Drop libevent and use our own event handling again.
[tinc] / src / route.c
1 /*
2     route.c -- routing
3     Copyright (C) 2000-2005 Ivo Timmermans,
4                   2000-2012 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         while(checksum >> 16)
79                 checksum = (checksum & 0xFFFF) + (checksum >> 16);
80
81         return ~checksum;
82 }
83
84 static bool ratelimit(int frequency) {
85         static time_t lasttime = 0;
86         static int count = 0;
87
88         if(lasttime == now.tv_sec) {
89                 if(count >= frequency)
90                         return true;
91         } else {
92                 lasttime = now.tv_sec;
93                 count = 0;
94         }
95
96         count++;
97         return false;
98 }
99
100 static bool checklength(node_t *source, vpn_packet_t *packet, length_t length) {
101         if(packet->len < length) {
102                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got too short packet from %s (%s)", source->name, source->hostname);
103                 return false;
104         } else
105                 return true;
106 }
107
108 static void clamp_mss(const node_t *source, const node_t *via, vpn_packet_t *packet) {
109         if(!source || !via || !(via->options & OPTION_CLAMP_MSS))
110                 return;
111
112         uint16_t mtu = source->mtu;
113         if(via != myself && via->mtu < mtu)
114                 mtu = via->mtu;
115
116         /* Find TCP header */
117         int start = ether_size;
118         uint16_t type = packet->data[12] << 8 | packet->data[13];
119
120         if(type == ETH_P_8021Q) {
121                 start += 4;
122                 type = packet->data[16] << 8 | packet->data[17];
123         }
124
125         if(type == ETH_P_IP && packet->data[start + 9] == 6)
126                 start += (packet->data[start] & 0xf) * 4;
127         else if(type == ETH_P_IPV6 && packet->data[start + 6] == 6)
128                 start += 40;
129         else
130                 return;
131
132         if(packet->len <= start + 20)
133                 return;
134
135         /* Use data offset field to calculate length of options field */
136         int len = ((packet->data[start + 12] >> 4) - 5) * 4;
137
138         if(packet->len < start + 20 + len)
139                 return;
140
141         /* Search for MSS option header */
142         for(int i = 0; i < len;) {
143                 if(packet->data[start + 20 + i] == 0)
144                         break;
145
146                 if(packet->data[start + 20 + i] == 1) {
147                         i++;
148                         continue;
149                 }
150
151                 if(i > len - 2 || i > len - packet->data[start + 21 + i])
152                         break;
153
154                 if(packet->data[start + 20 + i] != 2) {
155                         if(packet->data[start + 21 + i] < 2)
156                                 break;
157                         i += packet->data[start + 21 + i];
158                         continue;
159                 }
160
161                 if(packet->data[start + 21] != 4)
162                         break;
163
164                 /* Found it */
165                 uint16_t oldmss = packet->data[start + 22 + i] << 8 | packet->data[start + 23 + i];
166                 uint16_t newmss = mtu - start - 20;
167                 uint16_t csum = packet->data[start + 16] << 8 | packet->data[start + 17];
168
169                 if(oldmss <= newmss)
170                         break;
171
172                 logger(DEBUG_TRAFFIC, LOG_INFO, "Clamping MSS of packet from %s to %s to %d", source->name, via->name, newmss);
173
174                 /* Update the MSS value and the checksum */
175                 packet->data[start + 22 + i] = newmss >> 8;
176                 packet->data[start + 23 + i] = newmss & 0xff;
177                 csum ^= 0xffff;
178                 csum -= oldmss;
179                 csum += newmss;
180                 csum ^= 0xffff;
181                 packet->data[start + 16] = csum >> 8;
182                 packet->data[start + 17] = csum & 0xff;
183                 break;
184         }
185 }
186
187 static void swap_mac_addresses(vpn_packet_t *packet) {
188         mac_t tmp;
189         memcpy(&tmp, &packet->data[0], sizeof tmp);
190         memcpy(&packet->data[0], &packet->data[6], sizeof tmp);
191         memcpy(&packet->data[6], &tmp, sizeof tmp);
192 }
193
194 static void age_subnets(void *data) {
195         bool left = false;
196
197         for splay_each(subnet_t, s, myself->subnet_tree) {
198                 if(s->expires && s->expires < now.tv_sec) {
199                         if(debug_level >= DEBUG_TRAFFIC) {
200                                 char netstr[MAXNETSTR];
201                                 if(net2str(netstr, sizeof netstr, s))
202                                         logger(DEBUG_TRAFFIC, LOG_INFO, "Subnet %s expired", netstr);
203                         }
204
205                         for list_each(connection_t, c, connection_list)
206                                 if(c->status.active)
207                                         send_del_subnet(c, s);
208
209                         subnet_del(myself, s);
210                 } else {
211                         if(s->expires)
212                                 left = true;
213                 }
214         }
215
216         if(left)
217                 timeout_set(&age_subnets_timeout, &(struct timeval){10, rand() % 100000});
218 }
219
220 static void learn_mac(mac_t *address) {
221         subnet_t *subnet = lookup_subnet_mac(myself, address);
222
223         /* If we don't know this MAC address yet, store it */
224
225         if(!subnet) {
226                 logger(DEBUG_TRAFFIC, LOG_INFO, "Learned new MAC address %hx:%hx:%hx:%hx:%hx:%hx",
227                                    address->x[0], address->x[1], address->x[2], address->x[3],
228                                    address->x[4], address->x[5]);
229
230                 subnet = new_subnet();
231                 subnet->type = SUBNET_MAC;
232                 subnet->expires = time(NULL) + macexpire;
233                 subnet->net.mac.address = *address;
234                 subnet->weight = 10;
235                 subnet_add(myself, subnet);
236                 subnet_update(myself, subnet, true);
237
238                 /* And tell all other tinc daemons it's our MAC */
239
240                 for list_each(connection_t, c, connection_list)
241                         if(c->status.active)
242                                 send_add_subnet(c, subnet);
243
244                 timeout_add(&age_subnets_timeout, age_subnets, NULL, &(struct timeval){10, rand() % 100000});
245         } else {
246                 if(subnet->expires)
247                         subnet->expires = time(NULL) + macexpire;
248         }
249 }
250
251 /* RFC 792 */
252
253 static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
254         struct ip ip = {0};
255         struct icmp icmp = {0};
256
257         struct in_addr ip_src;
258         struct in_addr ip_dst;
259         uint32_t oldlen;
260
261         if(ratelimit(3))
262                 return;
263
264         /* Swap Ethernet source and destination addresses */
265
266         swap_mac_addresses(packet);
267
268         /* Copy headers from packet into properly aligned structs on the stack */
269
270         memcpy(&ip, packet->data + ether_size, ip_size);
271
272         /* Remember original source and destination */
273
274         ip_src = ip.ip_src;
275         ip_dst = ip.ip_dst;
276
277         oldlen = packet->len - ether_size;
278
279         if(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
280                 icmp.icmp_nextmtu = htons(packet->len - ether_size);
281
282         if(oldlen >= IP_MSS - ip_size - icmp_size)
283                 oldlen = IP_MSS - ip_size - icmp_size;
284
285         /* Copy first part of original contents to ICMP message */
286
287         memmove(packet->data + ether_size + ip_size + icmp_size, packet->data + ether_size, oldlen);
288
289         /* Fill in IPv4 header */
290
291         ip.ip_v = 4;
292         ip.ip_hl = ip_size / 4;
293         ip.ip_tos = 0;
294         ip.ip_len = htons(ip_size + icmp_size + oldlen);
295         ip.ip_id = 0;
296         ip.ip_off = 0;
297         ip.ip_ttl = 255;
298         ip.ip_p = IPPROTO_ICMP;
299         ip.ip_sum = 0;
300         ip.ip_src = ip_dst;
301         ip.ip_dst = ip_src;
302
303         ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
304
305         /* Fill in ICMP header */
306
307         icmp.icmp_type = type;
308         icmp.icmp_code = code;
309         icmp.icmp_cksum = 0;
310
311         icmp.icmp_cksum = inet_checksum(&icmp, icmp_size, ~0);
312         icmp.icmp_cksum = inet_checksum(packet->data + ether_size + ip_size + icmp_size, oldlen, icmp.icmp_cksum);
313
314         /* Copy structs on stack back to packet */
315
316         memcpy(packet->data + ether_size, &ip, ip_size);
317         memcpy(packet->data + ether_size + ip_size, &icmp, icmp_size);
318
319         packet->len = ether_size + ip_size + icmp_size + oldlen;
320
321         send_packet(source, packet);
322 }
323
324 /* RFC 791 */
325
326 static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet, length_t ether_size) {
327         struct ip ip;
328         vpn_packet_t fragment;
329         int len, maxlen, todo;
330         uint8_t *offset;
331         uint16_t ip_off, origf;
332
333         memcpy(&ip, packet->data + ether_size, ip_size);
334         fragment.priority = packet->priority;
335
336         if(ip.ip_hl != ip_size / 4)
337                 return;
338
339         todo = ntohs(ip.ip_len) - ip_size;
340
341         if(ether_size + ip_size + todo != packet->len) {
342                 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));
343                 return;
344         }
345
346         logger(DEBUG_TRAFFIC, LOG_INFO, "Fragmenting packet of %d bytes to %s (%s)", packet->len, dest->name, dest->hostname);
347
348         offset = packet->data + ether_size + ip_size;
349         maxlen = (dest->mtu - ether_size - ip_size) & ~0x7;
350         ip_off = ntohs(ip.ip_off);
351         origf = ip_off & ~IP_OFFMASK;
352         ip_off &= IP_OFFMASK;
353
354         while(todo) {
355                 len = todo > maxlen ? maxlen : todo;
356                 memcpy(fragment.data + ether_size + ip_size, offset, len);
357                 todo -= len;
358                 offset += len;
359
360                 ip.ip_len = htons(ip_size + len);
361                 ip.ip_off = htons(ip_off | origf | (todo ? IP_MF : 0));
362                 ip.ip_sum = 0;
363                 ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
364                 memcpy(fragment.data, packet->data, ether_size);
365                 memcpy(fragment.data + ether_size, &ip, ip_size);
366                 fragment.len = ether_size + ip_size + len;
367
368                 send_packet(dest, &fragment);
369
370                 ip_off += len / 8;
371         }
372 }
373
374 static void route_ipv4_unicast(node_t *source, vpn_packet_t *packet) {
375         subnet_t *subnet;
376         node_t *via;
377         ipv4_t dest;
378
379         memcpy(&dest, &packet->data[30], sizeof dest);
380         subnet = lookup_subnet_ipv4(&dest);
381
382         if(!subnet) {
383                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv4 destination address %d.%d.%d.%d",
384                                 source->name, source->hostname,
385                                 dest.x[0],
386                                 dest.x[1],
387                                 dest.x[2],
388                                 dest.x[3]);
389
390                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
391                 return;
392         }
393
394         if(subnet->owner == source) {
395                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
396                 return;
397         }
398
399         if(!subnet->owner->status.reachable)
400                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
401
402         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
403                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
404
405         if(priorityinheritance)
406                 packet->priority = packet->data[15];
407
408         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
409
410         if(via == source) {
411                 logger(DEBUG_TRAFFIC, LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
412                 return;
413         }
414
415         if(directonly && subnet->owner != via)
416                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
417
418         if(via && packet->len > MAX(via->mtu, 590) && via != myself) {
419                 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);
420                 if(packet->data[20] & 0x40) {
421                         packet->len = MAX(via->mtu, 590);
422                         route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
423                 } else {
424                         fragment_ipv4_packet(via, packet, ether_size);
425                 }
426
427                 return;
428         }
429
430         clamp_mss(source, via, packet);
431
432         send_packet(subnet->owner, packet);
433 }
434
435 static void route_ipv4(node_t *source, vpn_packet_t *packet) {
436         if(!checklength(source, packet, ether_size + ip_size))
437                 return;
438
439         if(broadcast_mode && (((packet->data[30] & 0xf0) == 0xe0) || (
440                         packet->data[30] == 255 &&
441                         packet->data[31] == 255 &&
442                         packet->data[32] == 255 &&
443                         packet->data[33] == 255)))
444                 broadcast_packet(source, packet);
445         else
446                 route_ipv4_unicast(source, packet);
447 }
448
449 /* RFC 2463 */
450
451 static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
452         struct ip6_hdr ip6;
453         struct icmp6_hdr icmp6 = {0};
454         uint16_t checksum;
455
456         struct {
457                 struct in6_addr ip6_src;        /* source address */
458                 struct in6_addr ip6_dst;        /* destination address */
459                 uint32_t length;
460                 uint32_t next;
461         } pseudo;
462
463         if(ratelimit(3))
464                 return;
465
466         /* Swap Ethernet source and destination addresses */
467
468         swap_mac_addresses(packet);
469
470         /* Copy headers from packet to structs on the stack */
471
472         memcpy(&ip6, packet->data + ether_size, ip6_size);
473
474         /* Remember original source and destination */
475
476         pseudo.ip6_src = ip6.ip6_dst;
477         pseudo.ip6_dst = ip6.ip6_src;
478
479         pseudo.length = packet->len - ether_size;
480
481         if(type == ICMP6_PACKET_TOO_BIG)
482                 icmp6.icmp6_mtu = htonl(pseudo.length);
483
484         if(pseudo.length >= IP_MSS - ip6_size - icmp6_size)
485                 pseudo.length = IP_MSS - ip6_size - icmp6_size;
486
487         /* Copy first part of original contents to ICMP message */
488
489         memmove(packet->data + ether_size + ip6_size + icmp6_size, packet->data + ether_size, pseudo.length);
490
491         /* Fill in IPv6 header */
492
493         ip6.ip6_flow = htonl(0x60000000UL);
494         ip6.ip6_plen = htons(icmp6_size + pseudo.length);
495         ip6.ip6_nxt = IPPROTO_ICMPV6;
496         ip6.ip6_hlim = 255;
497         ip6.ip6_src = pseudo.ip6_src;
498         ip6.ip6_dst = pseudo.ip6_dst;
499
500         /* Fill in ICMP header */
501
502         icmp6.icmp6_type = type;
503         icmp6.icmp6_code = code;
504         icmp6.icmp6_cksum = 0;
505
506         /* Create pseudo header */
507
508         pseudo.length = htonl(icmp6_size + pseudo.length);
509         pseudo.next = htonl(IPPROTO_ICMPV6);
510
511         /* Generate checksum */
512
513         checksum = inet_checksum(&pseudo, sizeof pseudo, ~0);
514         checksum = inet_checksum(&icmp6, icmp6_size, checksum);
515         checksum = inet_checksum(packet->data + ether_size + ip6_size + icmp6_size, ntohl(pseudo.length) - icmp6_size, checksum);
516
517         icmp6.icmp6_cksum = checksum;
518
519         /* Copy structs on stack back to packet */
520
521         memcpy(packet->data + ether_size, &ip6, ip6_size);
522         memcpy(packet->data + ether_size + ip6_size, &icmp6, icmp6_size);
523
524         packet->len = ether_size + ip6_size + ntohl(pseudo.length);
525
526         send_packet(source, packet);
527 }
528
529 static void route_ipv6_unicast(node_t *source, vpn_packet_t *packet) {
530         subnet_t *subnet;
531         node_t *via;
532         ipv6_t dest;
533
534         memcpy(&dest, &packet->data[38], sizeof dest);
535         subnet = lookup_subnet_ipv6(&dest);
536
537         if(!subnet) {
538                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv6 destination address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
539                                 source->name, source->hostname,
540                                 ntohs(dest.x[0]),
541                                 ntohs(dest.x[1]),
542                                 ntohs(dest.x[2]),
543                                 ntohs(dest.x[3]),
544                                 ntohs(dest.x[4]),
545                                 ntohs(dest.x[5]),
546                                 ntohs(dest.x[6]),
547                                 ntohs(dest.x[7]));
548
549                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
550                 return;
551         }
552
553         if(subnet->owner == source) {
554                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
555                 return;
556         }
557
558         if(!subnet->owner->status.reachable)
559                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
560
561         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
562                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
563
564         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
565
566         if(via == source) {
567                 logger(DEBUG_TRAFFIC, LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
568                 return;
569         }
570
571         if(directonly && subnet->owner != via)
572                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
573
574         if(via && packet->len > MAX(via->mtu, 1294) && via != myself) {
575                 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);
576                 packet->len = MAX(via->mtu, 1294);
577                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_PACKET_TOO_BIG, 0);
578                 return;
579         }
580
581         clamp_mss(source, via, packet);
582
583         send_packet(subnet->owner, packet);
584 }
585
586 /* RFC 2461 */
587
588 static void route_neighborsol(node_t *source, vpn_packet_t *packet) {
589         struct ip6_hdr ip6;
590         struct nd_neighbor_solicit ns;
591         struct nd_opt_hdr opt;
592         subnet_t *subnet;
593         uint16_t checksum;
594         bool has_opt;
595
596         struct {
597                 struct in6_addr ip6_src;
598                 struct in6_addr ip6_dst;
599                 uint32_t length;
600                 uint32_t next;
601         } pseudo;
602
603         if(!checklength(source, packet, ether_size + ip6_size + ns_size))
604                 return;
605
606         has_opt = packet->len >= ether_size + ip6_size + ns_size + opt_size + ETH_ALEN;
607
608         if(source != myself) {
609                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got neighbor solicitation request from %s (%s) while in router mode!", source->name, source->hostname);
610                 return;
611         }
612
613         /* Copy headers from packet to structs on the stack */
614
615         memcpy(&ip6, packet->data + ether_size, ip6_size);
616         memcpy(&ns, packet->data + ether_size + ip6_size, ns_size);
617         if(has_opt)
618                 memcpy(&opt, packet->data + ether_size + ip6_size + ns_size, opt_size);
619
620         /* First, snatch the source address from the neighbor solicitation packet */
621
622         if(overwrite_mac)
623                 memcpy(mymac.x, packet->data + ETH_ALEN, ETH_ALEN);
624
625         /* Check if this is a valid neighbor solicitation request */
626
627         if(ns.nd_ns_hdr.icmp6_type != ND_NEIGHBOR_SOLICIT ||
628            (has_opt && opt.nd_opt_type != ND_OPT_SOURCE_LINKADDR)) {
629                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: received unknown type neighbor solicitation request");
630                 return;
631         }
632
633         /* Create pseudo header */
634
635         pseudo.ip6_src = ip6.ip6_src;
636         pseudo.ip6_dst = ip6.ip6_dst;
637         if(has_opt)
638                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
639         else
640                 pseudo.length = htonl(ns_size);
641         pseudo.next = htonl(IPPROTO_ICMPV6);
642
643         /* Generate checksum */
644
645         checksum = inet_checksum(&pseudo, sizeof pseudo, ~0);
646         checksum = inet_checksum(&ns, ns_size, checksum);
647         if(has_opt) {
648                 checksum = inet_checksum(&opt, opt_size, checksum);
649                 checksum = inet_checksum(packet->data + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
650         }
651
652         if(checksum) {
653                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: checksum error for neighbor solicitation request");
654                 return;
655         }
656
657         /* Check if the IPv6 address exists on the VPN */
658
659         subnet = lookup_subnet_ipv6((ipv6_t *) &ns.nd_ns_target);
660
661         if(!subnet) {
662                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: neighbor solicitation request for unknown address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
663                                    ntohs(((uint16_t *) &ns.nd_ns_target)[0]),
664                                    ntohs(((uint16_t *) &ns.nd_ns_target)[1]),
665                                    ntohs(((uint16_t *) &ns.nd_ns_target)[2]),
666                                    ntohs(((uint16_t *) &ns.nd_ns_target)[3]),
667                                    ntohs(((uint16_t *) &ns.nd_ns_target)[4]),
668                                    ntohs(((uint16_t *) &ns.nd_ns_target)[5]),
669                                    ntohs(((uint16_t *) &ns.nd_ns_target)[6]),
670                                    ntohs(((uint16_t *) &ns.nd_ns_target)[7]));
671
672                 return;
673         }
674
675         /* Check if it is for our own subnet */
676
677         if(subnet->owner == myself)
678                 return;                                          /* silently ignore */
679
680         /* Create neighbor advertation reply */
681
682         memcpy(packet->data, packet->data + ETH_ALEN, ETH_ALEN); /* copy destination address */
683         packet->data[ETH_ALEN * 2 - 1] ^= 0xFF;                  /* mangle source address so it looks like it's not from us */
684
685         ip6.ip6_dst = ip6.ip6_src;                               /* swap destination and source protocoll address */
686         ip6.ip6_src = ns.nd_ns_target;
687
688         if(has_opt)
689                 memcpy(packet->data + ether_size + ip6_size + ns_size + opt_size, packet->data + ETH_ALEN, ETH_ALEN);   /* add fake source hard addr */
690
691         ns.nd_ns_cksum = 0;
692         ns.nd_ns_type = ND_NEIGHBOR_ADVERT;
693         ns.nd_ns_reserved = htonl(0x40000000UL);                 /* Set solicited flag */
694         opt.nd_opt_type = ND_OPT_TARGET_LINKADDR;
695
696         /* Create pseudo header */
697
698         pseudo.ip6_src = ip6.ip6_src;
699         pseudo.ip6_dst = ip6.ip6_dst;
700         if(has_opt)
701                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
702         else
703                 pseudo.length = htonl(ns_size);
704         pseudo.next = htonl(IPPROTO_ICMPV6);
705
706         /* Generate checksum */
707
708         checksum = inet_checksum(&pseudo, sizeof pseudo, ~0);
709         checksum = inet_checksum(&ns, ns_size, checksum);
710         if(has_opt) {
711                 checksum = inet_checksum(&opt, opt_size, checksum);
712                 checksum = inet_checksum(packet->data + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
713         }
714
715         ns.nd_ns_hdr.icmp6_cksum = checksum;
716
717         /* Copy structs on stack back to packet */
718
719         memcpy(packet->data + ether_size, &ip6, ip6_size);
720         memcpy(packet->data + ether_size + ip6_size, &ns, ns_size);
721         if(has_opt)
722                 memcpy(packet->data + ether_size + ip6_size + ns_size, &opt, opt_size);
723
724         send_packet(source, packet);
725 }
726
727 static void route_ipv6(node_t *source, vpn_packet_t *packet) {
728         if(!checklength(source, packet, ether_size + ip6_size))
729                 return;
730
731         if(packet->data[20] == IPPROTO_ICMPV6 && checklength(source, packet, ether_size + ip6_size + icmp6_size) && packet->data[54] == ND_NEIGHBOR_SOLICIT) {
732                 route_neighborsol(source, packet);
733                 return;
734         }
735
736         if(broadcast_mode && packet->data[38] == 255)
737                 broadcast_packet(source, packet);
738         else
739                 route_ipv6_unicast(source, packet);
740 }
741
742 /* RFC 826 */
743
744 static void route_arp(node_t *source, vpn_packet_t *packet) {
745         struct ether_arp arp;
746         subnet_t *subnet;
747         struct in_addr addr;
748
749         if(!checklength(source, packet, ether_size + arp_size))
750                 return;
751
752         if(source != myself) {
753                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got ARP request from %s (%s) while in router mode!", source->name, source->hostname);
754                 return;
755         }
756
757         /* First, snatch the source address from the ARP packet */
758
759         if(overwrite_mac)
760                 memcpy(mymac.x, packet->data + ETH_ALEN, ETH_ALEN);
761
762         /* Copy headers from packet to structs on the stack */
763
764         memcpy(&arp, packet->data + ether_size, arp_size);
765
766         /* Check if this is a valid ARP request */
767
768         if(ntohs(arp.arp_hrd) != ARPHRD_ETHER || ntohs(arp.arp_pro) != ETH_P_IP ||
769            arp.arp_hln != ETH_ALEN || arp.arp_pln != sizeof addr || ntohs(arp.arp_op) != ARPOP_REQUEST) {
770                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: received unknown type ARP request");
771                 return;
772         }
773
774         /* Check if the IPv4 address exists on the VPN */
775
776         subnet = lookup_subnet_ipv4((ipv4_t *) &arp.arp_tpa);
777
778         if(!subnet) {
779                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: ARP request for unknown address %d.%d.%d.%d",
780                                    arp.arp_tpa[0], arp.arp_tpa[1], arp.arp_tpa[2],
781                                    arp.arp_tpa[3]);
782                 return;
783         }
784
785         /* Check if it is for our own subnet */
786
787         if(subnet->owner == myself)
788                 return;                                          /* silently ignore */
789
790         memcpy(packet->data, packet->data + ETH_ALEN, ETH_ALEN); /* copy destination address */
791         packet->data[ETH_ALEN * 2 - 1] ^= 0xFF;                  /* mangle source address so it looks like it's not from us */
792
793         memcpy(&addr, arp.arp_tpa, sizeof addr);                 /* save protocol addr */
794         memcpy(arp.arp_tpa, arp.arp_spa, sizeof addr);           /* swap destination and source protocol address */
795         memcpy(arp.arp_spa, &addr, sizeof addr);                 /* ... */
796
797         memcpy(arp.arp_tha, arp.arp_sha, ETH_ALEN);              /* set target hard/proto addr */
798         memcpy(arp.arp_sha, packet->data + ETH_ALEN, ETH_ALEN);  /* add fake source hard addr */
799         arp.arp_op = htons(ARPOP_REPLY);
800
801         /* Copy structs on stack back to packet */
802
803         memcpy(packet->data + ether_size, &arp, arp_size);
804
805         send_packet(source, packet);
806 }
807
808 static void route_mac(node_t *source, vpn_packet_t *packet) {
809         subnet_t *subnet;
810         mac_t dest;
811
812         /* Learn source address */
813
814         if(source == myself) {
815                 mac_t src;
816                 memcpy(&src, &packet->data[6], sizeof src);
817                 learn_mac(&src);
818         }
819
820         /* Lookup destination address */
821
822         memcpy(&dest, &packet->data[0], sizeof dest);
823         subnet = lookup_subnet_mac(NULL, &dest);
824
825         if(!subnet) {
826                 broadcast_packet(source, packet);
827                 return;
828         }
829
830         if(subnet->owner == source) {
831                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
832                 return;
833         }
834
835         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
836                 return;
837
838         // Handle packets larger than PMTU
839
840         node_t *via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
841
842         if(directonly && subnet->owner != via)
843                 return;
844
845         if(via && packet->len > via->mtu && via != myself) {
846                 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);
847                 uint16_t type = packet->data[12] << 8 | packet->data[13];
848                 length_t ethlen = 14;
849
850                 if(type == ETH_P_8021Q) {
851                         type = packet->data[16] << 8 | packet->data[17];
852                         ethlen += 4;
853                 }
854
855                 if(type == ETH_P_IP && packet->len > 576 + ethlen) {
856                         if(packet->data[6 + ethlen] & 0x40) {
857                                 packet->len = via->mtu;
858                                 route_ipv4_unreachable(source, packet, ethlen, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
859                         } else {
860                                 fragment_ipv4_packet(via, packet, ethlen);
861                         }
862                         return;
863                 } else if(type == ETH_P_IPV6 && packet->len > 1280 + ethlen) {
864                         packet->len = via->mtu;
865                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_PACKET_TOO_BIG, 0);
866                         return;
867                 }
868         }
869
870         clamp_mss(source, via, packet);
871
872         send_packet(subnet->owner, packet);
873 }
874
875 static void send_pcap(vpn_packet_t *packet) {
876         pcap = false;
877
878         for list_each(connection_t, c, connection_list) {
879                 if(!c->status.pcap)
880                         continue;
881
882                 pcap = true;
883                 int len = packet->len;
884                 if(c->outmaclength && c->outmaclength < len)
885                         len = c->outmaclength;
886
887                 if(send_request(c, "%d %d %d", CONTROL, REQ_PCAP, len))
888                         send_meta(c, (char *)packet->data, len);
889         }
890 }
891
892 static bool do_decrement_ttl(node_t *source, vpn_packet_t *packet) {
893         uint16_t type = packet->data[12] << 8 | packet->data[13];
894         length_t ethlen = ether_size;
895
896         if(type == ETH_P_8021Q) {
897                 type = packet->data[16] << 8 | packet->data[17];
898                 ethlen += 4;
899         }
900
901         switch (type) {
902                 case ETH_P_IP:
903                         if(!checklength(source, packet, ethlen + ip_size))
904                                 return false;
905
906                         if(packet->data[ethlen + 8] < 1) {
907                                 if(packet->data[ethlen + 11] != IPPROTO_ICMP || packet->data[ethlen + 32] != ICMP_TIME_EXCEEDED)
908                                         route_ipv4_unreachable(source, packet, ethlen, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL);
909                                 return false;
910                         }
911
912                         uint16_t old = packet->data[ethlen + 8] << 8 | packet->data[ethlen + 9];
913                         packet->data[ethlen + 8]--;
914                         uint16_t new = packet->data[ethlen + 8] << 8 | packet->data[ethlen + 9];
915
916                         uint32_t checksum = packet->data[ethlen + 10] << 8 | packet->data[ethlen + 11];
917                         checksum += old + (~new & 0xFFFF);
918                         while(checksum >> 16)
919                                 checksum = (checksum & 0xFFFF) + (checksum >> 16);
920                         packet->data[ethlen + 10] = checksum >> 8;
921                         packet->data[ethlen + 11] = checksum & 0xff;
922
923                         return true;
924
925                 case ETH_P_IPV6:
926                         if(!checklength(source, packet, ethlen + ip6_size))
927                                 return false;
928
929                         if(packet->data[ethlen + 7] < 1) {
930                                 if(packet->data[ethlen + 6] != IPPROTO_ICMPV6 || packet->data[ethlen + 40] != ICMP6_TIME_EXCEEDED)
931                                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT);
932                                 return false;
933                         }
934
935                         packet->data[ethlen + 7]--;
936
937                         return true;
938
939                 default:
940                         return true;
941         }
942 }
943
944 void route(node_t *source, vpn_packet_t *packet) {
945         if(pcap)
946                 send_pcap(packet);
947
948         if(forwarding_mode == FMODE_KERNEL && source != myself) {
949                 send_packet(myself, packet);
950                 return;
951         }
952
953         if(!checklength(source, packet, ether_size))
954                 return;
955
956         if(decrement_ttl && source != myself)
957                 if(!do_decrement_ttl(source, packet))
958                         return;
959
960         uint16_t type = packet->data[12] << 8 | packet->data[13];
961
962         switch (routing_mode) {
963                 case RMODE_ROUTER:
964                         switch (type) {
965                                 case ETH_P_ARP:
966                                         route_arp(source, packet);
967                                         break;
968
969                                 case ETH_P_IP:
970                                         route_ipv4(source, packet);
971                                         break;
972
973                                 case ETH_P_IPV6:
974                                         route_ipv6(source, packet);
975                                         break;
976
977                                 default:
978                                         logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown type %hx", source->name, source->hostname, type);
979                                         break;
980                         }
981                         break;
982
983                 case RMODE_SWITCH:
984                         route_mac(source, packet);
985                         break;
986
987                 case RMODE_HUB:
988                         broadcast_packet(source, packet);
989                         break;
990         }
991 }