Fix DecrementTTL option.
[tinc] / src / route.c
1 /*
2     route.c -- routing
3     Copyright (C) 2000-2005 Ivo Timmermans,
4                   2000-2014 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 "avl_tree.h"
24 #include "connection.h"
25 #include "ethernet.h"
26 #include "ipv4.h"
27 #include "ipv6.h"
28 #include "logger.h"
29 #include "net.h"
30 #include "protocol.h"
31 #include "route.h"
32 #include "subnet.h"
33 #include "utils.h"
34
35 rmode_t routing_mode = RMODE_ROUTER;
36 fmode_t forwarding_mode = FMODE_INTERNAL;
37 bmode_t broadcast_mode = BMODE_MST;
38 bool decrement_ttl = false;
39 bool directonly = false;
40 bool priorityinheritance = false;
41 int macexpire = 600;
42 bool overwrite_mac = false;
43 mac_t mymac = {{0xFE, 0xFD, 0, 0, 0, 0}};
44
45 /* Sizes of various headers */
46
47 static const size_t ether_size = sizeof(struct ether_header);
48 static const size_t arp_size = sizeof(struct ether_arp);
49 static const size_t ip_size = sizeof(struct ip);
50 static const size_t icmp_size = sizeof(struct icmp) - sizeof(struct ip);
51 static const size_t ip6_size = sizeof(struct ip6_hdr);
52 static const size_t icmp6_size = sizeof(struct icmp6_hdr);
53 static const size_t ns_size = sizeof(struct nd_neighbor_solicit);
54 static const size_t opt_size = sizeof(struct nd_opt_hdr);
55
56 #ifndef MAX
57 #define MAX(a, b) ((a) > (b) ? (a) : (b))
58 #endif
59
60 /* RFC 1071 */
61
62 static uint16_t inet_checksum(void *data, int len, uint16_t prevsum) {
63         uint16_t *p = data;
64         uint32_t checksum = prevsum ^ 0xFFFF;
65
66         while(len >= 2) {
67                 checksum += *p++;
68                 len -= 2;
69         }
70         
71         if(len)
72                 checksum += *(uint8_t *)p;
73
74         while(checksum >> 16)
75                 checksum = (checksum & 0xFFFF) + (checksum >> 16);
76
77         return ~checksum;
78 }
79
80 static bool ratelimit(int frequency) {
81         static time_t lasttime = 0;
82         static int count = 0;
83         
84         if(lasttime == now) {
85                 if(count >= frequency)
86                         return true;
87         } else {
88                 lasttime = now;
89                 count = 0;
90         }
91
92         count++;
93         return false;
94 }
95
96 static bool checklength(node_t *source, vpn_packet_t *packet, length_t length) {
97         if(packet->len < length) {
98                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Got too short packet from %s (%s)", source->name, source->hostname);
99                 return false;
100         } else
101                 return true;
102 }
103
104 static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *packet) {
105         if(!source || !via || !(via->options & OPTION_CLAMP_MSS))
106                 return;
107
108         uint16_t mtu = source->mtu;
109         if(via != myself && via->mtu < mtu)
110                 mtu = via->mtu;
111
112         /* Find TCP header */
113         int start = ether_size;
114         uint16_t type = packet->data[12] << 8 | packet->data[13];
115
116         if(type == ETH_P_8021Q) {
117                 start += 4;
118                 type = packet->data[16] << 8 | packet->data[17];
119         }
120
121         if(type == ETH_P_IP && packet->data[start + 9] == 6)
122                 start += (packet->data[start] & 0xf) * 4;
123         else if(type == ETH_P_IPV6 && packet->data[start + 6] == 6)
124                 start += 40;
125         else
126                 return;
127
128         if(packet->len <= start + 20)
129                 return;
130
131         /* Use data offset field to calculate length of options field */
132         int len = ((packet->data[start + 12] >> 4) - 5) * 4;
133
134         if(packet->len < start + 20 + len)
135                 return;
136
137         /* Search for MSS option header */
138         for(int i = 0; i < len;) {
139                 if(packet->data[start + 20 + i] == 0)
140                         break;
141
142                 if(packet->data[start + 20 + i] == 1) {
143                         i++;
144                         continue;
145                 }
146
147                 if(i > len - 2 || i > len - packet->data[start + 21 + i])
148                         break;
149
150                 if(packet->data[start + 20 + i] != 2) {
151                         if(packet->data[start + 21 + i] < 2)
152                                 break;
153                         i += packet->data[start + 21 + i];
154                         continue;
155                 }
156
157                 if(packet->data[start + 21] != 4)
158                         break;
159
160                 /* Found it */
161                 uint16_t oldmss = packet->data[start + 22 + i] << 8 | packet->data[start + 23 + i];
162                 uint16_t newmss = mtu - start - 20;
163                 uint16_t csum = packet->data[start + 16] << 8 | packet->data[start + 17];
164
165                 if(oldmss <= newmss)
166                         break;
167                 
168                 ifdebug(TRAFFIC) logger(LOG_INFO, "Clamping MSS of packet from %s to %s to %d", source->name, via->name, newmss);
169
170                 /* Update the MSS value and the checksum */
171                 packet->data[start + 22 + i] = newmss >> 8;
172                 packet->data[start + 23 + i] = newmss & 0xff;
173                 csum ^= 0xffff;
174                 csum -= oldmss;
175                 csum += newmss;
176                 csum ^= 0xffff;
177                 packet->data[start + 16] = csum >> 8;
178                 packet->data[start + 17] = csum & 0xff;
179                 break;
180         }
181 }
182
183 static void swap_mac_addresses(vpn_packet_t *packet) {
184         mac_t tmp;
185         memcpy(&tmp, &packet->data[0], sizeof tmp);
186         memcpy(&packet->data[0], &packet->data[6], sizeof tmp);
187         memcpy(&packet->data[6], &tmp, sizeof tmp);
188 }
189         
190 static void learn_mac(mac_t *address) {
191         subnet_t *subnet;
192         avl_node_t *node;
193         connection_t *c;
194
195         subnet = lookup_subnet_mac(myself, address);
196
197         /* If we don't know this MAC address yet, store it */
198
199         if(!subnet) {
200                 ifdebug(TRAFFIC) logger(LOG_INFO, "Learned new MAC address %x:%x:%x:%x:%x:%x",
201                                    address->x[0], address->x[1], address->x[2], address->x[3],
202                                    address->x[4], address->x[5]);
203
204                 subnet = new_subnet();
205                 subnet->type = SUBNET_MAC;
206                 subnet->expires = now + macexpire;
207                 subnet->net.mac.address = *address;
208                 subnet->weight = 10;
209                 subnet_add(myself, subnet);
210                 subnet_update(myself, subnet, true);
211
212                 /* And tell all other tinc daemons it's our MAC */
213
214                 for(node = connection_tree->head; node; node = node->next) {
215                         c = node->data;
216                         if(c->status.active)
217                                 send_add_subnet(c, subnet);
218                 }
219         }
220
221         if(subnet->expires)
222                 subnet->expires = now + macexpire;
223 }
224
225 void age_subnets(void) {
226         subnet_t *s;
227         connection_t *c;
228         avl_node_t *node, *next, *node2;
229
230         for(node = myself->subnet_tree->head; node; node = next) {
231                 next = node->next;
232                 s = node->data;
233                 if(s->expires && s->expires <= now) {
234                         ifdebug(TRAFFIC) {
235                                 char netstr[MAXNETSTR];
236                                 if(net2str(netstr, sizeof netstr, s))
237                                         logger(LOG_INFO, "Subnet %s expired", netstr);
238                         }
239
240                         for(node2 = connection_tree->head; node2; node2 = node2->next) {
241                                 c = node2->data;
242                                 if(c->status.active)
243                                         send_del_subnet(c, s);
244                         }
245
246                         subnet_update(myself, s, false);
247                         subnet_del(myself, s);
248                 }
249         }
250 }
251
252 /* RFC 792 */
253
254 static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
255         struct ip ip = {0};
256         struct icmp icmp = {0};
257         
258         struct in_addr ip_src;
259         struct in_addr ip_dst;
260         uint32_t oldlen;
261
262         if(ratelimit(3))
263                 return;
264         
265         /* Swap Ethernet source and destination addresses */
266
267         swap_mac_addresses(packet);
268
269         /* Copy headers from packet into properly aligned structs on the stack */
270
271         memcpy(&ip, packet->data + ether_size, ip_size);
272
273         /* Remember original source and destination */
274         
275         ip_src = ip.ip_src;
276         ip_dst = ip.ip_dst;
277
278         oldlen = packet->len - ether_size;
279
280         if(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
281                 icmp.icmp_nextmtu = htons(packet->len - ether_size);
282
283         if(oldlen >= IP_MSS - ip_size - icmp_size)
284                 oldlen = IP_MSS - ip_size - icmp_size;
285         
286         /* Copy first part of original contents to ICMP message */
287         
288         memmove(packet->data + ether_size + ip_size + icmp_size, packet->data + ether_size, oldlen);
289
290         /* Fill in IPv4 header */
291         
292         ip.ip_v = 4;
293         ip.ip_hl = ip_size / 4;
294         ip.ip_tos = 0;
295         ip.ip_len = htons(ip_size + icmp_size + oldlen);
296         ip.ip_id = 0;
297         ip.ip_off = 0;
298         ip.ip_ttl = 255;
299         ip.ip_p = IPPROTO_ICMP;
300         ip.ip_sum = 0;
301         ip.ip_src = ip_dst;
302         ip.ip_dst = ip_src;
303
304         ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
305         
306         /* Fill in ICMP header */
307         
308         icmp.icmp_type = type;
309         icmp.icmp_code = code;
310         icmp.icmp_cksum = 0;
311         
312         icmp.icmp_cksum = inet_checksum(&icmp, icmp_size, ~0);
313         icmp.icmp_cksum = inet_checksum(packet->data + ether_size + ip_size + icmp_size, oldlen, icmp.icmp_cksum);
314
315         /* Copy structs on stack back to packet */
316
317         memcpy(packet->data + ether_size, &ip, ip_size);
318         memcpy(packet->data + ether_size + ip_size, &icmp, icmp_size);
319         
320         packet->len = ether_size + ip_size + icmp_size + oldlen;
321
322         send_packet(source, packet);
323 }
324
325 /* RFC 791 */
326
327 static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet, length_t ether_size) {
328         struct ip ip;
329         vpn_packet_t fragment;
330         int len, maxlen, todo;
331         uint8_t *offset;
332         uint16_t ip_off, origf;
333         
334         memcpy(&ip, packet->data + ether_size, ip_size);
335         fragment.priority = packet->priority;
336
337         if(ip.ip_hl != ip_size / 4)
338                 return;
339         
340         todo = ntohs(ip.ip_len) - ip_size;
341
342         if(ether_size + ip_size + todo != packet->len) {
343                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Length of packet (%d) doesn't match length in IPv4 header (%d)", packet->len, (int)(ether_size + ip_size + todo));
344                 return;
345         }
346
347         ifdebug(TRAFFIC) logger(LOG_INFO, "Fragmenting packet of %d bytes to %s (%s)", packet->len, dest->name, dest->hostname);
348
349         offset = packet->data + ether_size + ip_size;
350         maxlen = (dest->mtu - ether_size - ip_size) & ~0x7;
351         ip_off = ntohs(ip.ip_off);
352         origf = ip_off & ~IP_OFFMASK;
353         ip_off &= IP_OFFMASK;
354         
355         while(todo) {
356                 len = todo > maxlen ? maxlen : todo;
357                 memcpy(fragment.data + ether_size + ip_size, offset, len);
358                 todo -= len;
359                 offset += len;
360
361                 ip.ip_len = htons(ip_size + len);
362                 ip.ip_off = htons(ip_off | origf | (todo ? IP_MF : 0));
363                 ip.ip_sum = 0;
364                 ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
365                 memcpy(fragment.data, packet->data, ether_size);
366                 memcpy(fragment.data + ether_size, &ip, ip_size);
367                 fragment.len = ether_size + ip_size + len;
368
369                 send_packet(dest, &fragment);
370
371                 ip_off += len / 8;
372         }       
373 }
374
375 static void route_ipv4_unicast(node_t *source, vpn_packet_t *packet) {
376         subnet_t *subnet;
377         node_t *via;
378         ipv4_t dest;
379
380         memcpy(&dest, &packet->data[30], sizeof dest);
381         subnet = lookup_subnet_ipv4(&dest);
382
383         if(!subnet) {
384                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv4 destination address %d.%d.%d.%d",
385                                 source->name, source->hostname,
386                                 dest.x[0],
387                                 dest.x[1],
388                                 dest.x[2],
389                                 dest.x[3]);
390
391                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
392                 return;
393         }
394         
395         if(subnet->owner == source) {
396                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
397                 return;
398         }
399
400         if(!subnet->owner->status.reachable)
401                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
402
403         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
404                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
405
406         if(priorityinheritance)
407                 packet->priority = packet->data[15];
408
409         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
410
411         if(via == source) {
412                 ifdebug(TRAFFIC) logger(LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
413                 return;
414         }
415         
416         if(directonly && subnet->owner != via)
417                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
418
419         if(via && packet->len > MAX(via->mtu, 590) && via != myself) {
420                 ifdebug(TRAFFIC) logger(LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
421                 if(packet->data[20] & 0x40) {
422                         packet->len = MAX(via->mtu, 590);
423                         route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
424                 } else {
425                         fragment_ipv4_packet(via, packet, ether_size);
426                 }
427
428                 return;
429         }
430
431         clamp_mss(source, via, packet);
432  
433         send_packet(subnet->owner, packet);
434 }
435
436 static void route_ipv4(node_t *source, vpn_packet_t *packet) {
437         if(!checklength(source, packet, ether_size + ip_size))
438                 return;
439
440         if(broadcast_mode && (((packet->data[30] & 0xf0) == 0xe0) || (
441                         packet->data[30] == 255 &&
442                         packet->data[31] == 255 &&
443                         packet->data[32] == 255 &&
444                         packet->data[33] == 255)))
445                 broadcast_packet(source, packet);
446         else
447                 route_ipv4_unicast(source, packet);
448 }
449
450 /* RFC 2463 */
451
452 static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
453         struct ip6_hdr ip6;
454         struct icmp6_hdr icmp6 = {0};
455         uint16_t checksum;      
456
457         struct {
458                 struct in6_addr ip6_src;        /* source address */
459                 struct in6_addr ip6_dst;        /* destination address */
460                 uint32_t length;
461                 uint32_t next;
462         } pseudo;
463
464         if(ratelimit(3))
465                 return;
466         
467         /* Swap Ethernet source and destination addresses */
468
469         swap_mac_addresses(packet);
470
471         /* Copy headers from packet to structs on the stack */
472
473         memcpy(&ip6, packet->data + ether_size, ip6_size);
474
475         /* Remember original source and destination */
476         
477         pseudo.ip6_src = ip6.ip6_dst;
478         pseudo.ip6_dst = ip6.ip6_src;
479
480         pseudo.length = packet->len - ether_size;
481
482         if(type == ICMP6_PACKET_TOO_BIG)
483                 icmp6.icmp6_mtu = htonl(pseudo.length);
484         
485         if(pseudo.length >= IP_MSS - ip6_size - icmp6_size)
486                 pseudo.length = IP_MSS - ip6_size - icmp6_size;
487         
488         /* Copy first part of original contents to ICMP message */
489         
490         memmove(packet->data + ether_size + ip6_size + icmp6_size, packet->data + ether_size, pseudo.length);
491
492         /* Fill in IPv6 header */
493         
494         ip6.ip6_flow = htonl(0x60000000UL);
495         ip6.ip6_plen = htons(icmp6_size + pseudo.length);
496         ip6.ip6_nxt = IPPROTO_ICMPV6;
497         ip6.ip6_hlim = 255;
498         ip6.ip6_src = pseudo.ip6_src;
499         ip6.ip6_dst = pseudo.ip6_dst;
500
501         /* Fill in ICMP header */
502         
503         icmp6.icmp6_type = type;
504         icmp6.icmp6_code = code;
505         icmp6.icmp6_cksum = 0;
506
507         /* Create pseudo header */
508                 
509         pseudo.length = htonl(icmp6_size + pseudo.length);
510         pseudo.next = htonl(IPPROTO_ICMPV6);
511
512         /* Generate checksum */
513         
514         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
515         checksum = inet_checksum(&icmp6, icmp6_size, checksum);
516         checksum = inet_checksum(packet->data + ether_size + ip6_size + icmp6_size, ntohl(pseudo.length) - icmp6_size, checksum);
517
518         icmp6.icmp6_cksum = checksum;
519
520         /* Copy structs on stack back to packet */
521
522         memcpy(packet->data + ether_size, &ip6, ip6_size);
523         memcpy(packet->data + ether_size + ip6_size, &icmp6, icmp6_size);
524         
525         packet->len = ether_size + ip6_size + ntohl(pseudo.length);
526         
527         send_packet(source, packet);
528 }
529
530 static void route_ipv6_unicast(node_t *source, vpn_packet_t *packet) {
531         subnet_t *subnet;
532         node_t *via;
533         ipv6_t dest;
534
535         memcpy(&dest, &packet->data[38], sizeof dest);
536         subnet = lookup_subnet_ipv6(&dest);
537
538         if(!subnet) {
539                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv6 destination address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
540                                 source->name, source->hostname,
541                                 ntohs(dest.x[0]),
542                                 ntohs(dest.x[1]),
543                                 ntohs(dest.x[2]),
544                                 ntohs(dest.x[3]),
545                                 ntohs(dest.x[4]),
546                                 ntohs(dest.x[5]),
547                                 ntohs(dest.x[6]),
548                                 ntohs(dest.x[7]));
549
550                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
551                 return;
552         }
553
554         if(subnet->owner == source) {
555                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
556                 return;
557         }
558
559         if(!subnet->owner->status.reachable)
560                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
561
562         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
563                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
564
565         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
566         
567         if(via == source) {
568                 ifdebug(TRAFFIC) logger(LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
569                 return;
570         }
571         
572         if(directonly && subnet->owner != via)
573                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
574
575         if(via && packet->len > MAX(via->mtu, 1294) && via != myself) {
576                 ifdebug(TRAFFIC) logger(LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
577                 packet->len = MAX(via->mtu, 1294);
578                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_PACKET_TOO_BIG, 0);
579                 return;
580         }
581
582         clamp_mss(source, via, packet);
583  
584         send_packet(subnet->owner, packet);
585 }
586
587 /* RFC 2461 */
588
589 static void route_neighborsol(node_t *source, vpn_packet_t *packet) {
590         struct ip6_hdr ip6;
591         struct nd_neighbor_solicit ns;
592         struct nd_opt_hdr opt;
593         subnet_t *subnet;
594         uint16_t checksum;
595         bool has_opt;
596
597         struct {
598                 struct in6_addr ip6_src;        /* source address */
599                 struct in6_addr ip6_dst;        /* destination address */
600                 uint32_t length;
601                 uint32_t next;
602         } pseudo;
603
604         if(!checklength(source, packet, ether_size + ip6_size + ns_size))
605                 return;
606         
607         has_opt = packet->len >= ether_size + ip6_size + ns_size + opt_size + ETH_ALEN;
608         
609         if(source != myself) {
610                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Got neighbor solicitation request from %s (%s) while in router mode!", source->name, source->hostname);
611                 return;
612         }
613
614         /* Copy headers from packet to structs on the stack */
615
616         memcpy(&ip6, packet->data + ether_size, ip6_size);
617         memcpy(&ns, packet->data + ether_size + ip6_size, ns_size);
618         if(has_opt)
619                 memcpy(&opt, packet->data + ether_size + ip6_size + ns_size, opt_size);
620
621         /* First, snatch the source address from the neighbor solicitation packet */
622
623         if(overwrite_mac)
624                 memcpy(mymac.x, packet->data + ETH_ALEN, ETH_ALEN);
625
626         /* Check if this is a valid neighbor solicitation request */
627
628         if(ns.nd_ns_hdr.icmp6_type != ND_NEIGHBOR_SOLICIT ||
629            (has_opt && opt.nd_opt_type != ND_OPT_SOURCE_LINKADDR)) {
630                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: received unknown type neighbor solicitation request");
631                 return;
632         }
633
634         /* Create pseudo header */
635
636         pseudo.ip6_src = ip6.ip6_src;
637         pseudo.ip6_dst = ip6.ip6_dst;
638         if(has_opt)
639                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
640         else
641                 pseudo.length = htonl(ns_size);
642         pseudo.next = htonl(IPPROTO_ICMPV6);
643
644         /* Generate checksum */
645
646         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
647         checksum = inet_checksum(&ns, ns_size, checksum);
648         if(has_opt) {
649                 checksum = inet_checksum(&opt, opt_size, checksum);
650                 checksum = inet_checksum(packet->data + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
651         }
652
653         if(checksum) {
654                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: checksum error for neighbor solicitation request");
655                 return;
656         }
657
658         /* Check if the IPv6 address exists on the VPN */
659
660         subnet = lookup_subnet_ipv6((ipv6_t *) &ns.nd_ns_target);
661
662         if(!subnet) {
663                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: neighbor solicitation request for unknown address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
664                                    ntohs(((uint16_t *) &ns.nd_ns_target)[0]),
665                                    ntohs(((uint16_t *) &ns.nd_ns_target)[1]),
666                                    ntohs(((uint16_t *) &ns.nd_ns_target)[2]),
667                                    ntohs(((uint16_t *) &ns.nd_ns_target)[3]),
668                                    ntohs(((uint16_t *) &ns.nd_ns_target)[4]),
669                                    ntohs(((uint16_t *) &ns.nd_ns_target)[5]),
670                                    ntohs(((uint16_t *) &ns.nd_ns_target)[6]),
671                                    ntohs(((uint16_t *) &ns.nd_ns_target)[7]));
672
673                 return;
674         }
675
676         /* Check if it is for our own subnet */
677
678         if(subnet->owner == myself)
679                 return;                                 /* silently ignore */
680
681         /* Create neighbor advertation reply */
682
683         memcpy(packet->data, packet->data + ETH_ALEN, ETH_ALEN);        /* copy destination address */
684         packet->data[ETH_ALEN * 2 - 1] ^= 0xFF; /* mangle source address so it looks like it's not from us */
685
686         ip6.ip6_dst = ip6.ip6_src;                      /* swap destination and source protocoll address */
687         ip6.ip6_src = ns.nd_ns_target;
688
689         if(has_opt)
690                 memcpy(packet->data + ether_size + ip6_size + ns_size + opt_size, packet->data + ETH_ALEN, ETH_ALEN);   /* add fake source hard addr */
691
692         ns.nd_ns_cksum = 0;
693         ns.nd_ns_type = ND_NEIGHBOR_ADVERT;
694         ns.nd_ns_reserved = htonl(0x40000000UL);        /* Set solicited flag */
695         opt.nd_opt_type = ND_OPT_TARGET_LINKADDR;
696
697         /* Create pseudo header */
698
699         pseudo.ip6_src = ip6.ip6_src;
700         pseudo.ip6_dst = ip6.ip6_dst;
701         if(has_opt)
702                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
703         else
704                 pseudo.length = htonl(ns_size);
705         pseudo.next = htonl(IPPROTO_ICMPV6);
706
707         /* Generate checksum */
708
709         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
710         checksum = inet_checksum(&ns, ns_size, checksum);
711         if(has_opt) {
712                 checksum = inet_checksum(&opt, opt_size, checksum);
713                 checksum = inet_checksum(packet->data + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
714         }
715
716         ns.nd_ns_hdr.icmp6_cksum = checksum;
717
718         /* Copy structs on stack back to packet */
719
720         memcpy(packet->data + ether_size, &ip6, ip6_size);
721         memcpy(packet->data + ether_size + ip6_size, &ns, ns_size);
722         if(has_opt)
723                 memcpy(packet->data + ether_size + ip6_size + ns_size, &opt, opt_size);
724
725         send_packet(source, packet);
726 }
727
728 static void route_ipv6(node_t *source, vpn_packet_t *packet) {
729         if(!checklength(source, packet, ether_size + ip6_size))
730                 return;
731
732         if(packet->data[20] == IPPROTO_ICMPV6 && checklength(source, packet, ether_size + ip6_size + icmp6_size) && packet->data[54] == ND_NEIGHBOR_SOLICIT) {
733                 route_neighborsol(source, packet);
734                 return;
735         }
736
737         if(broadcast_mode && packet->data[38] == 255)
738                 broadcast_packet(source, packet);
739         else
740                 route_ipv6_unicast(source, packet);
741 }
742
743 /* RFC 826 */
744
745 static void route_arp(node_t *source, vpn_packet_t *packet) {
746         struct ether_arp arp;
747         subnet_t *subnet;
748         struct in_addr addr;
749
750         if(!checklength(source, packet, ether_size + arp_size))
751                 return;
752
753         if(source != myself) {
754                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Got ARP request from %s (%s) while in router mode!", source->name, source->hostname);
755                 return;
756         }
757
758         /* First, snatch the source address from the ARP packet */
759
760         if(overwrite_mac)
761                 memcpy(mymac.x, packet->data + ETH_ALEN, ETH_ALEN);
762
763         /* Copy headers from packet to structs on the stack */
764
765         memcpy(&arp, packet->data + ether_size, arp_size);
766
767         /* Check if this is a valid ARP request */
768
769         if(ntohs(arp.arp_hrd) != ARPHRD_ETHER || ntohs(arp.arp_pro) != ETH_P_IP ||
770            arp.arp_hln != ETH_ALEN || arp.arp_pln != sizeof(addr) || ntohs(arp.arp_op) != ARPOP_REQUEST) {
771                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: received unknown type ARP request");
772                 return;
773         }
774
775         /* Check if the IPv4 address exists on the VPN */
776
777         subnet = lookup_subnet_ipv4((ipv4_t *) &arp.arp_tpa);
778
779         if(!subnet) {
780                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: ARP request for unknown address %d.%d.%d.%d",
781                                    arp.arp_tpa[0], arp.arp_tpa[1], arp.arp_tpa[2],
782                                    arp.arp_tpa[3]);
783                 return;
784         }
785
786         /* Check if it is for our own subnet */
787
788         if(subnet->owner == myself)
789                 return;                                 /* silently ignore */
790
791         memcpy(packet->data, packet->data + ETH_ALEN, ETH_ALEN);        /* copy destination address */
792         packet->data[ETH_ALEN * 2 - 1] ^= 0xFF; /* mangle source address so it looks like it's not from us */
793
794         memcpy(&addr, arp.arp_tpa, sizeof(addr));       /* save protocol addr */
795         memcpy(arp.arp_tpa, arp.arp_spa, sizeof(addr)); /* swap destination and source protocol address */
796         memcpy(arp.arp_spa, &addr, sizeof(addr));       /* ... */
797
798         memcpy(arp.arp_tha, arp.arp_sha, ETH_ALEN);     /* set target hard/proto addr */
799         memcpy(arp.arp_sha, packet->data + ETH_ALEN, ETH_ALEN); /* add fake source hard addr */
800         arp.arp_op = htons(ARPOP_REPLY);
801
802         /* Copy structs on stack back to packet */
803
804         memcpy(packet->data + ether_size, &arp, arp_size);
805
806         send_packet(source, packet);
807 }
808
809 static void route_mac(node_t *source, vpn_packet_t *packet) {
810         subnet_t *subnet;
811         mac_t dest;
812
813         /* Learn source address */
814
815         if(source == myself) {
816                 mac_t src;
817                 memcpy(&src, &packet->data[6], sizeof src);
818                 learn_mac(&src);
819         }
820
821         /* Lookup destination address */
822
823         memcpy(&dest, &packet->data[0], sizeof dest);
824         subnet = lookup_subnet_mac(NULL, &dest);
825
826         if(!subnet) {
827                 broadcast_packet(source, packet);
828                 return;
829         }
830
831         if(subnet->owner == source) {
832                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
833                 return;
834         }
835
836         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
837                 return;
838
839         uint16_t type = packet->data[12] << 8 | packet->data[13];
840
841         if(priorityinheritance && type == ETH_P_IP && packet->len >= ether_size + ip_size)
842                 packet->priority = packet->data[15];
843
844         // Handle packets larger than PMTU
845
846         node_t *via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
847
848         if(directonly && subnet->owner != via)
849                 return;
850         
851         if(via && packet->len > via->mtu && via != myself) {
852                 ifdebug(TRAFFIC) logger(LOG_INFO, "Packet for %s (%s) length %d larger than MTU %d", subnet->owner->name, subnet->owner->hostname, packet->len, via->mtu);
853                 length_t ethlen = 14;
854
855                 if(type == ETH_P_8021Q) {
856                         type = packet->data[16] << 8 | packet->data[17];
857                         ethlen += 4;
858                 }
859
860                 if(type == ETH_P_IP && packet->len > 576 + ethlen) {
861                         if(packet->data[6 + ethlen] & 0x40) {
862                                 packet->len = via->mtu;
863                                 route_ipv4_unreachable(source, packet, ethlen, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
864                         } else {
865                                 fragment_ipv4_packet(via, packet, ethlen);
866                         }
867                         return;
868                 } else if(type == ETH_P_IPV6 && packet->len > 1280 + ethlen) {
869                         packet->len = via->mtu;
870                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_PACKET_TOO_BIG, 0);
871                         return;
872                 }
873         }
874
875         clamp_mss(source, via, packet);
876  
877         send_packet(subnet->owner, packet);
878 }
879
880 static bool do_decrement_ttl(node_t *source, vpn_packet_t *packet) {
881         uint16_t type = packet->data[12] << 8 | packet->data[13];
882         length_t ethlen = ether_size;
883
884         if(type == ETH_P_8021Q) {
885                 type = packet->data[16] << 8 | packet->data[17];
886                 ethlen += 4;
887         }
888
889         switch (type) {
890                 case ETH_P_IP:
891                         if(!checklength(source, packet, ethlen + ip_size))
892                                 return false;
893
894                         if(packet->data[ethlen + 8] <= 1) {
895                                 if(packet->data[ethlen + 11] != IPPROTO_ICMP || packet->data[ethlen + 32] != ICMP_TIME_EXCEEDED)
896                                         route_ipv4_unreachable(source, packet, ethlen, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL);
897                                 return false;
898                         }
899
900                         uint16_t old = packet->data[ethlen + 8] << 8 | packet->data[ethlen + 9];
901                         packet->data[ethlen + 8]--;
902                         uint16_t new = packet->data[ethlen + 8] << 8 | packet->data[ethlen + 9];
903
904                         uint32_t checksum = packet->data[ethlen + 10] << 8 | packet->data[ethlen + 11];
905                         checksum += old + (~new & 0xFFFF);
906                         while(checksum >> 16)
907                                 checksum = (checksum & 0xFFFF) + (checksum >> 16);
908                         packet->data[ethlen + 10] = checksum >> 8;
909                         packet->data[ethlen + 11] = checksum & 0xff;
910
911                         return true;
912
913                 case ETH_P_IPV6:
914                         if(!checklength(source, packet, ethlen + ip6_size))
915                                 return false;
916
917                         if(packet->data[ethlen + 7] <= 1) {
918                                 if(packet->data[ethlen + 6] != IPPROTO_ICMPV6 || packet->data[ethlen + 40] != ICMP6_TIME_EXCEEDED)
919                                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT);
920                                 return false;
921                         }
922
923                         packet->data[ethlen + 7]--;
924
925                         return true;
926
927                 default:
928                         return true;
929         }
930 }
931
932 void route(node_t *source, vpn_packet_t *packet) {
933         if(forwarding_mode == FMODE_KERNEL && source != myself) {
934                 send_packet(myself, packet);
935                 return;
936         }
937
938         if(!checklength(source, packet, ether_size))
939                 return;
940
941         if(decrement_ttl && source != myself)
942                 if(!do_decrement_ttl(source, packet))
943                         return;
944
945         switch (routing_mode) {
946                 case RMODE_ROUTER:
947                         {
948                                 uint16_t type = packet->data[12] << 8 | packet->data[13];
949
950                                 switch (type) {
951                                         case ETH_P_ARP:
952                                                 route_arp(source, packet);
953                                                 break;
954
955                                         case ETH_P_IP:
956                                                 route_ipv4(source, packet);
957                                                 break;
958
959                                         case ETH_P_IPV6:
960                                                 route_ipv6(source, packet);
961                                                 break;
962
963                                         default:
964                                                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown type %hx", source->name, source->hostname, type);
965                                                 break;
966                                 }
967                         }
968                         break;
969
970                 case RMODE_SWITCH:
971                         route_mac(source, packet);
972                         break;
973
974                 case RMODE_HUB:
975                         broadcast_packet(source, packet);
976                         break;
977         }
978 }