ec990ed5dff89006cfd1d50dcca3351616431e7f
[project/odhcpd.git] / src / ndp.c
1 /**
2 * Copyright (C) 2012-2013 Steven Barth <steven@midlink.org>
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License v2 as published by
6 * the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
12 *
13 */
14
15 #include <stdio.h>
16 #include <stdlib.h>
17 #include <signal.h>
18 #include <errno.h>
19
20 #include <fcntl.h>
21 #include <unistd.h>
22 #include <arpa/inet.h>
23 #include <sys/socket.h>
24 #include <net/ethernet.h>
25 #include <netinet/ip6.h>
26 #include <netinet/icmp6.h>
27 #include <netpacket/packet.h>
28
29 #include <linux/rtnetlink.h>
30 #include <linux/filter.h>
31
32 #include <netlink/msg.h>
33 #include <netlink/socket.h>
34 #include <netlink/attr.h>
35
36 #include "dhcpv6.h"
37 #include "odhcpd.h"
38
39 struct event_socket {
40 struct odhcpd_event ev;
41 struct nl_sock *sock;
42 int sock_bufsize;
43 };
44
45 static void handle_solicit(void *addr, void *data, size_t len,
46 struct interface *iface, void *dest);
47 static void handle_rtnl_event(struct odhcpd_event *ev);
48 static int cb_rtnl_valid(struct nl_msg *msg, void *arg);
49 static void catch_rtnl_err(struct odhcpd_event *e, int error);
50
51 static int ping_socket = -1;
52 static struct event_socket rtnl_event = {
53 .ev = {
54 .uloop = {.fd = - 1, },
55 .handle_dgram = NULL,
56 .handle_error = catch_rtnl_err,
57 .recv_msgs = handle_rtnl_event,
58 },
59 .sock = NULL,
60 .sock_bufsize = 133120,
61 };
62
63 // Filter ICMPv6 messages of type neighbor soliciation
64 static struct sock_filter bpf[] = {
65 BPF_STMT(BPF_LD | BPF_B | BPF_ABS, offsetof(struct ip6_hdr, ip6_nxt)),
66 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, IPPROTO_ICMPV6, 0, 3),
67 BPF_STMT(BPF_LD | BPF_B | BPF_ABS, sizeof(struct ip6_hdr) +
68 offsetof(struct icmp6_hdr, icmp6_type)),
69 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, ND_NEIGHBOR_SOLICIT, 0, 1),
70 BPF_STMT(BPF_RET | BPF_K, 0xffffffff),
71 BPF_STMT(BPF_RET | BPF_K, 0),
72 };
73 static const struct sock_fprog bpf_prog = {sizeof(bpf) / sizeof(*bpf), bpf};
74
75
76 // Initialize NDP-proxy
77 int init_ndp(void)
78 {
79 int val = 2;
80
81 rtnl_event.sock = odhcpd_create_nl_socket(NETLINK_ROUTE);
82 if (!rtnl_event.sock)
83 goto err;
84
85 rtnl_event.ev.uloop.fd = nl_socket_get_fd(rtnl_event.sock);
86
87 if (nl_socket_set_buffer_size(rtnl_event.sock, rtnl_event.sock_bufsize, 0))
88 goto err;
89
90 nl_socket_disable_seq_check(rtnl_event.sock);
91
92 nl_socket_modify_cb(rtnl_event.sock, NL_CB_VALID, NL_CB_CUSTOM,
93 cb_rtnl_valid, NULL);
94
95 // Receive IPv4 address, IPv6 address, IPv6 routes and neighbor events
96 if (nl_socket_add_memberships(rtnl_event.sock, RTNLGRP_IPV4_IFADDR,
97 RTNLGRP_IPV6_IFADDR, RTNLGRP_IPV6_ROUTE,
98 RTNLGRP_NEIGH, RTNLGRP_LINK, 0))
99 goto err;
100
101 odhcpd_register(&rtnl_event.ev);
102
103 // Open ICMPv6 socket
104 ping_socket = socket(AF_INET6, SOCK_RAW | SOCK_CLOEXEC, IPPROTO_ICMPV6);
105 if (ping_socket < 0) {
106 syslog(LOG_ERR, "Unable to open raw socket: %s", strerror(errno));
107 return -1;
108 }
109
110 setsockopt(ping_socket, IPPROTO_RAW, IPV6_CHECKSUM, &val, sizeof(val));
111
112 // This is required by RFC 4861
113 val = 255;
114 setsockopt(ping_socket, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, &val, sizeof(val));
115 setsockopt(ping_socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val));
116
117 // Filter all packages, we only want to send
118 struct icmp6_filter filt;
119 ICMP6_FILTER_SETBLOCKALL(&filt);
120 setsockopt(ping_socket, IPPROTO_ICMPV6, ICMP6_FILTER, &filt, sizeof(filt));
121
122 return 0;
123
124 err:
125 if (rtnl_event.sock) {
126 nl_socket_free(rtnl_event.sock);
127 rtnl_event.sock = NULL;
128 rtnl_event.ev.uloop.fd = -1;
129 }
130
131 return -1;
132 }
133
134 static void dump_neigh_table(const bool proxy)
135 {
136 struct nl_msg *msg;
137 struct ndmsg ndm = {
138 .ndm_family = AF_INET6,
139 .ndm_flags = proxy ? NTF_PROXY : 0,
140 };
141
142 msg = nlmsg_alloc_simple(RTM_GETNEIGH, NLM_F_REQUEST | NLM_F_DUMP);
143 if (!msg)
144 return;
145
146 nlmsg_append(msg, &ndm, sizeof(ndm), 0);
147
148 nl_send_auto_complete(rtnl_event.sock, msg);
149
150 nlmsg_free(msg);
151 }
152
153 static void dump_addr_table(bool v6)
154 {
155 struct nl_msg *msg;
156 struct ifaddrmsg ifa = {
157 .ifa_family = v6 ? AF_INET6 : AF_INET,
158 };
159
160 msg = nlmsg_alloc_simple(RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP);
161 if (!msg)
162 return;
163
164 nlmsg_append(msg, &ifa, sizeof(ifa), 0);
165
166 nl_send_auto_complete(rtnl_event.sock, msg);
167
168 nlmsg_free(msg);
169 }
170
171 int setup_ndp_interface(struct interface *iface, bool enable)
172 {
173 int ret = 0, procfd;
174 bool dump_neigh = false;
175 char procbuf[64];
176
177 snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
178 procfd = open(procbuf, O_WRONLY);
179
180 if (procfd < 0) {
181 ret = -1;
182 goto out;
183 }
184
185 if (iface->ndp_event.uloop.fd > 0) {
186 uloop_fd_delete(&iface->ndp_event.uloop);
187 close(iface->ndp_event.uloop.fd);
188 iface->ndp_event.uloop.fd = -1;
189
190 if (!enable || iface->ndp != MODE_RELAY)
191 if (write(procfd, "0\n", 2) < 0) {}
192
193 dump_neigh = true;
194 }
195
196 if (enable && iface->ndp == MODE_RELAY) {
197 if (write(procfd, "1\n", 2) < 0) {}
198
199 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
200 if (sock < 0) {
201 syslog(LOG_ERR, "Unable to open packet socket: %s",
202 strerror(errno));
203 ret = -1;
204 goto out;
205 }
206
207 #ifdef PACKET_RECV_TYPE
208 int pktt = 1 << PACKET_MULTICAST;
209 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
210 #endif
211
212 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
213 &bpf_prog, sizeof(bpf_prog))) {
214 syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
215 ret = -1;
216 goto out;
217 }
218
219 struct sockaddr_ll ll = {
220 .sll_family = AF_PACKET,
221 .sll_ifindex = iface->ifindex,
222 .sll_protocol = htons(ETH_P_IPV6),
223 .sll_hatype = 0,
224 .sll_pkttype = 0,
225 .sll_halen = 0,
226 .sll_addr = {0},
227 };
228 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
229
230 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
231 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
232
233 iface->ndp_event.uloop.fd = sock;
234 iface->ndp_event.handle_dgram = handle_solicit;
235 odhcpd_register(&iface->ndp_event);
236
237 // If we already were enabled dump is unnecessary, if not do dump
238 if (!dump_neigh)
239 dump_neigh_table(false);
240 else
241 dump_neigh = false;
242 }
243
244 if (dump_neigh)
245 dump_neigh_table(true);
246
247 out:
248 if (procfd >= 0)
249 close(procfd);
250
251 return ret;
252 }
253
254
255 // Send an ICMP-ECHO. This is less for actually pinging but for the
256 // neighbor cache to be kept up-to-date.
257 static void ping6(struct in6_addr *addr,
258 const struct interface *iface)
259 {
260 struct sockaddr_in6 dest = { .sin6_family = AF_INET6, .sin6_addr = *addr, .sin6_scope_id = iface->ifindex, };
261 struct icmp6_hdr echo = { .icmp6_type = ICMP6_ECHO_REQUEST };
262 struct iovec iov = { .iov_base = &echo, .iov_len = sizeof(echo) };
263 char ipbuf[INET6_ADDRSTRLEN];
264
265 inet_ntop(AF_INET6, addr, ipbuf, sizeof(ipbuf));
266 syslog(LOG_NOTICE, "Pinging for %s%%%s", ipbuf, iface->ifname);
267
268 odhcpd_setup_route(addr, 128, iface, NULL, 128, true);
269 odhcpd_send(ping_socket, &dest, &iov, 1, iface);
270 odhcpd_setup_route(addr, 128, iface, NULL, 128, false);
271 }
272
273 // Handle solicitations
274 static void handle_solicit(void *addr, void *data, size_t len,
275 struct interface *iface, _unused void *dest)
276 {
277 struct ip6_hdr *ip6 = data;
278 struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
279 struct sockaddr_ll *ll = addr;
280 char ipbuf[INET6_ADDRSTRLEN];
281 uint8_t mac[6];
282
283 // Solicitation is for duplicate address detection
284 bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
285
286 // Don't process solicit messages on non relay interfaces
287 // Don't forward any non-DAD solicitation for external ifaces
288 // TODO: check if we should even forward DADs for them
289 if (iface->ndp != MODE_RELAY || (iface->external && !ns_is_dad))
290 return;
291
292 if (len < sizeof(*ip6) + sizeof(*req))
293 return; // Invalid reqicitation
294
295 if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
296 IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
297 IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
298 return; // Invalid target
299
300 inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
301 syslog(LOG_DEBUG, "Got a NS for %s%%%s", ipbuf, iface->ifname);
302
303 odhcpd_get_mac(iface, mac);
304 if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
305 return; // Looped back
306
307 struct interface *c;
308 list_for_each_entry(c, &interfaces, head)
309 if (iface != c && c->ndp == MODE_RELAY &&
310 (ns_is_dad || !c->external))
311 ping6(&req->nd_ns_target, c);
312 }
313
314 // Use rtnetlink to modify kernel routes
315 static void setup_route(struct in6_addr *addr, struct interface *iface, bool add)
316 {
317 char ipbuf[INET6_ADDRSTRLEN];
318
319 inet_ntop(AF_INET6, addr, ipbuf, sizeof(ipbuf));
320 syslog(LOG_NOTICE, "%s about %s%%%s",
321 (add) ? "Learned" : "Forgot", ipbuf, iface->ifname);
322
323 if (iface->learn_routes)
324 odhcpd_setup_route(addr, 128, iface, NULL, 1024, add);
325 }
326
327 // Check address update
328 static void check_addr_updates(struct interface *iface)
329 {
330 struct odhcpd_ipaddr *addr = NULL;
331 ssize_t len = odhcpd_get_interface_addresses(iface->ifindex, false, &addr);
332
333 if (len < 0)
334 return;
335
336 bool change = len != (ssize_t)iface->addr4_len;
337 for (ssize_t i = 0; !change && i < len; ++i)
338 if (addr[i].addr.in.s_addr != iface->addr4[i].addr.in.s_addr)
339 change = true;
340
341 free(iface->addr4);
342 iface->addr4 = addr;
343 iface->addr4_len = len;
344 }
345
346 // Check v6 address update
347 static void check_addr6_updates(struct interface *iface)
348 {
349 struct odhcpd_ipaddr *addr = NULL;
350 ssize_t len = odhcpd_get_interface_addresses(iface->ifindex, true, &addr);
351
352 if (len < 0)
353 return;
354
355 bool change = len != (ssize_t)iface->ia_addr_len;
356 for (ssize_t i = 0; !change && i < len; ++i)
357 if (!IN6_ARE_ADDR_EQUAL(&addr[i].addr.in6, &iface->ia_addr[i].addr.in6) ||
358 (addr[i].preferred > 0) != (iface->ia_addr[i].preferred > 0) ||
359 addr[i].valid < iface->ia_addr[i].valid ||
360 addr[i].preferred < iface->ia_addr[i].preferred)
361 change = true;
362
363 if (change)
364 dhcpv6_ia_preupdate(iface);
365
366 free(iface->ia_addr);
367 iface->ia_addr = addr;
368 iface->ia_addr_len = len;
369
370 if (change) {
371 dhcpv6_ia_postupdate(iface);
372 syslog(LOG_INFO, "Raising SIGUSR1 due to address change on %s", iface->ifname);
373 raise(SIGUSR1);
374 }
375 }
376
377 static void setup_addr_for_relaying(struct in6_addr *addr, struct interface *iface, bool add)
378 {
379 struct interface *c;
380 char ipbuf[INET6_ADDRSTRLEN];
381
382 inet_ntop(AF_INET6, addr, ipbuf, sizeof(ipbuf));
383
384 list_for_each_entry(c, &interfaces, head) {
385 if (iface == c || (c->ndp != MODE_RELAY && !add))
386 continue;
387
388 bool neigh_add = (c->ndp == MODE_RELAY ? add : false);
389
390 if (odhcpd_setup_proxy_neigh(addr, c, neigh_add))
391 syslog(LOG_DEBUG, "Failed to %s proxy neighbour entry %s%%%s",
392 neigh_add ? "add" : "delete", ipbuf, c->ifname);
393 else
394 syslog(LOG_DEBUG, "%s proxy neighbour entry %s%%%s",
395 neigh_add ? "Added" : "Deleted", ipbuf, c->ifname);
396 }
397 }
398
399 static void handle_rtnl_event(struct odhcpd_event *e)
400 {
401 struct event_socket *ev_sock = container_of(e, struct event_socket, ev);
402
403 nl_recvmsgs_default(ev_sock->sock);
404 }
405
406
407 // Handler for neighbor cache entries from the kernel. This is our source
408 // to learn and unlearn hosts on interfaces.
409 static int cb_rtnl_valid(struct nl_msg *msg, _unused void *arg)
410 {
411 struct nlmsghdr *hdr = nlmsg_hdr(msg);
412 struct in6_addr *addr6 = NULL;
413 struct interface *iface = NULL;
414 bool add = false;
415 char ipbuf[INET6_ADDRSTRLEN];
416
417 switch (hdr->nlmsg_type) {
418 case RTM_NEWLINK: {
419 struct ifinfomsg *ifi = nlmsg_data(hdr);
420 struct nlattr *nla[__IFLA_MAX];
421
422 if (!nlmsg_valid_hdr(hdr, sizeof(*ifi)) ||
423 ifi->ifi_family != AF_UNSPEC)
424 return NL_SKIP;
425
426 nlmsg_parse(hdr, sizeof(struct ifinfomsg), nla, __IFLA_MAX - 1, NULL);
427 if (!nla[IFLA_IFNAME])
428 return NL_SKIP;
429
430 struct interface *iface = odhcpd_get_interface_by_name(nla_data(nla[IFLA_IFNAME]));
431 if (!iface)
432 return NL_SKIP;
433
434 if (iface->ifindex != ifi->ifi_index) {
435 iface->ifindex = ifi->ifi_index;
436 check_addr_updates(iface);
437 }
438 break;
439 }
440
441 case RTM_NEWROUTE:
442 case RTM_DELROUTE: {
443 struct rtmsg *rtm = nlmsg_data(hdr);
444
445 if (!nlmsg_valid_hdr(hdr, sizeof(*rtm)) ||
446 rtm->rtm_family != AF_INET6)
447 return NL_SKIP;
448
449 if (rtm->rtm_dst_len == 0) {
450 syslog(LOG_INFO, "Raising SIGUSR1 due to default route change");
451 raise(SIGUSR1);
452 }
453 break;
454 }
455
456 case RTM_NEWADDR:
457 add = true;
458 /* fall through */
459 case RTM_DELADDR: {
460 struct ifaddrmsg *ifa = nlmsg_data(hdr);
461 struct nlattr *nla[__IFA_MAX];
462
463 if (!nlmsg_valid_hdr(hdr, sizeof(*ifa)) ||
464 (ifa->ifa_family != AF_INET6 &&
465 ifa->ifa_family != AF_INET))
466 return NL_SKIP;
467
468 iface = odhcpd_get_interface_by_index(ifa->ifa_index);
469 if (!iface)
470 return NL_SKIP;
471
472 nlmsg_parse(hdr, sizeof(*ifa), nla, __IFA_MAX - 1, NULL);
473
474 if (ifa->ifa_family == AF_INET6) {
475 if (!nla[IFA_ADDRESS])
476 return NL_SKIP;
477
478 addr6 = nla_data(nla[IFA_ADDRESS]);
479 if (!addr6 || IN6_IS_ADDR_LINKLOCAL(addr6) ||
480 IN6_IS_ADDR_MULTICAST(addr6))
481 return NL_SKIP;
482
483 inet_ntop(AF_INET6, addr6, ipbuf, sizeof(ipbuf));
484 syslog(LOG_DEBUG, "Netlink %s %s%%%s", add ? "newaddr" : "deladdr",
485 ipbuf, iface->ifname);
486
487 check_addr6_updates(iface);
488
489 if (iface->ndp != MODE_RELAY)
490 break;
491
492 /* handle the relay logic below */
493 setup_addr_for_relaying(addr6, iface, add);
494
495 if (!add)
496 dump_neigh_table(false);
497 } else {
498 if (!nla[IFA_LOCAL])
499 return NL_SKIP;
500
501 struct in_addr *addr = nla_data(nla[IFA_ADDRESS]);
502
503 inet_ntop(AF_INET, addr, ipbuf, sizeof(ipbuf));
504 syslog(LOG_DEBUG, "Netlink %s %s%%%s", add ? "newaddr" : "deladdr",
505 ipbuf, iface->ifname);
506
507 check_addr_updates(iface);
508 }
509 break;
510 }
511
512 case RTM_NEWNEIGH:
513 add = true;
514 /* fall through */
515 case RTM_DELNEIGH: {
516 struct ndmsg *ndm = nlmsg_data(hdr);
517 struct nlattr *nla[__NDA_MAX];
518
519 if (!nlmsg_valid_hdr(hdr, sizeof(*ndm)) ||
520 ndm->ndm_family != AF_INET6)
521 return NL_SKIP;
522
523 iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex);
524 if (!iface || iface->ndp != MODE_RELAY)
525 return (iface ? NL_OK : NL_SKIP);
526
527 nlmsg_parse(hdr, sizeof(*ndm), nla, __NDA_MAX - 1, NULL);
528 if (!nla[NDA_DST])
529 return NL_SKIP;
530
531 addr6 = nla_data(nla[NDA_DST]);
532 if (!addr6 || IN6_IS_ADDR_LINKLOCAL(addr6) ||
533 IN6_IS_ADDR_MULTICAST(addr6))
534 return NL_SKIP;
535
536 inet_ntop(AF_INET6, addr6, ipbuf, sizeof(ipbuf));
537 syslog(LOG_DEBUG, "Netlink %s %s%%%s", true ? "newneigh" : "delneigh",
538 ipbuf, iface->ifname);
539
540 if (ndm->ndm_flags & NTF_PROXY) {
541 /* Dump and flush proxy entries */
542 if (hdr->nlmsg_type == RTM_NEWNEIGH) {
543 odhcpd_setup_proxy_neigh(addr6, iface, false);
544 setup_route(addr6, iface, false);
545 dump_neigh_table(false);
546 }
547
548 return NL_OK;
549 }
550
551 if (add && !(ndm->ndm_state &
552 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE |
553 NUD_PERMANENT | NUD_NOARP)))
554 return NL_OK;
555
556 setup_addr_for_relaying(addr6, iface, add);
557 setup_route(addr6, iface, add);
558
559 if (!add)
560 dump_neigh_table(false);
561 break;
562 }
563
564 default:
565 return NL_SKIP;
566 }
567
568 return NL_OK;
569 }
570
571 static void catch_rtnl_err(struct odhcpd_event *e, int error)
572 {
573 struct event_socket *ev_sock = container_of(e, struct event_socket, ev);
574
575 if (error != ENOBUFS)
576 goto err;
577
578 /* Double netlink event buffer size */
579 ev_sock->sock_bufsize *= 2;
580
581 if (nl_socket_set_buffer_size(ev_sock->sock, ev_sock->sock_bufsize, 0))
582 goto err;
583
584 dump_addr_table(true);
585 return;
586
587 err:
588 odhcpd_deregister(e);
589 }