f046e6e4abb161871511eef014a1fc7a818c3521
[oweals/tinc.git] / 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         int sockfd;
262
263         if(ratelimit(3))
264                 return;
265         
266         /* Swap Ethernet source and destination addresses */
267
268         swap_mac_addresses(packet);
269
270         /* Copy headers from packet into properly aligned structs on the stack */
271
272         memcpy(&ip, packet->data + ether_size, ip_size);
273
274         /* Remember original source and destination */
275         
276         ip_src = ip.ip_src;
277         ip_dst = ip.ip_dst;
278
279         /* Try to reply with an IP address assigned to the local machine */
280
281         sockfd = socket(AF_INET, SOCK_DGRAM, 0);
282         if (sockfd != -1) {
283                 struct sockaddr_in addr;
284                 memset(&addr, 0, sizeof(addr));
285                 addr.sin_family = AF_INET;
286                 addr.sin_addr = ip.ip_src;
287                 if (!connect(sockfd, (const struct sockaddr*) &addr, sizeof(addr))) {
288                         memset(&addr, 0, sizeof(addr));
289                         addr.sin_family = AF_INET;
290                         socklen_t addrlen = sizeof(addr);
291                         if (!getsockname(sockfd, (struct sockaddr*) &addr, &addrlen) && addrlen <= sizeof(addr)) {
292                                 ip_dst = addr.sin_addr;
293                         }
294                 }
295                 close(sockfd);
296         }
297
298         oldlen = packet->len - ether_size;
299
300         if(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
301                 icmp.icmp_nextmtu = htons(packet->len - ether_size);
302
303         if(oldlen >= IP_MSS - ip_size - icmp_size)
304                 oldlen = IP_MSS - ip_size - icmp_size;
305         
306         /* Copy first part of original contents to ICMP message */
307         
308         memmove(packet->data + ether_size + ip_size + icmp_size, packet->data + ether_size, oldlen);
309
310         /* Fill in IPv4 header */
311         
312         ip.ip_v = 4;
313         ip.ip_hl = ip_size / 4;
314         ip.ip_tos = 0;
315         ip.ip_len = htons(ip_size + icmp_size + oldlen);
316         ip.ip_id = 0;
317         ip.ip_off = 0;
318         ip.ip_ttl = 255;
319         ip.ip_p = IPPROTO_ICMP;
320         ip.ip_sum = 0;
321         ip.ip_src = ip_dst;
322         ip.ip_dst = ip_src;
323
324         ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
325         
326         /* Fill in ICMP header */
327         
328         icmp.icmp_type = type;
329         icmp.icmp_code = code;
330         icmp.icmp_cksum = 0;
331         
332         icmp.icmp_cksum = inet_checksum(&icmp, icmp_size, ~0);
333         icmp.icmp_cksum = inet_checksum(packet->data + ether_size + ip_size + icmp_size, oldlen, icmp.icmp_cksum);
334
335         /* Copy structs on stack back to packet */
336
337         memcpy(packet->data + ether_size, &ip, ip_size);
338         memcpy(packet->data + ether_size + ip_size, &icmp, icmp_size);
339         
340         packet->len = ether_size + ip_size + icmp_size + oldlen;
341
342         send_packet(source, packet);
343 }
344
345 /* RFC 791 */
346
347 static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet, length_t ether_size) {
348         struct ip ip;
349         vpn_packet_t fragment;
350         int len, maxlen, todo;
351         uint8_t *offset;
352         uint16_t ip_off, origf;
353         
354         memcpy(&ip, packet->data + ether_size, ip_size);
355         fragment.priority = packet->priority;
356
357         if(ip.ip_hl != ip_size / 4)
358                 return;
359         
360         todo = ntohs(ip.ip_len) - ip_size;
361
362         if(ether_size + ip_size + todo != packet->len) {
363                 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));
364                 return;
365         }
366
367         ifdebug(TRAFFIC) logger(LOG_INFO, "Fragmenting packet of %d bytes to %s (%s)", packet->len, dest->name, dest->hostname);
368
369         offset = packet->data + ether_size + ip_size;
370         maxlen = (dest->mtu - ether_size - ip_size) & ~0x7;
371         ip_off = ntohs(ip.ip_off);
372         origf = ip_off & ~IP_OFFMASK;
373         ip_off &= IP_OFFMASK;
374         
375         while(todo) {
376                 len = todo > maxlen ? maxlen : todo;
377                 memcpy(fragment.data + ether_size + ip_size, offset, len);
378                 todo -= len;
379                 offset += len;
380
381                 ip.ip_len = htons(ip_size + len);
382                 ip.ip_off = htons(ip_off | origf | (todo ? IP_MF : 0));
383                 ip.ip_sum = 0;
384                 ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
385                 memcpy(fragment.data, packet->data, ether_size);
386                 memcpy(fragment.data + ether_size, &ip, ip_size);
387                 fragment.len = ether_size + ip_size + len;
388
389                 send_packet(dest, &fragment);
390
391                 ip_off += len / 8;
392         }       
393 }
394
395 static void route_ipv4_unicast(node_t *source, vpn_packet_t *packet) {
396         subnet_t *subnet;
397         node_t *via;
398         ipv4_t dest;
399
400         memcpy(&dest, &packet->data[30], sizeof dest);
401         subnet = lookup_subnet_ipv4(&dest);
402
403         if(!subnet) {
404                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv4 destination address %d.%d.%d.%d",
405                                 source->name, source->hostname,
406                                 dest.x[0],
407                                 dest.x[1],
408                                 dest.x[2],
409                                 dest.x[3]);
410
411                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
412                 return;
413         }
414         
415         if(subnet->owner == source) {
416                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
417                 return;
418         }
419
420         if(!subnet->owner->status.reachable)
421                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
422
423         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
424                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
425
426         if(priorityinheritance)
427                 packet->priority = packet->data[15];
428
429         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
430
431         if(via == source) {
432                 ifdebug(TRAFFIC) logger(LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
433                 return;
434         }
435         
436         if(directonly && subnet->owner != via)
437                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
438
439         if(via && packet->len > MAX(via->mtu, 590) && via != myself) {
440                 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);
441                 if(packet->data[20] & 0x40) {
442                         packet->len = MAX(via->mtu, 590);
443                         route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
444                 } else {
445                         fragment_ipv4_packet(via, packet, ether_size);
446                 }
447
448                 return;
449         }
450
451         clamp_mss(source, via, packet);
452  
453         send_packet(subnet->owner, packet);
454 }
455
456 static void route_ipv4(node_t *source, vpn_packet_t *packet) {
457         if(!checklength(source, packet, ether_size + ip_size))
458                 return;
459
460         if(broadcast_mode && (((packet->data[30] & 0xf0) == 0xe0) || (
461                         packet->data[30] == 255 &&
462                         packet->data[31] == 255 &&
463                         packet->data[32] == 255 &&
464                         packet->data[33] == 255)))
465                 broadcast_packet(source, packet);
466         else
467                 route_ipv4_unicast(source, packet);
468 }
469
470 /* RFC 2463 */
471
472 static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
473         struct ip6_hdr ip6;
474         struct icmp6_hdr icmp6 = {0};
475         uint16_t checksum;      
476         int sockfd;
477
478         struct {
479                 struct in6_addr ip6_src;        /* source address */
480                 struct in6_addr ip6_dst;        /* destination address */
481                 uint32_t length;
482                 uint32_t next;
483         } pseudo;
484
485         if(ratelimit(3))
486                 return;
487         
488         /* Swap Ethernet source and destination addresses */
489
490         swap_mac_addresses(packet);
491
492         /* Copy headers from packet to structs on the stack */
493
494         memcpy(&ip6, packet->data + ether_size, ip6_size);
495
496         /* Remember original source and destination */
497         
498         pseudo.ip6_src = ip6.ip6_dst;
499         pseudo.ip6_dst = ip6.ip6_src;
500
501         /* Try to reply with an IP address assigned to the local machine */
502
503         sockfd = socket(AF_INET6, SOCK_DGRAM, 0);
504         if (sockfd != -1) {
505                 struct sockaddr_in6 addr;
506                 memset(&addr, 0, sizeof(addr));
507                 addr.sin6_family = AF_INET6;
508                 addr.sin6_addr = ip6.ip6_src;
509                 if (!connect(sockfd, (const struct sockaddr*) &addr, sizeof(addr))) {
510                         memset(&addr, 0, sizeof(addr));
511                         addr.sin6_family = AF_INET6;
512                         socklen_t addrlen = sizeof(addr);
513                         if (!getsockname(sockfd, (struct sockaddr*) &addr, &addrlen) && addrlen <= sizeof(addr)) {
514                                 pseudo.ip6_src = addr.sin6_addr;
515                         }
516                 }
517                 close(sockfd);
518         }
519
520         pseudo.length = packet->len - ether_size;
521
522         if(type == ICMP6_PACKET_TOO_BIG)
523                 icmp6.icmp6_mtu = htonl(pseudo.length);
524         
525         if(pseudo.length >= IP_MSS - ip6_size - icmp6_size)
526                 pseudo.length = IP_MSS - ip6_size - icmp6_size;
527         
528         /* Copy first part of original contents to ICMP message */
529         
530         memmove(packet->data + ether_size + ip6_size + icmp6_size, packet->data + ether_size, pseudo.length);
531
532         /* Fill in IPv6 header */
533         
534         ip6.ip6_flow = htonl(0x60000000UL);
535         ip6.ip6_plen = htons(icmp6_size + pseudo.length);
536         ip6.ip6_nxt = IPPROTO_ICMPV6;
537         ip6.ip6_hlim = 255;
538         ip6.ip6_src = pseudo.ip6_src;
539         ip6.ip6_dst = pseudo.ip6_dst;
540
541         /* Fill in ICMP header */
542         
543         icmp6.icmp6_type = type;
544         icmp6.icmp6_code = code;
545         icmp6.icmp6_cksum = 0;
546
547         /* Create pseudo header */
548                 
549         pseudo.length = htonl(icmp6_size + pseudo.length);
550         pseudo.next = htonl(IPPROTO_ICMPV6);
551
552         /* Generate checksum */
553         
554         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
555         checksum = inet_checksum(&icmp6, icmp6_size, checksum);
556         checksum = inet_checksum(packet->data + ether_size + ip6_size + icmp6_size, ntohl(pseudo.length) - icmp6_size, checksum);
557
558         icmp6.icmp6_cksum = checksum;
559
560         /* Copy structs on stack back to packet */
561
562         memcpy(packet->data + ether_size, &ip6, ip6_size);
563         memcpy(packet->data + ether_size + ip6_size, &icmp6, icmp6_size);
564         
565         packet->len = ether_size + ip6_size + ntohl(pseudo.length);
566         
567         send_packet(source, packet);
568 }
569
570 static void route_ipv6_unicast(node_t *source, vpn_packet_t *packet) {
571         subnet_t *subnet;
572         node_t *via;
573         ipv6_t dest;
574
575         memcpy(&dest, &packet->data[38], sizeof dest);
576         subnet = lookup_subnet_ipv6(&dest);
577
578         if(!subnet) {
579                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv6 destination address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
580                                 source->name, source->hostname,
581                                 ntohs(dest.x[0]),
582                                 ntohs(dest.x[1]),
583                                 ntohs(dest.x[2]),
584                                 ntohs(dest.x[3]),
585                                 ntohs(dest.x[4]),
586                                 ntohs(dest.x[5]),
587                                 ntohs(dest.x[6]),
588                                 ntohs(dest.x[7]));
589
590                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
591                 return;
592         }
593
594         if(subnet->owner == source) {
595                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
596                 return;
597         }
598
599         if(!subnet->owner->status.reachable)
600                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
601
602         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
603                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
604
605         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
606         
607         if(via == source) {
608                 ifdebug(TRAFFIC) logger(LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
609                 return;
610         }
611         
612         if(directonly && subnet->owner != via)
613                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
614
615         if(via && packet->len > MAX(via->mtu, 1294) && via != myself) {
616                 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);
617                 packet->len = MAX(via->mtu, 1294);
618                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_PACKET_TOO_BIG, 0);
619                 return;
620         }
621
622         clamp_mss(source, via, packet);
623  
624         send_packet(subnet->owner, packet);
625 }
626
627 /* RFC 2461 */
628
629 static void route_neighborsol(node_t *source, vpn_packet_t *packet) {
630         struct ip6_hdr ip6;
631         struct nd_neighbor_solicit ns;
632         struct nd_opt_hdr opt;
633         subnet_t *subnet;
634         uint16_t checksum;
635         bool has_opt;
636
637         struct {
638                 struct in6_addr ip6_src;        /* source address */
639                 struct in6_addr ip6_dst;        /* destination address */
640                 uint32_t length;
641                 uint32_t next;
642         } pseudo;
643
644         if(!checklength(source, packet, ether_size + ip6_size + ns_size))
645                 return;
646         
647         has_opt = packet->len >= ether_size + ip6_size + ns_size + opt_size + ETH_ALEN;
648         
649         if(source != myself) {
650                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Got neighbor solicitation request from %s (%s) while in router mode!", source->name, source->hostname);
651                 return;
652         }
653
654         /* Copy headers from packet to structs on the stack */
655
656         memcpy(&ip6, packet->data + ether_size, ip6_size);
657         memcpy(&ns, packet->data + ether_size + ip6_size, ns_size);
658         if(has_opt)
659                 memcpy(&opt, packet->data + ether_size + ip6_size + ns_size, opt_size);
660
661         /* First, snatch the source address from the neighbor solicitation packet */
662
663         if(overwrite_mac)
664                 memcpy(mymac.x, packet->data + ETH_ALEN, ETH_ALEN);
665
666         /* Check if this is a valid neighbor solicitation request */
667
668         if(ns.nd_ns_hdr.icmp6_type != ND_NEIGHBOR_SOLICIT ||
669            (has_opt && opt.nd_opt_type != ND_OPT_SOURCE_LINKADDR)) {
670                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: received unknown type neighbor solicitation request");
671                 return;
672         }
673
674         /* Create pseudo header */
675
676         pseudo.ip6_src = ip6.ip6_src;
677         pseudo.ip6_dst = ip6.ip6_dst;
678         if(has_opt)
679                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
680         else
681                 pseudo.length = htonl(ns_size);
682         pseudo.next = htonl(IPPROTO_ICMPV6);
683
684         /* Generate checksum */
685
686         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
687         checksum = inet_checksum(&ns, ns_size, checksum);
688         if(has_opt) {
689                 checksum = inet_checksum(&opt, opt_size, checksum);
690                 checksum = inet_checksum(packet->data + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
691         }
692
693         if(checksum) {
694                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: checksum error for neighbor solicitation request");
695                 return;
696         }
697
698         /* Check if the IPv6 address exists on the VPN */
699
700         subnet = lookup_subnet_ipv6((ipv6_t *) &ns.nd_ns_target);
701
702         if(!subnet) {
703                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: neighbor solicitation request for unknown address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
704                                    ntohs(((uint16_t *) &ns.nd_ns_target)[0]),
705                                    ntohs(((uint16_t *) &ns.nd_ns_target)[1]),
706                                    ntohs(((uint16_t *) &ns.nd_ns_target)[2]),
707                                    ntohs(((uint16_t *) &ns.nd_ns_target)[3]),
708                                    ntohs(((uint16_t *) &ns.nd_ns_target)[4]),
709                                    ntohs(((uint16_t *) &ns.nd_ns_target)[5]),
710                                    ntohs(((uint16_t *) &ns.nd_ns_target)[6]),
711                                    ntohs(((uint16_t *) &ns.nd_ns_target)[7]));
712
713                 return;
714         }
715
716         /* Check if it is for our own subnet */
717
718         if(subnet->owner == myself)
719                 return;                                 /* silently ignore */
720
721         /* Create neighbor advertation reply */
722
723         memcpy(packet->data, packet->data + ETH_ALEN, ETH_ALEN);        /* copy destination address */
724         packet->data[ETH_ALEN * 2 - 1] ^= 0xFF; /* mangle source address so it looks like it's not from us */
725
726         ip6.ip6_dst = ip6.ip6_src;                      /* swap destination and source protocoll address */
727         ip6.ip6_src = ns.nd_ns_target;
728
729         if(has_opt)
730                 memcpy(packet->data + ether_size + ip6_size + ns_size + opt_size, packet->data + ETH_ALEN, ETH_ALEN);   /* add fake source hard addr */
731
732         ns.nd_ns_cksum = 0;
733         ns.nd_ns_type = ND_NEIGHBOR_ADVERT;
734         ns.nd_ns_reserved = htonl(0x40000000UL);        /* Set solicited flag */
735         opt.nd_opt_type = ND_OPT_TARGET_LINKADDR;
736
737         /* Create pseudo header */
738
739         pseudo.ip6_src = ip6.ip6_src;
740         pseudo.ip6_dst = ip6.ip6_dst;
741         if(has_opt)
742                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
743         else
744                 pseudo.length = htonl(ns_size);
745         pseudo.next = htonl(IPPROTO_ICMPV6);
746
747         /* Generate checksum */
748
749         checksum = inet_checksum(&pseudo, sizeof(pseudo), ~0);
750         checksum = inet_checksum(&ns, ns_size, checksum);
751         if(has_opt) {
752                 checksum = inet_checksum(&opt, opt_size, checksum);
753                 checksum = inet_checksum(packet->data + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
754         }
755
756         ns.nd_ns_hdr.icmp6_cksum = checksum;
757
758         /* Copy structs on stack back to packet */
759
760         memcpy(packet->data + ether_size, &ip6, ip6_size);
761         memcpy(packet->data + ether_size + ip6_size, &ns, ns_size);
762         if(has_opt)
763                 memcpy(packet->data + ether_size + ip6_size + ns_size, &opt, opt_size);
764
765         send_packet(source, packet);
766 }
767
768 static void route_ipv6(node_t *source, vpn_packet_t *packet) {
769         if(!checklength(source, packet, ether_size + ip6_size))
770                 return;
771
772         if(packet->data[20] == IPPROTO_ICMPV6 && checklength(source, packet, ether_size + ip6_size + icmp6_size) && packet->data[54] == ND_NEIGHBOR_SOLICIT) {
773                 route_neighborsol(source, packet);
774                 return;
775         }
776
777         if(broadcast_mode && packet->data[38] == 255)
778                 broadcast_packet(source, packet);
779         else
780                 route_ipv6_unicast(source, packet);
781 }
782
783 /* RFC 826 */
784
785 static void route_arp(node_t *source, vpn_packet_t *packet) {
786         struct ether_arp arp;
787         subnet_t *subnet;
788         struct in_addr addr;
789
790         if(!checklength(source, packet, ether_size + arp_size))
791                 return;
792
793         if(source != myself) {
794                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Got ARP request from %s (%s) while in router mode!", source->name, source->hostname);
795                 return;
796         }
797
798         /* First, snatch the source address from the ARP packet */
799
800         if(overwrite_mac)
801                 memcpy(mymac.x, packet->data + ETH_ALEN, ETH_ALEN);
802
803         /* Copy headers from packet to structs on the stack */
804
805         memcpy(&arp, packet->data + ether_size, arp_size);
806
807         /* Check if this is a valid ARP request */
808
809         if(ntohs(arp.arp_hrd) != ARPHRD_ETHER || ntohs(arp.arp_pro) != ETH_P_IP ||
810            arp.arp_hln != ETH_ALEN || arp.arp_pln != sizeof(addr) || ntohs(arp.arp_op) != ARPOP_REQUEST) {
811                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: received unknown type ARP request");
812                 return;
813         }
814
815         /* Check if the IPv4 address exists on the VPN */
816
817         subnet = lookup_subnet_ipv4((ipv4_t *) &arp.arp_tpa);
818
819         if(!subnet) {
820                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet: ARP request for unknown address %d.%d.%d.%d",
821                                    arp.arp_tpa[0], arp.arp_tpa[1], arp.arp_tpa[2],
822                                    arp.arp_tpa[3]);
823                 return;
824         }
825
826         /* Check if it is for our own subnet */
827
828         if(subnet->owner == myself)
829                 return;                                 /* silently ignore */
830
831         memcpy(packet->data, packet->data + ETH_ALEN, ETH_ALEN);        /* copy destination address */
832         packet->data[ETH_ALEN * 2 - 1] ^= 0xFF; /* mangle source address so it looks like it's not from us */
833
834         memcpy(&addr, arp.arp_tpa, sizeof(addr));       /* save protocol addr */
835         memcpy(arp.arp_tpa, arp.arp_spa, sizeof(addr)); /* swap destination and source protocol address */
836         memcpy(arp.arp_spa, &addr, sizeof(addr));       /* ... */
837
838         memcpy(arp.arp_tha, arp.arp_sha, ETH_ALEN);     /* set target hard/proto addr */
839         memcpy(arp.arp_sha, packet->data + ETH_ALEN, ETH_ALEN); /* add fake source hard addr */
840         arp.arp_op = htons(ARPOP_REPLY);
841
842         /* Copy structs on stack back to packet */
843
844         memcpy(packet->data + ether_size, &arp, arp_size);
845
846         send_packet(source, packet);
847 }
848
849 static void route_mac(node_t *source, vpn_packet_t *packet) {
850         subnet_t *subnet;
851         mac_t dest;
852
853         /* Learn source address */
854
855         if(source == myself) {
856                 mac_t src;
857                 memcpy(&src, &packet->data[6], sizeof src);
858                 learn_mac(&src);
859         }
860
861         /* Lookup destination address */
862
863         memcpy(&dest, &packet->data[0], sizeof dest);
864         subnet = lookup_subnet_mac(NULL, &dest);
865
866         if(!subnet) {
867                 broadcast_packet(source, packet);
868                 return;
869         }
870
871         if(subnet->owner == source) {
872                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
873                 return;
874         }
875
876         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
877                 return;
878
879         uint16_t type = packet->data[12] << 8 | packet->data[13];
880
881         if(priorityinheritance && type == ETH_P_IP && packet->len >= ether_size + ip_size)
882                 packet->priority = packet->data[15];
883
884         // Handle packets larger than PMTU
885
886         node_t *via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
887
888         if(directonly && subnet->owner != via)
889                 return;
890         
891         if(via && packet->len > via->mtu && via != myself) {
892                 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);
893                 length_t ethlen = 14;
894
895                 if(type == ETH_P_8021Q) {
896                         type = packet->data[16] << 8 | packet->data[17];
897                         ethlen += 4;
898                 }
899
900                 if(type == ETH_P_IP && packet->len > 576 + ethlen) {
901                         if(packet->data[6 + ethlen] & 0x40) {
902                                 packet->len = via->mtu;
903                                 route_ipv4_unreachable(source, packet, ethlen, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
904                         } else {
905                                 fragment_ipv4_packet(via, packet, ethlen);
906                         }
907                         return;
908                 } else if(type == ETH_P_IPV6 && packet->len > 1280 + ethlen) {
909                         packet->len = via->mtu;
910                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_PACKET_TOO_BIG, 0);
911                         return;
912                 }
913         }
914
915         clamp_mss(source, via, packet);
916  
917         send_packet(subnet->owner, packet);
918 }
919
920 static bool do_decrement_ttl(node_t *source, vpn_packet_t *packet) {
921         uint16_t type = packet->data[12] << 8 | packet->data[13];
922         length_t ethlen = ether_size;
923
924         if(type == ETH_P_8021Q) {
925                 type = packet->data[16] << 8 | packet->data[17];
926                 ethlen += 4;
927         }
928
929         switch (type) {
930                 case ETH_P_IP:
931                         if(!checklength(source, packet, ethlen + ip_size))
932                                 return false;
933
934                         if(packet->data[ethlen + 8] <= 1) {
935                                 if(packet->data[ethlen + 11] != IPPROTO_ICMP || packet->data[ethlen + 32] != ICMP_TIME_EXCEEDED)
936                                         route_ipv4_unreachable(source, packet, ethlen, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL);
937                                 return false;
938                         }
939
940                         uint16_t old = packet->data[ethlen + 8] << 8 | packet->data[ethlen + 9];
941                         packet->data[ethlen + 8]--;
942                         uint16_t new = packet->data[ethlen + 8] << 8 | packet->data[ethlen + 9];
943
944                         uint32_t checksum = packet->data[ethlen + 10] << 8 | packet->data[ethlen + 11];
945                         checksum += old + (~new & 0xFFFF);
946                         while(checksum >> 16)
947                                 checksum = (checksum & 0xFFFF) + (checksum >> 16);
948                         packet->data[ethlen + 10] = checksum >> 8;
949                         packet->data[ethlen + 11] = checksum & 0xff;
950
951                         return true;
952
953                 case ETH_P_IPV6:
954                         if(!checklength(source, packet, ethlen + ip6_size))
955                                 return false;
956
957                         if(packet->data[ethlen + 7] <= 1) {
958                                 if(packet->data[ethlen + 6] != IPPROTO_ICMPV6 || packet->data[ethlen + 40] != ICMP6_TIME_EXCEEDED)
959                                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT);
960                                 return false;
961                         }
962
963                         packet->data[ethlen + 7]--;
964
965                         return true;
966
967                 default:
968                         return true;
969         }
970 }
971
972 void route(node_t *source, vpn_packet_t *packet) {
973         if(forwarding_mode == FMODE_KERNEL && source != myself) {
974                 send_packet(myself, packet);
975                 return;
976         }
977
978         if(!checklength(source, packet, ether_size))
979                 return;
980
981         if(decrement_ttl && source != myself)
982                 if(!do_decrement_ttl(source, packet))
983                         return;
984
985         switch (routing_mode) {
986                 case RMODE_ROUTER:
987                         {
988                                 uint16_t type = packet->data[12] << 8 | packet->data[13];
989
990                                 switch (type) {
991                                         case ETH_P_ARP:
992                                                 route_arp(source, packet);
993                                                 break;
994
995                                         case ETH_P_IP:
996                                                 route_ipv4(source, packet);
997                                                 break;
998
999                                         case ETH_P_IPV6:
1000                                                 route_ipv6(source, packet);
1001                                                 break;
1002
1003                                         default:
1004                                                 ifdebug(TRAFFIC) logger(LOG_WARNING, "Cannot route packet from %s (%s): unknown type %hx", source->name, source->hostname, type);
1005                                                 break;
1006                                 }
1007                         }
1008                         break;
1009
1010                 case RMODE_SWITCH:
1011                         route_mac(source, packet);
1012                         break;
1013
1014                 case RMODE_HUB:
1015                         broadcast_packet(source, packet);
1016                         break;
1017         }
1018 }