Fix source IP address for ICMP unreachable packets generated by tinc.
[tinc] / src / route.c
1 /*
2     route.c -- routing
3     Copyright (C) 2000-2005 Ivo Timmermans,
4                   2000-2013 Guus Sliepen <guus@tinc-vpn.org>
5
6     This program is free software; you can redistribute it and/or modify
7     it under the terms of the GNU General Public License as published by
8     the Free Software Foundation; either version 2 of the License, or
9     (at your option) any later version.
10
11     This program is distributed in the hope that it will be useful,
12     but WITHOUT ANY WARRANTY; without even the implied warranty of
13     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14     GNU General Public License for more details.
15
16     You should have received a copy of the GNU General Public License along
17     with this program; if not, write to the Free Software Foundation, Inc.,
18     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
19 */
20
21 #include "system.h"
22
23 #include "connection.h"
24 #include "control_common.h"
25 #include "ethernet.h"
26 #include "ipv4.h"
27 #include "ipv6.h"
28 #include "logger.h"
29 #include "meta.h"
30 #include "net.h"
31 #include "protocol.h"
32 #include "route.h"
33 #include "subnet.h"
34 #include "utils.h"
35
36 rmode_t routing_mode = RMODE_ROUTER;
37 fmode_t forwarding_mode = FMODE_INTERNAL;
38 bmode_t broadcast_mode = BMODE_MST;
39 bool decrement_ttl = false;
40 bool directonly = false;
41 bool priorityinheritance = false;
42 int macexpire = 600;
43 bool overwrite_mac = false;
44 mac_t mymac = {{0xFE, 0xFD, 0, 0, 0, 0}};
45 bool pcap = false;
46
47 /* Sizes of various headers */
48
49 static const size_t ether_size = sizeof(struct ether_header);
50 static const size_t arp_size = sizeof(struct ether_arp);
51 static const size_t ip_size = sizeof(struct ip);
52 static const size_t icmp_size = sizeof(struct icmp) - sizeof(struct ip);
53 static const size_t ip6_size = sizeof(struct ip6_hdr);
54 static const size_t icmp6_size = sizeof(struct icmp6_hdr);
55 static const size_t ns_size = sizeof(struct nd_neighbor_solicit);
56 static const size_t opt_size = sizeof(struct nd_opt_hdr);
57
58 #ifndef MAX
59 #define MAX(a, b) ((a) > (b) ? (a) : (b))
60 #endif
61
62 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 = DATA(packet)[12] << 8 | DATA(packet)[13];
119
120         if(type == ETH_P_8021Q) {
121                 start += 4;
122                 type = DATA(packet)[16] << 8 | DATA(packet)[17];
123         }
124
125         if(type == ETH_P_IP && DATA(packet)[start + 9] == 6)
126                 start += (DATA(packet)[start] & 0xf) * 4;
127         else if(type == ETH_P_IPV6 && DATA(packet)[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 = ((DATA(packet)[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(DATA(packet)[start + 20 + i] == 0)
144                         break;
145
146                 if(DATA(packet)[start + 20 + i] == 1) {
147                         i++;
148                         continue;
149                 }
150
151                 if(i > len - 2 || i > len - DATA(packet)[start + 21 + i])
152                         break;
153
154                 if(DATA(packet)[start + 20 + i] != 2) {
155                         if(DATA(packet)[start + 21 + i] < 2)
156                                 break;
157                         i += DATA(packet)[start + 21 + i];
158                         continue;
159                 }
160
161                 if(DATA(packet)[start + 21] != 4)
162                         break;
163
164                 /* Found it */
165                 uint16_t oldmss = DATA(packet)[start + 22 + i] << 8 | DATA(packet)[start + 23 + i];
166                 uint16_t newmss = mtu - start - 20;
167                 uint32_t csum = DATA(packet)[start + 16] << 8 | DATA(packet)[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                 DATA(packet)[start + 22 + i] = newmss >> 8;
176                 DATA(packet)[start + 23 + i] = newmss & 0xff;
177                 csum ^= 0xffff;
178                 csum += oldmss ^ 0xffff;
179                 csum += newmss;
180                 csum = (csum & 0xffff) + (csum >> 16);
181                 csum += csum >> 16;
182                 csum ^= 0xffff;
183                 DATA(packet)[start + 16] = csum >> 8;
184                 DATA(packet)[start + 17] = csum;
185                 break;
186         }
187 }
188
189 static void swap_mac_addresses(vpn_packet_t *packet) {
190         mac_t tmp;
191         memcpy(&tmp, &DATA(packet)[0], sizeof tmp);
192         memcpy(&DATA(packet)[0], &DATA(packet)[6], sizeof tmp);
193         memcpy(&DATA(packet)[6], &tmp, sizeof tmp);
194 }
195
196 static void age_subnets(void *data) {
197         bool left = false;
198
199         for splay_each(subnet_t, s, myself->subnet_tree) {
200                 if(s->expires && s->expires < now.tv_sec) {
201                         if(debug_level >= DEBUG_TRAFFIC) {
202                                 char netstr[MAXNETSTR];
203                                 if(net2str(netstr, sizeof netstr, s))
204                                         logger(DEBUG_TRAFFIC, LOG_INFO, "Subnet %s expired", netstr);
205                         }
206
207                         for list_each(connection_t, c, connection_list)
208                                 if(c->edge)
209                                         send_del_subnet(c, s);
210
211                         subnet_del(myself, s);
212                 } else {
213                         if(s->expires)
214                                 left = true;
215                 }
216         }
217
218         if(left)
219                 timeout_set(&age_subnets_timeout, &(struct timeval){10, rand() % 100000});
220 }
221
222 static void learn_mac(mac_t *address) {
223         subnet_t *subnet = lookup_subnet_mac(myself, address);
224
225         /* If we don't know this MAC address yet, store it */
226
227         if(!subnet) {
228                 logger(DEBUG_TRAFFIC, LOG_INFO, "Learned new MAC address %x:%x:%x:%x:%x:%x",
229                                    address->x[0], address->x[1], address->x[2], address->x[3],
230                                    address->x[4], address->x[5]);
231
232                 subnet = new_subnet();
233                 subnet->type = SUBNET_MAC;
234                 subnet->expires = now.tv_sec + macexpire;
235                 subnet->net.mac.address = *address;
236                 subnet->weight = 10;
237                 subnet_add(myself, subnet);
238                 subnet_update(myself, subnet, true);
239
240                 /* And tell all other tinc daemons it's our MAC */
241
242                 for list_each(connection_t, c, connection_list)
243                         if(c->edge)
244                                 send_add_subnet(c, subnet);
245
246                 timeout_add(&age_subnets_timeout, age_subnets, NULL, &(struct timeval){10, rand() % 100000});
247         } else {
248                 if(subnet->expires)
249                         subnet->expires = now.tv_sec + macexpire;
250         }
251 }
252
253 /* RFC 792 */
254
255 static void route_ipv4_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
256         struct ip ip = {0};
257         struct icmp icmp = {0};
258
259         struct in_addr ip_src;
260         struct in_addr ip_dst;
261         uint32_t oldlen;
262         int sockfd;
263
264         if(ratelimit(3))
265                 return;
266
267         /* Swap Ethernet source and destination addresses */
268
269         swap_mac_addresses(packet);
270
271         /* Copy headers from packet into properly aligned structs on the stack */
272
273         memcpy(&ip, DATA(packet) + ether_size, ip_size);
274
275         /* Remember original source and destination */
276
277         ip_src = ip.ip_src;
278         ip_dst = ip.ip_dst;
279
280         /* Try to reply with an IP address assigned to the local machine */
281
282         sockfd = socket(AF_INET, SOCK_DGRAM, 0);
283         if (sockfd != -1) {
284                 struct sockaddr_in addr;
285                 memset(&addr, 0, sizeof(addr));
286                 addr.sin_family = AF_INET;
287                 addr.sin_addr = ip.ip_src;
288                 if (!connect(sockfd, (const struct sockaddr*) &addr, sizeof(addr))) {
289                         memset(&addr, 0, sizeof(addr));
290                         addr.sin_family = AF_INET;
291                         socklen_t addrlen = sizeof(addr);
292                         if (!getsockname(sockfd, (struct sockaddr*) &addr, &addrlen) && addrlen <= sizeof(addr)) {
293                                 ip_dst = addr.sin_addr;
294                         }
295                 }
296                 close(sockfd);
297         }
298
299         oldlen = packet->len - ether_size;
300
301         if(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
302                 icmp.icmp_nextmtu = htons(packet->len - ether_size);
303
304         if(oldlen >= IP_MSS - ip_size - icmp_size)
305                 oldlen = IP_MSS - ip_size - icmp_size;
306
307         /* Copy first part of original contents to ICMP message */
308
309         memmove(DATA(packet) + ether_size + ip_size + icmp_size, DATA(packet) + ether_size, oldlen);
310
311         /* Fill in IPv4 header */
312
313         ip.ip_v = 4;
314         ip.ip_hl = ip_size / 4;
315         ip.ip_tos = 0;
316         ip.ip_len = htons(ip_size + icmp_size + oldlen);
317         ip.ip_id = 0;
318         ip.ip_off = 0;
319         ip.ip_ttl = 255;
320         ip.ip_p = IPPROTO_ICMP;
321         ip.ip_sum = 0;
322         ip.ip_src = ip_dst;
323         ip.ip_dst = ip_src;
324
325         ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
326
327         /* Fill in ICMP header */
328
329         icmp.icmp_type = type;
330         icmp.icmp_code = code;
331         icmp.icmp_cksum = 0;
332
333         icmp.icmp_cksum = inet_checksum(&icmp, icmp_size, ~0);
334         icmp.icmp_cksum = inet_checksum(DATA(packet) + ether_size + ip_size + icmp_size, oldlen, icmp.icmp_cksum);
335
336         /* Copy structs on stack back to packet */
337
338         memcpy(DATA(packet) + ether_size, &ip, ip_size);
339         memcpy(DATA(packet) + ether_size + ip_size, &icmp, icmp_size);
340
341         packet->len = ether_size + ip_size + icmp_size + oldlen;
342
343         send_packet(source, packet);
344 }
345
346 /* RFC 791 */
347
348 static void fragment_ipv4_packet(node_t *dest, vpn_packet_t *packet, length_t ether_size) {
349         struct ip ip;
350         vpn_packet_t fragment;
351         int len, maxlen, todo;
352         uint8_t *offset;
353         uint16_t ip_off, origf;
354
355         memcpy(&ip, DATA(packet) + ether_size, ip_size);
356         fragment.priority = packet->priority;
357         fragment.offset = DEFAULT_PACKET_OFFSET;
358
359         if(ip.ip_hl != ip_size / 4)
360                 return;
361
362         todo = ntohs(ip.ip_len) - ip_size;
363
364         if(ether_size + ip_size + todo != packet->len) {
365                 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));
366                 return;
367         }
368
369         logger(DEBUG_TRAFFIC, LOG_INFO, "Fragmenting packet of %d bytes to %s (%s)", packet->len, dest->name, dest->hostname);
370
371         offset = DATA(packet) + ether_size + ip_size;
372         maxlen = (dest->mtu - ether_size - ip_size) & ~0x7;
373         ip_off = ntohs(ip.ip_off);
374         origf = ip_off & ~IP_OFFMASK;
375         ip_off &= IP_OFFMASK;
376
377         while(todo) {
378                 len = todo > maxlen ? maxlen : todo;
379                 memcpy(DATA(&fragment) + ether_size + ip_size, offset, len);
380                 todo -= len;
381                 offset += len;
382
383                 ip.ip_len = htons(ip_size + len);
384                 ip.ip_off = htons(ip_off | origf | (todo ? IP_MF : 0));
385                 ip.ip_sum = 0;
386                 ip.ip_sum = inet_checksum(&ip, ip_size, ~0);
387                 memcpy(DATA(&fragment), DATA(packet), ether_size);
388                 memcpy(DATA(&fragment) + ether_size, &ip, ip_size);
389                 fragment.len = ether_size + ip_size + len;
390
391                 send_packet(dest, &fragment);
392
393                 ip_off += len / 8;
394         }
395 }
396
397 static void route_ipv4(node_t *source, vpn_packet_t *packet) {
398         if(!checklength(source, packet, ether_size + ip_size))
399                 return;
400
401         subnet_t *subnet;
402         node_t *via;
403         ipv4_t dest;
404
405         memcpy(&dest, &DATA(packet)[30], sizeof dest);
406         subnet = lookup_subnet_ipv4(&dest);
407
408         if(!subnet) {
409                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv4 destination address %d.%d.%d.%d",
410                                 source->name, source->hostname,
411                                 dest.x[0],
412                                 dest.x[1],
413                                 dest.x[2],
414                                 dest.x[3]);
415
416                 route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNKNOWN);
417                 return;
418         }
419
420         if (!subnet->owner) {
421                 broadcast_packet(source, packet);
422                 return;
423         }
424
425         if(subnet->owner == source) {
426                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
427                 return;
428         }
429
430         if(!subnet->owner->status.reachable)
431                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_UNREACH);
432
433         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
434                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
435
436         if(priorityinheritance)
437                 packet->priority = DATA(packet)[15];
438
439         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
440
441         if(via == source) {
442                 logger(DEBUG_TRAFFIC, LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
443                 return;
444         }
445
446         if(directonly && subnet->owner != via)
447                 return route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_NET_ANO);
448
449         if(via && packet->len > MAX(via->mtu, 590) && via != myself) {
450                 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);
451                 if(DATA(packet)[20] & 0x40) {
452                         packet->len = MAX(via->mtu, 590);
453                         route_ipv4_unreachable(source, packet, ether_size, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
454                 } else {
455                         fragment_ipv4_packet(via, packet, ether_size);
456                 }
457
458                 return;
459         }
460
461         clamp_mss(source, via, packet);
462
463         send_packet(subnet->owner, packet);
464 }
465
466 /* RFC 2463 */
467
468 static void route_ipv6_unreachable(node_t *source, vpn_packet_t *packet, length_t ether_size, uint8_t type, uint8_t code) {
469         struct ip6_hdr ip6;
470         struct icmp6_hdr icmp6 = {0};
471         uint16_t checksum;      
472         int sockfd;
473
474         struct {
475                 struct in6_addr ip6_src;        /* source address */
476                 struct in6_addr ip6_dst;        /* destination address */
477                 uint32_t length;
478                 uint32_t next;
479         } pseudo;
480
481         if(ratelimit(3))
482                 return;
483
484         /* Swap Ethernet source and destination addresses */
485
486         swap_mac_addresses(packet);
487
488         /* Copy headers from packet to structs on the stack */
489
490         memcpy(&ip6, DATA(packet) + ether_size, ip6_size);
491
492         /* Remember original source and destination */
493
494         pseudo.ip6_src = ip6.ip6_dst;
495         pseudo.ip6_dst = ip6.ip6_src;
496
497         /* Try to reply with an IP address assigned to the local machine */
498
499         sockfd = socket(AF_INET6, SOCK_DGRAM, 0);
500         if (sockfd != -1) {
501                 struct sockaddr_in6 addr;
502                 memset(&addr, 0, sizeof(addr));
503                 addr.sin6_family = AF_INET6;
504                 addr.sin6_addr = ip6.ip6_src;
505                 if (!connect(sockfd, (const struct sockaddr*) &addr, sizeof(addr))) {
506                         memset(&addr, 0, sizeof(addr));
507                         addr.sin6_family = AF_INET6;
508                         socklen_t addrlen = sizeof(addr);
509                         if (!getsockname(sockfd, (struct sockaddr*) &addr, &addrlen) && addrlen <= sizeof(addr)) {
510                                 pseudo.ip6_src = addr.sin6_addr;
511                         }
512                 }
513                 close(sockfd);
514         }
515
516         pseudo.length = packet->len - ether_size;
517
518         if(type == ICMP6_PACKET_TOO_BIG)
519                 icmp6.icmp6_mtu = htonl(pseudo.length);
520
521         if(pseudo.length >= IP_MSS - ip6_size - icmp6_size)
522                 pseudo.length = IP_MSS - ip6_size - icmp6_size;
523
524         /* Copy first part of original contents to ICMP message */
525
526         memmove(DATA(packet) + ether_size + ip6_size + icmp6_size, DATA(packet) + ether_size, pseudo.length);
527
528         /* Fill in IPv6 header */
529
530         ip6.ip6_flow = htonl(0x60000000UL);
531         ip6.ip6_plen = htons(icmp6_size + pseudo.length);
532         ip6.ip6_nxt = IPPROTO_ICMPV6;
533         ip6.ip6_hlim = 255;
534         ip6.ip6_src = pseudo.ip6_src;
535         ip6.ip6_dst = pseudo.ip6_dst;
536
537         /* Fill in ICMP header */
538
539         icmp6.icmp6_type = type;
540         icmp6.icmp6_code = code;
541         icmp6.icmp6_cksum = 0;
542
543         /* Create pseudo header */
544
545         pseudo.length = htonl(icmp6_size + pseudo.length);
546         pseudo.next = htonl(IPPROTO_ICMPV6);
547
548         /* Generate checksum */
549
550         checksum = inet_checksum(&pseudo, sizeof pseudo, ~0);
551         checksum = inet_checksum(&icmp6, icmp6_size, checksum);
552         checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + icmp6_size, ntohl(pseudo.length) - icmp6_size, checksum);
553
554         icmp6.icmp6_cksum = checksum;
555
556         /* Copy structs on stack back to packet */
557
558         memcpy(DATA(packet) + ether_size, &ip6, ip6_size);
559         memcpy(DATA(packet) + ether_size + ip6_size, &icmp6, icmp6_size);
560
561         packet->len = ether_size + ip6_size + ntohl(pseudo.length);
562
563         send_packet(source, packet);
564 }
565
566 static void route_neighborsol(node_t *source, vpn_packet_t *packet);
567
568 static void route_ipv6(node_t *source, vpn_packet_t *packet) {
569         if(!checklength(source, packet, ether_size + ip6_size))
570                 return;
571
572         if(DATA(packet)[20] == IPPROTO_ICMPV6 && checklength(source, packet, ether_size + ip6_size + icmp6_size) && DATA(packet)[54] == ND_NEIGHBOR_SOLICIT) {
573                 route_neighborsol(source, packet);
574                 return;
575         }
576
577         subnet_t *subnet;
578         node_t *via;
579         ipv6_t dest;
580
581         memcpy(&dest, &DATA(packet)[38], sizeof dest);
582         subnet = lookup_subnet_ipv6(&dest);
583
584         if(!subnet) {
585                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown IPv6 destination address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
586                                 source->name, source->hostname,
587                                 ntohs(dest.x[0]),
588                                 ntohs(dest.x[1]),
589                                 ntohs(dest.x[2]),
590                                 ntohs(dest.x[3]),
591                                 ntohs(dest.x[4]),
592                                 ntohs(dest.x[5]),
593                                 ntohs(dest.x[6]),
594                                 ntohs(dest.x[7]));
595
596                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR);
597                 return;
598         }
599
600         if (!subnet->owner) {
601                 broadcast_packet(source, packet);
602                 return;
603         }
604
605         if(subnet->owner == source) {
606                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
607                 return;
608         }
609
610         if(!subnet->owner->status.reachable)
611                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE);
612
613         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
614                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
615
616         via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
617
618         if(via == source) {
619                 logger(DEBUG_TRAFFIC, LOG_ERR, "Routing loop for packet from %s (%s)!", source->name, source->hostname);
620                 return;
621         }
622
623         if(directonly && subnet->owner != via)
624                 return route_ipv6_unreachable(source, packet, ether_size, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN);
625
626         if(via && packet->len > MAX(via->mtu, 1294) && via != myself) {
627                 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);
628                 packet->len = MAX(via->mtu, 1294);
629                 route_ipv6_unreachable(source, packet, ether_size, ICMP6_PACKET_TOO_BIG, 0);
630                 return;
631         }
632
633         clamp_mss(source, via, packet);
634
635         send_packet(subnet->owner, packet);
636 }
637
638 /* RFC 2461 */
639
640 static void route_neighborsol(node_t *source, vpn_packet_t *packet) {
641         struct ip6_hdr ip6;
642         struct nd_neighbor_solicit ns;
643         struct nd_opt_hdr opt;
644         subnet_t *subnet;
645         uint16_t checksum;
646         bool has_opt;
647
648         struct {
649                 struct in6_addr ip6_src;
650                 struct in6_addr ip6_dst;
651                 uint32_t length;
652                 uint32_t next;
653         } pseudo;
654
655         if(!checklength(source, packet, ether_size + ip6_size + ns_size))
656                 return;
657
658         has_opt = packet->len >= ether_size + ip6_size + ns_size + opt_size + ETH_ALEN;
659
660         if(source != myself) {
661                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got neighbor solicitation request from %s (%s) while in router mode!", source->name, source->hostname);
662                 return;
663         }
664
665         /* Copy headers from packet to structs on the stack */
666
667         memcpy(&ip6, DATA(packet) + ether_size, ip6_size);
668         memcpy(&ns, DATA(packet) + ether_size + ip6_size, ns_size);
669         if(has_opt)
670                 memcpy(&opt, DATA(packet) + ether_size + ip6_size + ns_size, opt_size);
671
672         /* First, snatch the source address from the neighbor solicitation packet */
673
674         if(overwrite_mac)
675                 memcpy(mymac.x, DATA(packet) + ETH_ALEN, ETH_ALEN);
676
677         /* Check if this is a valid neighbor solicitation request */
678
679         if(ns.nd_ns_hdr.icmp6_type != ND_NEIGHBOR_SOLICIT ||
680            (has_opt && opt.nd_opt_type != ND_OPT_SOURCE_LINKADDR)) {
681                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: received unknown type neighbor solicitation request");
682                 return;
683         }
684
685         /* Create pseudo header */
686
687         pseudo.ip6_src = ip6.ip6_src;
688         pseudo.ip6_dst = ip6.ip6_dst;
689         if(has_opt)
690                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
691         else
692                 pseudo.length = htonl(ns_size);
693         pseudo.next = htonl(IPPROTO_ICMPV6);
694
695         /* Generate checksum */
696
697         checksum = inet_checksum(&pseudo, sizeof pseudo, ~0);
698         checksum = inet_checksum(&ns, ns_size, checksum);
699         if(has_opt) {
700                 checksum = inet_checksum(&opt, opt_size, checksum);
701                 checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
702         }
703
704         if(checksum) {
705                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: checksum error for neighbor solicitation request");
706                 return;
707         }
708
709         /* Check if the IPv6 address exists on the VPN */
710
711         subnet = lookup_subnet_ipv6((ipv6_t *) &ns.nd_ns_target);
712
713         if(!subnet) {
714                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: neighbor solicitation request for unknown address %hx:%hx:%hx:%hx:%hx:%hx:%hx:%hx",
715                                    ntohs(((uint16_t *) &ns.nd_ns_target)[0]),
716                                    ntohs(((uint16_t *) &ns.nd_ns_target)[1]),
717                                    ntohs(((uint16_t *) &ns.nd_ns_target)[2]),
718                                    ntohs(((uint16_t *) &ns.nd_ns_target)[3]),
719                                    ntohs(((uint16_t *) &ns.nd_ns_target)[4]),
720                                    ntohs(((uint16_t *) &ns.nd_ns_target)[5]),
721                                    ntohs(((uint16_t *) &ns.nd_ns_target)[6]),
722                                    ntohs(((uint16_t *) &ns.nd_ns_target)[7]));
723
724                 return;
725         }
726
727         /* Check if it is for our own subnet */
728
729         if(subnet->owner == myself)
730                 return;                                          /* silently ignore */
731
732         /* Create neighbor advertation reply */
733
734         memcpy(DATA(packet), DATA(packet) + ETH_ALEN, ETH_ALEN); /* copy destination address */
735         DATA(packet)[ETH_ALEN * 2 - 1] ^= 0xFF;                  /* mangle source address so it looks like it's not from us */
736
737         ip6.ip6_dst = ip6.ip6_src;                               /* swap destination and source protocoll address */
738         ip6.ip6_src = ns.nd_ns_target;
739
740         if(has_opt)
741                 memcpy(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, DATA(packet) + ETH_ALEN, ETH_ALEN);   /* add fake source hard addr */
742
743         ns.nd_ns_cksum = 0;
744         ns.nd_ns_type = ND_NEIGHBOR_ADVERT;
745         ns.nd_ns_reserved = htonl(0x40000000UL);                 /* Set solicited flag */
746         opt.nd_opt_type = ND_OPT_TARGET_LINKADDR;
747
748         /* Create pseudo header */
749
750         pseudo.ip6_src = ip6.ip6_src;
751         pseudo.ip6_dst = ip6.ip6_dst;
752         if(has_opt)
753                 pseudo.length = htonl(ns_size + opt_size + ETH_ALEN);
754         else
755                 pseudo.length = htonl(ns_size);
756         pseudo.next = htonl(IPPROTO_ICMPV6);
757
758         /* Generate checksum */
759
760         checksum = inet_checksum(&pseudo, sizeof pseudo, ~0);
761         checksum = inet_checksum(&ns, ns_size, checksum);
762         if(has_opt) {
763                 checksum = inet_checksum(&opt, opt_size, checksum);
764                 checksum = inet_checksum(DATA(packet) + ether_size + ip6_size + ns_size + opt_size, ETH_ALEN, checksum);
765         }
766
767         ns.nd_ns_hdr.icmp6_cksum = checksum;
768
769         /* Copy structs on stack back to packet */
770
771         memcpy(DATA(packet) + ether_size, &ip6, ip6_size);
772         memcpy(DATA(packet) + ether_size + ip6_size, &ns, ns_size);
773         if(has_opt)
774                 memcpy(DATA(packet) + ether_size + ip6_size + ns_size, &opt, opt_size);
775
776         send_packet(source, packet);
777 }
778
779 /* RFC 826 */
780
781 static void route_arp(node_t *source, vpn_packet_t *packet) {
782         struct ether_arp arp;
783         subnet_t *subnet;
784         struct in_addr addr;
785
786         if(!checklength(source, packet, ether_size + arp_size))
787                 return;
788
789         if(source != myself) {
790                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Got ARP request from %s (%s) while in router mode!", source->name, source->hostname);
791                 return;
792         }
793
794         /* First, snatch the source address from the ARP packet */
795
796         if(overwrite_mac)
797                 memcpy(mymac.x, DATA(packet) + ETH_ALEN, ETH_ALEN);
798
799         /* Copy headers from packet to structs on the stack */
800
801         memcpy(&arp, DATA(packet) + ether_size, arp_size);
802
803         /* Check if this is a valid ARP request */
804
805         if(ntohs(arp.arp_hrd) != ARPHRD_ETHER || ntohs(arp.arp_pro) != ETH_P_IP ||
806            arp.arp_hln != ETH_ALEN || arp.arp_pln != sizeof addr || ntohs(arp.arp_op) != ARPOP_REQUEST) {
807                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: received unknown type ARP request");
808                 return;
809         }
810
811         /* Check if the IPv4 address exists on the VPN */
812
813         subnet = lookup_subnet_ipv4((ipv4_t *) &arp.arp_tpa);
814
815         if(!subnet) {
816                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet: ARP request for unknown address %d.%d.%d.%d",
817                                    arp.arp_tpa[0], arp.arp_tpa[1], arp.arp_tpa[2],
818                                    arp.arp_tpa[3]);
819                 return;
820         }
821
822         /* Check if it is for our own subnet */
823
824         if(subnet->owner == myself)
825                 return;                                          /* silently ignore */
826
827         memcpy(&addr, arp.arp_tpa, sizeof addr);                 /* save protocol addr */
828         memcpy(arp.arp_tpa, arp.arp_spa, sizeof addr);           /* swap destination and source protocol address */
829         memcpy(arp.arp_spa, &addr, sizeof addr);                 /* ... */
830
831         memcpy(arp.arp_tha, arp.arp_sha, ETH_ALEN);              /* set target hard/proto addr */
832         memcpy(arp.arp_sha, DATA(packet) + ETH_ALEN, ETH_ALEN);  /* set source hard/proto addr */
833         arp.arp_sha[ETH_ALEN - 1] ^= 0xFF;                       /* for consistency with route_packet() */
834         arp.arp_op = htons(ARPOP_REPLY);
835
836         /* Copy structs on stack back to packet */
837
838         memcpy(DATA(packet) + ether_size, &arp, arp_size);
839
840         send_packet(source, packet);
841 }
842
843 static void route_mac(node_t *source, vpn_packet_t *packet) {
844         subnet_t *subnet;
845         mac_t dest;
846
847         /* Learn source address */
848
849         if(source == myself) {
850                 mac_t src;
851                 memcpy(&src, &DATA(packet)[6], sizeof src);
852                 learn_mac(&src);
853         }
854
855         /* Lookup destination address */
856
857         memcpy(&dest, &DATA(packet)[0], sizeof dest);
858         subnet = lookup_subnet_mac(NULL, &dest);
859
860         if(!subnet || !subnet->owner) {
861                 broadcast_packet(source, packet);
862                 return;
863         }
864
865         if(subnet->owner == source) {
866                 logger(DEBUG_TRAFFIC, LOG_WARNING, "Packet looping back to %s (%s)!", source->name, source->hostname);
867                 return;
868         }
869
870         if(forwarding_mode == FMODE_OFF && source != myself && subnet->owner != myself)
871                 return;
872
873         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
874
875         if(priorityinheritance && type == ETH_P_IP && packet->len >= ether_size + ip_size)
876                 packet->priority = DATA(packet)[15];
877
878         // Handle packets larger than PMTU
879
880         node_t *via = (subnet->owner->via == myself) ? subnet->owner->nexthop : subnet->owner->via;
881
882         if(directonly && subnet->owner != via)
883                 return;
884
885         if(via && packet->len > via->mtu && via != myself) {
886                 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);
887                 length_t ethlen = 14;
888
889                 if(type == ETH_P_8021Q) {
890                         type = DATA(packet)[16] << 8 | DATA(packet)[17];
891                         ethlen += 4;
892                 }
893
894                 if(type == ETH_P_IP && packet->len > 576 + ethlen) {
895                         if(DATA(packet)[6 + ethlen] & 0x40) {
896                                 packet->len = via->mtu;
897                                 route_ipv4_unreachable(source, packet, ethlen, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED);
898                         } else {
899                                 fragment_ipv4_packet(via, packet, ethlen);
900                         }
901                         return;
902                 } else if(type == ETH_P_IPV6 && packet->len > 1280 + ethlen) {
903                         packet->len = via->mtu;
904                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_PACKET_TOO_BIG, 0);
905                         return;
906                 }
907         }
908
909         clamp_mss(source, via, packet);
910
911         send_packet(subnet->owner, packet);
912 }
913
914 static void send_pcap(vpn_packet_t *packet) {
915         pcap = false;
916
917         for list_each(connection_t, c, connection_list) {
918                 if(!c->status.pcap)
919                         continue;
920
921                 pcap = true;
922                 int len = packet->len;
923                 if(c->outmaclength && c->outmaclength < len)
924                         len = c->outmaclength;
925
926                 if(send_request(c, "%d %d %d", CONTROL, REQ_PCAP, len))
927                         send_meta(c, (char *)DATA(packet), len);
928         }
929 }
930
931 static bool do_decrement_ttl(node_t *source, vpn_packet_t *packet) {
932         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
933         length_t ethlen = ether_size;
934
935         if(type == ETH_P_8021Q) {
936                 type = DATA(packet)[16] << 8 | DATA(packet)[17];
937                 ethlen += 4;
938         }
939
940         switch (type) {
941                 case ETH_P_IP:
942                         if(!checklength(source, packet, ethlen + ip_size))
943                                 return false;
944
945                         if(DATA(packet)[ethlen + 8] <= 1) {
946                                 if(DATA(packet)[ethlen + 11] != IPPROTO_ICMP || DATA(packet)[ethlen + 32] != ICMP_TIME_EXCEEDED)
947                                         route_ipv4_unreachable(source, packet, ethlen, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL);
948                                 return false;
949                         }
950
951                         uint16_t old = DATA(packet)[ethlen + 8] << 8 | DATA(packet)[ethlen + 9];
952                         DATA(packet)[ethlen + 8]--;
953                         uint16_t new = DATA(packet)[ethlen + 8] << 8 | DATA(packet)[ethlen + 9];
954
955                         uint32_t checksum = DATA(packet)[ethlen + 10] << 8 | DATA(packet)[ethlen + 11];
956                         checksum += old + (~new & 0xFFFF);
957                         while(checksum >> 16)
958                                 checksum = (checksum & 0xFFFF) + (checksum >> 16);
959                         DATA(packet)[ethlen + 10] = checksum >> 8;
960                         DATA(packet)[ethlen + 11] = checksum & 0xff;
961
962                         return true;
963
964                 case ETH_P_IPV6:
965                         if(!checklength(source, packet, ethlen + ip6_size))
966                                 return false;
967
968                         if(DATA(packet)[ethlen + 7] <= 1) {
969                                 if(DATA(packet)[ethlen + 6] != IPPROTO_ICMPV6 || DATA(packet)[ethlen + 40] != ICMP6_TIME_EXCEEDED)
970                                         route_ipv6_unreachable(source, packet, ethlen, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT);
971                                 return false;
972                         }
973
974                         DATA(packet)[ethlen + 7]--;
975
976                         return true;
977
978                 default:
979                         return true;
980         }
981 }
982
983 void route(node_t *source, vpn_packet_t *packet) {
984         if(pcap)
985                 send_pcap(packet);
986
987         if(forwarding_mode == FMODE_KERNEL && source != myself) {
988                 send_packet(myself, packet);
989                 return;
990         }
991
992         if(!checklength(source, packet, ether_size))
993                 return;
994
995         if(decrement_ttl && source != myself)
996                 if(!do_decrement_ttl(source, packet))
997                         return;
998
999         uint16_t type = DATA(packet)[12] << 8 | DATA(packet)[13];
1000
1001         switch (routing_mode) {
1002                 case RMODE_ROUTER:
1003                         switch (type) {
1004                                 case ETH_P_ARP:
1005                                         route_arp(source, packet);
1006                                         break;
1007
1008                                 case ETH_P_IP:
1009                                         route_ipv4(source, packet);
1010                                         break;
1011
1012                                 case ETH_P_IPV6:
1013                                         route_ipv6(source, packet);
1014                                         break;
1015
1016                                 default:
1017                                         logger(DEBUG_TRAFFIC, LOG_WARNING, "Cannot route packet from %s (%s): unknown type %hx", source->name, source->hostname, type);
1018                                         break;
1019                         }
1020                         break;
1021
1022                 case RMODE_SWITCH:
1023                         route_mac(source, packet);
1024                         break;
1025
1026                 case RMODE_HUB:
1027                         broadcast_packet(source, packet);
1028                         break;
1029         }
1030 }