bgpd/bgp_label.c

   1 /* BGP carrying label information
   2  * Copyright (C) 2013 Cumulus Networks, Inc.
   3  *
   4  * This file is part of GNU Zebra.
   5  *
   6  * GNU Zebra is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the
   8  * Free Software Foundation; either version 2, or (at your option) any
   9  * later version.
  10  *
  11  * GNU Zebra is distributed in the hope that it will be useful, but
  12  * WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License along
  17  * with this program; see the file COPYING; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include <zebra.h>
  22
  23 #include "command.h"
  24 #include "thread.h"
  25 #include "prefix.h"
  26 #include "zclient.h"
  27 #include "stream.h"
  28 #include "network.h"
  29 #include "log.h"
  30 #include "memory.h"
  31 #include "nexthop.h"
  32 #include "mpls.h"
  33
  34 #include "bgpd/bgpd.h"
  35 #include "bgpd/bgp_table.h"
  36 #include "bgpd/bgp_route.h"
  37 #include "bgpd/bgp_attr.h"
  38 #include "bgpd/bgp_label.h"
  39 #include "bgpd/bgp_packet.h"
  40 #include "bgpd/bgp_debug.h"
  41 #include "bgpd/bgp_errors.h"
  42
  43 extern struct zclient *zclient;
  44
  45 int bgp_parse_fec_update(void)
  46 {
  47         struct stream *s;
  48         struct bgp_node *rn;
  49         struct bgp *bgp;
  50         struct bgp_table *table;
  51         struct prefix p;
  52         uint32_t label;
  53         afi_t afi;
  54         safi_t safi;
  55
  56         s = zclient->ibuf;
  57
  58         memset(&p, 0, sizeof(struct prefix));
  59         p.family = stream_getw(s);
  60         p.prefixlen = stream_getc(s);
  61         stream_get(p.u.val, s, PSIZE(p.prefixlen));
  62         label = stream_getl(s);
  63
  64         /* hack for the bgp instance & SAFI = have to send/receive it */
  65         afi = family2afi(p.family);
  66         safi = SAFI_UNICAST;
  67         bgp = bgp_get_default();
  68         if (!bgp) {
  69                 zlog_debug("no default bgp instance");
  70                 return -1;
  71         }
  72
  73         table = bgp->rib[afi][safi];
  74         if (!table) {
  75                 zlog_debug("no %u unicast table", p.family);
  76                 return -1;
  77         }
  78         rn = bgp_node_lookup(table, &p);
  79         if (!rn) {
  80                 zlog_debug("no node for the prefix");
  81                 return -1;
  82         }
  83
  84         /* treat it as implicit withdraw - the label is invalid */
  85         if (label == MPLS_INVALID_LABEL)
  86                 bgp_unset_valid_label(&rn->local_label);
  87         else {
  88                 label_ntop(label, 1, &rn->local_label);
  89                 bgp_set_valid_label(&rn->local_label);
  90         }
  91         SET_FLAG(rn->flags, BGP_NODE_LABEL_CHANGED);
  92         bgp_unlock_node(rn);
  93         bgp_process(bgp, rn, afi, safi);
  94         return 1;
  95 }
  96
  97 mpls_label_t bgp_adv_label(struct bgp_node *rn, struct bgp_path_info *pi,
  98                            struct peer *to, afi_t afi, safi_t safi)
  99 {
 100         struct peer *from;
 101         mpls_label_t remote_label;
 102         int reflect;
 103
 104         if (!rn || !pi || !to)
 105                 return MPLS_INVALID_LABEL;
 106
 107         remote_label = pi->extra ? pi->extra->label[0] : MPLS_INVALID_LABEL;
 108         from = pi->peer;
 109         reflect =
 110                 ((from->sort == BGP_PEER_IBGP) && (to->sort == BGP_PEER_IBGP));
 111
 112         if (reflect
 113             && !CHECK_FLAG(to->af_flags[afi][safi],
 114                            PEER_FLAG_FORCE_NEXTHOP_SELF))
 115                 return remote_label;
 116
 117         if (CHECK_FLAG(to->af_flags[afi][safi], PEER_FLAG_NEXTHOP_UNCHANGED))
 118                 return remote_label;
 119
 120         return rn->local_label;
 121 }
 122
 123 void bgp_reg_dereg_for_label(struct bgp_node *rn, struct bgp_path_info *pi,
 124                              int reg)
 125 {
 126         struct stream *s;
 127         struct prefix *p;
 128         int command;
 129         uint16_t flags = 0;
 130         size_t flags_pos = 0;
 131
 132         /* Check socket. */
 133         if (!zclient || zclient->sock < 0)
 134                 return;
 135
 136         p = &(rn->p);
 137         s = zclient->obuf;
 138         stream_reset(s);
 139         command = (reg) ? ZEBRA_FEC_REGISTER : ZEBRA_FEC_UNREGISTER;
 140         zclient_create_header(s, command, VRF_DEFAULT);
 141         flags_pos = stream_get_endp(s); /* save position of 'flags' */
 142         stream_putw(s, flags);          /* initial flags */
 143         stream_putw(s, PREFIX_FAMILY(p));
 144         stream_put_prefix(s, p);
 145         if (reg) {
 146                 assert(pi);
 147                 if (pi->attr->flag & ATTR_FLAG_BIT(BGP_ATTR_PREFIX_SID)) {
 148                         if (pi->attr->label_index != BGP_INVALID_LABEL_INDEX) {
 149                                 flags |= ZEBRA_FEC_REGISTER_LABEL_INDEX;
 150                                 stream_putl(s, pi->attr->label_index);
 151                         }
 152                 }
 153                 SET_FLAG(rn->flags, BGP_NODE_REGISTERED_FOR_LABEL);
 154         } else
 155                 UNSET_FLAG(rn->flags, BGP_NODE_REGISTERED_FOR_LABEL);
 156
 157         /* Set length and flags */
 158         stream_putw_at(s, 0, stream_get_endp(s));
 159
 160         /*
 161          * We only need to write new flags if this is a register
 162          */
 163         if (reg)
 164                 stream_putw_at(s, flags_pos, flags);
 165
 166         zclient_send_message(zclient);
 167 }
 168
 169 static int bgp_nlri_get_labels(struct peer *peer, uint8_t *pnt, uint8_t plen,
 170                                mpls_label_t *label)
 171 {
 172         uint8_t *data = pnt;
 173         uint8_t *lim = pnt + plen;
 174         uint8_t llen = 0;
 175         uint8_t label_depth = 0;
 176
 177         for (; data < lim; data += BGP_LABEL_BYTES) {
 178                 memcpy(label, data, BGP_LABEL_BYTES);
 179                 llen += BGP_LABEL_BYTES;
 180
 181                 bgp_set_valid_label(label);
 182                 label_depth += 1;
 183
 184                 if (bgp_is_withdraw_label(label) || label_bos(label))
 185                         break;
 186         }
 187
 188         /* If we RX multiple labels we will end up keeping only the last
 189          * one. We do not yet support a label stack greater than 1. */
 190         if (label_depth > 1)
 191                 zlog_info("%s rcvd UPDATE with label stack %d deep", peer->host,
 192                           label_depth);
 193
 194         if (!(bgp_is_withdraw_label(label) || label_bos(label)))
 195                 flog_warn(
 196                         EC_BGP_INVALID_LABEL_STACK,
 197                         "%s rcvd UPDATE with invalid label stack - no bottom of stack",
 198                         peer->host);
 199
 200         return llen;
 201 }
 202
 203 int bgp_nlri_parse_label(struct peer *peer, struct attr *attr,
 204                          struct bgp_nlri *packet)
 205 {
 206         uint8_t *pnt;
 207         uint8_t *lim;
 208         struct prefix p;
 209         int psize = 0;
 210         int prefixlen;
 211         afi_t afi;
 212         safi_t safi;
 213         int addpath_encoded;
 214         uint32_t addpath_id;
 215         mpls_label_t label = MPLS_INVALID_LABEL;
 216         uint8_t llen;
 217
 218         pnt = packet->nlri;
 219         lim = pnt + packet->length;
 220         afi = packet->afi;
 221         safi = packet->safi;
 222         addpath_id = 0;
 223
 224         addpath_encoded =
 225                 (CHECK_FLAG(peer->af_cap[afi][safi], PEER_CAP_ADDPATH_AF_RX_ADV)
 226                  && CHECK_FLAG(peer->af_cap[afi][safi],
 227                                PEER_CAP_ADDPATH_AF_TX_RCV));
 228
 229         for (; pnt < lim; pnt += psize) {
 230                 /* Clear prefix structure. */
 231                 memset(&p, 0, sizeof(struct prefix));
 232
 233                 if (addpath_encoded) {
 234
 235                         /* When packet overflow occurs return immediately. */
 236                         if (pnt + BGP_ADDPATH_ID_LEN > lim)
 237                                 return -1;
 238
 239                         addpath_id = ntohl(*((uint32_t *)pnt));
 240                         pnt += BGP_ADDPATH_ID_LEN;
 241                 }
 242
 243                 /* Fetch prefix length. */
 244                 prefixlen = *pnt++;
 245                 p.family = afi2family(packet->afi);
 246                 psize = PSIZE(prefixlen);
 247
 248                 /* sanity check against packet data */
 249                 if ((pnt + psize) > lim) {
 250                         flog_err(
 251                                 EC_BGP_UPDATE_RCV,
 252                                 "%s [Error] Update packet error / L-U (prefix length %d exceeds packet size %u)",
 253                                 peer->host, prefixlen, (uint)(lim - pnt));
 254                         return -1;
 255                 }
 256
 257                 /* Fill in the labels */
 258                 llen = bgp_nlri_get_labels(peer, pnt, psize, &label);
 259                 p.prefixlen = prefixlen - BSIZE(llen);
 260
 261                 /* There needs to be at least one label */
 262                 if (prefixlen < 24) {
 263                         flog_err(EC_BGP_UPDATE_RCV,
 264                                  "%s [Error] Update packet error"
 265                                  " (wrong label length %d)",
 266                                  peer->host, prefixlen);
 267                         bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR,
 268                                         BGP_NOTIFY_UPDATE_INVAL_NETWORK);
 269                         return -1;
 270                 }
 271
 272                 if ((afi == AFI_IP && p.prefixlen > 32)
 273                     || (afi == AFI_IP6 && p.prefixlen > 128))
 274                         return -1;
 275
 276                 /* Fetch prefix from NLRI packet */
 277                 memcpy(&p.u.prefix, pnt + llen, psize - llen);
 278
 279                 /* Check address. */
 280                 if (afi == AFI_IP && safi == SAFI_LABELED_UNICAST) {
 281                         if (IN_CLASSD(ntohl(p.u.prefix4.s_addr))) {
 282                                 /* From RFC4271 Section 6.3:
 283                                  *
 284                                  * If a prefix in the NLRI field is semantically
 285                                  * incorrect
 286                                  * (e.g., an unexpected multicast IP address),
 287                                  * an error SHOULD
 288                                  * be logged locally, and the prefix SHOULD be
 289                                  * ignored.
 290                                   */
 291                                 flog_err(
 292                                         EC_BGP_UPDATE_RCV,
 293                                         "%s: IPv4 labeled-unicast NLRI is multicast address %s, ignoring",
 294                                         peer->host, inet_ntoa(p.u.prefix4));
 295                                 continue;
 296                         }
 297                 }
 298
 299                 /* Check address. */
 300                 if (afi == AFI_IP6 && safi == SAFI_LABELED_UNICAST) {
 301                         if (IN6_IS_ADDR_LINKLOCAL(&p.u.prefix6)) {
 302                                 char buf[BUFSIZ];
 303
 304                                 flog_err(
 305                                         EC_BGP_UPDATE_RCV,
 306                                         "%s: IPv6 labeled-unicast NLRI is link-local address %s, ignoring",
 307                                         peer->host,
 308                                         inet_ntop(AF_INET6, &p.u.prefix6, buf,
 309                                                   BUFSIZ));
 310
 311                                 continue;
 312                         }
 313
 314                         if (IN6_IS_ADDR_MULTICAST(&p.u.prefix6)) {
 315                                 char buf[BUFSIZ];
 316
 317                                 flog_err(
 318                                         EC_BGP_UPDATE_RCV,
 319                                         "%s: IPv6 unicast NLRI is multicast address %s, ignoring",
 320                                         peer->host,
 321                                         inet_ntop(AF_INET6, &p.u.prefix6, buf,
 322                                                   BUFSIZ));
 323
 324                                 continue;
 325                         }
 326                 }
 327
 328                 if (attr) {
 329                         bgp_update(peer, &p, addpath_id, attr, packet->afi,
 330                                    SAFI_UNICAST, ZEBRA_ROUTE_BGP,
 331                                    BGP_ROUTE_NORMAL, NULL, &label, 1, 0, NULL);
 332                 } else {
 333                         bgp_withdraw(peer, &p, addpath_id, attr, packet->afi,
 334                                      SAFI_UNICAST, ZEBRA_ROUTE_BGP,
 335                                      BGP_ROUTE_NORMAL, NULL, &label, 1, NULL);
 336                 }
 337         }
 338
 339         /* Packet length consistency check. */
 340         if (pnt != lim) {
 341                 flog_err(
 342                         EC_BGP_UPDATE_RCV,
 343                         "%s [Error] Update packet error / L-U (%zu data remaining after parsing)",
 344                         peer->host, lim - pnt);
 345                 return -1;
 346         }
 347
 348         return 0;
 349 }