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         odhcpd_process(&rtnl_event);
 119 }
 120
 121
 122 int setup_ndp_interface(struct interface *iface, bool enable)
 123 {
 124         char procbuf[64];
 125         snprintf(procbuf, sizeof(procbuf), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface->ifname);
 126         int procfd = open(procbuf, O_WRONLY);
 127         bool dump_neigh = false;
 128
 129         if (iface->ndp_event.uloop.fd > 0) {
 130                 uloop_fd_delete(&iface->ndp_event.uloop);
 131                 close(iface->ndp_event.uloop.fd);
 132                 iface->ndp_event.uloop.fd = -1;
 133
 134                 if (!enable || iface->ndp != RELAYD_RELAY)
 135                         if (write(procfd, "0\n", 2) < 0) {}
 136
 137                 dump_neigh = true;
 138         }
 139
 140         if (enable && (iface->ra == RELAYD_SERVER ||
 141                         iface->dhcpv6 == RELAYD_SERVER || iface->ndp == RELAYD_RELAY)) {
 142                 // Synthesize initial address events
 143                 struct {
 144                         struct nlmsghdr nh;
 145                         struct ifaddrmsg ifa;
 146                 } req2 = {
 147                         {sizeof(req2), RTM_GETADDR, NLM_F_REQUEST | NLM_F_DUMP,
 148                                         ++rtnl_seqid, 0},
 149                         {.ifa_family = AF_INET6, .ifa_index = iface->ifindex}
 150                 };
 151                 send(rtnl_event.uloop.fd, &req2, sizeof(req2), MSG_DONTWAIT);
 152         }
 153
 154         if (enable && iface->ndp == RELAYD_RELAY) {
 155                 if (write(procfd, "1\n", 2) < 0) {}
 156                 close(procfd);
 157
 158                 int sock = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC, htons(ETH_P_IPV6));
 159                 if (sock < 0) {
 160                         syslog(LOG_ERR, "Unable to open packet socket: %s",
 161                                         strerror(errno));
 162                         return -1;
 163                 }
 164
 165 #ifdef PACKET_RECV_TYPE
 166                 int pktt = 1 << PACKET_MULTICAST;
 167                 setsockopt(sock, SOL_PACKET, PACKET_RECV_TYPE, &pktt, sizeof(pktt));
 168 #endif
 169
 170                 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER,
 171                                 &bpf_prog, sizeof(bpf_prog))) {
 172                         syslog(LOG_ERR, "Failed to set BPF: %s", strerror(errno));
 173                         return -1;
 174                 }
 175
 176                 struct sockaddr_ll ll = {
 177                         .sll_family = AF_PACKET,
 178                         .sll_ifindex = iface->ifindex,
 179                         .sll_protocol = htons(ETH_P_IPV6),
 180                         .sll_hatype = 0,
 181                         .sll_pkttype = 0,
 182                         .sll_halen = 0,
 183                         .sll_addr = {0},
 184                 };
 185                 bind(sock, (struct sockaddr*)&ll, sizeof(ll));
 186
 187                 struct packet_mreq mreq = {iface->ifindex, PACKET_MR_ALLMULTI, ETH_ALEN, {0}};
 188                 setsockopt(sock, SOL_PACKET, PACKET_ADD_MEMBERSHIP, &mreq, sizeof(mreq));
 189
 190                 iface->ndp_event.uloop.fd = sock;
 191                 iface->ndp_event.handle_dgram = handle_solicit;
 192                 odhcpd_register(&iface->ndp_event);
 193
 194                 // If we already were enabled dump is unnecessary, if not do dump
 195                 if (!dump_neigh)
 196                         dump_neigh_table(false);
 197                 else
 198                         dump_neigh = false;
 199         } else {
 200                 close(procfd);
 201         }
 202
 203         if (dump_neigh)
 204                 dump_neigh_table(true);
 205
 206         return 0;
 207 }
 208
 209
 210 // Send an ICMP-ECHO. This is less for actually pinging but for the
 211 // neighbor cache to be kept up-to-date.
 212 static ssize_t ping6(struct in6_addr *addr,
 213                 const struct interface *iface)
 214 {
 215         struct sockaddr_in6 dest = {AF_INET6, 0, 0, *addr, iface->ifindex};
 216         struct icmp6_hdr echo = {.icmp6_type = ICMP6_ECHO_REQUEST};
 217         struct iovec iov = {&echo, sizeof(echo)};
 218         return odhcpd_send(ping_socket, &dest, &iov, 1, iface);
 219 }
 220
 221
 222 // Handle solicitations
 223 static void handle_solicit(void *addr, void *data, size_t len,
 224                 struct interface *iface, _unused void *dest)
 225 {
 226         struct ip6_hdr *ip6 = data;
 227         struct nd_neighbor_solicit *req = (struct nd_neighbor_solicit*)&ip6[1];
 228         struct sockaddr_ll *ll = addr;
 229
 230         // Solicitation is for duplicate address detection
 231         bool ns_is_dad = IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src);
 232
 233         // Don't forward any non-DAD solicitation for external ifaces
 234         // TODO: check if we should even forward DADs for them
 235         if (iface->external && !ns_is_dad)
 236                 return;
 237
 238         if (len < sizeof(*ip6) + sizeof(*req))
 239                 return; // Invalid reqicitation
 240
 241         if (IN6_IS_ADDR_LINKLOCAL(&req->nd_ns_target) ||
 242                         IN6_IS_ADDR_LOOPBACK(&req->nd_ns_target) ||
 243                         IN6_IS_ADDR_MULTICAST(&req->nd_ns_target))
 244                 return; // Invalid target
 245
 246         char ipbuf[INET6_ADDRSTRLEN];
 247         inet_ntop(AF_INET6, &req->nd_ns_target, ipbuf, sizeof(ipbuf));
 248         syslog(LOG_DEBUG, "Got a NS for %s", ipbuf);
 249
 250         uint8_t mac[6];
 251         odhcpd_get_mac(iface, mac);
 252         if (!memcmp(ll->sll_addr, mac, sizeof(mac)))
 253                 return; // Looped back
 254
 255         struct interface *c;
 256         list_for_each_entry(c, &interfaces, head)
 257                 if (iface->ndp == RELAYD_RELAY && iface != c &&
 258                                 (ns_is_dad || !c->external))
 259                         ping6(&req->nd_ns_target, c);
 260 }
 261
 262
 263 void odhcpd_setup_route(const struct in6_addr *addr, int prefixlen,
 264                 const struct interface *iface, const struct in6_addr *gw, bool add)
 265 {
 266         struct req {
 267                 struct nlmsghdr nh;
 268                 struct rtmsg rtm;
 269                 struct rtattr rta_dst;
 270                 struct in6_addr dst_addr;
 271                 struct rtattr rta_oif;
 272                 uint32_t ifindex;
 273                 struct rtattr rta_table;
 274                 uint32_t table;
 275                 struct rtattr rta_gw;
 276                 struct in6_addr gw;
 277         } req = {
 278                 {sizeof(req), 0, NLM_F_REQUEST, ++rtnl_seqid, 0},
 279                 {AF_INET6, prefixlen, 0, 0, 0, 0, 0, 0, 0},
 280                 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_DST},
 281                 *addr,
 282                 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_OIF},
 283                 iface->ifindex,
 284                 {sizeof(struct rtattr) + sizeof(uint32_t), RTA_TABLE},
 285                 RT_TABLE_MAIN,
 286                 {sizeof(struct rtattr) + sizeof(struct in6_addr), RTA_GATEWAY},
 287                 IN6ADDR_ANY_INIT,
 288         };
 289
 290         if (gw)
 291                 req.gw = *gw;
 292
 293         if (add) {
 294                 req.nh.nlmsg_type = RTM_NEWROUTE;
 295                 req.nh.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
 296                 req.rtm.rtm_protocol = RTPROT_STATIC;
 297                 req.rtm.rtm_scope = (gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
 298                 req.rtm.rtm_type = RTN_UNICAST;
 299         } else {
 300                 req.nh.nlmsg_type = RTM_DELROUTE;
 301                 req.rtm.rtm_scope = RT_SCOPE_NOWHERE;
 302         }
 303
 304         req.nh.nlmsg_len = (gw) ? sizeof(req) : offsetof(struct req, rta_gw);
 305         send(rtnl_event.uloop.fd, &req, req.nh.nlmsg_len, MSG_DONTWAIT);
 306 }
 307
 308 // Use rtnetlink to modify kernel routes
 309 static void setup_route(struct in6_addr *addr, struct interface *iface, bool add)
 310 {
 311         char namebuf[INET6_ADDRSTRLEN];
 312         inet_ntop(AF_INET6, addr, namebuf, sizeof(namebuf));
 313         syslog(LOG_NOTICE, "%s about %s on %s",
 314                         (add) ? "Learned" : "Forgot", namebuf, iface->ifname);
 315
 316         if (iface->learn_routes)
 317                 odhcpd_setup_route(addr, 128, iface, NULL, add);
 318 }
 319
 320
 321 // Handler for neighbor cache entries from the kernel. This is our source
 322 // to learn and unlearn hosts on interfaces.
 323 static void handle_rtnetlink(_unused void *addr, void *data, size_t len,
 324                 _unused struct interface *iface, _unused void *dest)
 325 {
 326         bool dump_neigh = false;
 327         struct in6_addr last_solicited = IN6ADDR_ANY_INIT;
 328
 329         for (struct nlmsghdr *nh = data; NLMSG_OK(nh, len);
 330                         nh = NLMSG_NEXT(nh, len)) {
 331                 struct ndmsg *ndm = NLMSG_DATA(nh);
 332                 struct rtmsg *rtm = NLMSG_DATA(nh);
 333
 334                 bool is_addr = (nh->nlmsg_type == RTM_NEWADDR
 335                                 || nh->nlmsg_type == RTM_DELADDR);
 336                 bool is_route = (nh->nlmsg_type == RTM_NEWROUTE
 337                                 || nh->nlmsg_type == RTM_DELROUTE);
 338                 bool is_neigh = (nh->nlmsg_type == RTM_NEWNEIGH
 339                                 || nh->nlmsg_type == RTM_DELNEIGH);
 340
 341                 // Family and ifindex are on the same offset for NEIGH and ADDR
 342                 if ((!is_addr && !is_route && !is_neigh)
 343                                 || NLMSG_PAYLOAD(nh, 0) < sizeof(*ndm)
 344                                 || ndm->ndm_family != AF_INET6)
 345                         continue;
 346
 347                 // Inform about a change in default route
 348                 if (is_route && rtm->rtm_dst_len == 0)
 349                         raise(SIGUSR1);
 350                 else if (is_route && rtm->rtm_dst_len == 128)
 351                         continue;
 352
 353                 // Data to retrieve
 354                 size_t rta_offset = (is_route) ? sizeof(*rtm) : (is_addr) ?
 355                                 sizeof(struct ifaddrmsg) : sizeof(*ndm);
 356                 uint16_t atype = (is_route) ? RTA_DST : (is_addr) ? IFA_ADDRESS : NDA_DST;
 357                 ssize_t alen = NLMSG_PAYLOAD(nh, rta_offset);
 358                 struct in6_addr *addr = NULL;
 359                 int *ifindex = (!is_route) ? &ndm->ndm_ifindex : NULL;
 360                 int *metric = NULL;
 361
 362                 for (struct rtattr *rta = (void*)(((uint8_t*)ndm) + rta_offset);
 363                                 RTA_OK(rta, alen); rta = RTA_NEXT(rta, alen)) {
 364                         if (rta->rta_type == atype &&
 365                                         RTA_PAYLOAD(rta) >= sizeof(*addr)) {
 366                                 addr = RTA_DATA(rta);
 367                         } else if (is_route && rta->rta_type == RTA_OIF &&
 368                                         RTA_PAYLOAD(rta) == sizeof(int)) {
 369                                 ifindex = (int*)RTA_DATA(rta);
 370                         } else if (is_route && rta->rta_type == RTA_GATEWAY) {
 371                                 ifindex = NULL;
 372                                 break;
 373                         } else if (is_route && rta->rta_type == RTA_PRIORITY) {
 374                                 metric = (int*)RTA_DATA(rta);
 375                         }
 376                 }
 377
 378                 // Lookup interface
 379                 struct interface *iface = ifindex ? odhcpd_get_interface_by_index(*ifindex) : NULL;
 380                 if (!iface)
 381                         continue;
 382
 383                 // Keep-alive neighbor entries for RA sending
 384                 if (nh->nlmsg_type == RTM_DELNEIGH && !(ndm->ndm_state & NUD_FAILED) &&
 385                                 addr && IN6_IS_ADDR_LINKLOCAL(addr) && iface->ra == RELAYD_SERVER)
 386                         ping6(addr, iface);
 387
 388                 // Address not specified or unrelated
 389                 if (!addr || IN6_IS_ADDR_LINKLOCAL(addr) ||
 390                                 IN6_IS_ADDR_MULTICAST(addr))
 391                         continue;
 392
 393                 // Check for states
 394                 bool add;
 395                 if (is_addr)
 396                         add = (nh->nlmsg_type == RTM_NEWADDR);
 397                 else
 398                         add = (nh->nlmsg_type == RTM_NEWNEIGH && (ndm->ndm_state &
 399                                 (NUD_REACHABLE | NUD_STALE | NUD_DELAY | NUD_PROBE
 400                                                 | NUD_PERMANENT | NUD_NOARP)));
 401
 402                 if (iface->ndp == RELAYD_RELAY && !is_route) {
 403                         // Replay change to all neighbor cache
 404                         struct {
 405                                 struct nlmsghdr nh;
 406                                 struct ndmsg ndm;
 407                                 struct nlattr nla_dst;
 408                                 struct in6_addr dst;
 409                         } req = {
 410                                 {sizeof(req), RTM_DELNEIGH, NLM_F_REQUEST,
 411                                                 ++rtnl_seqid, 0},
 412                                 {.ndm_family = AF_INET6, .ndm_flags = NTF_PROXY},
 413                                 {sizeof(struct nlattr) + sizeof(struct in6_addr), NDA_DST},
 414                                 *addr
 415                         };
 416
 417                         if (ndm->ndm_flags & NTF_PROXY) {
 418                                 // Dump & flush proxy entries
 419                                 if (nh->nlmsg_type == RTM_NEWNEIGH) {
 420                                         req.ndm.ndm_ifindex = iface->ifindex;
 421                                         send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 422                                         setup_route(addr, iface, false);
 423                                         dump_neigh = true;
 424                                 }
 425                         } else if (add) {
 426                                 struct interface *c;
 427                                 list_for_each_entry(c, &interfaces, head) {
 428                                         if (iface == c)
 429                                                 continue;
 430
 431                                         if (c->ndp == RELAYD_RELAY) {
 432                                                 req.nh.nlmsg_type = RTM_NEWNEIGH;
 433                                                 req.nh.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
 434
 435                                                 req.ndm.ndm_ifindex = c->ifindex;
 436                                                 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 437                                         } else { // Delete NDP cache from interfaces without relay
 438                                                 req.nh.nlmsg_type = RTM_DELNEIGH;
 439                                                 req.nh.nlmsg_flags &= ~(NLM_F_CREATE | NLM_F_REPLACE);
 440
 441                                                 req.ndm.ndm_ifindex = c->ifindex;
 442                                                 send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 443                                         }
 444                                 }
 445
 446                                 setup_route(addr, iface, true);
 447                         } else {
 448                                 if (nh->nlmsg_type == RTM_NEWNEIGH) {
 449                                         // might be locally originating
 450                                         if (!IN6_ARE_ADDR_EQUAL(&last_solicited, addr)) {
 451                                                 last_solicited = *addr;
 452
 453                                                 struct interface *c;
 454                                                 list_for_each_entry(c, &interfaces, head)
 455                                                         if (iface->ndp == RELAYD_RELAY && iface != c &&
 456                                                                         !c->external == false)
 457                                                                 ping6(addr, c);
 458                                         }
 459                                 } else {
 460                                         struct interface *c;
 461                                         list_for_each_entry(c, &interfaces, head) {
 462                                                 if (c->ndp == RELAYD_RELAY && iface != c) {
 463                                                         req.ndm.ndm_ifindex = c->ifindex;
 464                                                         send(rtnl_event.uloop.fd, &req, sizeof(req), MSG_DONTWAIT);
 465                                                 }
 466                                         }
 467                                         setup_route(addr, iface, false);
 468
 469                                         // also: dump to add proxies back in case it moved elsewhere
 470                                         dump_neigh = true;
 471                                 }
 472                         }
 473                 }
 474
 475                 if (is_addr) {
 476                         if (iface->ra == RELAYD_SERVER)
 477                                 raise(SIGUSR1); // Inform about a change in addresses
 478
 479                         if (iface->dhcpv6 == RELAYD_SERVER)
 480                                 iface->ia_reconf = true;
 481                 } else if (is_route) {
 482                         if (iface->ndp == RELAYD_RELAY && iface->master) {
 483                                 // Replay on-link route changes on all slave interfaces
 484                                 nh->nlmsg_flags = NLM_F_REQUEST;
 485
 486                                 if (nh->nlmsg_type == RTM_NEWROUTE)
 487                                         nh->nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
 488
 489                                 struct interface *c;
 490                                 list_for_each_entry(c, &interfaces, head) {
 491                                         if (c->ndp == RELAYD_RELAY && !c->master) {
 492                                                 *ifindex = c->ifindex;
 493                                                 *metric = (*metric & 0xffff) | (c->ifindex << 16);
 494                                                 send(rtnl_event.uloop.fd, nh, nh->nlmsg_len, MSG_DONTWAIT);
 495                                         }
 496                                 }
 497                         }
 498                 }
 499         }
 500
 501         if (dump_neigh)
 502                 dump_neigh_table(false);
 503 }