2 * Copyright (C) 2012-2013 Steven Barth <steven@midlink.org>
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.
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.
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>
29 #include <linux/rtnetlink.h>
30 #include <linux/filter.h>
36 static void handle_solicit(void *addr, void *data, size_t len,
37 struct interface *iface, void *dest);
38 static void handle_rtnetlink(void *addr, void *data, size_t len,
39 struct interface *iface, void *dest);
40 static ssize_t ping6(struct in6_addr *addr,
41 const struct interface *iface);
43 static uint32_t rtnl_seqid = 0;
44 static int ping_socket = -1;
45 static struct odhcpd_event rtnl_event = {{.fd = -1}, handle_rtnetlink};
48 // Filter ICMPv6 messages of type neighbor soliciation
49 static struct sock_filter bpf[] = {
50 BPF_STMT(BPF_LD | BPF_B | BPF_ABS, offsetof(struct ip6_hdr, ip6_nxt)),
51 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, IPPROTO_ICMPV6, 0, 3),
52 BPF_STMT(BPF_LD | BPF_B | BPF_ABS, sizeof(struct ip6_hdr) +
53 offsetof(struct icmp6_hdr, icmp6_type)),
54 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, ND_NEIGHBOR_SOLICIT, 0, 1),
55 BPF_STMT(BPF_RET | BPF_K, 0xffffffff),
56 BPF_STMT(BPF_RET | BPF_K, 0),
58 static const struct sock_fprog bpf_prog = {sizeof(bpf) / sizeof(*bpf), bpf};
61 // Initialize NDP-proxy
64 // Setup netlink socket
65 if ((rtnl_event.uloop.fd = odhcpd_open_rtnl()) < 0)
68 // Receive netlink neighbor and ip-address events
69 uint32_t group = RTNLGRP_IPV6_IFADDR;
70 setsockopt(rtnl_event.uloop.fd, SOL_NETLINK,
71 NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
72 group = RTNLGRP_IPV6_ROUTE;
73 setsockopt(rtnl_event.uloop.fd, SOL_NETLINK,
74 NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
76 odhcpd_register(&rtnl_event);
79 ping_socket = socket(AF_INET6, SOCK_RAW | SOCK_CLOEXEC, IPPROTO_ICMPV6);
80 if (ping_socket < 0) {
81 syslog(LOG_ERR, "Unable to open raw socket: %s", strerror(errno));
86 setsockopt(ping_socket, IPPROTO_RAW, IPV6_CHECKSUM, &val, sizeof(val));
88 // This is required by RFC 4861
90 setsockopt(ping_socket, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, &val, sizeof(val));
91 setsockopt(ping_socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val));
93 // Filter all packages, we only want to send
94 struct icmp6_filter filt;
95 ICMP6_FILTER_SETBLOCKALL(&filt);
96 setsockopt(ping_socket, IPPROTO_ICMPV6, ICMP6_FILTER, &filt, sizeof(filt));
99 // Netlink socket, continued...
100 group = RTNLGRP_NEIGH;
101 setsockopt(rtnl_event.uloop.fd, SOL_NETLINK, NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
107 static void dump_neigh_table(bool proxy)
113 {sizeof(req), RTM_GETNEIGH, NLM_F_REQUEST | NLM_F_DUMP,
115 {.ndm_family = AF_INET6, .ndm_flags = (proxy) ? NTF_PROXY : 0}
117 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
121 int setup_ndp_interface(struct interface *iface, bool enable)
124 snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
125 int procfd = open(procbuf, O_WRONLY);
126 bool dump_neigh = false;
128 if (iface->ndp_event.uloop.fd > 0) {
129 uloop_fd_delete(&iface->ndp_event.uloop);
130 close(iface->ndp_event.uloop.fd);
131 iface->ndp_event.uloop.fd = -1;
133 if (!enable || iface->ndp != RELAYD_RELAY)
134 write(procfd, "0\n", 2);
139 if (enable && (iface->ra == RELAYD_SERVER ||
140 iface->dhcpv6 == RELAYD_SERVER || iface->ndp == RELAYD_RELAY)) {
141 // Synthesize initial address events
144 struct ifaddrmsg ifa;
146 {sizeof(req2), RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP,
148 {.ifa_family = AF_INET6, .ifa_index = iface->ifindex}
150 send(rtnl_event.uloop.fd, &req2, sizeof(req2), MSG_DONTWAIT);
153 if (enable && iface->ndp == RELAYD_RELAY) {
154 write(procfd, "1\n", 2);
157 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
159 syslog(LOG_ERR, "Unable to open packet socket: %s",
164 #ifdef PACKET_RECV_TYPE
165 int pktt = 1 << PACKET_MULTICAST;
166 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
169 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
170 &bpf_prog, sizeof(bpf_prog))) {
171 syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
175 struct sockaddr_ll ll = {
176 .sll_family = AF_PACKET,
177 .sll_ifindex = iface->ifindex,
178 .sll_protocol = htons(ETH_P_IPV6),
184 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
186 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
187 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
189 iface->ndp_event.uloop.fd = sock;
190 iface->ndp_event.handle_dgram = handle_solicit;
191 odhcpd_register(&iface->ndp_event);
193 // If we already were enabled dump is unnecessary, if not do dump
195 dump_neigh_table(false);
203 dump_neigh_table(true);
209 // Send an ICMP-ECHO. This is less for actually pinging but for the
210 // neighbor cache to be kept up-to-date.
211 static ssize_t ping6(struct in6_addr *addr,
212 const struct interface *iface)
214 struct sockaddr_in6 dest = {AF_INET6, 0, 0, *addr, 0};
215 struct icmp6_hdr echo = {.icmp6_type = ICMP6_ECHO_REQUEST};
216 struct iovec iov = {&echo, sizeof(echo)};
218 // Linux seems to not honor IPV6_PKTINFO on raw-sockets, so work around
219 setsockopt(ping_socket, SOL_SOCKET, SO_BINDTODEVICE,
220 iface->ifname, sizeof(iface->ifname));
221 return odhcpd_send(ping_socket, &dest, &iov, 1, iface);
225 // Handle solicitations
226 static void handle_solicit(void *addr, void *data, size_t len,
227 struct interface *iface, _unused void *dest)
229 struct ip6_hdr *ip6 = data;
230 struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
231 struct sockaddr_ll *ll = addr;
233 // Solicitation is for duplicate address detection
234 bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
236 // Don't forward any non-DAD solicitation for external ifaces
237 // TODO: check if we should even forward DADs for them
238 if (iface->external && !ns_is_dad)
241 if (len < sizeof(*ip6) + sizeof(*req))
242 return; // Invalid reqicitation
244 if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
245 IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
246 IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
247 return; // Invalid target
249 char ipbuf[INET6_ADDRSTRLEN];
250 inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
251 syslog(LOG_DEBUG, "Got a NS for %s", ipbuf);
254 odhcpd_get_mac(iface, mac);
255 if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
256 return; // Looped back
259 list_for_each_entry(c, &interfaces, head)
260 if (iface->ndp == RELAYD_RELAY && iface != c &&
261 (ns_is_dad || !c->external))
262 ping6(&req->nd_ns_target, c);
266 void odhcpd_setup_route(const struct in6_addr *addr, int prefixlen,
267 const struct interface *iface, const struct in6_addr *gw, bool add)
272 struct rtattr rta_dst;
273 struct in6_addr dst_addr;
274 struct rtattr rta_oif;
276 struct rtattr rta_table;
278 struct rtattr rta_gw;
281 {sizeof(req), 0, NLM_F_REQUEST, ++rtnl_seqid, 0},
282 {AF_INET6, prefixlen, 0, 0, 0, 0, 0, 0, 0},
283 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_DST},
285 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_OIF},
287 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_TABLE},
289 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_GATEWAY},
297 req.nh.nlmsg_type = RTM_NEWROUTE;
298 req.nh.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
299 req.rtm.rtm_protocol = RTPROT_STATIC;
300 req.rtm.rtm_scope = (gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
301 req.rtm.rtm_type = RTN_UNICAST;
303 req.nh.nlmsg_type = RTM_DELROUTE;
304 req.rtm.rtm_scope = RT_SCOPE_NOWHERE;
307 req.nh.nlmsg_len = (gw) ? sizeof(req) : offsetof(struct req, rta_gw);
308 send(rtnl_event.uloop.fd, &req, req.nh.nlmsg_len, MSG_DONTWAIT);
311 // Use rtnetlink to modify kernel routes
312 static void setup_route(struct in6_addr *addr, struct interface *iface,
315 char namebuf[INET6_ADDRSTRLEN];
316 inet_ntop(AF_INET6, addr, namebuf, sizeof(namebuf));
317 syslog(LOG_NOTICE, "%s about %s on %s", (add) ? "Learned" : "Forgot",
318 namebuf, (iface) ? iface->ifname : "<pending>");
320 if (!iface || !iface->learn_routes)
323 odhcpd_setup_route(addr, 128, iface, NULL, add);
327 // Handler for neighbor cache entries from the kernel. This is our source
328 // to learn and unlearn hosts on interfaces.
329 static void handle_rtnetlink(_unused void *addr, void *data, size_t len,
330 _unused struct interface *iface, _unused void *dest)
332 bool dump_neigh = false;
333 struct in6_addr last_solicited = IN6ADDR_ANY_INIT;
335 for (struct nlmsghdr *nh = data; NLMSG_OK(nh, len);
336 nh = NLMSG_NEXT(nh, len)) {
337 struct rtmsg *rtm = NLMSG_DATA(nh);
338 if ((nh->nlmsg_type == RTM_NEWROUTE ||
339 nh->nlmsg_type == RTM_DELROUTE) &&
340 rtm->rtm_dst_len == 0)
341 raise(SIGUSR1); // Inform about a change in default route
343 struct ndmsg *ndm = NLMSG_DATA(nh);
344 struct ifaddrmsg *ifa = NLMSG_DATA(nh);
345 if (nh->nlmsg_type != RTM_NEWNEIGH
346 && nh->nlmsg_type != RTM_DELNEIGH
347 && nh->nlmsg_type != RTM_NEWADDR
348 && nh->nlmsg_type != RTM_DELADDR)
349 continue; // Unrelated message type
350 bool is_addr = (nh->nlmsg_type == RTM_NEWADDR
351 || nh->nlmsg_type == RTM_DELADDR);
353 // Family and ifindex are on the same offset for NEIGH and ADDR
354 if (NLMSG_PAYLOAD(nh, 0) < sizeof(*ndm)
355 || ndm->ndm_family != AF_INET6)
359 struct interface *iface;
360 if (!(iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex)))
364 size_t rta_offset = (is_addr) ? sizeof(*ifa) : sizeof(*ndm);
365 uint16_t atype = (is_addr) ? IFA_ADDRESS : NDA_DST;
366 ssize_t alen = NLMSG_PAYLOAD(nh, rta_offset);
367 struct in6_addr *addr = NULL;
369 for (struct rtattr *rta = (void*)(((uint8_t*)ndm) + rta_offset);
370 RTA_OK(rta, alen); rta = RTA_NEXT(rta, alen))
371 if (rta->rta_type == atype &&
372 RTA_PAYLOAD(rta) >= sizeof(*addr))
373 addr = RTA_DATA(rta);
375 // Address not specified or unrelated
376 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
377 IN6_IS_ADDR_MULTICAST(addr))
383 add = (nh->nlmsg_type == RTM_NEWADDR);
385 add = (nh->nlmsg_type == RTM_NEWNEIGH && (ndm->ndm_state &
386 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE
387 | NUD_PERMANENT | NUD_NOARP)));
389 if (iface->ndp == RELAYD_RELAY) {
390 // Replay change to all neighbor cache
394 struct nlattr nla_dst;
397 {sizeof(req), RTM_DELNEIGH, NLM_F_REQUEST,
399 {.ndm_family = AF_INET6, .ndm_flags = NTF_PROXY},
400 {sizeof(struct nlattr) + sizeof(struct in6_addr), NDA_DST},
404 if (ndm->ndm_flags & NTF_PROXY) {
405 // Dump & flush proxy entries
406 if (nh->nlmsg_type == RTM_NEWNEIGH) {
407 req.ndm.ndm_ifindex = iface->ifindex;
408 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
413 list_for_each_entry(c, &interfaces, head) {
417 if (c->ndp == RELAYD_RELAY) {
418 req.nh.nlmsg_type = RTM_NEWNEIGH;
419 req.nh.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
421 req.ndm.ndm_ifindex = c->ifindex;
422 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
423 } else { // Delete NDP cache from interfaces without relay
424 req.nh.nlmsg_type = RTM_DELNEIGH;
425 req.nh.nlmsg_flags &= ~(NLM_F_CREATE | NLM_F_REPLACE);
427 req.ndm.ndm_ifindex = c->ifindex;
428 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
432 setup_route(addr, iface, true);
434 if (nh->nlmsg_type == RTM_NEWNEIGH) {
435 // might be locally originating
436 if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
437 last_solicited = *addr;
440 list_for_each_entry(c, &interfaces, head)
441 if (iface->ndp == RELAYD_RELAY && iface != c &&
442 !c->external == false)
447 list_for_each_entry(c, &interfaces, head) {
448 if (c->ndp == RELAYD_RELAY && iface != c) {
449 req.ndm.ndm_ifindex = c->ifindex;
450 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
453 setup_route(addr, iface, false);
455 // also: dump to add proxies back in case it moved elsewhere
462 if (iface->ra == RELAYD_SERVER)
463 raise(SIGUSR1); // Inform about a change in addresses
465 if (iface->dhcpv6 == RELAYD_SERVER)
466 iface->ia_reconf = true;
468 if (iface->ndp == RELAYD_RELAY && iface->master) {
469 // Replay address changes on all slave interfaces
470 nh->nlmsg_flags = NLM_F_REQUEST;
472 if (nh->nlmsg_type == RTM_NEWADDR)
473 nh->nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
476 list_for_each_entry(c, &interfaces, head) {
477 if (c->ndp == RELAYD_RELAY && !c->master) {
478 ifa->ifa_index = c->ifindex;
479 send(rtnl_event.uloop.fd, nh, nh->nlmsg_len, MSG_DONTWAIT);
487 dump_neigh_table(false);