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_info *ri,
  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 || !ri || !to)
 105                 return MPLS_INVALID_LABEL;
 106
 107         remote_label = ri->extra ? ri->extra->label[0] : MPLS_INVALID_LABEL;
 108         from = ri->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_info *ri, int reg)
 124 {
 125         struct stream *s;
 126         struct prefix *p;
 127         int command;
 128         uint16_t flags = 0;
 129         size_t flags_pos = 0;
 130
 131         /* Check socket. */
 132         if (!zclient || zclient->sock < 0)
 133                 return;
 134
 135         p = &(rn->p);
 136         s = zclient->obuf;
 137         stream_reset(s);
 138         command = (reg) ? ZEBRA_FEC_REGISTER : ZEBRA_FEC_UNREGISTER;
 139         zclient_create_header(s, command, VRF_DEFAULT);
 140         flags_pos = stream_get_endp(s); /* save position of 'flags' */
 141         stream_putw(s, flags);          /* initial flags */
 142         stream_putw(s, PREFIX_FAMILY(p));
 143         stream_put_prefix(s, p);
 144         if (reg) {
 145                 assert(ri);
 146                 if (ri->attr->flag & ATTR_FLAG_BIT(BGP_ATTR_PREFIX_SID)) {
 147                         if (ri->attr->label_index != BGP_INVALID_LABEL_INDEX) {
 148                                 flags |= ZEBRA_FEC_REGISTER_LABEL_INDEX;
 149                                 stream_putl(s, ri->attr->label_index);
 150                         }
 151                 }
 152                 SET_FLAG(rn->flags, BGP_NODE_REGISTERED_FOR_LABEL);
 153         } else
 154                 UNSET_FLAG(rn->flags, BGP_NODE_REGISTERED_FOR_LABEL);
 155
 156         /* Set length and flags */
 157         stream_putw_at(s, 0, stream_get_endp(s));
 158
 159         /*
 160          * We only need to write new flags if this is a register
 161          */
 162         if (reg)
 163                 stream_putw_at(s, flags_pos, flags);
 164
 165         zclient_send_message(zclient);
 166 }
 167
 168 static int bgp_nlri_get_labels(struct peer *peer, uint8_t *pnt, uint8_t plen,
 169                                mpls_label_t *label)
 170 {
 171         uint8_t *data = pnt;
 172         uint8_t *lim = pnt + plen;
 173         uint8_t llen = 0;
 174         uint8_t label_depth = 0;
 175
 176         for (; data < lim; data += BGP_LABEL_BYTES) {
 177                 memcpy(label, data, BGP_LABEL_BYTES);
 178                 llen += BGP_LABEL_BYTES;
 179
 180                 bgp_set_valid_label(label);
 181                 label_depth += 1;
 182
 183                 if (bgp_is_withdraw_label(label) || label_bos(label))
 184                         break;
 185         }
 186
 187         /* If we RX multiple labels we will end up keeping only the last
 188          * one. We do not yet support a label stack greater than 1. */
 189         if (label_depth > 1)
 190                 zlog_info("%s rcvd UPDATE with label stack %d deep", peer->host,
 191                           label_depth);
 192
 193         if (!(bgp_is_withdraw_label(label) || label_bos(label)))
 194                 flog_warn(
 195                         BGP_WARN_INVALID_LABEL_STACK,
 196                         "%s rcvd UPDATE with invalid label stack - no bottom of stack",
 197                         peer->host);
 198
 199         return llen;
 200 }
 201
 202 int bgp_nlri_parse_label(struct peer *peer, struct attr *attr,
 203                          struct bgp_nlri *packet)
 204 {
 205         uint8_t *pnt;
 206         uint8_t *lim;
 207         struct prefix p;
 208         int psize = 0;
 209         int prefixlen;
 210         afi_t afi;
 211         safi_t safi;
 212         int addpath_encoded;
 213         uint32_t addpath_id;
 214         mpls_label_t label = MPLS_INVALID_LABEL;
 215         uint8_t llen;
 216
 217         pnt = packet->nlri;
 218         lim = pnt + packet->length;
 219         afi = packet->afi;
 220         safi = packet->safi;
 221         addpath_id = 0;
 222
 223         addpath_encoded =
 224                 (CHECK_FLAG(peer->af_cap[afi][safi], PEER_CAP_ADDPATH_AF_RX_ADV)
 225                  && CHECK_FLAG(peer->af_cap[afi][safi],
 226                                PEER_CAP_ADDPATH_AF_TX_RCV));
 227
 228         for (; pnt < lim; pnt += psize) {
 229                 /* Clear prefix structure. */
 230                 memset(&p, 0, sizeof(struct prefix));
 231
 232                 if (addpath_encoded) {
 233
 234                         /* When packet overflow occurs return immediately. */
 235                         if (pnt + BGP_ADDPATH_ID_LEN > lim)
 236                                 return -1;
 237
 238                         addpath_id = ntohl(*((uint32_t *)pnt));
 239                         pnt += BGP_ADDPATH_ID_LEN;
 240                 }
 241
 242                 /* Fetch prefix length. */
 243                 prefixlen = *pnt++;
 244                 p.family = afi2family(packet->afi);
 245                 psize = PSIZE(prefixlen);
 246
 247                 /* sanity check against packet data */
 248                 if ((pnt + psize) > lim) {
 249                         flog_err(
 250                                 BGP_ERR_UPDATE_RCV,
 251                                 "%s [Error] Update packet error / L-U (prefix length %d exceeds packet size %u)",
 252                                 peer->host, prefixlen, (uint)(lim - pnt));
 253                         return -1;
 254                 }
 255
 256                 /* Fill in the labels */
 257                 llen = bgp_nlri_get_labels(peer, pnt, psize, &label);
 258                 p.prefixlen = prefixlen - BSIZE(llen);
 259
 260                 /* There needs to be at least one label */
 261                 if (prefixlen < 24) {
 262                         flog_err(BGP_ERR_UPDATE_RCV,
 263                                   "%s [Error] Update packet error"
 264                                   " (wrong label length %d)",
 265                                   peer->host, prefixlen);
 266                         bgp_notify_send(peer, BGP_NOTIFY_UPDATE_ERR,
 267                                         BGP_NOTIFY_UPDATE_INVAL_NETWORK);
 268                         return -1;
 269                 }
 270
 271                 if ((afi == AFI_IP && p.prefixlen > 32)
 272                     || (afi == AFI_IP6 && p.prefixlen > 128))
 273                         return -1;
 274
 275                 /* Fetch prefix from NLRI packet */
 276                 memcpy(&p.u.prefix, pnt + llen, psize - llen);
 277
 278                 /* Check address. */
 279                 if (afi == AFI_IP && safi == SAFI_LABELED_UNICAST) {
 280                         if (IN_CLASSD(ntohl(p.u.prefix4.s_addr))) {
 281                                 /* From RFC4271 Section 6.3:
 282                                  *
 283                                  * If a prefix in the NLRI field is semantically
 284                                  * incorrect
 285                                  * (e.g., an unexpected multicast IP address),
 286                                  * an error SHOULD
 287                                  * be logged locally, and the prefix SHOULD be
 288                                  * ignored.
 289                                   */
 290                                 flog_err(
 291                                         BGP_ERR_UPDATE_RCV,
 292                                         "%s: IPv4 labeled-unicast NLRI is multicast address %s, ignoring",
 293                                         peer->host, inet_ntoa(p.u.prefix4));
 294                                 continue;
 295                         }
 296                 }
 297
 298                 /* Check address. */
 299                 if (afi == AFI_IP6 && safi == SAFI_LABELED_UNICAST) {
 300                         if (IN6_IS_ADDR_LINKLOCAL(&p.u.prefix6)) {
 301                                 char buf[BUFSIZ];
 302
 303                                 flog_err(
 304                                         BGP_ERR_UPDATE_RCV,
 305                                         "%s: IPv6 labeled-unicast NLRI is link-local address %s, ignoring",
 306                                         peer->host,
 307                                         inet_ntop(AF_INET6, &p.u.prefix6, buf,
 308                                                   BUFSIZ));
 309
 310                                 continue;
 311                         }
 312
 313                         if (IN6_IS_ADDR_MULTICAST(&p.u.prefix6)) {
 314                                 char buf[BUFSIZ];
 315
 316                                 flog_err(
 317                                         BGP_ERR_UPDATE_RCV,
 318                                         "%s: IPv6 unicast NLRI is multicast address %s, ignoring",
 319                                         peer->host,
 320                                         inet_ntop(AF_INET6, &p.u.prefix6, buf,
 321                                                   BUFSIZ));
 322
 323                                 continue;
 324                         }
 325                 }
 326
 327                 if (attr) {
 328                         bgp_update(peer, &p, addpath_id, attr, packet->afi,
 329                                    SAFI_UNICAST, ZEBRA_ROUTE_BGP,
 330                                    BGP_ROUTE_NORMAL, NULL, &label, 1, 0, NULL);
 331                 } else {
 332                         bgp_withdraw(peer, &p, addpath_id, attr, packet->afi,
 333                                      SAFI_UNICAST, ZEBRA_ROUTE_BGP,
 334                                      BGP_ROUTE_NORMAL, NULL, &label, 1, NULL);
 335                 }
 336         }
 337
 338         /* Packet length consistency check. */
 339         if (pnt != lim) {
 340                 flog_err(
 341                         BGP_ERR_UPDATE_RCV,
 342                         "%s [Error] Update packet error / L-U (%zu data remaining after parsing)",
 343                         peer->host, lim - pnt);
 344                 return -1;
 345         }
 346
 347         return 0;
 348 }