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