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

   1 /*
   2     route.c -- routing
   3     Copyright (C) 2000-2005 Ivo Timmermans,
   4                   2000-2013 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 volatile int dummy;
  63 static timeout_t age_subnets_timeout;
  64
  65 /* RFC 1071 */
  66
  67 static uint16_t inet_checksum(void *data, int len, uint16_t prevsum) {
  68         uint16_t *p = data;
  69         uint32_t checksum = prevsum ^ 0xFFFF;
  70
  71         while(len >= 2) {
  72                 checksum += *p++;
  73                 len -= 2;
  74         }
  75
  76         if(len) {
  77                 checksum += *(uint8_t *)p;
  78         }
  79
  80         while(checksum >> 16) {
  81                 checksum = (checksum & 0xFFFF) + (checksum >> 16);
  82         }
  83
  84         // Work around a compiler optimization bug.
  85         if(checksum) {
  86                 dummy = 1;
  87         }
  88
  89         return ~checksum;
  90 }
  91
  92 static bool ratelimit(int frequency) {
  93         static time_t lasttime = 0;
  94         static int count = 0;
  95
  96         if(lasttime == now.tv_sec) {
  97                 if(count >= frequency) {
  98                         return true;
  99                 }
 100         } else {
 101                 lasttime = now.tv_sec;
 102                 count = 0;
 103         }
 104
 105         count++;
 106         return false;
 107 }
 108
 109 static bool checklength(node_t *source, vpn_packet_t *packet, length_t length) {
 110         if(packet->len < length) {
 111                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got too short packet from %s (%s)", source->name, source->hostname);
 112                 return false;
 113         } else {
 114                 return true;
 115         }
 116 }
 117
 118 static void swap_mac_addresses(vpn_packet_t *packet) {
 119         mac_t tmp;
 120         memcpy(&tmp, &DATA(packet)[0], sizeof(tmp));
 121         memcpy(&DATA(packet)[0], &DATA(packet)[6], sizeof(tmp));
 122         memcpy(&DATA(packet)[6], &tmp, sizeof(tmp));
 123 }
 124
 125 /* RFC 792 */
 126
 127 static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
 128         struct ip ip = {0};
 129         struct icmp icmp = {0};
 130
 131         struct in_addr ip_src;
 132         struct in_addr ip_dst;
 133         uint32_t oldlen;
 134
 135         if(ratelimit(3)) {
 136                 return;
 137         }
 138
 139         /* Swap Ethernet source and destination addresses */
 140
 141         swap_mac_addresses(packet);
 142
 143         /* Copy headers from packet into properly aligned structs on the stack */
 144
 145         memcpy(&ip, DATA(packet) + ether_size, ip_size);
 146
 147         /* Remember original source and destination */
 148
 149         ip_src = ip.ip_src;
 150         ip_dst = ip.ip_dst;
 151
 152         /* Try to reply with an IP address assigned to the local machine */
 153
 154         if(type == ICMP_TIME_EXCEEDED && code == ICMP_EXC_TTL) {
 155                 int sockfd = socket(AF_INET, SOCK_DGRAM, 0);
 156
 157                 if(sockfd != -1) {
 158                         struct sockaddr_in addr;
 159                         memset(&addr, 0, sizeof(addr));
 160                         addr.sin_family = AF_INET;
 161                         addr.sin_addr = ip.ip_src;
 162
 163                         if(!connect(sockfd, (const struct sockaddr *) &addr, sizeof(addr))) {
 164                                 memset(&addr, 0, sizeof(addr));
 165                                 addr.sin_family = AF_INET;
 166                                 socklen_t addrlen = sizeof(addr);
 167
 168                                 if(!getsockname(sockfd, (struct sockaddr *) &addr, &addrlen) && addrlen <= sizeof(addr)) {
 169                                         ip_dst = addr.sin_addr;
 170                                 }
 171                         }
 172
 173                         close(sockfd);
 174                 }
 175         }
 176
 177         oldlen = packet->len - ether_size;
 178
 179         if(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 180                 icmp.icmp_nextmtu = htons(packet->len - ether_size);
 181         }
 182
 183         if(oldlen >= IP_MSS - ip_size - icmp_size) {
 184                 oldlen = IP_MSS - ip_size - icmp_size;
 185         }
 186
 187         /* Copy first part of original contents to ICMP message */
 188
 189         memmove(DATA(packet) + ether_size + ip_size + icmp_size, DATA(packet) + ether_size, oldlen);
 190
 191         /* Fill in IPv4 header */
 192
 193         ip.ip_v = 4;
 194         ip.ip_hl = ip_size / 4;
 195         ip.ip_tos = 0;
 196         ip.ip_len = htons(ip_size + icmp_size + oldlen);
 197         ip.ip_id = 0;
 198         ip.ip_off = 0;
 199         ip.ip_ttl = 255;
 200         ip.ip_p = IPPROTO_ICMP;
 201         ip.ip_sum = 0;
 202         ip.ip_src = ip_dst;
 203         ip.ip_dst = ip_src;
 204
 205         ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
 206
 207         /* Fill in ICMP header */
 208
 209         icmp.icmp_type = type;
 210         icmp.icmp_code = code;
 211         icmp.icmp_cksum = 0;
 212
 213         icmp.icmp_cksum = inet_checksum(&icmp, icmp_size, ~0);
 214         icmp.icmp_cksum = inet_checksum(DATA(packet) + ether_size + ip_size + icmp_size, oldlen, icmp.icmp_cksum);
 215
 216         /* Copy structs on stack back to packet */
 217
 218         memcpy(DATA(packet) + ether_size, &ip, ip_size);
 219         memcpy(DATA(packet) + ether_size + ip_size, &icmp, icmp_size);
 220
 221         packet->len = ether_size + ip_size + icmp_size + oldlen;
 222
 223         send_packet(source, packet);
 224 }
 225
 226 /* RFC 2463 */
 227
 228 static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
 229         struct ip6_hdr ip6;
 230         struct icmp6_hdr icmp6 = {0};
 231         uint16_t checksum;
 232
 233         struct {
 234                 struct in6_addr ip6_src;        /* source address */
 235                 struct in6_addr ip6_dst;        /* destination address */
 236                 uint32_t length;
 237                 uint32_t next;
 238         } pseudo;
 239
 240         if(ratelimit(3)) {
 241                 return;
 242         }
 243
 244         /* Swap Ethernet source and destination addresses */
 245
 246         swap_mac_addresses(packet);
 247
 248         /* Copy headers from packet to structs on the stack */
 249
 250         memcpy(&ip6, DATA(packet) + ether_size, ip6_size);
 251
 252         /* Remember original source and destination */
 253
 254         pseudo.ip6_src = ip6.ip6_dst;
 255         pseudo.ip6_dst = ip6.ip6_src;
 256
 257         /* Try to reply with an IP address assigned to the local machine */
 258
 259         if(type == ICMP6_TIME_EXCEEDED && code == ICMP6_TIME_EXCEED_TRANSIT) {
 260                 int sockfd = socket(AF_INET6, SOCK_DGRAM, 0);
 261
 262                 if(sockfd != -1) {
 263                         struct sockaddr_in6 addr;
 264                         memset(&addr, 0, sizeof(addr));
 265                         addr.sin6_family = AF_INET6;
 266                         addr.sin6_addr = ip6.ip6_src;
 267
 268                         if(!connect(sockfd, (const struct sockaddr *) &addr, sizeof(addr))) {
 269                                 memset(&addr, 0, sizeof(addr));
 270                                 addr.sin6_family = AF_INET6;
 271                                 socklen_t addrlen = sizeof(addr);
 272
 273                                 if(!getsockname(sockfd, (struct sockaddr *) &addr, &addrlen) && addrlen <= sizeof(addr)) {
 274                                         pseudo.ip6_src = addr.sin6_addr;
 275                                 }
 276                         }
 277
 278                         close(sockfd);
 279                 }
 280         }
 281
 282         pseudo.length = packet->len - ether_size;
 283
 284         if(type == ICMP6_PACKET_TOO_BIG) {
 285                 icmp6.icmp6_mtu = htonl(pseudo.length);
 286         }
 287
 288         if(pseudo.length >= IP_MSS - ip6_size - icmp6_size) {
 289                 pseudo.length = IP_MSS - ip6_size - icmp6_size;
 290         }
 291
 292         /* Copy first part of original contents to ICMP message */
 293
 294         memmove(DATA(packet) + ether_size + ip6_size + icmp6_size, DATA(packet) + ether_size, pseudo.length);
 295
 296         /* Fill in IPv6 header */
 297
 298         ip6.ip6_flow = htonl(0x60000000UL);
 299         ip6.ip6_plen = htons(icmp6_size + pseudo.length);
 300         ip6.ip6_nxt = IPPROTO_ICMPV6;
 301         ip6.ip6_hlim = 255;
 302         ip6.ip6_src = pseudo.ip6_src;
 303         ip6.ip6_dst = pseudo.ip6_dst;
 304
 305         /* Fill in ICMP header */
 306
 307         icmp6.icmp6_type = type;
 308         icmp6.icmp6_code = code;
 309         icmp6.icmp6_cksum = 0;
 310
 311         /* Create pseudo header */
 312
 313         pseudo.length = htonl(icmp6_size + pseudo.length);
 314         pseudo.next = htonl(IPPROTO_ICMPV6);
 315
 316         /* Generate checksum */
 317
 318         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
 319         checksum = inet_checksum(&icmp6, icmp6_size, checksum);
 320         checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + icmp6_size, ntohl(pseudo.length) - icmp6_size, checksum);
 321
 322         icmp6.icmp6_cksum = checksum;
 323
 324         /* Copy structs on stack back to packet */
 325
 326         memcpy(DATA(packet) + ether_size, &ip6, ip6_size);
 327         memcpy(DATA(packet) + ether_size + ip6_size, &icmp6, icmp6_size);
 328
 329         packet->len = ether_size + ip6_size + ntohl(pseudo.length);
 330
 331         send_packet(source, packet);
 332 }
 333
 334 static bool do_decrement_ttl(node_t *source, vpn_packet_t *packet) {
 335         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
 336         length_t ethlen = ether_size;
 337
 338         if(type == ETH_P_8021Q) {
 339                 type = DATA(packet)[16] << 8 | DATA(packet)[17];
 340                 ethlen += 4;
 341         }
 342
 343         switch(type) {
 344         case ETH_P_IP:
 345                 if(!checklength(source, packet, ethlen + ip_size)) {
 346                         return false;
 347                 }
 348
 349                 if(DATA(packet)[ethlen + 8] <= 1) {
 350                         if(DATA(packet)[ethlen + 11] != IPPROTO_ICMP || DATA(packet)[ethlen + 32] != ICMP_TIME_EXCEEDED) {
 351                                 route_ipv4_unreachable(source, packet, ethlen, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL);
 352                         }
 353
 354                         return false;
 355                 }
 356
 357                 uint16_t old = DATA(packet)[ethlen + 8] << 8 | DATA(packet)[ethlen + 9];
 358                 DATA(packet)[ethlen + 8]--;
 359                 uint16_t new = DATA(packet)[ethlen + 8] << 8 | DATA(packet)[ethlen + 9];
 360
 361                 uint32_t checksum = DATA(packet)[ethlen + 10] << 8 | DATA(packet)[ethlen + 11];
 362                 checksum += old + (~new & 0xFFFF);
 363
 364                 while(checksum >> 16) {
 365                         checksum = (checksum & 0xFFFF) + (checksum >> 16);
 366                 }
 367
 368                 DATA(packet)[ethlen + 10] = checksum >> 8;
 369                 DATA(packet)[ethlen + 11] = checksum & 0xff;
 370
 371                 return true;
 372
 373         case ETH_P_IPV6:
 374                 if(!checklength(source, packet, ethlen + ip6_size)) {
 375                         return false;
 376                 }
 377
 378                 if(DATA(packet)[ethlen + 7] <= 1) {
 379                         if(DATA(packet)[ethlen + 6] != IPPROTO_ICMPV6 || DATA(packet)[ethlen + 40] != ICMP6_TIME_EXCEEDED) {
 380                                 route_ipv6_unreachable(source, packet, ethlen, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT);
 381                         }
 382
 383                         return false;
 384                 }
 385
 386                 DATA(packet)[ethlen + 7]--;
 387
 388                 return true;
 389
 390         default:
 391                 return true;
 392         }
 393 }
 394
 395 static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *packet) {
 396         if(!source || !via || !(via->options & OPTION_CLAMP_MSS)) {
 397                 return;
 398         }
 399
 400         uint16_t mtu = source->mtu;
 401
 402         if(via != myself && via->mtu < mtu) {
 403                 mtu = via->mtu;
 404         }
 405
 406         /* Find TCP header */
 407         int start = ether_size;
 408         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
 409
 410         if(type == ETH_P_8021Q) {
 411                 start += 4;
 412                 type = DATA(packet)[16] << 8 | DATA(packet)[17];
 413         }
 414
 415         /* IP in IP (RFC 2003) packet */
 416         if(type == ETH_P_IP && DATA(packet)[start + 9] == 4) {
 417                 start += 20;
 418         }
 419
 420         if(packet->len <= start + 20) {
 421                 return;
 422         }
 423
 424         if(type == ETH_P_IP && DATA(packet)[start + 9] == 6) {
 425                 start += (DATA(packet)[start] & 0xf) * 4;
 426         } else if(type == ETH_P_IPV6 && DATA(packet)[start + 6] == 6) {
 427                 start += 40;
 428         } else {
 429                 return;
 430         }
 431
 432         if(packet->len <= start + 20) {
 433                 return;
 434         }
 435
 436         /* Use data offset field to calculate length of options field */
 437         int len = ((DATA(packet)[start + 12] >> 4) - 5) * 4;
 438
 439         if(packet->len < start + 20 + len) {
 440                 return;
 441         }
 442
 443         /* Search for MSS option header */
 444         for(int i = 0; i < len;) {
 445                 if(DATA(packet)[start + 20 + i] == 0) {
 446                         break;
 447                 }
 448
 449                 if(DATA(packet)[start + 20 + i] == 1) {
 450                         i++;
 451                         continue;
 452                 }
 453
 454                 if(i > len - 2 || i > len - DATA(packet)[start + 21 + i]) {
 455                         break;
 456                 }
 457
 458                 if(DATA(packet)[start + 20 + i] != 2) {
 459                         if(DATA(packet)[start + 21 + i] < 2) {
 460                                 break;
 461                         }
 462
 463                         i += DATA(packet)[start + 21 + i];
 464                         continue;
 465                 }
 466
 467                 if(DATA(packet)[start + 21] != 4) {
 468                         break;
 469                 }
 470
 471                 /* Found it */
 472                 uint16_t oldmss = DATA(packet)[start + 22 + i] << 8 | DATA(packet)[start + 23 + i];
 473                 uint16_t newmss = mtu - start - 20;
 474                 uint32_t csum = DATA(packet)[start + 16] << 8 | DATA(packet)[start + 17];
 475
 476                 if(oldmss <= newmss) {
 477                         break;
 478                 }
 479
 480                 logger(DEBUG_TRAFFIC, LOG_INFO, "Clamping MSS of packet from %s to %s to %d", source->name, via->name, newmss);
 481
 482                 /* Update the MSS value and the checksum */
 483                 DATA(packet)[start + 22 + i] = newmss >> 8;
 484                 DATA(packet)[start + 23 + i] = newmss & 0xff;
 485                 csum ^= 0xffff;
 486                 csum += oldmss ^ 0xffff;
 487                 csum += newmss;
 488                 csum = (csum & 0xffff) + (csum >> 16);
 489                 csum += csum >> 16;
 490                 csum ^= 0xffff;
 491                 DATA(packet)[start + 16] = csum >> 8;
 492                 DATA(packet)[start + 17] = csum;
 493                 break;
 494         }
 495 }
 496
 497 static void age_subnets(void *data) {
 498         bool left = false;
 499
 500         for splay_each(subnet_t, s, myself->subnet_tree) {
 501                 if(s->expires && s->expires < now.tv_sec) {
 502                         if(debug_level >= DEBUG_TRAFFIC) {
 503                                 char netstr[MAXNETSTR];
 504
 505                                 if(net2str(netstr, sizeof(netstr), s)) {
 506                                         logger(DEBUG_TRAFFIC, LOG_INFO, "Subnet %s expired", netstr);
 507                                 }
 508                         }
 509
 510                         for list_each(connection_t, c, connection_list)
 511                                 if(c->edge) {
 512                                         send_del_subnet(c, s);
 513                                 }
 514
 515                         subnet_del(myself, s);
 516                 } else {
 517                         if(s->expires) {
 518                                 left = true;
 519                         }
 520                 }
 521         }
 522
 523         if(left)
 524                 timeout_set(&age_subnets_timeout, &(struct timeval) {
 525                 10, rand() % 100000
 526         });
 527 }
 528
 529 static void learn_mac(mac_t *address) {
 530         subnet_t *subnet = lookup_subnet_mac(myself, address);
 531
 532         /* If we don't know this MAC address yet, store it */
 533
 534         if(!subnet) {
 535                 logger(DEBUG_TRAFFIC, LOG_INFO, "Learned new MAC address %x:%x:%x:%x:%x:%x",
 536                        address->x[0], address->x[1], address->x[2], address->x[3],
 537                        address->x[4], address->x[5]);
 538
 539                 subnet = new_subnet();
 540                 subnet->type = SUBNET_MAC;
 541                 subnet->expires = now.tv_sec + macexpire;
 542                 subnet->net.mac.address = *address;
 543                 subnet->weight = 10;
 544                 subnet_add(myself, subnet);
 545                 subnet_update(myself, subnet, true);
 546
 547                 /* And tell all other tinc daemons it's our MAC */
 548
 549                 for list_each(connection_t, c, connection_list)
 550                         if(c->edge) {
 551                                 send_add_subnet(c, subnet);
 552                         }
 553
 554                 timeout_add(&age_subnets_timeout, age_subnets, NULL, &(struct timeval) {
 555                         10, rand() % 100000
 556                 });
 557         } else {
 558                 if(subnet->expires) {
 559                         subnet->expires = now.tv_sec + macexpire;
 560                 }
 561         }
 562 }
 563
 564 static void route_broadcast(node_t *source, vpn_packet_t *packet) {
 565         if(decrement_ttl && source != myself)
 566                 if(!do_decrement_ttl(source, packet)) {
 567                         return;
 568                 }
 569
 570         broadcast_packet(source, packet);
 571 }
 572
 573 /* RFC 791 */
 574
 575 static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet, length_t ether_size) {
 576         struct ip ip;
 577         vpn_packet_t fragment;
 578         int maxlen, todo;
 579         uint8_t *offset;
 580         uint16_t ip_off, origf;
 581
 582         memcpy(&ip, DATA(packet) + ether_size, ip_size);
 583         fragment.priority = packet->priority;
 584         fragment.offset = DEFAULT_PACKET_OFFSET;
 585
 586         if(ip.ip_hl != ip_size / 4) {
 587                 return;
 588         }
 589
 590         todo = ntohs(ip.ip_len) - ip_size;
 591
 592         if(ether_size + ip_size + todo != packet->len) {
 593                 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));
 594                 return;
 595         }
 596
 597         logger(DEBUG_TRAFFIC, LOG_INFO, "Fragmenting packet of %d bytes to %s (%s)", packet->len, dest->name, dest->hostname);
 598
 599         offset = DATA(packet) + ether_size + ip_size;
 600         maxlen = (dest->mtu - ether_size - ip_size) & ~0x7;
 601         ip_off = ntohs(ip.ip_off);
 602         origf = ip_off & ~IP_OFFMASK;
 603         ip_off &= IP_OFFMASK;
 604
 605         while(todo) {
 606                 int len = todo > maxlen ? maxlen : todo;
 607                 memcpy(DATA(&fragment) + ether_size + ip_size, offset, len);
 608                 todo -= len;
 609                 offset += len;
 610
 611                 ip.ip_len = htons(ip_size + len);
 612                 ip.ip_off = htons(ip_off | origf | (todo ? IP_MF : 0));
 613                 ip.ip_sum = 0;
 614                 ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
 615                 memcpy(DATA(&fragment), DATA(packet), ether_size);
 616                 memcpy(DATA(&fragment) + ether_size, &ip, ip_size);
 617                 fragment.len = ether_size + ip_size + len;
 618
 619                 send_packet(dest, &fragment);
 620
 621                 ip_off += len / 8;
 622         }
 623 }
 624
 625 static void route_ipv4(node_t *source, vpn_packet_t *packet) {
 626         if(!checklength(source, packet, ether_size + ip_size)) {
 627                 return;
 628         }
 629
 630         subnet_t *subnet;
 631         node_t *via;
 632         ipv4_t dest;
 633
 634         memcpy(&dest, &DATA(packet)[30], sizeof(dest));
 635         subnet = lookup_subnet_ipv4(&dest);
 636
 637         if(!subnet) {
 638                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv4 destination address %d.%d.%d.%d",
 639                        source->name, source->hostname,
 640                        dest.x[0],
 641                        dest.x[1],
 642                        dest.x[2],
 643                        dest.x[3]);
 644
 645                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
 646                 return;
 647         }
 648
 649         if(!subnet->owner) {
 650                 route_broadcast(source, packet);
 651                 return;
 652         }
 653
 654         if(subnet->owner == source) {
 655                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
 656                 return;
 657         }
 658
 659         if(!subnet->owner->status.reachable) {
 660                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
 661                 return;
 662         }
 663
 664         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself) {
 665                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
 666                 return;
 667         }
 668
 669         if(decrement_ttl && source != myself && subnet->owner != myself)
 670                 if(!do_decrement_ttl(source, packet)) {
 671                         return;
 672                 }
 673
 674         if(priorityinheritance) {
 675                 packet->priority = DATA(packet)[15];
 676         }
 677
 678         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
 679
 680         if(via == source) {
 681                 logger(DEBUG_TRAFFIC, LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
 682                 return;
 683         }
 684
 685         if(directonly && subnet->owner != via) {
 686                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
 687                 return;
 688         }
 689
 690         if(via && packet->len > MAX(via->mtu, 590) && via != myself) {
 691                 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);
 692
 693                 if(DATA(packet)[20] & 0x40) {
 694                         packet->len = MAX(via->mtu, 590);
 695                         route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
 696                 } else {
 697                         fragment_ipv4_packet(via, packet, ether_size);
 698                 }
 699
 700                 return;
 701         }
 702
 703         clamp_mss(source, via, packet);
 704
 705         send_packet(subnet->owner, packet);
 706 }
 707
 708 static void route_neighborsol(node_t *source, vpn_packet_t *packet);
 709
 710 static void route_ipv6(node_t *source, vpn_packet_t *packet) {
 711         if(!checklength(source, packet, ether_size + ip6_size)) {
 712                 return;
 713         }
 714
 715         if(DATA(packet)[20] == IPPROTO_ICMPV6 && checklength(source, packet, ether_size + ip6_size + icmp6_size) && DATA(packet)[54] == ND_NEIGHBOR_SOLICIT) {
 716                 route_neighborsol(source, packet);
 717                 return;
 718         }
 719
 720         subnet_t *subnet;
 721         node_t *via;
 722         ipv6_t dest;
 723
 724         memcpy(&dest, &DATA(packet)[38], sizeof(dest));
 725         subnet = lookup_subnet_ipv6(&dest);
 726
 727         if(!subnet) {
 728                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv6 destination address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
 729                        source->name, source->hostname,
 730                        ntohs(dest.x[0]),
 731                        ntohs(dest.x[1]),
 732                        ntohs(dest.x[2]),
 733                        ntohs(dest.x[3]),
 734                        ntohs(dest.x[4]),
 735                        ntohs(dest.x[5]),
 736                        ntohs(dest.x[6]),
 737                        ntohs(dest.x[7]));
 738
 739                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
 740                 return;
 741         }
 742
 743         if(!subnet->owner) {
 744                 route_broadcast(source, packet);
 745                 return;
 746         }
 747
 748         if(subnet->owner == source) {
 749                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
 750                 return;
 751         }
 752
 753         if(!subnet->owner->status.reachable) {
 754                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
 755                 return;
 756         }
 757
 758         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself) {
 759                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
 760                 return;
 761         }
 762
 763         if(decrement_ttl && source != myself && subnet->owner != myself)
 764                 if(!do_decrement_ttl(source, packet)) {
 765                         return;
 766                 }
 767
 768         if(priorityinheritance) {
 769                 packet->priority = ((DATA(packet)[14] & 0x0f) << 4) | (DATA(packet)[15] >> 4);
 770         }
 771
 772         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
 773
 774         if(via == source) {
 775                 logger(DEBUG_TRAFFIC, LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
 776                 return;
 777         }
 778
 779         if(directonly && subnet->owner != via) {
 780                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
 781                 return;
 782         }
 783
 784         if(via && packet->len > MAX(via->mtu, 1294) && via != myself) {
 785                 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);
 786                 packet->len = MAX(via->mtu, 1294);
 787                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_PACKET_TOO_BIG, 0);
 788                 return;
 789         }
 790
 791         clamp_mss(source, via, packet);
 792
 793         send_packet(subnet->owner, packet);
 794 }
 795
 796 /* RFC 2461 */
 797
 798 static void route_neighborsol(node_t *source, vpn_packet_t *packet) {
 799         struct ip6_hdr ip6;
 800         struct nd_neighbor_solicit ns;
 801         struct nd_opt_hdr opt;
 802         subnet_t *subnet;
 803         uint16_t checksum;
 804         bool has_opt;
 805
 806         struct {
 807                 struct in6_addr ip6_src;
 808                 struct in6_addr ip6_dst;
 809                 uint32_t length;
 810                 uint32_t next;
 811         } pseudo;
 812
 813         if(!checklength(source, packet, ether_size + ip6_size + ns_size)) {
 814                 return;
 815         }
 816
 817         has_opt = packet->len >= ether_size + ip6_size + ns_size + opt_size + ETH_ALEN;
 818
 819         if(source != myself) {
 820                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got neighbor solicitation request from %s (%s) while in router mode!", source->name, source->hostname);
 821                 return;
 822         }
 823
 824         /* Copy headers from packet to structs on the stack */
 825
 826         memcpy(&ip6, DATA(packet) + ether_size, ip6_size);
 827         memcpy(&ns, DATA(packet) + ether_size + ip6_size, ns_size);
 828
 829         if(has_opt) {
 830                 memcpy(&opt, DATA(packet) + ether_size + ip6_size + ns_size, opt_size);
 831         }
 832
 833         /* First, snatch the source address from the neighbor solicitation packet */
 834
 835         if(overwrite_mac) {
 836                 memcpy(mymac.x, DATA(packet) + ETH_ALEN, ETH_ALEN);
 837         }
 838
 839         /* Check if this is a valid neighbor solicitation request */
 840
 841         if(ns.nd_ns_hdr.icmp6_type != ND_NEIGHBOR_SOLICIT ||
 842                         (has_opt && opt.nd_opt_type != ND_OPT_SOURCE_LINKADDR)) {
 843                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: received unknown type neighbor solicitation request");
 844                 return;
 845         }
 846
 847         /* Create pseudo header */
 848
 849         pseudo.ip6_src = ip6.ip6_src;
 850         pseudo.ip6_dst = ip6.ip6_dst;
 851
 852         if(has_opt) {
 853                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
 854         } else {
 855                 pseudo.length = htonl(ns_size);
 856         }
 857
 858         pseudo.next = htonl(IPPROTO_ICMPV6);
 859
 860         /* Generate checksum */
 861
 862         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
 863         checksum = inet_checksum(&ns, ns_size, checksum);
 864
 865         if(has_opt) {
 866                 checksum = inet_checksum(&opt, opt_size, checksum);
 867                 checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
 868         }
 869
 870         if(checksum) {
 871                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: checksum error for neighbor solicitation request");
 872                 return;
 873         }
 874
 875         /* Check if the IPv6 address exists on the VPN */
 876
 877         subnet = lookup_subnet_ipv6((ipv6_t *) &ns.nd_ns_target);
 878
 879         if(!subnet) {
 880                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: neighbor solicitation request for unknown address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
 881                        ntohs(((uint16_t *) &ns.nd_ns_target)[0]),
 882                        ntohs(((uint16_t *) &ns.nd_ns_target)[1]),
 883                        ntohs(((uint16_t *) &ns.nd_ns_target)[2]),
 884                        ntohs(((uint16_t *) &ns.nd_ns_target)[3]),
 885                        ntohs(((uint16_t *) &ns.nd_ns_target)[4]),
 886                        ntohs(((uint16_t *) &ns.nd_ns_target)[5]),
 887                        ntohs(((uint16_t *) &ns.nd_ns_target)[6]),
 888                        ntohs(((uint16_t *) &ns.nd_ns_target)[7]));
 889
 890                 return;
 891         }
 892
 893         /* Check if it is for our own subnet */
 894
 895         if(subnet->owner == myself) {
 896                 return;        /* silently ignore */
 897         }
 898
 899         if(decrement_ttl)
 900                 if(!do_decrement_ttl(source, packet)) {
 901                         return;
 902                 }
 903
 904         /* Create neighbor advertation reply */
 905
 906         memcpy(DATA(packet), DATA(packet) + ETH_ALEN, ETH_ALEN); /* copy destination address */
 907         DATA(packet)[ETH_ALEN * 2 - 1] ^= 0xFF;                  /* mangle source address so it looks like it's not from us */
 908
 909         ip6.ip6_dst = ip6.ip6_src;                               /* swap destination and source protocoll address */
 910         ip6.ip6_src = ns.nd_ns_target;
 911
 912         if(has_opt) {
 913                 memcpy(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, DATA(packet) + ETH_ALEN, ETH_ALEN);        /* add fake source hard addr */
 914         }
 915
 916         ns.nd_ns_cksum = 0;
 917         ns.nd_ns_type = ND_NEIGHBOR_ADVERT;
 918         ns.nd_ns_reserved = htonl(0x40000000UL);                 /* Set solicited flag */
 919         opt.nd_opt_type = ND_OPT_TARGET_LINKADDR;
 920
 921         /* Create pseudo header */
 922
 923         pseudo.ip6_src = ip6.ip6_src;
 924         pseudo.ip6_dst = ip6.ip6_dst;
 925
 926         if(has_opt) {
 927                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
 928         } else {
 929                 pseudo.length = htonl(ns_size);
 930         }
 931
 932         pseudo.next = htonl(IPPROTO_ICMPV6);
 933
 934         /* Generate checksum */
 935
 936         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
 937         checksum = inet_checksum(&ns, ns_size, checksum);
 938
 939         if(has_opt) {
 940                 checksum = inet_checksum(&opt, opt_size, checksum);
 941                 checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
 942         }
 943
 944         ns.nd_ns_hdr.icmp6_cksum = checksum;
 945
 946         /* Copy structs on stack back to packet */
 947
 948         memcpy(DATA(packet) + ether_size, &ip6, ip6_size);
 949         memcpy(DATA(packet) + ether_size + ip6_size, &ns, ns_size);
 950
 951         if(has_opt) {
 952                 memcpy(DATA(packet) + ether_size + ip6_size + ns_size, &opt, opt_size);
 953         }
 954
 955         send_packet(source, packet);
 956 }
 957
 958 /* RFC 826 */
 959
 960 static void route_arp(node_t *source, vpn_packet_t *packet) {
 961         struct ether_arp arp;
 962         subnet_t *subnet;
 963         struct in_addr addr;
 964
 965         if(!checklength(source, packet, ether_size + arp_size)) {
 966                 return;
 967         }
 968
 969         if(source != myself) {
 970                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got ARP request from %s (%s) while in router mode!", source->name, source->hostname);
 971                 return;
 972         }
 973
 974         /* First, snatch the source address from the ARP packet */
 975
 976         if(overwrite_mac) {
 977                 memcpy(mymac.x, DATA(packet) + ETH_ALEN, ETH_ALEN);
 978         }
 979
 980         /* Copy headers from packet to structs on the stack */
 981
 982         memcpy(&arp, DATA(packet) + ether_size, arp_size);
 983
 984         /* Check if this is a valid ARP request */
 985
 986         if(ntohs(arp.arp_hrd) != ARPHRD_ETHER || ntohs(arp.arp_pro) != ETH_P_IP ||
 987                         arp.arp_hln != ETH_ALEN || arp.arp_pln != sizeof(addr) || ntohs(arp.arp_op) != ARPOP_REQUEST) {
 988                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: received unknown type ARP request");
 989                 return;
 990         }
 991
 992         /* Check if the IPv4 address exists on the VPN */
 993
 994         subnet = lookup_subnet_ipv4((ipv4_t *) &arp.arp_tpa);
 995
 996         if(!subnet) {
 997                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: ARP request for unknown address %d.%d.%d.%d",
 998                        arp.arp_tpa[0], arp.arp_tpa[1], arp.arp_tpa[2],
 999                        arp.arp_tpa[3]);
1000                 return;
1001         }
1002
1003         /* Check if it is for our own subnet */
1004
1005         if(subnet->owner == myself) {
1006                 return;        /* silently ignore */
1007         }
1008
1009         if(decrement_ttl)
1010                 if(!do_decrement_ttl(source, packet)) {
1011                         return;
1012                 }
1013
1014         memcpy(&addr, arp.arp_tpa, sizeof(addr));                 /* save protocol addr */
1015         memcpy(arp.arp_tpa, arp.arp_spa, sizeof(addr));           /* swap destination and source protocol address */
1016         memcpy(arp.arp_spa, &addr, sizeof(addr));                 /* ... */
1017
1018         memcpy(arp.arp_tha, arp.arp_sha, ETH_ALEN);              /* set target hard/proto addr */
1019         memcpy(arp.arp_sha, DATA(packet) + ETH_ALEN, ETH_ALEN);  /* set source hard/proto addr */
1020         arp.arp_sha[ETH_ALEN - 1] ^= 0xFF;                       /* for consistency with route_packet() */
1021         arp.arp_op = htons(ARPOP_REPLY);
1022
1023         /* Copy structs on stack back to packet */
1024
1025         memcpy(DATA(packet) + ether_size, &arp, arp_size);
1026
1027         send_packet(source, packet);
1028 }
1029
1030 static void route_mac(node_t *source, vpn_packet_t *packet) {
1031         subnet_t *subnet;
1032         mac_t dest;
1033
1034         /* Learn source address */
1035
1036         if(source == myself) {
1037                 mac_t src;
1038                 memcpy(&src, &DATA(packet)[6], sizeof(src));
1039                 learn_mac(&src);
1040         }
1041
1042         /* Lookup destination address */
1043
1044         memcpy(&dest, &DATA(packet)[0], sizeof(dest));
1045         subnet = lookup_subnet_mac(NULL, &dest);
1046
1047         if(!subnet || !subnet->owner) {
1048                 route_broadcast(source, packet);
1049                 return;
1050         }
1051
1052         if(subnet->owner == source) {
1053                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
1054                 return;
1055         }
1056
1057         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself) {
1058                 return;
1059         }
1060
1061         if(decrement_ttl && source != myself && subnet->owner != myself)
1062                 if(!do_decrement_ttl(source, packet)) {
1063                         return;
1064                 }
1065
1066         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
1067
1068         if(priorityinheritance) {
1069                 if(type == ETH_P_IP && packet->len >= ether_size + ip_size) {
1070                         packet->priority = DATA(packet)[15];
1071                 } else if(type == ETH_P_IPV6 && packet->len >= ether_size + ip6_size) {
1072                         packet->priority = ((DATA(packet)[14] & 0x0f) << 4) | (DATA(packet)[15] >> 4);
1073                 }
1074         }
1075
1076         // Handle packets larger than PMTU
1077
1078         node_t *via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
1079
1080         if(directonly && subnet->owner != via) {
1081                 return;
1082         }
1083
1084         if(via && packet->len > via->mtu && via != myself) {
1085                 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);
1086                 length_t ethlen = 14;
1087
1088                 if(type == ETH_P_8021Q) {
1089                         type = DATA(packet)[16] << 8 | DATA(packet)[17];
1090                         ethlen += 4;
1091                 }
1092
1093                 if(type == ETH_P_IP && packet->len > 576 + ethlen) {
1094                         if(DATA(packet)[6 + ethlen] & 0x40) {
1095                                 packet->len = via->mtu;
1096                                 route_ipv4_unreachable(source, packet, ethlen, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
1097                         } else {
1098                                 fragment_ipv4_packet(via, packet, ethlen);
1099                         }
1100
1101                         return;
1102                 } else if(type == ETH_P_IPV6 && packet->len > 1280 + ethlen) {
1103                         packet->len = via->mtu;
1104                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_PACKET_TOO_BIG, 0);
1105                         return;
1106                 }
1107         }
1108
1109         clamp_mss(source, via, packet);
1110
1111         send_packet(subnet->owner, packet);
1112 }
1113
1114 static void send_pcap(vpn_packet_t *packet) {
1115         pcap = false;
1116
1117         for list_each(connection_t, c, connection_list) {
1118                 if(!c->status.pcap) {
1119                         continue;
1120                 }
1121
1122                 pcap = true;
1123                 int len = packet->len;
1124
1125                 if(c->outmaclength && c->outmaclength < len) {
1126                         len = c->outmaclength;
1127                 }
1128
1129                 if(send_request(c, "%d %d %d", CONTROL, REQ_PCAP, len)) {
1130                         send_meta(c, (char *)DATA(packet), len);
1131                 }
1132         }
1133 }
1134
1135 void route(node_t *source, vpn_packet_t *packet) {
1136         if(pcap) {
1137                 send_pcap(packet);
1138         }
1139
1140         if(forwarding_mode == FMODE_KERNEL && source != myself) {
1141                 send_packet(myself, packet);
1142                 return;
1143         }
1144
1145         if(!checklength(source, packet, ether_size)) {
1146                 return;
1147         }
1148
1149         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
1150
1151         switch(routing_mode) {
1152         case RMODE_ROUTER:
1153                 switch(type) {
1154                 case ETH_P_ARP:
1155                         route_arp(source, packet);
1156                         break;
1157
1158                 case ETH_P_IP:
1159                         route_ipv4(source, packet);
1160                         break;
1161
1162                 case ETH_P_IPV6:
1163                         route_ipv6(source, packet);
1164                         break;
1165
1166                 default:
1167                         logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown type %hx", source->name, source->hostname, type);
1168                         break;
1169                 }
1170
1171                 break;
1172
1173         case RMODE_SWITCH:
1174                 route_mac(source, packet);
1175                 break;
1176
1177         case RMODE_HUB:
1178                 route_broadcast(source, packet);
1179                 break;
1180         }
1181 }