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
  32 #include <netlink/msg.h>
  33 #include <netlink/socket.h>
  34 #include <netlink/attr.h>
  35
  36 #include "router.h"
  37 #include "dhcpv6.h"
  38 #include "ndp.h"
  39
  40 struct event_socket {
  41         struct odhcpd_event ev;
  42         struct nl_sock *sock;
  43 };
  44
  45 static void handle_solicit(void *addr, void *data, size_t len,
  46                 struct interface *iface, void *dest);
  47 static void handle_rtnl_event(struct odhcpd_event *ev);
  48 static int cb_rtnl_valid(struct nl_msg *msg, void *arg);
  49 static void catch_rtnetlink(int error);
  50
  51 static int ping_socket = -1;
  52 static struct event_socket rtnl_event = {
  53         .ev = {
  54                 .uloop = {.fd = - 1, },
  55                 .handle_dgram = NULL,
  56                 .handle_error = catch_rtnetlink,
  57                 .recv_msgs = handle_rtnl_event,
  58         },
  59         .sock = NULL,
  60 };
  61
  62 // Filter ICMPv6 messages of type neighbor soliciation
  63 static struct sock_filter bpf[] = {
  64         BPF_STMT(BPF_LD | BPF_B | BPF_ABS, offsetof(struct ip6_hdr, ip6_nxt)),
  65         BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, IPPROTO_ICMPV6, 0, 3),
  66         BPF_STMT(BPF_LD | BPF_B | BPF_ABS, sizeof(struct ip6_hdr) +
  67                         offsetof(struct icmp6_hdr, icmp6_type)),
  68         BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, ND_NEIGHBOR_SOLICIT, 0, 1),
  69         BPF_STMT(BPF_RET | BPF_K, 0xffffffff),
  70         BPF_STMT(BPF_RET | BPF_K, 0),
  71 };
  72 static const struct sock_fprog bpf_prog = {sizeof(bpf) / sizeof(*bpf), bpf};
  73
  74
  75 // Initialize NDP-proxy
  76 int init_ndp(void)
  77 {
  78         int val = 256 * 1024;
  79
  80         rtnl_event.sock = odhcpd_create_nl_socket(NETLINK_ROUTE);
  81         if (!rtnl_event.sock)
  82                 goto err;
  83
  84         rtnl_event.ev.uloop.fd = nl_socket_get_fd(rtnl_event.sock);
  85
  86         if (nl_socket_set_buffer_size(rtnl_event.sock, val, 0))
  87                 goto err;
  88
  89         nl_socket_disable_seq_check(rtnl_event.sock);
  90
  91         nl_socket_modify_cb(rtnl_event.sock, NL_CB_VALID, NL_CB_CUSTOM,
  92                         cb_rtnl_valid, NULL);
  93
  94         // Receive IPv6 address, IPv6 routes and neighbor events
  95         if (nl_socket_add_memberships(rtnl_event.sock, RTNLGRP_IPV6_IFADDR,
  96                                 RTNLGRP_IPV6_ROUTE, RTNLGRP_NEIGH, 0))
  97                 goto err;
  98
  99         odhcpd_register(&rtnl_event.ev);
 100
 101         // Open ICMPv6 socket
 102         ping_socket = socket(AF_INET6, SOCK_RAW | SOCK_CLOEXEC, IPPROTO_ICMPV6);
 103         if (ping_socket < 0) {
 104                 syslog(LOG_ERR, "Unable to open raw socket: %s", strerror(errno));
 105                         return -1;
 106         }
 107
 108         val = 2;
 109         setsockopt(ping_socket, IPPROTO_RAW, IPV6_CHECKSUM, &val, sizeof(val));
 110
 111         // This is required by RFC 4861
 112         val = 255;
 113         setsockopt(ping_socket, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, &val, sizeof(val));
 114         setsockopt(ping_socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val));
 115
 116         // Filter all packages, we only want to send
 117         struct icmp6_filter filt;
 118         ICMP6_FILTER_SETBLOCKALL(&filt);
 119         setsockopt(ping_socket, IPPROTO_ICMPV6, ICMP6_FILTER, &filt, sizeof(filt));
 120
 121         return 0;
 122
 123 err:
 124         if (rtnl_event.sock) {
 125                 nl_socket_free(rtnl_event.sock);
 126                 rtnl_event.sock = NULL;
 127                 rtnl_event.ev.uloop.fd = -1;
 128         }
 129
 130         return -1;
 131 }
 132
 133 static void dump_neigh_table(const bool proxy)
 134 {
 135         struct nl_msg *msg;
 136         struct ndmsg ndm = {
 137                 .ndm_family = AF_INET6,
 138                 .ndm_flags = proxy ? NTF_PROXY : 0,
 139         };
 140
 141         msg = nlmsg_alloc_simple(RTM_GETNEIGH, NLM_F_REQUEST | NLM_F_DUMP);
 142         if (!msg)
 143                 return;
 144
 145         nlmsg_append(msg, &ndm, sizeof(ndm), 0);
 146
 147         nl_send_auto_complete(rtnl_event.sock, msg);
 148
 149         nlmsg_free(msg);
 150 }
 151
 152 static void dump_addr_table(void)
 153 {
 154         struct nl_msg *msg;
 155         struct ifaddrmsg ifa = {
 156                 .ifa_family = AF_INET6,
 157         };
 158
 159         msg = nlmsg_alloc_simple(RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP);
 160         if (!msg)
 161                 return;
 162
 163         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
 164
 165         nl_send_auto_complete(rtnl_event.sock, msg);
 166
 167         nlmsg_free(msg);
 168 }
 169
 170 int setup_ndp_interface(struct interface *iface, bool enable)
 171 {
 172         char procbuf[64];
 173         snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
 174         int procfd = open(procbuf, O_WRONLY);
 175         bool dump_neigh = false;
 176
 177         if (iface->ndp_event.uloop.fd > 0) {
 178                 uloop_fd_delete(&iface->ndp_event.uloop);
 179                 close(iface->ndp_event.uloop.fd);
 180                 iface->ndp_event.uloop.fd = -1;
 181
 182                 if (!enable || iface->ndp != RELAYD_RELAY)
 183                         if (write(procfd, "0\n", 2) < 0) {}
 184
 185                 dump_neigh = true;
 186         }
 187
 188         if (enable && (iface->ra == RELAYD_SERVER ||
 189                         iface->dhcpv6 == RELAYD_SERVER || iface->ndp == RELAYD_RELAY))
 190                 dump_addr_table();
 191
 192         if (enable && iface->ndp == RELAYD_RELAY) {
 193                 if (write(procfd, "1\n", 2) < 0) {}
 194                 close(procfd);
 195
 196                 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
 197                 if (sock < 0) {
 198                         syslog(LOG_ERR, "Unable to open packet socket: %s",
 199                                         strerror(errno));
 200                         return -1;
 201                 }
 202
 203 #ifdef PACKET_RECV_TYPE
 204                 int pktt = 1 << PACKET_MULTICAST;
 205                 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
 206 #endif
 207
 208                 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
 209                                 &bpf_prog, sizeof(bpf_prog))) {
 210                         syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
 211                         return -1;
 212                 }
 213
 214                 struct sockaddr_ll ll = {
 215                         .sll_family = AF_PACKET,
 216                         .sll_ifindex = iface->ifindex,
 217                         .sll_protocol = htons(ETH_P_IPV6),
 218                         .sll_hatype = 0,
 219                         .sll_pkttype = 0,
 220                         .sll_halen = 0,
 221                         .sll_addr = {0},
 222                 };
 223                 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
 224
 225                 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
 226                 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
 227
 228                 iface->ndp_event.uloop.fd = sock;
 229                 iface->ndp_event.handle_dgram = handle_solicit;
 230                 odhcpd_register(&iface->ndp_event);
 231
 232                 // If we already were enabled dump is unnecessary, if not do dump
 233                 if (!dump_neigh)
 234                         dump_neigh_table(false);
 235                 else
 236                         dump_neigh = false;
 237         } else
 238                 close(procfd);
 239
 240         if (dump_neigh)
 241                 dump_neigh_table(true);
 242
 243         return 0;
 244 }
 245
 246
 247 // Send an ICMP-ECHO. This is less for actually pinging but for the
 248 // neighbor cache to be kept up-to-date.
 249 static void ping6(struct in6_addr *addr,
 250                 const struct interface *iface)
 251 {
 252         struct sockaddr_in6 dest = { .sin6_family = AF_INET6, .sin6_addr = *addr, .sin6_scope_id = iface->ifindex, };
 253         struct icmp6_hdr echo = { .icmp6_type = ICMP6_ECHO_REQUEST };
 254         struct iovec iov = { .iov_base = &echo, .iov_len = sizeof(echo) };
 255
 256         odhcpd_setup_route(addr, 128, iface, NULL, 128, true);
 257         odhcpd_send(ping_socket, &dest, &iov, 1, iface);
 258         odhcpd_setup_route(addr, 128, iface, NULL, 128, false);
 259 }
 260
 261
 262 // Handle solicitations
 263 static void handle_solicit(void *addr, void *data, size_t len,
 264                 struct interface *iface, _unused void *dest)
 265 {
 266         struct ip6_hdr *ip6 = data;
 267         struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
 268         struct sockaddr_ll *ll = addr;
 269
 270         // Solicitation is for duplicate address detection
 271         bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
 272
 273         // Don't process solicit messages on non relay interfaces
 274         // Don't forward any non-DAD solicitation for external ifaces
 275         // TODO: check if we should even forward DADs for them
 276         if (iface->ndp != RELAYD_RELAY || (iface->external && !ns_is_dad))
 277                 return;
 278
 279         if (len < sizeof(*ip6) + sizeof(*req))
 280                 return; // Invalid reqicitation
 281
 282         if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
 283                         IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
 284                         IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
 285                 return; // Invalid target
 286
 287         char ipbuf[INET6_ADDRSTRLEN];
 288         inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
 289         syslog(LOG_DEBUG, "Got a NS for %s", ipbuf);
 290
 291         uint8_t mac[6];
 292         odhcpd_get_mac(iface, mac);
 293         if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
 294                 return; // Looped back
 295
 296         struct interface *c;
 297         list_for_each_entry(c, &interfaces, head)
 298                 if (iface != c && c->ndp == RELAYD_RELAY &&
 299                                 (ns_is_dad || !c->external))
 300                         ping6(&req->nd_ns_target, c);
 301 }
 302
 303 // Use rtnetlink to modify kernel routes
 304 static void setup_route(struct in6_addr *addr, struct interface *iface, bool add)
 305 {
 306         char namebuf[INET6_ADDRSTRLEN];
 307         inet_ntop(AF_INET6, addr, namebuf, sizeof(namebuf));
 308         syslog(LOG_NOTICE, "%s about %s on %s",
 309                         (add) ? "Learned" : "Forgot", namebuf, iface->ifname);
 310
 311         if (iface->learn_routes)
 312                 odhcpd_setup_route(addr, 128, iface, NULL, 1024, add);
 313 }
 314
 315 // compare prefixes
 316 static int prefixcmp(const void *va, const void *vb)
 317 {
 318         const struct odhcpd_ipaddr *a = va, *b = vb;
 319         uint32_t a_pref = ((a->addr.s6_addr[0] & 0xfe) != 0xfc) ? a->preferred : 1;
 320         uint32_t b_pref = ((b->addr.s6_addr[0] & 0xfe) != 0xfc) ? b->preferred : 1;
 321         return (a_pref < b_pref) ? 1 : (a_pref > b_pref) ? -1 : 0;
 322 }
 323
 324 // Check address update
 325 static void check_addr_updates(struct interface *iface)
 326 {
 327         struct odhcpd_ipaddr addr[RELAYD_MAX_ADDRS] = {{IN6ADDR_ANY_INIT, 0, 0, 0, 0}};
 328         time_t now = odhcpd_time();
 329         ssize_t len = odhcpd_get_interface_addresses(iface->ifindex, addr, ARRAY_SIZE(addr));
 330
 331         if (len < 0)
 332                 return;
 333
 334         qsort(addr, len, sizeof(*addr), prefixcmp);
 335
 336         for (int i = 0; i < len; ++i) {
 337                 addr[i].addr.s6_addr32[3] = 0;
 338
 339                 if (addr[i].preferred < UINT32_MAX - now)
 340                         addr[i].preferred += now;
 341
 342                 if (addr[i].valid < UINT32_MAX - now)
 343                         addr[i].valid += now;
 344         }
 345
 346         bool change = len != (ssize_t)iface->ia_addr_len;
 347         for (ssize_t i = 0; !change && i < len; ++i)
 348                 if (!IN6_ARE_ADDR_EQUAL(&addr[i].addr, &iface->ia_addr[i].addr) ||
 349                                 (addr[i].preferred > 0) != (iface->ia_addr[i].preferred > 0) ||
 350                                 addr[i].valid < iface->ia_addr[i].valid ||
 351                                 addr[i].preferred < iface->ia_addr[i].preferred)
 352                         change = true;
 353
 354         if (change)
 355                 dhcpv6_ia_preupdate(iface);
 356
 357         memcpy(iface->ia_addr, addr, len * sizeof(*addr));
 358         iface->ia_addr_len = len;
 359
 360         if (change)
 361                 dhcpv6_ia_postupdate(iface, now);
 362
 363         if (change) {
 364                 syslog(LOG_INFO, "Raising SIGUSR1 due to address change on %s", iface->ifname);
 365                 raise(SIGUSR1);
 366         }
 367 }
 368
 369 void setup_addr_for_relaying(struct in6_addr *addr, struct interface *iface, bool add)
 370 {
 371         struct interface *c;
 372
 373         list_for_each_entry(c, &interfaces, head) {
 374                 if (iface == c || (c->ndp != RELAYD_RELAY && !add))
 375                         continue;
 376
 377                 odhcpd_setup_proxy_neigh(addr, c, c->ndp == RELAYD_RELAY ? add : false);
 378         }
 379 }
 380
 381 void setup_ping6(struct in6_addr *addr, struct interface *iface)
 382 {
 383         struct interface *c;
 384
 385         list_for_each_entry(c, &interfaces, head) {
 386                 if (iface == c || c->ndp != RELAYD_RELAY ||
 387                                 c->external == true)
 388                         continue;
 389
 390                 ping6(addr, c);
 391         }
 392 }
 393
 394 static struct in6_addr last_solicited;
 395
 396 static void handle_rtnl_event(struct odhcpd_event *e)
 397 {
 398         struct event_socket *ev_sock = container_of(e, struct event_socket, ev);
 399
 400         nl_recvmsgs_default(ev_sock->sock);
 401 }
 402
 403
 404 // Handler for neighbor cache entries from the kernel. This is our source
 405 // to learn and unlearn hosts on interfaces.
 406 static int cb_rtnl_valid(struct nl_msg *msg, _unused void *arg)
 407 {
 408         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 409         struct in6_addr *addr = NULL;
 410         struct interface *iface = NULL;
 411         bool add = false;
 412
 413         switch (hdr->nlmsg_type) {
 414         case RTM_NEWROUTE:
 415         case RTM_DELROUTE: {
 416                 struct rtmsg *rtm = nlmsg_data(hdr);
 417
 418                 if (!nlmsg_valid_hdr(hdr, sizeof(*rtm)) ||
 419                                 rtm->rtm_family != AF_INET6)
 420                         return NL_SKIP;
 421
 422                 if (rtm->rtm_dst_len == 0) {
 423                         syslog(LOG_INFO, "Raising SIGUSR1 due to default route change");
 424                         raise(SIGUSR1);
 425                 }
 426                 return NL_OK;
 427         }
 428
 429         case RTM_NEWADDR:
 430                 add = true;
 431         case RTM_DELADDR: {
 432                 struct ifaddrmsg *ifa = nlmsg_data(hdr);
 433                 struct nlattr *nla[__IFA_MAX];
 434
 435                 if (!nlmsg_valid_hdr(hdr, sizeof(*ifa)) ||
 436                                 ifa->ifa_family != AF_INET6)
 437                         return NL_SKIP;
 438
 439                 iface = odhcpd_get_interface_by_index(ifa->ifa_index);
 440                 if (!iface)
 441                         return NL_SKIP;
 442
 443                 nlmsg_parse(hdr, sizeof(*ifa), nla, __IFA_MAX - 1, NULL);
 444                 if (!nla[IFA_ADDRESS])
 445                         return NL_SKIP;
 446
 447                 addr = nla_data(nla[IFA_ADDRESS]);
 448                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 449                                 IN6_IS_ADDR_MULTICAST(addr))
 450                         return NL_SKIP;
 451
 452                 check_addr_updates(iface);
 453
 454                 if (iface->ndp != RELAYD_RELAY)
 455                         break;
 456
 457                 /* handle the relay logic below */
 458                 setup_addr_for_relaying(addr, iface, add);
 459
 460                 if (!add)
 461                         dump_neigh_table(false);
 462                 break;
 463         }
 464
 465         case RTM_NEWNEIGH:
 466                 add = true;
 467         case RTM_DELNEIGH: {
 468                 struct ndmsg *ndm = nlmsg_data(hdr);
 469                 struct nlattr *nla[__NDA_MAX];
 470
 471                 if (!nlmsg_valid_hdr(hdr, sizeof(*ndm)) ||
 472                                 ndm->ndm_family != AF_INET6)
 473                         return NL_SKIP;
 474
 475                 iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex);
 476                 if (!iface || iface->ndp != RELAYD_RELAY)
 477                         return (iface ? NL_OK : NL_SKIP);
 478
 479                 nlmsg_parse(hdr, sizeof(*ndm), nla, __NDA_MAX - 1, NULL);
 480                 if (!nla[NDA_DST])
 481                         return NL_SKIP;
 482
 483                 addr = nla_data(nla[NDA_DST]);
 484                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 485                                 IN6_IS_ADDR_MULTICAST(addr))
 486                         return NL_SKIP;
 487
 488                 if (ndm->ndm_flags & NTF_PROXY) {
 489                         /* Dump and flush proxy entries */
 490                         if (hdr->nlmsg_type == RTM_NEWNEIGH) {
 491                                 odhcpd_setup_proxy_neigh(addr, iface, false);
 492                                 setup_route(addr, iface, false);
 493                                 dump_neigh_table(false);
 494                         }
 495
 496                         return NL_OK;
 497                 }
 498
 499                 if (add && !(ndm->ndm_state &
 500                                 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE |
 501                                  NUD_PERMANENT | NUD_NOARP))) {
 502                         if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
 503                                 last_solicited = *addr;
 504                                 setup_ping6(addr, iface);
 505                         }
 506
 507                         return NL_OK;
 508                 }
 509
 510                 setup_addr_for_relaying(addr, iface, add);
 511                 setup_route(addr, iface, add);
 512
 513                 if (!add)
 514                         dump_neigh_table(false);
 515                 break;
 516         }
 517
 518         default:
 519                 return NL_SKIP;
 520         }
 521
 522         return NL_OK;
 523 }
 524
 525 static void catch_rtnetlink(int error)
 526 {
 527         if (error == ENOBUFS)
 528                 dump_addr_table();
 529 }