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 #include "router.h"
  32 #include "ndp.h"
  33
  34
  35
  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);
  42
  43 static uint32_t rtnl_seqid = 0;
  44 static int ping_socket = -1;
  45 static struct odhcpd_event rtnl_event = {{.fd = -1}, handle_rtnetlink};
  46
  47
  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),
  57 };
  58 static const struct sock_fprog bpf_prog = {sizeof(bpf) / sizeof(*bpf), bpf};
  59
  60
  61 // Initialize NDP-proxy
  62 int init_ndp(void)
  63 {
  64         // Setup netlink socket
  65         if ((rtnl_event.uloop.fd = odhcpd_open_rtnl()) < 0)
  66                 return -1;
  67
  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));
  75
  76         // Synthesize initial address events
  77         struct {
  78                 struct nlmsghdr nh;
  79                 struct ifaddrmsg ifa;
  80         } req2 = {
  81                 {sizeof(req2), RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP,
  82                                 ++rtnl_seqid, 0},
  83                 {.ifa_family = AF_INET6}
  84         };
  85         send(rtnl_event.uloop.fd, &req2, sizeof(req2), MSG_DONTWAIT);
  86         odhcpd_register(&rtnl_event);
  87
  88         // Open ICMPv6 socket
  89         ping_socket = socket(AF_INET6, SOCK_RAW | SOCK_CLOEXEC, IPPROTO_ICMPV6);
  90         if (ping_socket < 0) {
  91                 syslog(LOG_ERR, "Unable to open raw socket: %s", strerror(errno));
  92                         return -1;
  93         }
  94
  95         int val = 2;
  96         setsockopt(ping_socket, IPPROTO_RAW, IPV6_CHECKSUM, &val, sizeof(val));
  97
  98         // This is required by RFC 4861
  99         val = 255;
 100         setsockopt(ping_socket, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, &val, sizeof(val));
 101         setsockopt(ping_socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val));
 102
 103         // Filter all packages, we only want to send
 104         struct icmp6_filter filt;
 105         ICMP6_FILTER_SETBLOCKALL(&filt);
 106         setsockopt(ping_socket, IPPROTO_ICMPV6, ICMP6_FILTER, &filt, sizeof(filt));
 107
 108
 109         // Netlink socket, continued...
 110         group = RTNLGRP_NEIGH;
 111         setsockopt(rtnl_event.uloop.fd, SOL_NETLINK, NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
 112
 113         return 0;
 114 }
 115
 116
 117 static void dump_neigh_table(bool proxy)
 118 {
 119         struct {
 120                 struct nlmsghdr nh;
 121                 struct ndmsg ndm;
 122         } req = {
 123                 {sizeof(req), RTM_GETNEIGH, NLM_F_REQUEST | NLM_F_DUMP,
 124                                 ++rtnl_seqid, 0},
 125                 {.ndm_family = AF_INET6, .ndm_flags = (proxy) ? NTF_PROXY : 0}
 126         };
 127         send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 128 }
 129
 130
 131 int setup_ndp_interface(struct interface *iface, bool enable)
 132 {
 133         char procbuf[64];
 134         snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
 135         int procfd = open(procbuf, O_WRONLY);
 136         bool dump_neigh = false;
 137
 138         if (iface->ndp_event.uloop.fd >= 0) {
 139                 uloop_fd_delete(&iface->ndp_event.uloop);
 140                 close(iface->ndp_event.uloop.fd);
 141                 iface->ndp_event.uloop.fd = -1;
 142
 143                 write(procfd, "0\n", 2);
 144                 dump_neigh = true;
 145         }
 146
 147         if (enable && iface->ndp == RELAYD_RELAY) {
 148                 write(procfd, "1\n", 2);
 149
 150                 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
 151                 if (sock < 0) {
 152                         syslog(LOG_ERR, "Unable to open packet socket: %s",
 153                                         strerror(errno));
 154                         return -1;
 155                 }
 156
 157 #ifdef PACKET_RECV_TYPE
 158                 int pktt = 1 << PACKET_MULTICAST;
 159                 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
 160 #endif
 161
 162                 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
 163                                 &bpf_prog, sizeof(bpf_prog))) {
 164                         syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
 165                         return -1;
 166                 }
 167
 168                 struct sockaddr_ll ll = {
 169                         .sll_family = AF_PACKET,
 170                         .sll_ifindex = iface->ifindex,
 171                         .sll_protocol = htons(ETH_P_IPV6),
 172                 };
 173                 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
 174
 175                 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
 176                 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
 177
 178                 iface->ndp_event.uloop.fd = sock;
 179                 iface->ndp_event.handle_dgram = handle_solicit;
 180                 odhcpd_register(&iface->ndp_event);
 181
 182                 // If we already were enabled dump is unnecessary, if not do dump
 183                 if (!dump_neigh)
 184                         dump_neigh_table(false);
 185                 else
 186                         dump_neigh = false;
 187         }
 188         close(procfd);
 189
 190         if (dump_neigh)
 191                 dump_neigh_table(true);
 192
 193         return 0;
 194 }
 195
 196
 197 // Send an ICMP-ECHO. This is less for actually pinging but for the
 198 // neighbor cache to be kept up-to-date.
 199 static ssize_t ping6(struct in6_addr *addr,
 200                 const struct interface *iface)
 201 {
 202         struct sockaddr_in6 dest = {AF_INET6, 0, 0, *addr, 0};
 203         struct icmp6_hdr echo = {.icmp6_type = ICMP6_ECHO_REQUEST};
 204         struct iovec iov = {&echo, sizeof(echo)};
 205
 206         // Linux seems to not honor IPV6_PKTINFO on raw-sockets, so work around
 207         setsockopt(ping_socket, SOL_SOCKET, SO_BINDTODEVICE,
 208                         iface->ifname, sizeof(iface->ifname));
 209         return odhcpd_send(ping_socket, &dest, &iov, 1, iface);
 210 }
 211
 212
 213 // Handle solicitations
 214 static void handle_solicit(void *addr, void *data, size_t len,
 215                 struct interface *iface, _unused void *dest)
 216 {
 217         struct ip6_hdr *ip6 = data;
 218         struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
 219         struct sockaddr_ll *ll = addr;
 220
 221         // Solicitation is for duplicate address detection
 222         bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
 223
 224         // Don't forward any non-DAD solicitation for external ifaces
 225         // TODO: check if we should even forward DADs for them
 226         if (iface->external && !ns_is_dad)
 227                 return;
 228
 229         if (len < sizeof(*ip6) + sizeof(*req))
 230                 return; // Invalid reqicitation
 231
 232         if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
 233                         IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
 234                         IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
 235                 return; // Invalid target
 236
 237         char ipbuf[INET6_ADDRSTRLEN];
 238         inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
 239         syslog(LOG_DEBUG, "Got a NS for %s", ipbuf);
 240
 241         uint8_t mac[6];
 242         odhcpd_get_mac(iface, mac);
 243         if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
 244                 return; // Looped back
 245
 246         struct interface *c;
 247         list_for_each_entry(c, &interfaces, head)
 248                 if (iface->ndp == RELAYD_RELAY && iface != c &&
 249                                 (!ns_is_dad || !c->external == false))
 250                         ping6(&req->nd_ns_target, c);
 251 }
 252
 253
 254 void odhcpd_setup_route(const struct in6_addr *addr, int prefixlen,
 255                 const struct interface *iface, const struct in6_addr *gw, bool add)
 256 {
 257         struct req {
 258                 struct nlmsghdr nh;
 259                 struct rtmsg rtm;
 260                 struct rtattr rta_dst;
 261                 struct in6_addr dst_addr;
 262                 struct rtattr rta_oif;
 263                 uint32_t ifindex;
 264                 struct rtattr rta_table;
 265                 uint32_t table;
 266                 struct rtattr rta_gw;
 267                 struct in6_addr gw;
 268         } req = {
 269                 {sizeof(req), 0, NLM_F_REQUEST, ++rtnl_seqid, 0},
 270                 {AF_INET6, prefixlen, 0, 0, 0, 0, 0, 0, 0},
 271                 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_DST},
 272                 *addr,
 273                 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_OIF},
 274                 iface->ifindex,
 275                 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_TABLE},
 276                 RT_TABLE_MAIN,
 277                 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_GATEWAY},
 278                 IN6ADDR_ANY_INIT,
 279         };
 280
 281         if (gw)
 282                 req.gw = *gw;
 283
 284         if (add) {
 285                 req.nh.nlmsg_type = RTM_NEWROUTE;
 286                 req.nh.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
 287                 req.rtm.rtm_protocol = RTPROT_BOOT;
 288                 req.rtm.rtm_scope = (gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
 289                 req.rtm.rtm_type = RTN_UNICAST;
 290         } else {
 291                 req.nh.nlmsg_type = RTM_DELROUTE;
 292                 req.rtm.rtm_scope = RT_SCOPE_NOWHERE;
 293         }
 294
 295         size_t reqlen = (gw) ? sizeof(req) : offsetof(struct req, rta_gw);
 296         send(rtnl_event.uloop.fd, &req, reqlen, MSG_DONTWAIT);
 297 }
 298
 299 // Use rtnetlink to modify kernel routes
 300 static void setup_route(struct in6_addr *addr, struct interface *iface,
 301                 bool add)
 302 {
 303         char namebuf[INET6_ADDRSTRLEN];
 304         inet_ntop(AF_INET6, addr, namebuf, sizeof(namebuf));
 305         syslog(LOG_NOTICE, "%s about %s on %s", (add) ? "Learned" : "Forgot",
 306                         namebuf, (iface) ? iface->ifname : "<pending>");
 307
 308         if (!iface || !iface->learn_routes)
 309                 return;
 310
 311         odhcpd_setup_route(addr, 128, iface, NULL, add);
 312 }
 313
 314
 315 // Handler for neighbor cache entries from the kernel. This is our source
 316 // to learn and unlearn hosts on interfaces.
 317 static void handle_rtnetlink(_unused void *addr, void *data, size_t len,
 318                 _unused struct interface *iface, _unused void *dest)
 319 {
 320         bool dump_neigh = false;
 321         struct in6_addr last_solicited = IN6ADDR_ANY_INIT;
 322
 323         for (struct nlmsghdr *nh = data; NLMSG_OK(nh, len);
 324                         nh = NLMSG_NEXT(nh, len)) {
 325                 struct rtmsg *rtm = NLMSG_DATA(nh);
 326                 if ((nh->nlmsg_type == RTM_NEWROUTE ||
 327                                 nh->nlmsg_type == RTM_DELROUTE) &&
 328                                 rtm->rtm_dst_len == 0)
 329                         raise(SIGUSR1); // Inform about a change in default route
 330
 331                 struct ndmsg *ndm = NLMSG_DATA(nh);
 332                 struct ifaddrmsg *ifa = NLMSG_DATA(nh);
 333                 if (nh->nlmsg_type != RTM_NEWNEIGH
 334                                 && nh->nlmsg_type != RTM_DELNEIGH
 335                                 && nh->nlmsg_type != RTM_NEWADDR
 336                                 && nh->nlmsg_type != RTM_DELADDR)
 337                         continue; // Unrelated message type
 338                 bool is_addr = (nh->nlmsg_type == RTM_NEWADDR
 339                                 || nh->nlmsg_type == RTM_DELADDR);
 340
 341                 // Family and ifindex are on the same offset for NEIGH and ADDR
 342                 if (NLMSG_PAYLOAD(nh, 0) < sizeof(*ndm)
 343                                 || ndm->ndm_family != AF_INET6)
 344                         continue; //
 345
 346                 // Lookup interface
 347                 struct interface *iface;
 348                 if (!(iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex)))
 349                         continue;
 350
 351                 // Data to retrieve
 352                 size_t rta_offset = (is_addr) ? sizeof(*ifa) : sizeof(*ndm);
 353                 uint16_t atype = (is_addr) ? IFA_ADDRESS : NDA_DST;
 354                 ssize_t alen = NLMSG_PAYLOAD(nh, rta_offset);
 355                 struct in6_addr *addr = NULL;
 356
 357                 for (struct rtattr *rta = (void*)(((uint8_t*)ndm) + rta_offset);
 358                                 RTA_OK(rta, alen); rta = RTA_NEXT(rta, alen))
 359                         if (rta->rta_type == atype &&
 360                                         RTA_PAYLOAD(rta) >= sizeof(*addr))
 361                                 addr = RTA_DATA(rta);
 362
 363                 // Address not specified or unrelated
 364                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 365                                 IN6_IS_ADDR_MULTICAST(addr))
 366                         continue;
 367
 368                 // Check for states
 369                 bool add;
 370                 if (is_addr)
 371                         add = (nh->nlmsg_type == RTM_NEWADDR);
 372                 else
 373                         add = (nh->nlmsg_type == RTM_NEWNEIGH && (ndm->ndm_state &
 374                                 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE
 375                                                 | NUD_PERMANENT | NUD_NOARP)));
 376
 377                 if (iface->ndp == RELAYD_RELAY) {
 378                         // Replay change to all neighbor cache
 379                         struct {
 380                                 struct nlmsghdr nh;
 381                                 struct ndmsg ndm;
 382                                 struct nlattr nla_dst;
 383                                 struct in6_addr dst;
 384                         } req = {
 385                                 {sizeof(req), RTM_DELNEIGH, NLM_F_REQUEST,
 386                                                 ++rtnl_seqid, 0},
 387                                 {.ndm_family = AF_INET6, .ndm_flags = NTF_PROXY},
 388                                 {sizeof(struct nlattr) + sizeof(struct in6_addr), NDA_DST},
 389                                 *addr
 390                         };
 391
 392                         if (ndm->ndm_flags & NTF_PROXY) {
 393                                 // Dump & flush proxy entries
 394                                 if (nh->nlmsg_type == RTM_NEWNEIGH) {
 395                                         req.ndm.ndm_ifindex = iface->ifindex;
 396                                         send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 397                                         dump_neigh = true;
 398                                 }
 399                         } else if (add) {
 400                                 struct interface *c;
 401                                 list_for_each_entry(c, &interfaces, head) {
 402                                         if (iface == c)
 403                                                 continue;
 404
 405                                         if (c->ndp == RELAYD_RELAY) {
 406                                                 req.nh.nlmsg_type = RTM_NEWNEIGH;
 407                                                 req.nh.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
 408
 409                                                 req.ndm.ndm_ifindex = c->ifindex;
 410                                                 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 411                                         } else { // Delete NDP cache from interfaces without relay
 412                                                 req.nh.nlmsg_type = RTM_DELNEIGH;
 413                                                 req.nh.nlmsg_flags &= ~(NLM_F_CREATE | NLM_F_REPLACE);
 414
 415                                                 req.ndm.ndm_ifindex = c->ifindex;
 416                                                 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 417                                         }
 418                                 }
 419
 420                                 setup_route(addr, iface, true);
 421                         } else {
 422                                 if (nh->nlmsg_type == RTM_NEWNEIGH) {
 423                                         // might be locally originating
 424                                         if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
 425                                                 last_solicited = *addr;
 426
 427                                                 struct interface *c;
 428                                                 list_for_each_entry(c, &interfaces, head)
 429                                                         if (iface->ndp == RELAYD_RELAY && iface != c &&
 430                                                                         !c->external == false)
 431                                                                 ping6(addr, c);
 432                                         }
 433                                 } else {
 434                                         struct interface *c;
 435                                         list_for_each_entry(c, &interfaces, head) {
 436                                                 if (c->ndp == RELAYD_RELAY && iface != c) {
 437                                                         req.ndm.ndm_ifindex = c->ifindex;
 438                                                         send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 439                                                 }
 440                                         }
 441                                         setup_route(addr, iface, false);
 442
 443                                         // also: dump to add proxies back in case it moved elsewhere
 444                                         dump_neigh = true;
 445                                 }
 446                         }
 447                 }
 448
 449                 if (is_addr) {
 450                         if (iface->ra == RELAYD_SERVER)
 451                                 raise(SIGUSR1); // Inform about a change in addresses
 452
 453                         if (iface->dhcpv6 == RELAYD_SERVER)
 454                                 iface->ia_reconf = true;
 455
 456                         if (iface->ndp == RELAYD_RELAY && iface->master) {
 457                                 // Replay address changes on all slave interfaces
 458                                 nh->nlmsg_flags = NLM_F_REQUEST;
 459
 460                                 if (nh->nlmsg_type == RTM_NEWADDR)
 461                                         nh->nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
 462
 463                                 struct interface *c;
 464                                 list_for_each_entry(c, &interfaces, head) {
 465                                         if (c->ndp == RELAYD_RELAY && !c->master) {
 466                                                 ifa->ifa_index = c->ifindex;
 467                                                 send(rtnl_event.uloop.fd, nh, nh->nlmsg_len, MSG_DONTWAIT);
 468                                         }
 469                                 }
 470                         }
 471                 }
 472         }
 473
 474         if (dump_neigh)
 475                 dump_neigh_table(false);
 476 }