bgpd/bgp_nht.c

   1 /* BGP Nexthop tracking
   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 "vrf.h"
  33 #include "filter.h"
  34
  35 #include "bgpd/bgpd.h"
  36 #include "bgpd/bgp_table.h"
  37 #include "bgpd/bgp_route.h"
  38 #include "bgpd/bgp_attr.h"
  39 #include "bgpd/bgp_nexthop.h"
  40 #include "bgpd/bgp_debug.h"
  41 #include "bgpd/bgp_errors.h"
  42 #include "bgpd/bgp_nht.h"
  43 #include "bgpd/bgp_fsm.h"
  44 #include "bgpd/bgp_zebra.h"
  45 #include "bgpd/bgp_flowspec_util.h"
  46
  47 extern struct zclient *zclient;
  48
  49 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
  50                                int is_bgp_static_route);
  51 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
  52                                  int is_bgp_static_route);
  53 static void evaluate_paths(struct bgp_nexthop_cache *bnc);
  54 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p);
  55
  56 static int bgp_isvalid_nexthop(struct bgp_nexthop_cache *bnc)
  57 {
  58         return (bgp_zebra_num_connects() == 0
  59                 || (bnc && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID)));
  60 }
  61
  62 static int bgp_isvalid_labeled_nexthop(struct bgp_nexthop_cache *bnc)
  63 {
  64         return (bgp_zebra_num_connects() == 0
  65                 || (bnc && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_LABELED_VALID)));
  66 }
  67
  68 int bgp_find_nexthop(struct bgp_path_info *path, int connected)
  69 {
  70         struct bgp_nexthop_cache *bnc = path->nexthop;
  71
  72         if (!bnc)
  73                 return 0;
  74
  75         /*
  76          * We are cheating here.  Views have no associated underlying
  77          * ability to detect nexthops.  So when we have a view
  78          * just tell everyone the nexthop is valid
  79          */
  80         if (path->peer && path->peer->bgp->inst_type == BGP_INSTANCE_TYPE_VIEW)
  81                 return 1;
  82
  83         if (connected && !(CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)))
  84                 return 0;
  85
  86         return (bgp_isvalid_nexthop(bnc));
  87 }
  88
  89 static void bgp_unlink_nexthop_check(struct bgp_nexthop_cache *bnc)
  90 {
  91         if (LIST_EMPTY(&(bnc->paths)) && !bnc->nht_info) {
  92                 if (BGP_DEBUG(nht, NHT)) {
  93                         char buf[PREFIX2STR_BUFFER];
  94                         zlog_debug("bgp_unlink_nexthop: freeing bnc %s",
  95                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER));
  96                 }
  97                 unregister_zebra_rnh(bnc,
  98                                      CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE));
  99                 bgp_node_set_bgp_nexthop_info(bnc->node, NULL);
 100                 bgp_unlock_node(bnc->node);
 101                 bnc->node = NULL;
 102                 bnc_free(bnc);
 103         }
 104 }
 105
 106 void bgp_unlink_nexthop(struct bgp_path_info *path)
 107 {
 108         struct bgp_nexthop_cache *bnc = path->nexthop;
 109
 110         if (!bnc)
 111                 return;
 112
 113         path_nh_map(path, NULL, false);
 114
 115         bgp_unlink_nexthop_check(bnc);
 116 }
 117
 118 void bgp_unlink_nexthop_by_peer(struct peer *peer)
 119 {
 120         struct prefix p;
 121         struct bgp_node *rn;
 122         struct bgp_nexthop_cache *bnc;
 123         afi_t afi = family2afi(peer->su.sa.sa_family);
 124
 125         if (!sockunion2hostprefix(&peer->su, &p))
 126                 return;
 127
 128         rn = bgp_node_get(peer->bgp->nexthop_cache_table[afi], &p);
 129
 130         bnc = bgp_node_get_bgp_nexthop_info(rn);
 131         if (!bnc)
 132                 return;
 133
 134         /* cleanup the peer reference */
 135         bnc->nht_info = NULL;
 136
 137         bgp_unlink_nexthop_check(bnc);
 138 }
 139
 140 /*
 141  * A route and its nexthop might belong to different VRFs. Therefore,
 142  * we need both the bgp_route and bgp_nexthop pointers.
 143  */
 144 int bgp_find_or_add_nexthop(struct bgp *bgp_route, struct bgp *bgp_nexthop,
 145                             afi_t afi, struct bgp_path_info *pi,
 146                             struct peer *peer, int connected)
 147 {
 148         struct bgp_node *rn;
 149         struct bgp_nexthop_cache *bnc;
 150         struct prefix p;
 151         int is_bgp_static_route = 0;
 152
 153         if (pi) {
 154                 is_bgp_static_route = ((pi->type == ZEBRA_ROUTE_BGP)
 155                                        && (pi->sub_type == BGP_ROUTE_STATIC))
 156                                               ? 1
 157                                               : 0;
 158
 159                 /* Since Extended Next-hop Encoding (RFC5549) support, we want
 160                    to derive
 161                    address-family from the next-hop. */
 162                 if (!is_bgp_static_route)
 163                         afi = BGP_ATTR_NEXTHOP_AFI_IP6(pi->attr) ? AFI_IP6
 164                                                                  : AFI_IP;
 165
 166                 /* This will return TRUE if the global IPv6 NH is a link local
 167                  * addr */
 168                 if (make_prefix(afi, pi, &p) < 0)
 169                         return 1;
 170         } else if (peer) {
 171                 if (!sockunion2hostprefix(&peer->su, &p)) {
 172                         if (BGP_DEBUG(nht, NHT)) {
 173                                 zlog_debug(
 174                                         "%s: Attempting to register with unknown AFI %d (not %d or %d)",
 175                                         __FUNCTION__, afi, AFI_IP, AFI_IP6);
 176                         }
 177                         return 0;
 178                 }
 179         } else
 180                 return 0;
 181
 182         if (is_bgp_static_route)
 183                 rn = bgp_node_get(bgp_nexthop->import_check_table[afi], &p);
 184         else
 185                 rn = bgp_node_get(bgp_nexthop->nexthop_cache_table[afi], &p);
 186
 187         bnc = bgp_node_get_bgp_nexthop_info(rn);
 188         if (!bnc) {
 189                 bnc = bnc_new();
 190                 bgp_node_set_bgp_nexthop_info(rn, bnc);
 191                 bnc->node = rn;
 192                 bnc->bgp = bgp_nexthop;
 193                 bgp_lock_node(rn);
 194                 if (BGP_DEBUG(nht, NHT)) {
 195                         char buf[PREFIX2STR_BUFFER];
 196
 197                         zlog_debug("Allocated bnc %s peer %p",
 198                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER), peer);
 199                 }
 200         }
 201
 202         bgp_unlock_node(rn);
 203         if (is_bgp_static_route) {
 204                 SET_FLAG(bnc->flags, BGP_STATIC_ROUTE);
 205
 206                 /* If we're toggling the type, re-register */
 207                 if ((bgp_flag_check(bgp_route, BGP_FLAG_IMPORT_CHECK))
 208                     && !CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)) {
 209                         SET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 210                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 211                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 212                 } else if ((!bgp_flag_check(bgp_route, BGP_FLAG_IMPORT_CHECK))
 213                            && CHECK_FLAG(bnc->flags,
 214                                          BGP_STATIC_ROUTE_EXACT_MATCH)) {
 215                         UNSET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 216                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 217                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 218                 }
 219         }
 220         /* When nexthop is already known, but now requires 'connected'
 221          * resolution,
 222          * re-register it. The reverse scenario where the nexthop currently
 223          * requires
 224          * 'connected' resolution does not need a re-register (i.e., we treat
 225          * 'connected-required' as an override) except in the scenario where
 226          * this
 227          * is actually a case of tracking a peer for connectivity (e.g., after
 228          * disable connected-check).
 229          * NOTE: We don't track the number of paths separately for 'connected-
 230          * required' vs 'connected-not-required' as this change is not a common
 231          * scenario.
 232          */
 233         else if (connected && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 234                 SET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 235                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 236                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 237         } else if (peer && !connected
 238                    && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 239                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 240                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 241                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 242         }
 243         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW) {
 244                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 245                 SET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 246         } else if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED) &&
 247                    !is_default_host_route(&bnc->node->p))
 248                 register_zebra_rnh(bnc, is_bgp_static_route);
 249
 250         if (pi && pi->nexthop != bnc) {
 251                 /* Unlink from existing nexthop cache, if any. This will also
 252                  * free
 253                  * the nexthop cache entry, if appropriate.
 254                  */
 255                 bgp_unlink_nexthop(pi);
 256
 257                 /* updates NHT pi list reference */
 258                 path_nh_map(pi, bnc, true);
 259
 260                 if (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID) && bnc->metric)
 261                         (bgp_path_info_extra_get(pi))->igpmetric = bnc->metric;
 262                 else if (pi->extra)
 263                         pi->extra->igpmetric = 0;
 264         } else if (peer)
 265                 bnc->nht_info = (void *)peer; /* NHT peer reference */
 266
 267         /*
 268          * We are cheating here.  Views have no associated underlying
 269          * ability to detect nexthops.  So when we have a view
 270          * just tell everyone the nexthop is valid
 271          */
 272         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW)
 273                 return 1;
 274         else
 275                 return (bgp_isvalid_nexthop(bnc));
 276 }
 277
 278 void bgp_delete_connected_nexthop(afi_t afi, struct peer *peer)
 279 {
 280         struct bgp_node *rn;
 281         struct bgp_nexthop_cache *bnc;
 282         struct prefix p;
 283
 284         if (!peer)
 285                 return;
 286
 287         if (!sockunion2hostprefix(&peer->su, &p))
 288                 return;
 289
 290         rn = bgp_node_lookup(
 291                 peer->bgp->nexthop_cache_table[family2afi(p.family)], &p);
 292         if (!rn) {
 293                 if (BGP_DEBUG(nht, NHT))
 294                         zlog_debug("Cannot find connected NHT node for peer %s",
 295                                    peer->host);
 296                 return;
 297         }
 298
 299         bnc = bgp_node_get_bgp_nexthop_info(rn);
 300         if (!bnc) {
 301                 if (BGP_DEBUG(nht, NHT))
 302                         zlog_debug("Cannot find connected NHT node for peer %s on route_node as expected",
 303                                    peer->host);
 304                 bgp_unlock_node(rn);
 305                 return;
 306         }
 307         bgp_unlock_node(rn);
 308
 309         if (bnc->nht_info != peer) {
 310                 if (BGP_DEBUG(nht, NHT))
 311                         zlog_debug(
 312                                 "Connected NHT %p node for peer %s points to %p",
 313                                 bnc, peer->host, bnc->nht_info);
 314                 return;
 315         }
 316
 317         bnc->nht_info = NULL;
 318
 319         if (LIST_EMPTY(&(bnc->paths))) {
 320                 if (BGP_DEBUG(nht, NHT))
 321                         zlog_debug("Freeing connected NHT node %p for peer %s",
 322                                    bnc, peer->host);
 323                 unregister_zebra_rnh(bnc, 0);
 324                 bgp_node_set_bgp_nexthop_info(bnc->node, NULL);
 325                 bgp_unlock_node(bnc->node);
 326                 bnc_free(bnc);
 327         }
 328 }
 329
 330 void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 331 {
 332         struct bgp_node *rn = NULL;
 333         struct bgp_nexthop_cache *bnc;
 334         struct nexthop *nexthop;
 335         struct nexthop *oldnh;
 336         struct nexthop *nhlist_head = NULL;
 337         struct nexthop *nhlist_tail = NULL;
 338         int i;
 339         struct bgp *bgp;
 340         struct zapi_route nhr;
 341
 342         bgp = bgp_lookup_by_vrf_id(vrf_id);
 343         if (!bgp) {
 344                 flog_err(
 345                         EC_BGP_NH_UPD,
 346                         "parse nexthop update: instance not found for vrf_id %u",
 347                         vrf_id);
 348                 return;
 349         }
 350
 351         if (!zapi_nexthop_update_decode(zclient->ibuf, &nhr)) {
 352                 if (BGP_DEBUG(nht, NHT))
 353                         zlog_debug("%s: Failure to decode nexthop update",
 354                                    __PRETTY_FUNCTION__);
 355                 return;
 356         }
 357
 358         if (command == ZEBRA_NEXTHOP_UPDATE)
 359                 rn = bgp_node_lookup(
 360                         bgp->nexthop_cache_table[family2afi(nhr.prefix.family)],
 361                         &nhr.prefix);
 362         else if (command == ZEBRA_IMPORT_CHECK_UPDATE)
 363                 rn = bgp_node_lookup(
 364                         bgp->import_check_table[family2afi(nhr.prefix.family)],
 365                         &nhr.prefix);
 366
 367         if (!rn) {
 368                 if (BGP_DEBUG(nht, NHT)) {
 369                         char buf[PREFIX2STR_BUFFER];
 370                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 371                         zlog_debug("parse nexthop update(%s): rn not found",
 372                                    buf);
 373                 }
 374                 return;
 375         }
 376
 377         bnc = bgp_node_get_bgp_nexthop_info(rn);
 378         if (!bnc) {
 379                 if (BGP_DEBUG(nht, NHT)) {
 380                         char buf[PREFIX2STR_BUFFER];
 381
 382                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 383                         zlog_debug("parse nexthop update(%s): bnc node info not found",
 384                                    buf);
 385                 }
 386                 bgp_unlock_node(rn);
 387                 return;
 388         }
 389
 390         bgp_unlock_node(rn);
 391         bnc->last_update = bgp_clock();
 392         bnc->change_flags = 0;
 393
 394         /* debug print the input */
 395         if (BGP_DEBUG(nht, NHT)) {
 396                 char buf[PREFIX2STR_BUFFER];
 397                 prefix2str(&nhr.prefix, buf, sizeof(buf));
 398                 zlog_debug(
 399                         "%u: Rcvd NH update %s - metric %d/%d #nhops %d/%d flags 0x%x",
 400                         vrf_id, buf, nhr.metric, bnc->metric, nhr.nexthop_num,
 401                         bnc->nexthop_num, bnc->flags);
 402         }
 403
 404         if (nhr.metric != bnc->metric)
 405                 bnc->change_flags |= BGP_NEXTHOP_METRIC_CHANGED;
 406
 407         if (nhr.nexthop_num != bnc->nexthop_num)
 408                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 409
 410         if (nhr.nexthop_num) {
 411                 struct peer *peer = bnc->nht_info;
 412
 413                 /* notify bgp fsm if nbr ip goes from invalid->valid */
 414                 if (!bnc->nexthop_num)
 415                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 416
 417                 bnc->flags |= BGP_NEXTHOP_VALID;
 418                 bnc->metric = nhr.metric;
 419                 bnc->nexthop_num = nhr.nexthop_num;
 420
 421                 bnc->flags &= ~BGP_NEXTHOP_LABELED_VALID; /* check below */
 422
 423                 for (i = 0; i < nhr.nexthop_num; i++) {
 424                         int num_labels = 0;
 425
 426                         nexthop = nexthop_from_zapi_nexthop(&nhr.nexthops[i]);
 427
 428                         /*
 429                          * Turn on RA for the v6 nexthops
 430                          * we receive from bgp.  This is to allow us
 431                          * to work with v4 routing over v6 nexthops
 432                          */
 433                         if (peer && !peer->ifp
 434                             && CHECK_FLAG(peer->flags,
 435                                           PEER_FLAG_CAPABILITY_ENHE)
 436                             && nhr.prefix.family == AF_INET6) {
 437                                 struct interface *ifp;
 438
 439                                 ifp = if_lookup_by_index(nexthop->ifindex,
 440                                                          nexthop->vrf_id);
 441                                 zclient_send_interface_radv_req(
 442                                         zclient, nexthop->vrf_id, ifp, true,
 443                                         BGP_UNNUM_DEFAULT_RA_INTERVAL);
 444                         }
 445                         /* There is at least one label-switched path */
 446                         if (nexthop->nh_label &&
 447                                 nexthop->nh_label->num_labels) {
 448
 449                                 bnc->flags |= BGP_NEXTHOP_LABELED_VALID;
 450                                 num_labels = nexthop->nh_label->num_labels;
 451                         }
 452
 453                         if (BGP_DEBUG(nht, NHT)) {
 454                                 char buf[NEXTHOP_STRLEN];
 455                                 zlog_debug(
 456                                         "    nhop via %s (%d labels)",
 457                                         nexthop2str(nexthop, buf, sizeof(buf)),
 458                                         num_labels);
 459                         }
 460
 461                         if (nhlist_tail) {
 462                                 nhlist_tail->next = nexthop;
 463                                 nhlist_tail = nexthop;
 464                         } else {
 465                                 nhlist_tail = nexthop;
 466                                 nhlist_head = nexthop;
 467                         }
 468
 469                         /* No need to evaluate the nexthop if we have already
 470                          * determined
 471                          * that there has been a change.
 472                          */
 473                         if (bnc->change_flags & BGP_NEXTHOP_CHANGED)
 474                                 continue;
 475
 476                         for (oldnh = bnc->nexthop; oldnh; oldnh = oldnh->next)
 477                                 if (nexthop_same_no_recurse(oldnh, nexthop) &&
 478                                     nexthop_labels_match(oldnh, nexthop))
 479                                         break;
 480
 481                         if (!oldnh)
 482                                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 483                 }
 484                 bnc_nexthop_free(bnc);
 485                 bnc->nexthop = nhlist_head;
 486         } else {
 487                 bnc->flags &= ~BGP_NEXTHOP_VALID;
 488                 bnc->nexthop_num = nhr.nexthop_num;
 489
 490                 /* notify bgp fsm if nbr ip goes from valid->invalid */
 491                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 492
 493                 bnc_nexthop_free(bnc);
 494                 bnc->nexthop = NULL;
 495         }
 496
 497         evaluate_paths(bnc);
 498 }
 499
 500 /*
 501  * Cleanup nexthop registration and status information for BGP nexthops
 502  * pertaining to this VRF. This is invoked upon VRF deletion.
 503  */
 504 void bgp_cleanup_nexthops(struct bgp *bgp)
 505 {
 506         afi_t afi;
 507         struct bgp_node *rn;
 508         struct bgp_nexthop_cache *bnc;
 509
 510         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 511                 if (!bgp->nexthop_cache_table[afi])
 512                         continue;
 513
 514                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 515                      rn = bgp_route_next(rn)) {
 516                         bnc = bgp_node_get_bgp_nexthop_info(rn);
 517                         if (!bnc)
 518                                 continue;
 519
 520                         /* Clear relevant flags. */
 521                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 522                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 523                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 524                 }
 525         }
 526 }
 527
 528 /**
 529  * make_prefix - make a prefix structure from the path (essentially
 530  * path's node.
 531  */
 532 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p)
 533 {
 534
 535         int is_bgp_static = ((pi->type == ZEBRA_ROUTE_BGP)
 536                              && (pi->sub_type == BGP_ROUTE_STATIC))
 537                                     ? 1
 538                                     : 0;
 539         struct bgp_node *net = pi->net;
 540         struct prefix *p_orig = &net->p;
 541
 542         if (p_orig->family == AF_FLOWSPEC) {
 543                 if (!pi->peer)
 544                         return -1;
 545                 return bgp_flowspec_get_first_nh(pi->peer->bgp,
 546                                                  pi, p);
 547         }
 548         memset(p, 0, sizeof(struct prefix));
 549         switch (afi) {
 550         case AFI_IP:
 551                 p->family = AF_INET;
 552                 if (is_bgp_static) {
 553                         p->u.prefix4 = pi->net->p.u.prefix4;
 554                         p->prefixlen = pi->net->p.prefixlen;
 555                 } else {
 556                         p->u.prefix4 = pi->attr->nexthop;
 557                         p->prefixlen = IPV4_MAX_BITLEN;
 558                 }
 559                 break;
 560         case AFI_IP6:
 561                 p->family = AF_INET6;
 562
 563                 if (is_bgp_static) {
 564                         p->u.prefix6 = pi->net->p.u.prefix6;
 565                         p->prefixlen = pi->net->p.prefixlen;
 566                 } else {
 567                         p->u.prefix6 = pi->attr->mp_nexthop_global;
 568                         p->prefixlen = IPV6_MAX_BITLEN;
 569                 }
 570                 break;
 571         default:
 572                 if (BGP_DEBUG(nht, NHT)) {
 573                         zlog_debug(
 574                                 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
 575                                 __FUNCTION__, afi, AFI_IP, AFI_IP6);
 576                 }
 577                 break;
 578         }
 579         return 0;
 580 }
 581
 582 /**
 583  * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
 584  *   command to Zebra.
 585  * ARGUMENTS:
 586  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 587  *   int command -- command to send to zebra
 588  * RETURNS:
 589  *   void.
 590  */
 591 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache *bnc, int command)
 592 {
 593         struct prefix *p;
 594         bool exact_match = false;
 595         int ret;
 596
 597         if (!zclient)
 598                 return;
 599
 600         /* Don't try to register if Zebra doesn't know of this instance. */
 601         if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc->bgp)) {
 602                 if (BGP_DEBUG(zebra, ZEBRA))
 603                         zlog_debug("%s: No zebra instance to talk to, not installing NHT entry",
 604                                    __PRETTY_FUNCTION__);
 605                 return;
 606         }
 607
 608         if (!bgp_zebra_num_connects()) {
 609                 if (BGP_DEBUG(zebra, ZEBRA))
 610                         zlog_debug("%s: We have not connected yet, cannot send nexthops",
 611                                    __PRETTY_FUNCTION__);
 612         }
 613         p = &(bnc->node->p);
 614         if ((command == ZEBRA_NEXTHOP_REGISTER
 615              || command == ZEBRA_IMPORT_ROUTE_REGISTER)
 616             && (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)
 617                 || CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)))
 618                 exact_match = true;
 619
 620         if (BGP_DEBUG(zebra, ZEBRA)) {
 621                 char buf[PREFIX2STR_BUFFER];
 622
 623                 prefix2str(p, buf, PREFIX2STR_BUFFER);
 624                 zlog_debug("%s: sending cmd %s for %s (vrf %s)",
 625                         __func__, zserv_command_string(command), buf,
 626                         bnc->bgp->name);
 627         }
 628
 629         ret = zclient_send_rnh(zclient, command, p, exact_match,
 630                                bnc->bgp->vrf_id);
 631         /* TBD: handle the failure */
 632         if (ret < 0)
 633                 flog_warn(EC_BGP_ZEBRA_SEND,
 634                           "sendmsg_nexthop: zclient_send_message() failed");
 635
 636         if ((command == ZEBRA_NEXTHOP_REGISTER)
 637             || (command == ZEBRA_IMPORT_ROUTE_REGISTER))
 638                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 639         else if ((command == ZEBRA_NEXTHOP_UNREGISTER)
 640                  || (command == ZEBRA_IMPORT_ROUTE_UNREGISTER))
 641                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 642         return;
 643 }
 644
 645 /**
 646  * register_zebra_rnh - register a NH/route with Zebra for notification
 647  *    when the route or the route to the nexthop changes.
 648  * ARGUMENTS:
 649  *   struct bgp_nexthop_cache *bnc
 650  * RETURNS:
 651  *   void.
 652  */
 653 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
 654                                int is_bgp_import_route)
 655 {
 656         /* Check if we have already registered */
 657         if (bnc->flags & BGP_NEXTHOP_REGISTERED)
 658                 return;
 659         if (is_bgp_import_route)
 660                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_REGISTER);
 661         else
 662                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_REGISTER);
 663 }
 664
 665 /**
 666  * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
 667  * ARGUMENTS:
 668  *   struct bgp_nexthop_cache *bnc
 669  * RETURNS:
 670  *   void.
 671  */
 672 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
 673                                  int is_bgp_import_route)
 674 {
 675         /* Check if we have already registered */
 676         if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 677                 return;
 678
 679         if (is_bgp_import_route)
 680                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_UNREGISTER);
 681         else
 682                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_UNREGISTER);
 683 }
 684
 685 /**
 686  * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
 687  * ARGUMENTS:
 688  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 689  * RETURNS:
 690  *   void.
 691  */
 692 static void evaluate_paths(struct bgp_nexthop_cache *bnc)
 693 {
 694         struct bgp_node *rn;
 695         struct bgp_path_info *path;
 696         int afi;
 697         struct peer *peer = (struct peer *)bnc->nht_info;
 698         struct bgp_table *table;
 699         safi_t safi;
 700         struct bgp *bgp_path;
 701
 702         if (BGP_DEBUG(nht, NHT)) {
 703                 char buf[PREFIX2STR_BUFFER];
 704                 bnc_str(bnc, buf, PREFIX2STR_BUFFER);
 705                 zlog_debug(
 706                         "NH update for %s - flags 0x%x chgflags 0x%x - evaluate paths",
 707                         buf, bnc->flags, bnc->change_flags);
 708         }
 709
 710         LIST_FOREACH (path, &(bnc->paths), nh_thread) {
 711                 if (!(path->type == ZEBRA_ROUTE_BGP
 712                       && ((path->sub_type == BGP_ROUTE_NORMAL)
 713                           || (path->sub_type == BGP_ROUTE_STATIC)
 714                           || (path->sub_type == BGP_ROUTE_IMPORTED))))
 715                         continue;
 716
 717                 rn = path->net;
 718                 assert(rn && bgp_node_table(rn));
 719                 afi = family2afi(rn->p.family);
 720                 table = bgp_node_table(rn);
 721                 safi = table->safi;
 722
 723                 /*
 724                  * handle routes from other VRFs (they can have a
 725                  * nexthop in THIS VRF). bgp_path is the bgp instance
 726                  * that owns the route referencing this nexthop.
 727                  */
 728                 bgp_path = table->bgp;
 729
 730                 /*
 731                  * Path becomes valid/invalid depending on whether the nexthop
 732                  * reachable/unreachable.
 733                  *
 734                  * In case of unicast routes that were imported from vpn
 735                  * and that have labels, they are valid only if there are
 736                  * nexthops with labels
 737                  */
 738
 739                 int bnc_is_valid_nexthop = 0;
 740
 741                 if (safi == SAFI_UNICAST &&
 742                         path->sub_type == BGP_ROUTE_IMPORTED &&
 743                         path->extra &&
 744                         path->extra->num_labels) {
 745
 746                         bnc_is_valid_nexthop =
 747                                 bgp_isvalid_labeled_nexthop(bnc) ? 1 : 0;
 748                 } else {
 749                         bnc_is_valid_nexthop =
 750                                 bgp_isvalid_nexthop(bnc) ? 1 : 0;
 751                 }
 752
 753                 if (BGP_DEBUG(nht, NHT)) {
 754                         char buf[PREFIX_STRLEN];
 755
 756                         prefix2str(&rn->p, buf, PREFIX_STRLEN);
 757                         zlog_debug("%s: prefix %s (vrf %s) %svalid",
 758                                 __func__, buf, bgp_path->name,
 759                                 (bnc_is_valid_nexthop ? "" : "not "));
 760                 }
 761
 762                 if ((CHECK_FLAG(path->flags, BGP_PATH_VALID) ? 1 : 0)
 763                     != bnc_is_valid_nexthop) {
 764                         if (CHECK_FLAG(path->flags, BGP_PATH_VALID)) {
 765                                 bgp_aggregate_decrement(bgp_path, &rn->p,
 766                                                         path, afi, safi);
 767                                 bgp_path_info_unset_flag(rn, path,
 768                                                          BGP_PATH_VALID);
 769                         } else {
 770                                 bgp_path_info_set_flag(rn, path,
 771                                                        BGP_PATH_VALID);
 772                                 bgp_aggregate_increment(bgp_path, &rn->p,
 773                                                         path, afi, safi);
 774                         }
 775                 }
 776
 777                 /* Copy the metric to the path. Will be used for bestpath
 778                  * computation */
 779                 if (bgp_isvalid_nexthop(bnc) && bnc->metric)
 780                         (bgp_path_info_extra_get(path))->igpmetric =
 781                                 bnc->metric;
 782                 else if (path->extra)
 783                         path->extra->igpmetric = 0;
 784
 785                 if (CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_METRIC_CHANGED)
 786                     || CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_CHANGED))
 787                         SET_FLAG(path->flags, BGP_PATH_IGP_CHANGED);
 788
 789                 bgp_process(bgp_path, rn, afi, safi);
 790         }
 791
 792         if (peer && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED)) {
 793                 if (BGP_DEBUG(nht, NHT))
 794                         zlog_debug("%s: Updating peer (%s) status with NHT",
 795                                    __FUNCTION__, peer->host);
 796                 bgp_fsm_event_update(peer, bgp_isvalid_nexthop(bnc));
 797                 SET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 798         }
 799
 800         RESET_FLAG(bnc->change_flags);
 801 }
 802
 803 /**
 804  * path_nh_map - make or break path-to-nexthop association.
 805  * ARGUMENTS:
 806  *   path - pointer to the path structure
 807  *   bnc - pointer to the nexthop structure
 808  *   make - if set, make the association. if unset, just break the existing
 809  *          association.
 810  */
 811 void path_nh_map(struct bgp_path_info *path, struct bgp_nexthop_cache *bnc,
 812                  bool make)
 813 {
 814         if (path->nexthop) {
 815                 LIST_REMOVE(path, nh_thread);
 816                 path->nexthop->path_count--;
 817                 path->nexthop = NULL;
 818         }
 819         if (make) {
 820                 LIST_INSERT_HEAD(&(bnc->paths), path, nh_thread);
 821                 path->nexthop = bnc;
 822                 path->nexthop->path_count++;
 823         }
 824 }
 825
 826 /*
 827  * This function is called to register nexthops to zebra
 828  * as that we may have tried to install the nexthops
 829  * before we actually have a zebra connection
 830  */
 831 void bgp_nht_register_nexthops(struct bgp *bgp)
 832 {
 833         struct bgp_node *rn;
 834         struct bgp_nexthop_cache *bnc;
 835         afi_t afi;
 836
 837         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 838                 if (!bgp->nexthop_cache_table[afi])
 839                         continue;
 840
 841                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 842                      rn = bgp_route_next(rn)) {
 843                         bnc = bgp_node_get_bgp_nexthop_info(rn);
 844
 845                         if (!bnc)
 846                                 continue;
 847
 848                         register_zebra_rnh(bnc, 0);
 849                 }
 850         }
 851 }
 852
 853 void bgp_nht_register_enhe_capability_interfaces(struct peer *peer)
 854 {
 855         struct bgp *bgp;
 856         struct bgp_node *rn;
 857         struct bgp_nexthop_cache *bnc;
 858         struct nexthop *nhop;
 859         struct interface *ifp;
 860         struct prefix p;
 861
 862         if (peer->ifp)
 863                 return;
 864
 865         bgp = peer->bgp;
 866
 867         if (!bgp->nexthop_cache_table[AFI_IP6])
 868                 return;
 869
 870         if (!sockunion2hostprefix(&peer->su, &p)) {
 871                 if (BGP_DEBUG(nht, NHT))
 872                         zlog_debug("%s: Unable to convert prefix to sockunion",
 873                                    __PRETTY_FUNCTION__);
 874                 return;
 875         }
 876
 877         if (p.family != AF_INET6)
 878                 return;
 879         rn = bgp_node_lookup(bgp->nexthop_cache_table[AFI_IP6], &p);
 880         if (!rn)
 881                 return;
 882
 883         bnc = bgp_node_get_bgp_nexthop_info(rn);
 884         if (!bnc)
 885                 return;
 886
 887         if (peer != bnc->nht_info)
 888                 return;
 889
 890         for (nhop = bnc->nexthop; nhop; nhop = nhop->next) {
 891                 ifp = if_lookup_by_index(nhop->ifindex,
 892                                          nhop->vrf_id);
 893                 zclient_send_interface_radv_req(zclient,
 894                                                 nhop->vrf_id,
 895                                                 ifp, true,
 896                                                 BGP_UNNUM_DEFAULT_RA_INTERVAL);
 897         }
 898 }