origf;
 579
 580         memcpy(&ip, DATA(packet) + ether_size, ip_size);
 581         fragment.priority = packet->priority;
 582         fragment.offset = DEFAULT_PACKET_OFFSET;
 583
 584         if(ip.ip_hl != ip_size / 4) {
 585                 return;
 586         }
 587
 588         todo = ntohs(ip.ip_len) - ip_size;
 589
 590         if(ether_size + ip_size + todo != packet->len) {
 591                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Length of packet (%d) doesn't match length in IPv4 header (%d)", packet->len, (int)(ether_size + ip_size + todo));
 592                 return;
 593         }
 594
 595         logger(DEBUG_TRAFFIC, LOG_INFO, "Fragmenting packet of %d bytes to %s (%s)", packet->len, dest->name, dest->hostname);
 596
 597         offset = DATA(packet) + ether_size + ip_size;
 598         maxlen = (MAX(dest->mtu, 590) - ether_size - ip_size) & ~0x7;
 599         ip_off = ntohs(ip.ip_off);
 600         origf = ip_off & ~IP_OFFMASK;
 601         ip_off &= IP_OFFMASK;
 602
 603         while(todo) {
 604                 int len = todo > maxlen ? maxlen : todo;
 605                 memcpy(DATA(&fragment) + ether_size + ip_size, offset, len);
 606                 todo -= len;
 607                 offset += len;
 608
 609                 ip.ip_len = htons(ip_size + len);
 610                 ip.ip_off = htons(ip_off | origf | (todo ? IP_MF : 0));
 611                 ip.ip_sum = 0;
 612                 ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
 613                 memcpy(DATA(&fragment), DATA(packet), ether_size);
 614                 memcpy(DATA(&fragment) + ether_size, &ip, ip_size);
 615                 fragment.len = ether_size + ip_size + len;
 616
 617                 send_packet(dest, &fragment);
 618
 619                 ip_off += len / 8;
 620         }
 621 }
 622
 623 static void route_ipv4(node_t *source, vpn_packet_t *packet) {
 624         if(!checklength(source, packet, ether_size + ip_size)) {
 625                 return;
 626         }
 627
 628         subnet_t *subnet;
 629         node_t *via;
 630         ipv4_t dest;
 631
 632         memcpy(&dest, &DATA(packet)[30], sizeof(dest));
 633         subnet = lookup_subnet_ipv4(&dest);
 634
 635         if(!subnet) {
 636                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv4 destination address %d.%d.%d.%d",
 637                        source->name, source->hostname,
 638                        dest.x[0],
 639                        dest.x[1],
 640                        dest.x[2],
 641                        dest.x[3]);
 642
 643                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
 644                 return;
 645         }
 646
 647         if(!subnet->owner) {
 648                 route_broadcast(source, packet);
 649                 return;
 650         }
 651
 652         if(subnet->owner == source) {
 653                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
 654                 return;
 655         }
 656
 657         if(!subnet->owner->status.reachable) {
 658                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
 659                 return;
 660         }
 661
 662         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself) {
 663                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
 664                 return;
 665         }
 666
 667         if(decrement_ttl && source != myself && subnet->owner != myself)
 668                 if(!do_decrement_ttl(source, packet)) {
 669                         return;
 670                 }
 671
 672         if(priorityinheritance) {
 673                 packet->priority = DATA(packet)[15];
 674         }
 675
 676         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
 677
 678         if(via == source) {
 679                 logger(DEBUG_TRAFFIC, LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
 680                 return;
 681         }
 682
 683         if(directonly && subnet->owner != via) {
 684                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
 685                 return;
 686         }
 687
 688         if(via && packet->len > MAX(via->mtu, 590) && via != myself) {
 689                 logger(DEBUG_TRAFFIC, LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
 690
 691                 if(DATA(packet)[20] & 0x40) {
 692                         packet->len = MAX(via->mtu, 590);
 693                         route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
 694                 } else {
 695                         fragment_ipv4_packet(via, packet, ether_size);
 696                 }
 697
 698                 return;
 699         }
 700
 701         clamp_mss(source, via, packet);
 702
 703         send_packet(subnet->owner, packet);
 704 }
 705
 706 static void route_neighborsol(node_t *source, vpn_packet_t *packet);
 707
 708 static void route_ipv6(node_t *source, vpn_packet_t *packet) {
 709         if(!checklength(source, packet, ether_size + ip6_size)) {
 710                 return;
 711         }
 712
 713         if(DATA(packet)[20] == IPPROTO_ICMPV6 && checklength(source, packet, ether_size + ip6_size + icmp6_size) && DATA(packet)[54] == ND_NEIGHBOR_SOLICIT) {
 714                 route_neighborsol(source, packet);
 715                 return;
 716         }
 717
 718         subnet_t *subnet;
 719         node_t *via;
 720         ipv6_t dest;
 721
 722         memcpy(&dest, &DATA(packet)[38], sizeof(dest));
 723         subnet = lookup_subnet_ipv6(&dest);
 724
 725         if(!subnet) {
 726                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv6 destination address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
 727                        source->name, source->hostname,
 728                        ntohs(dest.x[0]),
 729                        ntohs(dest.x[1]),
 730                        ntohs(dest.x[2]),
 731                        ntohs(dest.x[3]),
 732                        ntohs(dest.x[4]),
 733                        ntohs(dest.x[5]),
 734                        ntohs(dest.x[6]),
 735                        ntohs(dest.x[7]));
 736
 737                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
 738                 return;
 739         }
 740
 741         if(!subnet->owner) {
 742                 route_broadcast(source, packet);
 743                 return;
 744         }
 745
 746         if(subnet->owner == source) {
 747                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
 748                 return;
 749         }
 750
 751         if(!subnet->owner->status.reachable) {
 752                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
 753                 return;
 754         }
 755
 756         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself) {
 757                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
 758                 return;
 759         }
 760
 761         if(decrement_ttl && source != myself && subnet->owner != myself)
 762                 if(!do_decrement_ttl(source, packet)) {
 763                         return;
 764                 }
 765
 766         if(priorityinheritance) {
 767                 packet->priority = ((DATA(packet)[14] & 0x0f) << 4) | (DATA(packet)[15] >> 4);
 768         }
 769
 770         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
 771
 772         if(via == source) {
 773                 logger(DEBUG_TRAFFIC, LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
 774                 return;
 775         }
 776
 777         if(directonly && subnet->owner != via) {
 778                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
 779                 return;
 780         }
 781
 782         if(via && packet->len > MAX(via->mtu, 1294) && via != myself) {
 783                 logger(DEBUG_TRAFFIC, LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
 784                 packet->len = MAX(via->mtu, 1294);
 785                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_PACKET_TOO_BIG, 0);
 786                 return;
 787         }
 788
 789         clamp_mss(source, via, packet);
 790
 791         send_packet(subnet->owner, packet);
 792 }
 793
 794 /* RFC 2461 */
 795
 796 static void route_neighborsol(node_t *source, vpn_packet_t *packet) {
 797         struct ip6_hdr ip6;
 798         struct nd_neighbor_solicit ns;
 799         struct nd_opt_hdr opt;
 800         subnet_t *subnet;
 801         uint16_t checksum;
 802         bool has_opt;
 803
 804         struct {
 805                 struct in6_addr ip6_src;
 806                 struct in6_addr ip6_dst;
 807                 uint32_t length;
 808                 uint32_t next;
 809         } pseudo;
 810
 811         if(!checklength(source, packet, ether_size + ip6_size + ns_size)) {
 812                 return;
 813         }
 814
 815         has_opt = packet->len >= ether_size + ip6_size + ns_size + opt_size + ETH_ALEN;
 816
 817         if(source != myself) {
 818                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got neighbor solicitation request from %s (%s) while in router mode!", source->name, source->hostname);
 819                 return;
 820         }
 821
 822         /* Copy headers from packet to structs on the stack */
 823
 824         memcpy(&ip6, DATA(packet) + ether_size, ip6_size);
 825         memcpy(&ns, DATA(packet) + ether_size + ip6_size, ns_size);
 826
 827         if(has_opt) {
 828                 memcpy(&opt, DATA(packet) + ether_size + ip6_size + ns_size, opt_size);
 829         }
 830
 831         /* First, snatch the source address from the neighbor solicitation packet */
 832
 833         if(overwrite_mac) {
 834                 memcpy(mymac.x, DATA(packet) + ETH_ALEN, ETH_ALEN);
 835         }
 836
 837         /* Check if this is a valid neighbor solicitation request */
 838
 839         if(ns.nd_ns_hdr.icmp6_type != ND_NEIGHBOR_SOLICIT ||
 840                         (has_opt && opt.nd_opt_type != ND_OPT_SOURCE_LINKADDR)) {
 841                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: received unknown type neighbor solicitation request");
 842                 return;
 843         }
 844
 845         /* Create pseudo header */
 846
 847         pseudo.ip6_src = ip6.ip6_src;
 848         pseudo.ip6_dst = ip6.ip6_dst;
 849
 850         if(has_opt) {
 851                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
 852         } else {
 853                 pseudo.length = htonl(ns_size);
 854         }
 855
 856         pseudo.next = htonl(IPPROTO_ICMPV6);
 857
 858         /* Generate checksum */
 859
 860         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
 861         checksum = inet_checksum(&ns, ns_size, checksum);
 862
 863         if(has_opt) {
 864                 checksum = inet_checksum(&opt, opt_size, checksum);
 865                 checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
 866         }
 867
 868         if(checksum) {
 869                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: checksum error for neighbor solicitation request");
 870                 return;
 871         }
 872
 873         /* Check if the IPv6 address exists on the VPN */
 874
 875         subnet = lookup_subnet_ipv6((ipv6_t *) &ns.nd_ns_target);
 876
 877         if(!subnet) {
 878                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: neighbor solicitation request for unknown address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
 879                        ntohs(((uint16_t *) &ns.nd_ns_target)[0]),
 880                        ntohs(((uint16_t *) &ns.nd_ns_target)[1]),
 881                        ntohs(((uint16_t *) &ns.nd_ns_target)[2]),
 882                        ntohs(((uint16_t *) &ns.nd_ns_target)[3]),
 883                        ntohs(((uint16_t *) &ns.nd_ns_target)[4]),
 884                        ntohs(((uint16_t *) &ns.nd_ns_target)[5]),
 885                        ntohs(((uint16_t *) &ns.nd_ns_target)[6]),
 886                        ntohs(((uint16_t *) &ns.nd_ns_target)[7]));
 887
 888                 return;
 889         }
 890
 891         /* Check if it is for our own subnet */
 892
 893         if(subnet->owner == myself) {
 894                 return;        /* silently ignore */
 895         }
 896
 897         if(decrement_ttl)
 898                 if(!do_decrement_ttl(source, packet)) {
 899                         return;
 900                 }
 901
 902         /* Create neighbor advertation reply */
 903
 904         memcpy(DATA(packet), DATA(packet) + ETH_ALEN, ETH_ALEN); /* copy destination address */
 905         DATA(packet)[ETH_ALEN * 2 - 1] ^= 0xFF;                  /* mangle source address so it looks like it's not from us */
 906
 907         ip6.ip6_dst = ip6.ip6_src;                               /* swap destination and source protocol address */
 908         ip6.ip6_src = ns.nd_ns_target;
 909
 910         if(has_opt) {
 911                 memcpy(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, DATA(packet) + ETH_ALEN, ETH_ALEN);        /* add fake source hard addr */
 912         }
 913
 914         ns.nd_ns_cksum = 0;
 915         ns.nd_ns_type = ND_NEIGHBOR_ADVERT;
 916         ns.nd_ns_reserved = htonl(0x40000000UL);                 /* Set solicited flag */
 917         opt.nd_opt_type = ND_OPT_TARGET_LINKADDR;
 918
 919         /* Create pseudo header */
 920
 921         pseudo.ip6_src = ip6.ip6_src;
 922         pseudo.ip6_dst = ip6.ip6_dst;
 923
 924         if(has_opt) {
 925                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
 926         } else {
 927                 pseudo.length = htonl(ns_size);
 928         }
 929
 930         pseudo.next = htonl(IPPROTO_ICMPV6);
 931
 932         /* Generate checksum */
 933
 934         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
 935         checksum = inet_checksum(&ns, ns_size, checksum);
 936
 937         if(has_opt) {
 938                 checksum = inet_checksum(&opt, opt_size, checksum);
 939                 checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
 940         }
 941
 942         ns.nd_ns_hdr.icmp6_cksum = checksum;
 943
 944         /* Copy structs on stack back to packet */
 945
 946         memcpy(DATA(packet) + ether_size, &ip6, ip6_size);
 947         memcpy(DATA(packet) + ether_size + ip6_size, &ns, ns_size);
 948
 949         if(has_opt) {
 950                 memcpy(DATA(packet) + ether_size + ip6_size + ns_size, &opt, opt_size);
 951         }
 952
 953         send_packet(source, packet);
 954 }
 955
 956 /* RFC 826 */
 957
 958 static void route_arp(node_t *source, vpn_packet_t *packet) {
 959         struct ether_arp arp;
 960         subnet_t *subnet;
 961         struct in_addr addr;
 962
 963         if(!checklength(source, packet, ether_size + arp_size)) {
 964                 return;
 965         }
 966
 967         if(source != myself) {
 968                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got ARP request from %s (%s) while in router mode!", source->name, source->hostname);
 969                 return;
 970         }
 971
 972         /* First, snatch the source address from the ARP packet */
 973
 974         if(overwrite_mac) {
 975                 memcpy(mymac.x, DATA(packet) + ETH_ALEN, ETH_ALEN);
 976         }
 977
 978         /* Copy headers from packet to structs on the stack */
 979
 980         memcpy(&arp, DATA(packet) + ether_size, arp_size);
 981
 982         /* Check if this is a valid ARP request */
 983
 984         if(ntohs(arp.arp_hrd) != ARPHRD_ETHER || ntohs(arp.arp_pro) != ETH_P_IP ||
 985                         arp.arp_hln != ETH_ALEN || arp.arp_pln != sizeof(addr) || ntohs(arp.arp_op) != ARPOP_REQUEST) {
 986                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: received unknown type ARP request");
 987                 return;
 988         }
 989
 990         /* Check if the IPv4 address exists on the VPN */
 991
 992         subnet = lookup_subnet_ipv4((ipv4_t *) &arp.arp_tpa);
 993
 994         if(!subnet) {
 995                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: ARP request for unknown address %d.%d.%d.%d",
 996                        arp.arp_tpa[0], arp.arp_tpa[1], arp.arp_tpa[2],
 997                        arp.arp_tpa[3]);
 998                 return;
 999         }
1000
1001         /* Check if it is for our own subnet */
1002
1003         if(subnet->owner == myself) {
1004                 return;        /* silently ignore */
1005         }
1006
1007         if(decrement_ttl)
1008                 if(!do_decrement_ttl(source, packet)) {
1009                         return;
1010                 }
1011
1012         memcpy(&addr, arp.arp_tpa, sizeof(addr));                 /* save protocol addr */
1013         memcpy(arp.arp_tpa, arp.arp_spa, sizeof(addr));           /* swap destination and source protocol address */
1014         memcpy(arp.arp_spa, &addr, sizeof(addr));                 /* ... */
1015
1016         memcpy(arp.arp_tha, arp.arp_sha, ETH_ALEN);              /* set target hard/proto addr */
1017         memcpy(arp.arp_sha, DATA(packet) + ETH_ALEN, ETH_ALEN);  /* set source hard/proto addr */
1018         arp.arp_sha[ETH_ALEN - 1] ^= 0xFF;                       /* for consistency with route_packet() */
1019         arp.arp_op = htons(ARPOP_REPLY);
1020
1021         /* Copy structs on stack back to packet */
1022
1023         memcpy(DATA(packet) + ether_size, &arp, arp_size);
1024
1025         send_packet(source, packet);
1026 }
1027
1028 static void route_mac(node_t *source, vpn_packet_t *packet) {
1029         subnet_t *subnet;
1030         mac_t dest;
1031
1032         /* Learn source address */
1033
1034         if(source == myself) {
1035                 mac_t src;
1036                 memcpy(&src, &DATA(packet)[6], sizeof(src));
1037                 learn_mac(&src);
1038         }
1039
1040         /* Lookup destination address */
1041
1042         memcpy(&dest, &DATA(packet)[0], sizeof(dest));
1043         subnet = lookup_subnet_mac(NULL, &dest);
1044
1045         if(!subnet || !subnet->owner) {
1046                 route_broadcast(source, packet);
1047                 return;
1048         }
1049
1050         if(subnet->owner == source) {
1051                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
1052                 return;
1053         }
1054
1055         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself) {
1056                 return;
1057         }
1058
1059         if(decrement_ttl && source != myself && subnet->owner != myself)
1060                 if(!do_decrement_ttl(source, packet)) {
1061                         return;
1062                 }
1063
1064         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
1065
1066         if(priorityinheritance) {
1067                 if(type == ETH_P_IP && packet->len >= ether_size + ip_size) {
1068                         packet->priority = DATA(packet)[15];
1069                 } else if(type == ETH_P_IPV6 && packet->len >= ether_size + ip6_size) {
1070                         packet->priority = ((DATA(packet)[14] & 0x0f) << 4) | (DATA(packet)[15] >> 4);
1071                 }
1072         }
1073
1074         // Handle packets larger than PMTU
1075
1076         node_t *via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
1077
1078         if(directonly && subnet->owner != via) {
1079                 return;
1080         }
1081
1082         if(via && packet->len > via->mtu && via != myself) {
1083                 logger(DEBUG_TRAFFIC, LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
1084                 length_t ethlen = 14;
1085
1086                 if(type == ETH_P_8021Q) {
1087                         type = DATA(packet)[16] << 8 | DATA(packet)[17];
1088                         ethlen += 4;
1089                 }
1090
1091                 if(type == ETH_P_IP && packet->len > 576 + ethlen) {
1092                         if(DATA(packet)[6 + ethlen] & 0x40) {
1093                                 packet->len = via->mtu;
1094                                 route_ipv4_unreachable(source, packet, ethlen, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
1095                         } else {
1096                                 fragment_ipv4_packet(via, packet, ethlen);
1097                         }
1098
1099                         return;
1100                 } else if(type == ETH_P_IPV6 && packet->len > 1280 + ethlen) {
1101                         packet->len = via->mtu;
1102                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_PACKET_TOO_BIG, 0);
1103                         return;
1104                 }
1105         }
1106
1107         clamp_mss(source, via, packet);
1108
1109         send_packet(subnet->owner, packet);
1110 }
1111
1112 static void send_pcap(vpn_packet_t *packet) {
1113         pcap = false;
1114
1115         for list_each(connection_t, c, connection_list) {
1116                 if(!c->status.pcap) {
1117                         continue;
1118                 }
1119
1120                 pcap = true;
1121                 int len = packet->len;
1122
1123                 if(c->outmaclength && c->outmaclength < len) {
1124                         len = c->outmaclength;
1125                 }
1126
1127                 if(send_request(c, "%d %d %d", CONTROL, REQ_PCAP, len)) {
1128                         send_meta(c, (char *)DATA(packet), len);
1129                 }
1130         }
1131 }
1132
1133 void route(node_t *source, vpn_packet_t *packet) {
1134         if(pcap) {
1135                 send_pcap(packet);
1136         }
1137
1138         if(forwarding_mode == FMODE_KERNEL && source != myself) {
1139                 send_packet(myself, packet);
1140                 return;
1141         }
1142
1143         if(!checklength(source, packet, ether_size)) {
1144                 return;
1145         }
1146
1147         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
1148
1149         switch(routing_mode) {
1150         case RMODE_ROUTER:
1151                 switch(type) {
1152                 case ETH_P_ARP:
1153                         route_arp(source, packet);
1154                         break;
1155
1156                 case ETH_P_IP:
1157                         route_ipv4(source, packet);
1158                         break;
1159
1160                 case ETH_P_IPV6:
1161                         route_ipv6(source, packet);
1162                         break;
1163
1164                 default:
1165                         logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown type %hx", source->name, source->hostname, type);
1166                         break;
1167                 }
1168
1169                 break;
1170
1171         case RMODE_SWITCH:
1172                 route_mac(source, packet);
1173                 break;
1174
1175         case RMODE_HUB:
1176                 route_broadcast(source, packet);
1177                 break;
1178         }
1179 }