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);
118 odhcpd_process(&rtnl_event);
122 int setup_ndp_interface(struct interface *iface, bool enable)
125 snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
126 int procfd = open(procbuf, O_WRONLY);
127 bool dump_neigh = false;
129 if (iface->ndp_event.uloop.fd > 0) {
130 uloop_fd_delete(&iface->ndp_event.uloop);
131 close(iface->ndp_event.uloop.fd);
132 iface->ndp_event.uloop.fd = -1;
134 if (!enable || iface->ndp != RELAYD_RELAY)
135 if (write(procfd, "0\n", 2) < 0) {}
140 if (enable && (iface->ra == RELAYD_SERVER ||
141 iface->dhcpv6 == RELAYD_SERVER || iface->ndp == RELAYD_RELAY)) {
142 // Synthesize initial address events
145 struct ifaddrmsg ifa;
147 {sizeof(req2), RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP,
149 {.ifa_family = AF_INET6, .ifa_index = iface->ifindex}
151 send(rtnl_event.uloop.fd, &req2, sizeof(req2), MSG_DONTWAIT);
154 if (enable && iface->ndp == RELAYD_RELAY) {
155 if (write(procfd, "1\n", 2) < 0) {}
158 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
160 syslog(LOG_ERR, "Unable to open packet socket: %s",
165 #ifdef PACKET_RECV_TYPE
166 int pktt = 1 << PACKET_MULTICAST;
167 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
170 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
171 &bpf_prog, sizeof(bpf_prog))) {
172 syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
176 struct sockaddr_ll ll = {
177 .sll_family = AF_PACKET,
178 .sll_ifindex = iface->ifindex,
179 .sll_protocol = htons(ETH_P_IPV6),
185 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
187 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
188 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
190 iface->ndp_event.uloop.fd = sock;
191 iface->ndp_event.handle_dgram = handle_solicit;
192 odhcpd_register(&iface->ndp_event);
194 // If we already were enabled dump is unnecessary, if not do dump
196 dump_neigh_table(false);
204 dump_neigh_table(true);
210 // Send an ICMP-ECHO. This is less for actually pinging but for the
211 // neighbor cache to be kept up-to-date.
212 static ssize_t ping6(struct in6_addr *addr,
213 const struct interface *iface)
215 struct sockaddr_in6 dest = {AF_INET6, 0, 0, *addr, 0};
216 struct icmp6_hdr echo = {.icmp6_type = ICMP6_ECHO_REQUEST};
217 struct iovec iov = {&echo, sizeof(echo)};
219 // Linux seems to not honor IPV6_PKTINFO on raw-sockets, so work around
220 setsockopt(ping_socket, SOL_SOCKET, SO_BINDTODEVICE,
221 iface->ifname, sizeof(iface->ifname));
222 return odhcpd_send(ping_socket, &dest, &iov, 1, iface);
226 // Handle solicitations
227 static void handle_solicit(void *addr, void *data, size_t len,
228 struct interface *iface, _unused void *dest)
230 struct ip6_hdr *ip6 = data;
231 struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
232 struct sockaddr_ll *ll = addr;
234 // Solicitation is for duplicate address detection
235 bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
237 // Don't forward any non-DAD solicitation for external ifaces
238 // TODO: check if we should even forward DADs for them
239 if (iface->external && !ns_is_dad)
242 if (len < sizeof(*ip6) + sizeof(*req))
243 return; // Invalid reqicitation
245 if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
246 IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
247 IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
248 return; // Invalid target
250 char ipbuf[INET6_ADDRSTRLEN];
251 inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
252 syslog(LOG_DEBUG, "Got a NS for %s", ipbuf);
255 odhcpd_get_mac(iface, mac);
256 if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
257 return; // Looped back
260 list_for_each_entry(c, &interfaces, head)
261 if (iface->ndp == RELAYD_RELAY && iface != c &&
262 (ns_is_dad || !c->external))
263 ping6(&req->nd_ns_target, c);
267 void odhcpd_setup_route(const struct in6_addr *addr, int prefixlen,
268 const struct interface *iface, const struct in6_addr *gw, bool add)
273 struct rtattr rta_dst;
274 struct in6_addr dst_addr;
275 struct rtattr rta_oif;
277 struct rtattr rta_table;
279 struct rtattr rta_gw;
282 {sizeof(req), 0, NLM_F_REQUEST, ++rtnl_seqid, 0},
283 {AF_INET6, prefixlen, 0, 0, 0, 0, 0, 0, 0},
284 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_DST},
286 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_OIF},
288 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_TABLE},
290 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_GATEWAY},
298 req.nh.nlmsg_type = RTM_NEWROUTE;
299 req.nh.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
300 req.rtm.rtm_protocol = RTPROT_STATIC;
301 req.rtm.rtm_scope = (gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
302 req.rtm.rtm_type = RTN_UNICAST;
304 req.nh.nlmsg_type = RTM_DELROUTE;
305 req.rtm.rtm_scope = RT_SCOPE_NOWHERE;
308 req.nh.nlmsg_len = (gw) ? sizeof(req) : offsetof(struct req, rta_gw);
309 send(rtnl_event.uloop.fd, &req, req.nh.nlmsg_len, MSG_DONTWAIT);
312 // Use rtnetlink to modify kernel routes
313 static void setup_route(struct in6_addr *addr, struct interface *iface, bool add)
315 char namebuf[INET6_ADDRSTRLEN];
316 inet_ntop(AF_INET6, addr, namebuf, sizeof(namebuf));
317 syslog(LOG_NOTICE, "%s about %s on %s",
318 (add) ? "Learned" : "Forgot", namebuf, iface->ifname);
320 if (iface->learn_routes)
321 odhcpd_setup_route(addr, 128, iface, NULL, add);
325 // Handler for neighbor cache entries from the kernel. This is our source
326 // to learn and unlearn hosts on interfaces.
327 static void handle_rtnetlink(_unused void *addr, void *data, size_t len,
328 _unused struct interface *iface, _unused void *dest)
330 bool dump_neigh = false;
331 struct in6_addr last_solicited = IN6ADDR_ANY_INIT;
333 for (struct nlmsghdr *nh = data; NLMSG_OK(nh, len);
334 nh = NLMSG_NEXT(nh, len)) {
335 struct rtmsg *rtm = NLMSG_DATA(nh);
336 if ((nh->nlmsg_type == RTM_NEWROUTE ||
337 nh->nlmsg_type == RTM_DELROUTE) &&
338 rtm->rtm_dst_len == 0)
339 raise(SIGUSR1); // Inform about a change in default route
341 struct ndmsg *ndm = NLMSG_DATA(nh);
342 struct ifaddrmsg *ifa = NLMSG_DATA(nh);
343 if (nh->nlmsg_type != RTM_NEWNEIGH
344 && nh->nlmsg_type != RTM_DELNEIGH
345 && nh->nlmsg_type != RTM_NEWADDR
346 && nh->nlmsg_type != RTM_DELADDR)
347 continue; // Unrelated message type
348 bool is_addr = (nh->nlmsg_type == RTM_NEWADDR
349 || nh->nlmsg_type == RTM_DELADDR);
351 // Family and ifindex are on the same offset for NEIGH and ADDR
352 if (NLMSG_PAYLOAD(nh, 0) < sizeof(*ndm)
353 || ndm->ndm_family != AF_INET6)
357 struct interface *iface;
358 if (!(iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex)))
362 size_t rta_offset = (is_addr) ? sizeof(*ifa) : sizeof(*ndm);
363 uint16_t atype = (is_addr) ? IFA_ADDRESS : NDA_DST;
364 ssize_t alen = NLMSG_PAYLOAD(nh, rta_offset);
365 struct in6_addr *addr = NULL;
367 for (struct rtattr *rta = (void*)(((uint8_t*)ndm) + rta_offset);
368 RTA_OK(rta, alen); rta = RTA_NEXT(rta, alen))
369 if (rta->rta_type == atype &&
370 RTA_PAYLOAD(rta) >= sizeof(*addr))
371 addr = RTA_DATA(rta);
373 // Address not specified or unrelated
374 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
375 IN6_IS_ADDR_MULTICAST(addr))
381 add = (nh->nlmsg_type == RTM_NEWADDR);
383 add = (nh->nlmsg_type == RTM_NEWNEIGH && (ndm->ndm_state &
384 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE
385 | NUD_PERMANENT | NUD_NOARP)));
387 if (iface->ndp == RELAYD_RELAY) {
388 // Replay change to all neighbor cache
392 struct nlattr nla_dst;
395 {sizeof(req), RTM_DELNEIGH, NLM_F_REQUEST,
397 {.ndm_family = AF_INET6, .ndm_flags = NTF_PROXY},
398 {sizeof(struct nlattr) + sizeof(struct in6_addr), NDA_DST},
402 if (ndm->ndm_flags & NTF_PROXY) {
403 // Dump & flush proxy entries
404 if (nh->nlmsg_type == RTM_NEWNEIGH) {
405 req.ndm.ndm_ifindex = iface->ifindex;
406 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
407 setup_route(addr, iface, false);
412 list_for_each_entry(c, &interfaces, head) {
416 if (c->ndp == RELAYD_RELAY) {
417 req.nh.nlmsg_type = RTM_NEWNEIGH;
418 req.nh.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
420 req.ndm.ndm_ifindex = c->ifindex;
421 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
422 } else { // Delete NDP cache from interfaces without relay
423 req.nh.nlmsg_type = RTM_DELNEIGH;
424 req.nh.nlmsg_flags &= ~(NLM_F_CREATE | NLM_F_REPLACE);
426 req.ndm.ndm_ifindex = c->ifindex;
427 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
431 setup_route(addr, iface, true);
433 if (nh->nlmsg_type == RTM_NEWNEIGH) {
434 // might be locally originating
435 if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
436 last_solicited = *addr;
439 list_for_each_entry(c, &interfaces, head)
440 if (iface->ndp == RELAYD_RELAY && iface != c &&
441 !c->external == false)
446 list_for_each_entry(c, &interfaces, head) {
447 if (c->ndp == RELAYD_RELAY && iface != c) {
448 req.ndm.ndm_ifindex = c->ifindex;
449 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
452 setup_route(addr, iface, false);
454 // also: dump to add proxies back in case it moved elsewhere
461 if (iface->ra == RELAYD_SERVER)
462 raise(SIGUSR1); // Inform about a change in addresses
464 if (iface->dhcpv6 == RELAYD_SERVER)
465 iface->ia_reconf = true;
467 if (iface->ndp == RELAYD_RELAY && iface->master) {
468 // Replay address changes on all slave interfaces
469 nh->nlmsg_flags = NLM_F_REQUEST;
471 if (nh->nlmsg_type == RTM_NEWADDR)
472 nh->nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
475 list_for_each_entry(c, &interfaces, head) {
476 if (c->ndp == RELAYD_RELAY && !c->master) {
477 ifa->ifa_index = c->ifindex;
478 send(rtnl_event.uloop.fd, nh, nh->nlmsg_len, MSG_DONTWAIT);
486 dump_neigh_table(false);