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         odhcpd_register(&rtnl_event);
  77
  78         // Open ICMPv6 socket
  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));
  82                         return -1;
  83         }
  84
  85         int val = 2;
  86         setsockopt(ping_socket, IPPROTO_RAW, IPV6_CHECKSUM, &val, sizeof(val));
  87
  88         // This is required by RFC 4861
  89         val = 255;
  90         setsockopt(ping_socket, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, &val, sizeof(val));
  91         setsockopt(ping_socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val));
  92
  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));
  97
  98
  99         // Netlink socket, continued...
 100         group = RTNLGRP_NEIGH;
 101         setsockopt(rtnl_event.uloop.fd, SOL_NETLINK, NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
 102
 103         return 0;
 104 }
 105
 106
 107 static void dump_neigh_table(bool proxy)
 108 {
 109         struct {
 110                 struct nlmsghdr nh;
 111                 struct ndmsg ndm;
 112         } req = {
 113                 {sizeof(req), RTM_GETNEIGH, NLM_F_REQUEST | NLM_F_DUMP,
 114                                 ++rtnl_seqid, 0},
 115                 {.ndm_family = AF_INET6, .ndm_flags = (proxy) ? NTF_PROXY : 0}
 116         };
 117         send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 118 }
 119
 120
 121 int setup_ndp_interface(struct interface *iface, bool enable)
 122 {
 123         char procbuf[64];
 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;
 127
 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;
 132
 133                 write(procfd, "0\n", 2);
 134                 dump_neigh = true;
 135         }
 136
 137         if (enable && (iface->ra == RELAYD_SERVER || iface->dhcpv6 == RELAYD_SERVER)) {
 138                 // Synthesize initial address events
 139                 struct {
 140                         struct nlmsghdr nh;
 141                         struct ifaddrmsg ifa;
 142                 } req2 = {
 143                         {sizeof(req2), RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP,
 144                                         ++rtnl_seqid, 0},
 145                         {.ifa_family = AF_INET6, .ifa_index = iface->ifindex}
 146                 };
 147                 send(rtnl_event.uloop.fd, &req2, sizeof(req2), MSG_DONTWAIT);
 148         }
 149
 150         if (enable && iface->ndp == RELAYD_RELAY) {
 151                 write(procfd, "1\n", 2);
 152
 153                 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
 154                 if (sock < 0) {
 155                         syslog(LOG_ERR, "Unable to open packet socket: %s",
 156                                         strerror(errno));
 157                         return -1;
 158                 }
 159
 160 #ifdef PACKET_RECV_TYPE
 161                 int pktt = 1 << PACKET_MULTICAST;
 162                 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
 163 #endif
 164
 165                 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
 166                                 &bpf_prog, sizeof(bpf_prog))) {
 167                         syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
 168                         return -1;
 169                 }
 170
 171                 struct sockaddr_ll ll = {
 172                         .sll_family = AF_PACKET,
 173                         .sll_ifindex = iface->ifindex,
 174                         .sll_protocol = htons(ETH_P_IPV6),
 175                         .sll_hatype = 0,
 176                         .sll_pkttype = 0,
 177                         .sll_halen = 0,
 178                         .sll_addr = {0},
 179                 };
 180                 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
 181
 182                 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
 183                 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
 184
 185                 iface->ndp_event.uloop.fd = sock;
 186                 iface->ndp_event.handle_dgram = handle_solicit;
 187                 odhcpd_register(&iface->ndp_event);
 188
 189                 // If we already were enabled dump is unnecessary, if not do dump
 190                 if (!dump_neigh)
 191                         dump_neigh_table(false);
 192                 else
 193                         dump_neigh = false;
 194         }
 195         close(procfd);
 196
 197         if (dump_neigh)
 198                 dump_neigh_table(true);
 199
 200         return 0;
 201 }
 202
 203
 204 // Send an ICMP-ECHO. This is less for actually pinging but for the
 205 // neighbor cache to be kept up-to-date.
 206 static ssize_t ping6(struct in6_addr *addr,
 207                 const struct interface *iface)
 208 {
 209         struct sockaddr_in6 dest = {AF_INET6, 0, 0, *addr, 0};
 210         struct icmp6_hdr echo = {.icmp6_type = ICMP6_ECHO_REQUEST};
 211         struct iovec iov = {&echo, sizeof(echo)};
 212
 213         // Linux seems to not honor IPV6_PKTINFO on raw-sockets, so work around
 214         setsockopt(ping_socket, SOL_SOCKET, SO_BINDTODEVICE,
 215                         iface->ifname, sizeof(iface->ifname));
 216         return odhcpd_send(ping_socket, &dest, &iov, 1, iface);
 217 }
 218
 219
 220 // Handle solicitations
 221 static void handle_solicit(void *addr, void *data, size_t len,
 222                 struct interface *iface, _unused void *dest)
 223 {
 224         struct ip6_hdr *ip6 = data;
 225         struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
 226         struct sockaddr_ll *ll = addr;
 227
 228         // Solicitation is for duplicate address detection
 229         bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
 230
 231         // Don't forward any non-DAD solicitation for external ifaces
 232         // TODO: check if we should even forward DADs for them
 233         if (iface->external && !ns_is_dad)
 234                 return;
 235
 236         if (len < sizeof(*ip6) + sizeof(*req))
 237                 return; // Invalid reqicitation
 238
 239         if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
 240                         IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
 241                         IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
 242                 return; // Invalid target
 243
 244         char ipbuf[INET6_ADDRSTRLEN];
 245         inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
 246         syslog(LOG_DEBUG, "Got a NS for %s", ipbuf);
 247
 248         uint8_t mac[6];
 249         odhcpd_get_mac(iface, mac);
 250         if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
 251                 return; // Looped back
 252
 253         struct interface *c;
 254         list_for_each_entry(c, &interfaces, head)
 255                 if (iface->ndp == RELAYD_RELAY && iface != c &&
 256                                 (!ns_is_dad || !c->external == false))
 257                         ping6(&req->nd_ns_target, c);
 258 }
 259
 260
 261 void odhcpd_setup_route(const struct in6_addr *addr, int prefixlen,
 262                 const struct interface *iface, const struct in6_addr *gw, bool add)
 263 {
 264         struct req {
 265                 struct nlmsghdr nh;
 266                 struct rtmsg rtm;
 267                 struct rtattr rta_dst;
 268                 struct in6_addr dst_addr;
 269                 struct rtattr rta_oif;
 270                 uint32_t ifindex;
 271                 struct rtattr rta_table;
 272                 uint32_t table;
 273                 struct rtattr rta_gw;
 274                 struct in6_addr gw;
 275         } req = {
 276                 {sizeof(req), 0, NLM_F_REQUEST, ++rtnl_seqid, 0},
 277                 {AF_INET6, prefixlen, 0, 0, 0, 0, 0, 0, 0},
 278                 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_DST},
 279                 *addr,
 280                 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_OIF},
 281                 iface->ifindex,
 282                 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_TABLE},
 283                 RT_TABLE_MAIN,
 284                 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_GATEWAY},
 285                 IN6ADDR_ANY_INIT,
 286         };
 287
 288         if (gw)
 289                 req.gw = *gw;
 290
 291         if (add) {
 292                 req.nh.nlmsg_type = RTM_NEWROUTE;
 293                 req.nh.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
 294                 req.rtm.rtm_protocol = RTPROT_BOOT;
 295                 req.rtm.rtm_scope = (gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
 296                 req.rtm.rtm_type = RTN_UNICAST;
 297         } else {
 298                 req.nh.nlmsg_type = RTM_DELROUTE;
 299                 req.rtm.rtm_scope = RT_SCOPE_NOWHERE;
 300         }
 301
 302         size_t reqlen = (gw) ? sizeof(req) : offsetof(struct req, rta_gw);
 303         send(rtnl_event.uloop.fd, &req, reqlen, MSG_DONTWAIT);
 304 }
 305
 306 // Use rtnetlink to modify kernel routes
 307 static void setup_route(struct in6_addr *addr, struct interface *iface,
 308                 bool add)
 309 {
 310         char namebuf[INET6_ADDRSTRLEN];
 311         inet_ntop(AF_INET6, addr, namebuf, sizeof(namebuf));
 312         syslog(LOG_NOTICE, "%s about %s on %s", (add) ? "Learned" : "Forgot",
 313                         namebuf, (iface) ? iface->ifname : "<pending>");
 314
 315         if (!iface || !iface->learn_routes)
 316                 return;
 317
 318         odhcpd_setup_route(addr, 128, iface, NULL, add);
 319 }
 320
 321
 322 // Handler for neighbor cache entries from the kernel. This is our source
 323 // to learn and unlearn hosts on interfaces.
 324 static void handle_rtnetlink(_unused void *addr, void *data, size_t len,
 325                 _unused struct interface *iface, _unused void *dest)
 326 {
 327         bool dump_neigh = false;
 328         struct in6_addr last_solicited = IN6ADDR_ANY_INIT;
 329
 330         for (struct nlmsghdr *nh = data; NLMSG_OK(nh, len);
 331                         nh = NLMSG_NEXT(nh, len)) {
 332                 struct rtmsg *rtm = NLMSG_DATA(nh);
 333                 if ((nh->nlmsg_type == RTM_NEWROUTE ||
 334                                 nh->nlmsg_type == RTM_DELROUTE) &&
 335                                 rtm->rtm_dst_len == 0)
 336                         raise(SIGUSR1); // Inform about a change in default route
 337
 338                 struct ndmsg *ndm = NLMSG_DATA(nh);
 339                 struct ifaddrmsg *ifa = NLMSG_DATA(nh);
 340                 if (nh->nlmsg_type != RTM_NEWNEIGH
 341                                 && nh->nlmsg_type != RTM_DELNEIGH
 342                                 && nh->nlmsg_type != RTM_NEWADDR
 343                                 && nh->nlmsg_type != RTM_DELADDR)
 344                         continue; // Unrelated message type
 345                 bool is_addr = (nh->nlmsg_type == RTM_NEWADDR
 346                                 || nh->nlmsg_type == RTM_DELADDR);
 347
 348                 // Family and ifindex are on the same offset for NEIGH and ADDR
 349                 if (NLMSG_PAYLOAD(nh, 0) < sizeof(*ndm)
 350                                 || ndm->ndm_family != AF_INET6)
 351                         continue; //
 352
 353                 // Lookup interface
 354                 struct interface *iface;
 355                 if (!(iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex)))
 356                         continue;
 357
 358                 // Data to retrieve
 359                 size_t rta_offset = (is_addr) ? sizeof(*ifa) : sizeof(*ndm);
 360                 uint16_t atype = (is_addr) ? IFA_ADDRESS : NDA_DST;
 361                 ssize_t alen = NLMSG_PAYLOAD(nh, rta_offset);
 362                 struct in6_addr *addr = NULL;
 363
 364                 for (struct rtattr *rta = (void*)(((uint8_t*)ndm) + rta_offset);
 365                                 RTA_OK(rta, alen); rta = RTA_NEXT(rta, alen))
 366                         if (rta->rta_type == atype &&
 367                                         RTA_PAYLOAD(rta) >= sizeof(*addr))
 368                                 addr = RTA_DATA(rta);
 369
 370                 // Address not specified or unrelated
 371                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 372                                 IN6_IS_ADDR_MULTICAST(addr))
 373                         continue;
 374
 375                 // Check for states
 376                 bool add;
 377                 if (is_addr)
 378                         add = (nh->nlmsg_type == RTM_NEWADDR);
 379                 else
 380                         add = (nh->nlmsg_type == RTM_NEWNEIGH && (ndm->ndm_state &
 381                                 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE
 382                                                 | NUD_PERMANENT | NUD_NOARP)));
 383
 384                 if (iface->ndp == RELAYD_RELAY) {
 385                         // Replay change to all neighbor cache
 386                         struct {
 387                                 struct nlmsghdr nh;
 388                                 struct ndmsg ndm;
 389                                 struct nlattr nla_dst;
 390                                 struct in6_addr dst;
 391                         } req = {
 392                                 {sizeof(req), RTM_DELNEIGH, NLM_F_REQUEST,
 393                                                 ++rtnl_seqid, 0},
 394                                 {.ndm_family = AF_INET6, .ndm_flags = NTF_PROXY},
 395                                 {sizeof(struct nlattr) + sizeof(struct in6_addr), NDA_DST},
 396                                 *addr
 397                         };
 398
 399                         if (ndm->ndm_flags & NTF_PROXY) {
 400                                 // Dump & flush proxy entries
 401                                 if (nh->nlmsg_type == RTM_NEWNEIGH) {
 402                                         req.ndm.ndm_ifindex = iface->ifindex;
 403                                         send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 404                                         dump_neigh = true;
 405                                 }
 406                         } else if (add) {
 407                                 struct interface *c;
 408                                 list_for_each_entry(c, &interfaces, head) {
 409                                         if (iface == c)
 410                                                 continue;
 411
 412                                         if (c->ndp == RELAYD_RELAY) {
 413                                                 req.nh.nlmsg_type = RTM_NEWNEIGH;
 414                                                 req.nh.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
 415
 416                                                 req.ndm.ndm_ifindex = c->ifindex;
 417                                                 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 418                                         } else { // Delete NDP cache from interfaces without relay
 419                                                 req.nh.nlmsg_type = RTM_DELNEIGH;
 420                                                 req.nh.nlmsg_flags &= ~(NLM_F_CREATE | NLM_F_REPLACE);
 421
 422                                                 req.ndm.ndm_ifindex = c->ifindex;
 423                                                 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 424                                         }
 425                                 }
 426
 427                                 setup_route(addr, iface, true);
 428                         } else {
 429                                 if (nh->nlmsg_type == RTM_NEWNEIGH) {
 430                                         // might be locally originating
 431                                         if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
 432                                                 last_solicited = *addr;
 433
 434                                                 struct interface *c;
 435                                                 list_for_each_entry(c, &interfaces, head)
 436                                                         if (iface->ndp == RELAYD_RELAY && iface != c &&
 437                                                                         !c->external == false)
 438                                                                 ping6(addr, c);
 439                                         }
 440                                 } else {
 441                                         struct interface *c;
 442                                         list_for_each_entry(c, &interfaces, head) {
 443                                                 if (c->ndp == RELAYD_RELAY && iface != c) {
 444                                                         req.ndm.ndm_ifindex = c->ifindex;
 445                                                         send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 446                                                 }
 447                                         }
 448                                         setup_route(addr, iface, false);
 449
 450                                         // also: dump to add proxies back in case it moved elsewhere
 451                                         dump_neigh = true;
 452                                 }
 453                         }
 454                 }
 455
 456                 if (is_addr) {
 457                         if (iface->ra == RELAYD_SERVER)
 458                                 raise(SIGUSR1); // Inform about a change in addresses
 459
 460                         if (iface->dhcpv6 == RELAYD_SERVER)
 461                                 iface->ia_reconf = true;
 462
 463                         if (iface->ndp == RELAYD_RELAY && iface->master) {
 464                                 // Replay address changes on all slave interfaces
 465                                 nh->nlmsg_flags = NLM_F_REQUEST;
 466
 467                                 if (nh->nlmsg_type == RTM_NEWADDR)
 468                                         nh->nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
 469
 470                                 struct interface *c;
 471                                 list_for_each_entry(c, &interfaces, head) {
 472                                         if (c->ndp == RELAYD_RELAY && !c->master) {
 473                                                 ifa->ifa_index = c->ifindex;
 474                                                 send(rtnl_event.uloop.fd, nh, nh->nlmsg_len, MSG_DONTWAIT);
 475                                         }
 476                                 }
 477                         }
 478                 }
 479         }
 480
 481         if (dump_neigh)
 482                 dump_neigh_table(false);
 483 }