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 "dhcpv6.h"
  37 #include "odhcpd.h"
  38
  39 struct event_socket {
  40         struct odhcpd_event ev;
  41         struct nl_sock *sock;
  42         int sock_bufsize;
  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_rtnl_err(struct odhcpd_event *e, int error);
  50
  51 static int addr6_dump_rqs = 0;
  52 static int ping_socket = -1;
  53 static struct event_socket rtnl_event = {
  54         .ev = {
  55                 .uloop = {.fd = - 1, },
  56                 .handle_dgram = NULL,
  57                 .handle_error = catch_rtnl_err,
  58                 .recv_msgs = handle_rtnl_event,
  59         },
  60         .sock = NULL,
  61         .sock_bufsize = 133120,
  62 };
  63
  64 // Filter ICMPv6 messages of type neighbor soliciation
  65 static struct sock_filter bpf[] = {
  66         BPF_STMT(BPF_LD | BPF_B | BPF_ABS, offsetof(struct ip6_hdr, ip6_nxt)),
  67         BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, IPPROTO_ICMPV6, 0, 3),
  68         BPF_STMT(BPF_LD | BPF_B | BPF_ABS, sizeof(struct ip6_hdr) +
  69                         offsetof(struct icmp6_hdr, icmp6_type)),
  70         BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, ND_NEIGHBOR_SOLICIT, 0, 1),
  71         BPF_STMT(BPF_RET | BPF_K, 0xffffffff),
  72         BPF_STMT(BPF_RET | BPF_K, 0),
  73 };
  74 static const struct sock_fprog bpf_prog = {sizeof(bpf) / sizeof(*bpf), bpf};
  75
  76
  77 // Initialize NDP-proxy
  78 int init_ndp(void)
  79 {
  80         int val = 2;
  81
  82         rtnl_event.sock = odhcpd_create_nl_socket(NETLINK_ROUTE);
  83         if (!rtnl_event.sock)
  84                 goto err;
  85
  86         rtnl_event.ev.uloop.fd = nl_socket_get_fd(rtnl_event.sock);
  87
  88         if (nl_socket_set_buffer_size(rtnl_event.sock, rtnl_event.sock_bufsize, 0))
  89                 goto err;
  90
  91         nl_socket_disable_seq_check(rtnl_event.sock);
  92
  93         nl_socket_modify_cb(rtnl_event.sock, NL_CB_VALID, NL_CB_CUSTOM,
  94                         cb_rtnl_valid, NULL);
  95
  96         // Receive IPv6 address, IPv6 routes and neighbor events
  97         if (nl_socket_add_memberships(rtnl_event.sock, RTNLGRP_IPV6_IFADDR,
  98                                 RTNLGRP_IPV6_ROUTE, RTNLGRP_NEIGH, 0))
  99                 goto err;
 100
 101         odhcpd_register(&rtnl_event.ev);
 102
 103         // Open ICMPv6 socket
 104         ping_socket = socket(AF_INET6, SOCK_RAW | SOCK_CLOEXEC, IPPROTO_ICMPV6);
 105         if (ping_socket < 0) {
 106                 syslog(LOG_ERR, "Unable to open raw socket: %s", strerror(errno));
 107                         return -1;
 108         }
 109
 110         setsockopt(ping_socket, IPPROTO_RAW, IPV6_CHECKSUM, &val, sizeof(val));
 111
 112         // This is required by RFC 4861
 113         val = 255;
 114         setsockopt(ping_socket, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, &val, sizeof(val));
 115         setsockopt(ping_socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val));
 116
 117         // Filter all packages, we only want to send
 118         struct icmp6_filter filt;
 119         ICMP6_FILTER_SETBLOCKALL(&filt);
 120         setsockopt(ping_socket, IPPROTO_ICMPV6, ICMP6_FILTER, &filt, sizeof(filt));
 121
 122         return 0;
 123
 124 err:
 125         if (rtnl_event.sock) {
 126                 nl_socket_free(rtnl_event.sock);
 127                 rtnl_event.sock = NULL;
 128                 rtnl_event.ev.uloop.fd = -1;
 129         }
 130
 131         return -1;
 132 }
 133
 134 static void dump_neigh_table(const bool proxy)
 135 {
 136         struct nl_msg *msg;
 137         struct ndmsg ndm = {
 138                 .ndm_family = AF_INET6,
 139                 .ndm_flags = proxy ? NTF_PROXY : 0,
 140         };
 141
 142         msg = nlmsg_alloc_simple(RTM_GETNEIGH, NLM_F_REQUEST | NLM_F_DUMP);
 143         if (!msg)
 144                 return;
 145
 146         nlmsg_append(msg, &ndm, sizeof(ndm), 0);
 147
 148         nl_send_auto_complete(rtnl_event.sock, msg);
 149
 150         nlmsg_free(msg);
 151 }
 152
 153 static void dump_addr6_table(void)
 154 {
 155         struct nl_msg *msg;
 156         struct ifaddrmsg ifa = {
 157                 .ifa_family = AF_INET6,
 158         };
 159
 160         msg = nlmsg_alloc_simple(RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP);
 161         if (!msg)
 162                 return;
 163
 164         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
 165
 166         nl_send_auto_complete(rtnl_event.sock, msg);
 167
 168         nlmsg_free(msg);
 169 }
 170
 171 void ndp_handle_addr6_dump(void)
 172 {
 173         if (!addr6_dump_rqs)
 174                 return;
 175
 176         dump_addr6_table();
 177         addr6_dump_rqs = 0;
 178 }
 179
 180 inline void ndp_rqs_addr6_dump(void)
 181 {
 182         addr6_dump_rqs++;
 183 }
 184
 185 int setup_ndp_interface(struct interface *iface, bool enable)
 186 {
 187         int ret = 0, procfd;
 188         bool dump_neigh = false;
 189         char procbuf[64];
 190
 191         snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
 192         procfd = open(procbuf, O_WRONLY);
 193
 194         if (procfd < 0) {
 195                 ret = -1;
 196                 goto out;
 197         }
 198
 199         if (iface->ndp_event.uloop.fd > 0) {
 200                 uloop_fd_delete(&iface->ndp_event.uloop);
 201                 close(iface->ndp_event.uloop.fd);
 202                 iface->ndp_event.uloop.fd = -1;
 203
 204                 if (!enable || iface->ndp != RELAYD_RELAY)
 205                         if (write(procfd, "0\n", 2) < 0) {}
 206
 207                 dump_neigh = true;
 208         }
 209
 210         if (enable && iface->ndp == RELAYD_RELAY) {
 211                 if (write(procfd, "1\n", 2) < 0) {}
 212
 213                 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
 214                 if (sock < 0) {
 215                         syslog(LOG_ERR, "Unable to open packet socket: %s",
 216                                         strerror(errno));
 217                         ret = -1;
 218                         goto out;
 219                 }
 220
 221 #ifdef PACKET_RECV_TYPE
 222                 int pktt = 1 << PACKET_MULTICAST;
 223                 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
 224 #endif
 225
 226                 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
 227                                 &bpf_prog, sizeof(bpf_prog))) {
 228                         syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
 229                         ret = -1;
 230                         goto out;
 231                 }
 232
 233                 struct sockaddr_ll ll = {
 234                         .sll_family = AF_PACKET,
 235                         .sll_ifindex = iface->ifindex,
 236                         .sll_protocol = htons(ETH_P_IPV6),
 237                         .sll_hatype = 0,
 238                         .sll_pkttype = 0,
 239                         .sll_halen = 0,
 240                         .sll_addr = {0},
 241                 };
 242                 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
 243
 244                 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
 245                 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
 246
 247                 iface->ndp_event.uloop.fd = sock;
 248                 iface->ndp_event.handle_dgram = handle_solicit;
 249                 odhcpd_register(&iface->ndp_event);
 250
 251                 // If we already were enabled dump is unnecessary, if not do dump
 252                 if (!dump_neigh)
 253                         dump_neigh_table(false);
 254                 else
 255                         dump_neigh = false;
 256
 257                 ndp_rqs_addr6_dump();
 258         }
 259
 260         if (dump_neigh)
 261                 dump_neigh_table(true);
 262
 263 out:
 264         if (procfd >= 0)
 265                 close(procfd);
 266
 267         return ret;
 268 }
 269
 270
 271 // Send an ICMP-ECHO. This is less for actually pinging but for the
 272 // neighbor cache to be kept up-to-date.
 273 static void ping6(struct in6_addr *addr,
 274                 const struct interface *iface)
 275 {
 276         struct sockaddr_in6 dest = { .sin6_family = AF_INET6, .sin6_addr = *addr, .sin6_scope_id = iface->ifindex, };
 277         struct icmp6_hdr echo = { .icmp6_type = ICMP6_ECHO_REQUEST };
 278         struct iovec iov = { .iov_base = &echo, .iov_len = sizeof(echo) };
 279         char ipbuf[INET6_ADDRSTRLEN];
 280
 281         inet_ntop(AF_INET6, addr, ipbuf, sizeof(ipbuf));
 282         syslog(LOG_NOTICE, "Pinging for %s%%%s", ipbuf, iface->ifname);
 283
 284         odhcpd_setup_route(addr, 128, iface, NULL, 128, true);
 285         odhcpd_send(ping_socket, &dest, &iov, 1, iface);
 286         odhcpd_setup_route(addr, 128, iface, NULL, 128, false);
 287 }
 288
 289 // Handle solicitations
 290 static void handle_solicit(void *addr, void *data, size_t len,
 291                 struct interface *iface, _unused void *dest)
 292 {
 293         struct ip6_hdr *ip6 = data;
 294         struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
 295         struct sockaddr_ll *ll = addr;
 296         char ipbuf[INET6_ADDRSTRLEN];
 297         uint8_t mac[6];
 298
 299         // Solicitation is for duplicate address detection
 300         bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
 301
 302         // Don't process solicit messages on non relay interfaces
 303         // Don't forward any non-DAD solicitation for external ifaces
 304         // TODO: check if we should even forward DADs for them
 305         if (iface->ndp != RELAYD_RELAY || (iface->external && !ns_is_dad))
 306                 return;
 307
 308         if (len < sizeof(*ip6) + sizeof(*req))
 309                 return; // Invalid reqicitation
 310
 311         if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
 312                         IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
 313                         IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
 314                 return; // Invalid target
 315
 316         inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
 317         syslog(LOG_DEBUG, "Got a NS for %s%%%s", ipbuf, iface->ifname);
 318
 319         odhcpd_get_mac(iface, mac);
 320         if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
 321                 return; // Looped back
 322
 323         struct interface *c;
 324         list_for_each_entry(c, &interfaces, head)
 325                 if (iface != c && c->ndp == RELAYD_RELAY &&
 326                                 (ns_is_dad || !c->external))
 327                         ping6(&req->nd_ns_target, c);
 328 }
 329
 330 // Use rtnetlink to modify kernel routes
 331 static void setup_route(struct in6_addr *addr, struct interface *iface, bool add)
 332 {
 333         char ipbuf[INET6_ADDRSTRLEN];
 334
 335         inet_ntop(AF_INET6, addr, ipbuf, sizeof(ipbuf));
 336         syslog(LOG_NOTICE, "%s about %s%%%s",
 337                         (add) ? "Learned" : "Forgot", ipbuf, iface->ifname);
 338
 339         if (iface->learn_routes)
 340                 odhcpd_setup_route(addr, 128, iface, NULL, 1024, add);
 341 }
 342
 343 // compare prefixes
 344 static int prefixcmp(const void *va, const void *vb)
 345 {
 346         const struct odhcpd_ipaddr *a = va, *b = vb;
 347         uint32_t a_pref = IN6_IS_ADDR_ULA(&a->addr) ? 1 : a->preferred;
 348         uint32_t b_pref = IN6_IS_ADDR_ULA(&b->addr) ? 1 : b->preferred;
 349         return (a_pref < b_pref) ? 1 : (a_pref > b_pref) ? -1 : 0;
 350 }
 351
 352 // Check address update
 353 static void check_addr_updates(struct interface *iface)
 354 {
 355         struct odhcpd_ipaddr addr[RELAYD_MAX_ADDRS] = {{IN6ADDR_ANY_INIT, 0, 0, 0, 0}};
 356         time_t now = odhcpd_time();
 357         ssize_t len = odhcpd_get_interface_addresses(iface->ifindex, addr, ARRAY_SIZE(addr));
 358
 359         if (len < 0)
 360                 return;
 361
 362         qsort(addr, len, sizeof(*addr), prefixcmp);
 363
 364         for (int i = 0; i < len; ++i) {
 365                 addr[i].addr.s6_addr32[3] = 0;
 366
 367                 if (addr[i].preferred < UINT32_MAX - now)
 368                         addr[i].preferred += now;
 369
 370                 if (addr[i].valid < UINT32_MAX - now)
 371                         addr[i].valid += now;
 372         }
 373
 374         bool change = len != (ssize_t)iface->ia_addr_len;
 375         for (ssize_t i = 0; !change && i < len; ++i)
 376                 if (!IN6_ARE_ADDR_EQUAL(&addr[i].addr, &iface->ia_addr[i].addr) ||
 377                                 (addr[i].preferred > 0) != (iface->ia_addr[i].preferred > 0) ||
 378                                 addr[i].valid < iface->ia_addr[i].valid ||
 379                                 addr[i].preferred < iface->ia_addr[i].preferred)
 380                         change = true;
 381
 382         if (change)
 383                 dhcpv6_ia_preupdate(iface);
 384
 385         memcpy(iface->ia_addr, addr, len * sizeof(*addr));
 386         iface->ia_addr_len = len;
 387
 388         if (change)
 389                 dhcpv6_ia_postupdate(iface, now);
 390
 391         if (change) {
 392                 syslog(LOG_INFO, "Raising SIGUSR1 due to address change on %s", iface->ifname);
 393                 raise(SIGUSR1);
 394         }
 395 }
 396
 397 static void setup_addr_for_relaying(struct in6_addr *addr, struct interface *iface, bool add)
 398 {
 399         struct interface *c;
 400         char ipbuf[INET6_ADDRSTRLEN];
 401
 402         inet_ntop(AF_INET6, addr, ipbuf, sizeof(ipbuf));
 403
 404         list_for_each_entry(c, &interfaces, head) {
 405                 if (iface == c || (c->ndp != RELAYD_RELAY && !add))
 406                         continue;
 407
 408                 add = (c->ndp == RELAYD_RELAY ? add : false);
 409
 410                 if (odhcpd_setup_proxy_neigh(addr, c, add))
 411                         syslog(LOG_DEBUG, "Failed to %s proxy neighbour entry %s%%%s",
 412                                 add ? "add" : "delete", ipbuf, iface->ifname);
 413                 else
 414                         syslog(LOG_DEBUG, "%s proxy neighbour entry %s%%%s",
 415                                 add ? "Added" : "Deleted", ipbuf, iface->ifname);
 416         }
 417 }
 418
 419 static void setup_ping6(struct in6_addr *addr, struct interface *iface)
 420 {
 421         struct interface *c;
 422
 423         list_for_each_entry(c, &interfaces, head) {
 424                 if (iface == c || c->ndp != RELAYD_RELAY ||
 425                                 c->external == true)
 426                         continue;
 427
 428                 ping6(addr, c);
 429         }
 430 }
 431
 432 static struct in6_addr last_solicited;
 433
 434 static void handle_rtnl_event(struct odhcpd_event *e)
 435 {
 436         struct event_socket *ev_sock = container_of(e, struct event_socket, ev);
 437
 438         nl_recvmsgs_default(ev_sock->sock);
 439 }
 440
 441
 442 // Handler for neighbor cache entries from the kernel. This is our source
 443 // to learn and unlearn hosts on interfaces.
 444 static int cb_rtnl_valid(struct nl_msg *msg, _unused void *arg)
 445 {
 446         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 447         struct in6_addr *addr = NULL;
 448         struct interface *iface = NULL;
 449         bool add = false;
 450         char ipbuf[INET6_ADDRSTRLEN];
 451
 452         switch (hdr->nlmsg_type) {
 453         case RTM_NEWROUTE:
 454         case RTM_DELROUTE: {
 455                 struct rtmsg *rtm = nlmsg_data(hdr);
 456
 457                 if (!nlmsg_valid_hdr(hdr, sizeof(*rtm)) ||
 458                                 rtm->rtm_family != AF_INET6)
 459                         return NL_SKIP;
 460
 461                 if (rtm->rtm_dst_len == 0) {
 462                         syslog(LOG_INFO, "Raising SIGUSR1 due to default route change");
 463                         raise(SIGUSR1);
 464                 }
 465                 return NL_OK;
 466         }
 467
 468         case RTM_NEWADDR:
 469                 add = true;
 470         case RTM_DELADDR: {
 471                 struct ifaddrmsg *ifa = nlmsg_data(hdr);
 472                 struct nlattr *nla[__IFA_MAX];
 473
 474                 if (!nlmsg_valid_hdr(hdr, sizeof(*ifa)) ||
 475                                 ifa->ifa_family != AF_INET6)
 476                         return NL_SKIP;
 477
 478                 iface = odhcpd_get_interface_by_index(ifa->ifa_index);
 479                 if (!iface)
 480                         return NL_SKIP;
 481
 482                 nlmsg_parse(hdr, sizeof(*ifa), nla, __IFA_MAX - 1, NULL);
 483                 if (!nla[IFA_ADDRESS])
 484                         return NL_SKIP;
 485
 486                 addr = nla_data(nla[IFA_ADDRESS]);
 487                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 488                                 IN6_IS_ADDR_MULTICAST(addr))
 489                         return NL_SKIP;
 490
 491                 inet_ntop(AF_INET6, addr, ipbuf, sizeof(ipbuf));
 492                 syslog(LOG_DEBUG, "Netlink %s %s%%%s", true ? "newaddr" : "deladdr",
 493                         ipbuf, iface->ifname);
 494
 495                 check_addr_updates(iface);
 496
 497                 if (iface->ndp != RELAYD_RELAY)
 498                         break;
 499
 500                 /* handle the relay logic below */
 501                 setup_addr_for_relaying(addr, iface, add);
 502
 503                 if (!add)
 504                         dump_neigh_table(false);
 505                 break;
 506         }
 507
 508         case RTM_NEWNEIGH:
 509                 add = true;
 510         case RTM_DELNEIGH: {
 511                 struct ndmsg *ndm = nlmsg_data(hdr);
 512                 struct nlattr *nla[__NDA_MAX];
 513
 514                 if (!nlmsg_valid_hdr(hdr, sizeof(*ndm)) ||
 515                                 ndm->ndm_family != AF_INET6)
 516                         return NL_SKIP;
 517
 518                 iface = odhcpd_get_interface_by_index(ndm->ndm_ifindex);
 519                 if (!iface || iface->ndp != RELAYD_RELAY)
 520                         return (iface ? NL_OK : NL_SKIP);
 521
 522                 nlmsg_parse(hdr, sizeof(*ndm), nla, __NDA_MAX - 1, NULL);
 523                 if (!nla[NDA_DST])
 524                         return NL_SKIP;
 525
 526                 addr = nla_data(nla[NDA_DST]);
 527                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 528                                 IN6_IS_ADDR_MULTICAST(addr))
 529                         return NL_SKIP;
 530
 531                 inet_ntop(AF_INET6, addr, ipbuf, sizeof(ipbuf));
 532                 syslog(LOG_DEBUG, "Netlink %s %s%%%s", true ? "newneigh" : "delneigh",
 533                         ipbuf, iface->ifname);
 534
 535                 if (ndm->ndm_flags & NTF_PROXY) {
 536                         /* Dump and flush proxy entries */
 537                         if (hdr->nlmsg_type == RTM_NEWNEIGH) {
 538                                 odhcpd_setup_proxy_neigh(addr, iface, false);
 539                                 setup_route(addr, iface, false);
 540                                 dump_neigh_table(false);
 541                         }
 542
 543                         return NL_OK;
 544                 }
 545
 546                 if (add && !(ndm->ndm_state &
 547                                 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE |
 548                                  NUD_PERMANENT | NUD_NOARP))) {
 549                         if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
 550                                 last_solicited = *addr;
 551                                 setup_ping6(addr, iface);
 552                         }
 553
 554                         return NL_OK;
 555                 }
 556
 557                 setup_addr_for_relaying(addr, iface, add);
 558                 setup_route(addr, iface, add);
 559
 560                 if (!add)
 561                         dump_neigh_table(false);
 562                 break;
 563         }
 564
 565         default:
 566                 return NL_SKIP;
 567         }
 568
 569         return NL_OK;
 570 }
 571
 572 static void catch_rtnl_err(struct odhcpd_event *e, int error)
 573 {
 574         struct event_socket *ev_sock = container_of(e, struct event_socket, ev);
 575
 576         if (error != ENOBUFS)
 577                 goto err;
 578
 579         /* Double netlink event buffer size */
 580         ev_sock->sock_bufsize *= 2;
 581
 582         if (nl_socket_set_buffer_size(ev_sock->sock, ev_sock->sock_bufsize, 0))
 583                 goto err;
 584
 585         dump_addr6_table();
 586         return;
 587
 588 err:
 589         odhcpd_deregister(e);
 590 }