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>
37 static void handle_solicit(void *addr, void *data, size_t len,
38 struct interface *iface, void *dest);
39 static void handle_rtnetlink(void *addr, void *data, size_t len,
40 struct interface *iface, void *dest);
41 static void catch_rtnetlink(int error);
43 static uint32_t rtnl_seqid = 0;
44 static int ping_socket = -1;
45 static struct odhcpd_event rtnl_event = {{.fd = -1}, handle_rtnetlink, catch_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
66 // Setup netlink socket
67 if ((rtnl_event.uloop.fd = odhcpd_open_rtnl()) < 0)
70 if (setsockopt(rtnl_event.uloop.fd, SOL_SOCKET, SO_RCVBUF, &val, sizeof(val)))
71 setsockopt(rtnl_event.uloop.fd, SOL_SOCKET, SO_RCVBUFFORCE, &val, sizeof(val));
73 // Receive netlink neighbor and ip-address events
74 uint32_t group = RTNLGRP_IPV6_IFADDR;
75 setsockopt(rtnl_event.uloop.fd, SOL_NETLINK,
76 NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
77 group = RTNLGRP_IPV6_ROUTE;
78 setsockopt(rtnl_event.uloop.fd, SOL_NETLINK,
79 NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
81 odhcpd_register(&rtnl_event);
84 ping_socket = socket(AF_INET6, SOCK_RAW | SOCK_CLOEXEC, IPPROTO_ICMPV6);
85 if (ping_socket < 0) {
86 syslog(LOG_ERR, "Unable to open raw socket: %s", strerror(errno));
91 setsockopt(ping_socket, IPPROTO_RAW, IPV6_CHECKSUM, &val, sizeof(val));
93 // This is required by RFC 4861
95 setsockopt(ping_socket, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, &val, sizeof(val));
96 setsockopt(ping_socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val));
98 // Filter all packages, we only want to send
99 struct icmp6_filter filt;
100 ICMP6_FILTER_SETBLOCKALL(&filt);
101 setsockopt(ping_socket, IPPROTO_ICMPV6, ICMP6_FILTER, &filt, sizeof(filt));
104 // Netlink socket, continued...
105 group = RTNLGRP_NEIGH;
106 setsockopt(rtnl_event.uloop.fd, SOL_NETLINK, NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
112 static void dump_neigh_table(bool proxy)
118 {sizeof(req), RTM_GETNEIGH, NLM_F_REQUEST | NLM_F_DUMP,
120 {.ndm_family = AF_INET6, .ndm_flags = (proxy) ? NTF_PROXY : 0}
122 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
123 odhcpd_process(&rtnl_event);
127 int setup_ndp_interface(struct interface *iface, bool enable)
130 snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
131 int procfd = open(procbuf, O_WRONLY);
132 bool dump_neigh = false;
134 if (iface->ndp_event.uloop.fd > 0) {
135 uloop_fd_delete(&iface->ndp_event.uloop);
136 close(iface->ndp_event.uloop.fd);
137 iface->ndp_event.uloop.fd = -1;
139 if (!enable || iface->ndp != RELAYD_RELAY)
140 if (write(procfd, "0\n", 2) < 0) {}
145 if (enable && (iface->ra == RELAYD_SERVER ||
146 iface->dhcpv6 == RELAYD_SERVER || iface->ndp == RELAYD_RELAY)) {
147 // Synthesize initial address events
150 struct ifaddrmsg ifa;
152 {sizeof(req2), RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP,
154 {.ifa_family = AF_INET6, .ifa_index = iface->ifindex}
156 send(rtnl_event.uloop.fd, &req2, sizeof(req2), MSG_DONTWAIT);
159 if (enable && iface->ndp == RELAYD_RELAY) {
160 if (write(procfd, "1\n", 2) < 0) {}
163 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
165 syslog(LOG_ERR, "Unable to open packet socket: %s",
170 #ifdef PACKET_RECV_TYPE
171 int pktt = 1 << PACKET_MULTICAST;
172 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
175 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
176 &bpf_prog, sizeof(bpf_prog))) {
177 syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
181 struct sockaddr_ll ll = {
182 .sll_family = AF_PACKET,
183 .sll_ifindex = iface->ifindex,
184 .sll_protocol = htons(ETH_P_IPV6),
190 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
192 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
193 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
195 iface->ndp_event.uloop.fd = sock;
196 iface->ndp_event.handle_dgram = handle_solicit;
197 odhcpd_register(&iface->ndp_event);
199 // If we already were enabled dump is unnecessary, if not do dump
201 dump_neigh_table(false);
209 dump_neigh_table(true);
215 // Send an ICMP-ECHO. This is less for actually pinging but for the
216 // neighbor cache to be kept up-to-date.
217 static void ping6(struct in6_addr *addr,
218 const struct interface *iface)
220 struct sockaddr_in6 dest = {AF_INET6, 0, 0, *addr, iface->ifindex};
221 struct icmp6_hdr echo = {.icmp6_type = ICMP6_ECHO_REQUEST};
222 struct iovec iov = {&echo, sizeof(echo)};
224 odhcpd_setup_route(addr, 128, iface, NULL, 128, true);
225 odhcpd_send(ping_socket, &dest, &iov, 1, iface);
226 odhcpd_setup_route(addr, 128, iface, NULL, 128, false);
230 // Handle solicitations
231 static void handle_solicit(void *addr, void *data, size_t len,
232 struct interface *iface, _unused void *dest)
234 struct ip6_hdr *ip6 = data;
235 struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
236 struct sockaddr_ll *ll = addr;
238 // Solicitation is for duplicate address detection
239 bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
241 // Don't forward any non-DAD solicitation for external ifaces
242 // TODO: check if we should even forward DADs for them
243 if (iface->external && !ns_is_dad)
246 if (len < sizeof(*ip6) + sizeof(*req))
247 return; // Invalid reqicitation
249 if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
250 IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
251 IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
252 return; // Invalid target
254 char ipbuf[INET6_ADDRSTRLEN];
255 inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
256 syslog(LOG_DEBUG, "Got a NS for %s", ipbuf);
259 odhcpd_get_mac(iface, mac);
260 if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
261 return; // Looped back
264 list_for_each_entry(c, &interfaces, head)
265 if (iface->ndp == RELAYD_RELAY && iface != c &&
266 (ns_is_dad || !c->external))
267 ping6(&req->nd_ns_target, c);
270 // Use rtnetlink to modify kernel routes
271 static void setup_route(struct in6_addr *addr, struct interface *iface, bool add)
273 char namebuf[INET6_ADDRSTRLEN];
274 inet_ntop(AF_INET6, addr, namebuf, sizeof(namebuf));
275 syslog(LOG_NOTICE, "%s about %s on %s",
276 (add) ? "Learned" : "Forgot", namebuf, iface->ifname);
278 if (iface->learn_routes)
279 odhcpd_setup_route(addr, 128, iface, NULL, 1024, add);
283 static int prefixcmp(const void *va, const void *vb)
285 const struct odhcpd_ipaddr *a = va, *b = vb;
286 uint32_t a_pref = ((a->addr.s6_addr[0] & 0xfe) != 0xfc) ? a->preferred : 1;
287 uint32_t b_pref = ((b->addr.s6_addr[0] & 0xfe) != 0xfc) ? b->preferred : 1;
288 return (a_pref < b_pref) ? 1 : (a_pref > b_pref) ? -1 : 0;
291 // Check address update
292 static void check_updates(struct interface *iface)
294 struct odhcpd_ipaddr addr[RELAYD_MAX_ADDRS] = {{IN6ADDR_ANY_INIT, 0, 0, 0, 0}};
295 time_t now = odhcpd_time();
296 ssize_t len = odhcpd_get_interface_addresses(iface->ifindex, addr, ARRAY_SIZE(addr));
301 qsort(addr, len, sizeof(*addr), prefixcmp);
303 for (int i = 0; i < len; ++i) {
304 addr[i].addr.s6_addr32[3] = 0;
306 if (addr[i].preferred < UINT32_MAX - now)
307 addr[i].preferred += now;
309 if (addr[i].valid < UINT32_MAX - now)
310 addr[i].valid += now;
313 bool change = len != (ssize_t)iface->ia_addr_len;
314 for (ssize_t i = 0; !change && i < len; ++i)
315 if (!IN6_ARE_ADDR_EQUAL(&addr[i].addr, &iface->ia_addr[i].addr) ||
316 (addr[i].preferred > 0) != (iface->ia_addr[i].preferred > 0) ||
317 addr[i].valid < iface->ia_addr[i].valid ||
318 addr[i].preferred < iface->ia_addr[i].preferred)
322 dhcpv6_ia_preupdate(iface);
324 memcpy(iface->ia_addr, addr, len * sizeof(*addr));
325 iface->ia_addr_len = len;
328 dhcpv6_ia_postupdate(iface, now);
331 syslog(LOG_INFO, "Raising SIGUSR1 due to address change on %s", iface->ifname);
337 // Handler for neighbor cache entries from the kernel. This is our source
338 // to learn and unlearn hosts on interfaces.
339 static void handle_rtnetlink(_unused void *addr, void *data, size_t len,
340 _unused struct interface *iface, _unused void *dest)
342 bool dump_neigh = false;
343 struct in6_addr last_solicited = IN6ADDR_ANY_INIT;
345 for (struct nlmsghdr *nh = data; NLMSG_OK(nh, len);
346 nh = NLMSG_NEXT(nh, len)) {
347 struct ndmsg *ndm = NLMSG_DATA(nh);
348 struct rtmsg *rtm = NLMSG_DATA(nh);
350 bool is_addr = (nh->nlmsg_type == RTM_NEWADDR
351 || nh->nlmsg_type == RTM_DELADDR);
352 bool is_route = (nh->nlmsg_type == RTM_NEWROUTE
353 || nh->nlmsg_type == RTM_DELROUTE);
354 bool is_neigh = (nh->nlmsg_type == RTM_NEWNEIGH
355 || nh->nlmsg_type == RTM_DELNEIGH);
357 // Family and ifindex are on the same offset for NEIGH and ADDR
358 if ((!is_addr && !is_route && !is_neigh)
359 || NLMSG_PAYLOAD(nh, 0) < sizeof(*ndm)
360 || ndm->ndm_family != AF_INET6)
364 // Inform about a change in default route
365 if (rtm->rtm_dst_len == 0) {
366 syslog(LOG_INFO, "Raising SIGUSR1 due to default route change");
374 size_t rta_offset = (is_addr) ? sizeof(struct ifaddrmsg) : sizeof(*ndm);
375 uint16_t atype = (is_addr) ? IFA_ADDRESS : NDA_DST;
376 ssize_t alen = NLMSG_PAYLOAD(nh, rta_offset);
377 struct in6_addr *addr = NULL;
379 for (struct rtattr *rta = (void*)(((uint8_t*)ndm) + rta_offset);
380 RTA_OK(rta, alen); rta = RTA_NEXT(rta, alen)) {
381 if (rta->rta_type == atype &&
382 RTA_PAYLOAD(rta) >= sizeof(*addr)) {
383 addr = RTA_DATA(rta);
388 struct interface *iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex);
392 // Address not specified or unrelated
393 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
394 IN6_IS_ADDR_MULTICAST(addr))
400 add = (nh->nlmsg_type == RTM_NEWADDR);
402 add = (nh->nlmsg_type == RTM_NEWNEIGH && (ndm->ndm_state &
403 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE
404 | NUD_PERMANENT | NUD_NOARP)));
406 if (iface->ndp == RELAYD_RELAY) {
407 // Replay change to all neighbor cache
411 struct nlattr nla_dst;
414 {sizeof(req), RTM_DELNEIGH, NLM_F_REQUEST,
416 {.ndm_family = AF_INET6, .ndm_flags = NTF_PROXY},
417 {sizeof(struct nlattr) + sizeof(struct in6_addr), NDA_DST},
421 if (ndm->ndm_flags & NTF_PROXY) {
422 // Dump & flush proxy entries
423 if (nh->nlmsg_type == RTM_NEWNEIGH) {
424 req.ndm.ndm_ifindex = iface->ifindex;
425 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
426 setup_route(addr, iface, false);
431 list_for_each_entry(c, &interfaces, head) {
435 if (c->ndp == RELAYD_RELAY) {
436 req.nh.nlmsg_type = RTM_NEWNEIGH;
437 req.nh.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
439 req.ndm.ndm_ifindex = c->ifindex;
440 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
441 } else { // Delete NDP cache from interfaces without relay
442 req.nh.nlmsg_type = RTM_DELNEIGH;
443 req.nh.nlmsg_flags &= ~(NLM_F_CREATE | NLM_F_REPLACE);
445 req.ndm.ndm_ifindex = c->ifindex;
446 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
450 setup_route(addr, iface, true);
452 if (nh->nlmsg_type == RTM_NEWNEIGH) {
453 // might be locally originating
454 if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
455 last_solicited = *addr;
458 list_for_each_entry(c, &interfaces, head)
459 if (iface->ndp == RELAYD_RELAY && iface != c &&
460 !c->external == false)
465 list_for_each_entry(c, &interfaces, head) {
466 if (c->ndp == RELAYD_RELAY && iface != c) {
467 req.ndm.ndm_ifindex = c->ifindex;
468 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
471 setup_route(addr, iface, false);
473 // also: dump to add proxies back in case it moved elsewhere
480 check_updates(iface);
482 if (iface->ndp == RELAYD_RELAY && iface->master) {
483 // Replay address changes on all slave interfaces
484 nh->nlmsg_flags = NLM_F_REQUEST;
486 if (nh->nlmsg_type == RTM_NEWADDR)
487 nh->nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
490 list_for_each_entry(c, &interfaces, head) {
491 if (c->ndp == RELAYD_RELAY && !c->master) {
492 ndm->ndm_ifindex = c->ifindex;
493 send(rtnl_event.uloop.fd, nh, nh->nlmsg_len, MSG_DONTWAIT);
501 dump_neigh_table(false);
504 static void catch_rtnetlink(int error)
506 if (error == ENOBUFS) {
509 struct ifaddrmsg ifa;
511 {sizeof(req2), RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP,
513 {.ifa_family = AF_INET6}
515 send(rtnl_event.uloop.fd, &req2, sizeof(req2), MSG_DONTWAIT);