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         int ret = 0, procfd;
 173         bool dump_neigh = false;
 174         char procbuf[64];
 175
 176         snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
 177         procfd = open(procbuf, O_WRONLY);
 178
 179         if (procfd < 0) {
 180                 ret = -1;
 181                 goto out;
 182         }
 183
 184         if (iface->ndp_event.uloop.fd > 0) {
 185                 uloop_fd_delete(&iface->ndp_event.uloop);
 186                 close(iface->ndp_event.uloop.fd);
 187                 iface->ndp_event.uloop.fd = -1;
 188
 189                 if (!enable || iface->ndp != RELAYD_RELAY)
 190                         if (write(procfd, "0\n", 2) < 0) {}
 191
 192                 dump_neigh = true;
 193         }
 194
 195         if (enable && (iface->ra == RELAYD_SERVER ||
 196                         iface->dhcpv6 == RELAYD_SERVER || iface->ndp == RELAYD_RELAY))
 197                 dump_addr_table();
 198
 199         if (enable && iface->ndp == RELAYD_RELAY) {
 200                 if (write(procfd, "1\n", 2) < 0) {}
 201
 202                 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
 203                 if (sock < 0) {
 204                         syslog(LOG_ERR, "Unable to open packet socket: %s",
 205                                         strerror(errno));
 206                         ret = -1;
 207                         goto out;
 208                 }
 209
 210 #ifdef PACKET_RECV_TYPE
 211                 int pktt = 1 << PACKET_MULTICAST;
 212                 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
 213 #endif
 214
 215                 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
 216                                 &bpf_prog, sizeof(bpf_prog))) {
 217                         syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
 218                         ret = -1;
 219                         goto out;
 220                 }
 221
 222                 struct sockaddr_ll ll = {
 223                         .sll_family = AF_PACKET,
 224                         .sll_ifindex = iface->ifindex,
 225                         .sll_protocol = htons(ETH_P_IPV6),
 226                         .sll_hatype = 0,
 227                         .sll_pkttype = 0,
 228                         .sll_halen = 0,
 229                         .sll_addr = {0},
 230                 };
 231                 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
 232
 233                 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
 234                 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
 235
 236                 iface->ndp_event.uloop.fd = sock;
 237                 iface->ndp_event.handle_dgram = handle_solicit;
 238                 odhcpd_register(&iface->ndp_event);
 239
 240                 // If we already were enabled dump is unnecessary, if not do dump
 241                 if (!dump_neigh)
 242                         dump_neigh_table(false);
 243                 else
 244                         dump_neigh = false;
 245         }
 246
 247         if (dump_neigh)
 248                 dump_neigh_table(true);
 249
 250 out:
 251         if (procfd >= 0)
 252                 close(procfd);
 253
 254         return ret;
 255 }
 256
 257
 258 // Send an ICMP-ECHO. This is less for actually pinging but for the
 259 // neighbor cache to be kept up-to-date.
 260 static void ping6(struct in6_addr *addr,
 261                 const struct interface *iface)
 262 {
 263         struct sockaddr_in6 dest = { .sin6_family = AF_INET6, .sin6_addr = *addr, .sin6_scope_id = iface->ifindex, };
 264         struct icmp6_hdr echo = { .icmp6_type = ICMP6_ECHO_REQUEST };
 265         struct iovec iov = { .iov_base = &echo, .iov_len = sizeof(echo) };
 266
 267         odhcpd_setup_route(addr, 128, iface, NULL, 128, true);
 268         odhcpd_send(ping_socket, &dest, &iov, 1, iface);
 269         odhcpd_setup_route(addr, 128, iface, NULL, 128, false);
 270 }
 271
 272
 273 // Handle solicitations
 274 static void handle_solicit(void *addr, void *data, size_t len,
 275                 struct interface *iface, _unused void *dest)
 276 {
 277         struct ip6_hdr *ip6 = data;
 278         struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
 279         struct sockaddr_ll *ll = addr;
 280
 281         // Solicitation is for duplicate address detection
 282         bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
 283
 284         // Don't process solicit messages on non relay interfaces
 285         // Don't forward any non-DAD solicitation for external ifaces
 286         // TODO: check if we should even forward DADs for them
 287         if (iface->ndp != RELAYD_RELAY || (iface->external && !ns_is_dad))
 288                 return;
 289
 290         if (len < sizeof(*ip6) + sizeof(*req))
 291                 return; // Invalid reqicitation
 292
 293         if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
 294                         IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
 295                         IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
 296                 return; // Invalid target
 297
 298         char ipbuf[INET6_ADDRSTRLEN];
 299         inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
 300         syslog(LOG_DEBUG, "Got a NS for %s", ipbuf);
 301
 302         uint8_t mac[6];
 303         odhcpd_get_mac(iface, mac);
 304         if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
 305                 return; // Looped back
 306
 307         struct interface *c;
 308         list_for_each_entry(c, &interfaces, head)
 309                 if (iface != c && c->ndp == RELAYD_RELAY &&
 310                                 (ns_is_dad || !c->external))
 311                         ping6(&req->nd_ns_target, c);
 312 }
 313
 314 // Use rtnetlink to modify kernel routes
 315 static void setup_route(struct in6_addr *addr, struct interface *iface, bool add)
 316 {
 317         char namebuf[INET6_ADDRSTRLEN];
 318         inet_ntop(AF_INET6, addr, namebuf, sizeof(namebuf));
 319         syslog(LOG_NOTICE, "%s about %s on %s",
 320                         (add) ? "Learned" : "Forgot", namebuf, iface->ifname);
 321
 322         if (iface->learn_routes)
 323                 odhcpd_setup_route(addr, 128, iface, NULL, 1024, add);
 324 }
 325
 326 // compare prefixes
 327 static int prefixcmp(const void *va, const void *vb)
 328 {
 329         const struct odhcpd_ipaddr *a = va, *b = vb;
 330         uint32_t a_pref = ((a->addr.s6_addr[0] & 0xfe) != 0xfc) ? a->preferred : 1;
 331         uint32_t b_pref = ((b->addr.s6_addr[0] & 0xfe) != 0xfc) ? b->preferred : 1;
 332         return (a_pref < b_pref) ? 1 : (a_pref > b_pref) ? -1 : 0;
 333 }
 334
 335 // Check address update
 336 static void check_addr_updates(struct interface *iface)
 337 {
 338         struct odhcpd_ipaddr addr[RELAYD_MAX_ADDRS] = {{IN6ADDR_ANY_INIT, 0, 0, 0, 0}};
 339         time_t now = odhcpd_time();
 340         ssize_t len = odhcpd_get_interface_addresses(iface->ifindex, addr, ARRAY_SIZE(addr));
 341
 342         if (len < 0)
 343                 return;
 344
 345         qsort(addr, len, sizeof(*addr), prefixcmp);
 346
 347         for (int i = 0; i < len; ++i) {
 348                 addr[i].addr.s6_addr32[3] = 0;
 349
 350                 if (addr[i].preferred < UINT32_MAX - now)
 351                         addr[i].preferred += now;
 352
 353                 if (addr[i].valid < UINT32_MAX - now)
 354                         addr[i].valid += now;
 355         }
 356
 357         bool change = len != (ssize_t)iface->ia_addr_len;
 358         for (ssize_t i = 0; !change && i < len; ++i)
 359                 if (!IN6_ARE_ADDR_EQUAL(&addr[i].addr, &iface->ia_addr[i].addr) ||
 360                                 (addr[i].preferred > 0) != (iface->ia_addr[i].preferred > 0) ||
 361                                 addr[i].valid < iface->ia_addr[i].valid ||
 362                                 addr[i].preferred < iface->ia_addr[i].preferred)
 363                         change = true;
 364
 365         if (change)
 366                 dhcpv6_ia_preupdate(iface);
 367
 368         memcpy(iface->ia_addr, addr, len * sizeof(*addr));
 369         iface->ia_addr_len = len;
 370
 371         if (change)
 372                 dhcpv6_ia_postupdate(iface, now);
 373
 374         if (change) {
 375                 syslog(LOG_INFO, "Raising SIGUSR1 due to address change on %s", iface->ifname);
 376                 raise(SIGUSR1);
 377         }
 378 }
 379
 380 void setup_addr_for_relaying(struct in6_addr *addr, struct interface *iface, bool add)
 381 {
 382         struct interface *c;
 383
 384         list_for_each_entry(c, &interfaces, head) {
 385                 if (iface == c || (c->ndp != RELAYD_RELAY && !add))
 386                         continue;
 387
 388                 odhcpd_setup_proxy_neigh(addr, c, c->ndp == RELAYD_RELAY ? add : false);
 389         }
 390 }
 391
 392 void setup_ping6(struct in6_addr *addr, struct interface *iface)
 393 {
 394         struct interface *c;
 395
 396         list_for_each_entry(c, &interfaces, head) {
 397                 if (iface == c || c->ndp != RELAYD_RELAY ||
 398                                 c->external == true)
 399                         continue;
 400
 401                 ping6(addr, c);
 402         }
 403 }
 404
 405 static struct in6_addr last_solicited;
 406
 407 static void handle_rtnl_event(struct odhcpd_event *e)
 408 {
 409         struct event_socket *ev_sock = container_of(e, struct event_socket, ev);
 410
 411         nl_recvmsgs_default(ev_sock->sock);
 412 }
 413
 414
 415 // Handler for neighbor cache entries from the kernel. This is our source
 416 // to learn and unlearn hosts on interfaces.
 417 static int cb_rtnl_valid(struct nl_msg *msg, _unused void *arg)
 418 {
 419         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 420         struct in6_addr *addr = NULL;
 421         struct interface *iface = NULL;
 422         bool add = false;
 423
 424         switch (hdr->nlmsg_type) {
 425         case RTM_NEWROUTE:
 426         case RTM_DELROUTE: {
 427                 struct rtmsg *rtm = nlmsg_data(hdr);
 428
 429                 if (!nlmsg_valid_hdr(hdr, sizeof(*rtm)) ||
 430                                 rtm->rtm_family != AF_INET6)
 431                         return NL_SKIP;
 432
 433                 if (rtm->rtm_dst_len == 0) {
 434                         syslog(LOG_INFO, "Raising SIGUSR1 due to default route change");
 435                         raise(SIGUSR1);
 436                 }
 437                 return NL_OK;
 438         }
 439
 440         case RTM_NEWADDR:
 441                 add = true;
 442         case RTM_DELADDR: {
 443                 struct ifaddrmsg *ifa = nlmsg_data(hdr);
 444                 struct nlattr *nla[__IFA_MAX];
 445
 446                 if (!nlmsg_valid_hdr(hdr, sizeof(*ifa)) ||
 447                                 ifa->ifa_family != AF_INET6)
 448                         return NL_SKIP;
 449
 450                 iface = odhcpd_get_interface_by_index(ifa->ifa_index);
 451                 if (!iface)
 452                         return NL_SKIP;
 453
 454                 nlmsg_parse(hdr, sizeof(*ifa), nla, __IFA_MAX - 1, NULL);
 455                 if (!nla[IFA_ADDRESS])
 456                         return NL_SKIP;
 457
 458                 addr = nla_data(nla[IFA_ADDRESS]);
 459                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 460                                 IN6_IS_ADDR_MULTICAST(addr))
 461                         return NL_SKIP;
 462
 463                 check_addr_updates(iface);
 464
 465                 if (iface->ndp != RELAYD_RELAY)
 466                         break;
 467
 468                 /* handle the relay logic below */
 469                 setup_addr_for_relaying(addr, iface, add);
 470
 471                 if (!add)
 472                         dump_neigh_table(false);
 473                 break;
 474         }
 475
 476         case RTM_NEWNEIGH:
 477                 add = true;
 478         case RTM_DELNEIGH: {
 479                 struct ndmsg *ndm = nlmsg_data(hdr);
 480                 struct nlattr *nla[__NDA_MAX];
 481
 482                 if (!nlmsg_valid_hdr(hdr, sizeof(*ndm)) ||
 483                                 ndm->ndm_family != AF_INET6)
 484                         return NL_SKIP;
 485
 486                 iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex);
 487                 if (!iface || iface->ndp != RELAYD_RELAY)
 488                         return (iface ? NL_OK : NL_SKIP);
 489
 490                 nlmsg_parse(hdr, sizeof(*ndm), nla, __NDA_MAX - 1, NULL);
 491                 if (!nla[NDA_DST])
 492                         return NL_SKIP;
 493
 494                 addr = nla_data(nla[NDA_DST]);
 495                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 496                                 IN6_IS_ADDR_MULTICAST(addr))
 497                         return NL_SKIP;
 498
 499                 if (ndm->ndm_flags & NTF_PROXY) {
 500                         /* Dump and flush proxy entries */
 501                         if (hdr->nlmsg_type == RTM_NEWNEIGH) {
 502                                 odhcpd_setup_proxy_neigh(addr, iface, false);
 503                                 setup_route(addr, iface, false);
 504                                 dump_neigh_table(false);
 505                         }
 506
 507                         return NL_OK;
 508                 }
 509
 510                 if (add && !(ndm->ndm_state &
 511                                 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE |
 512                                  NUD_PERMANENT | NUD_NOARP))) {
 513                         if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
 514                                 last_solicited = *addr;
 515                                 setup_ping6(addr, iface);
 516                         }
 517
 518                         return NL_OK;
 519                 }
 520
 521                 setup_addr_for_relaying(addr, iface, add);
 522                 setup_route(addr, iface, add);
 523
 524                 if (!add)
 525                         dump_neigh_table(false);
 526                 break;
 527         }
 528
 529         default:
 530                 return NL_SKIP;
 531         }
 532
 533         return NL_OK;
 534 }
 535
 536 static void catch_rtnetlink(int error)
 537 {
 538         if (error == ENOBUFS)
 539                 dump_addr_table();
 540 }