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
  17  * along with GNU Zebra; see the file COPYING.  If not, write to the Free
  18  * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
  19  * 02111-1307, USA.
  20  */
  21
  22 #include <zebra.h>
  23
  24 #include "command.h"
  25 #include "thread.h"
  26 #include "prefix.h"
  27 #include "zclient.h"
  28 #include "stream.h"
  29 #include "network.h"
  30 #include "log.h"
  31 #include "memory.h"
  32 #include "nexthop.h"
  33 #include "vrf.h"
  34 #include "filter.h"
  35
  36 #include "bgpd/bgpd.h"
  37 #include "bgpd/bgp_table.h"
  38 #include "bgpd/bgp_route.h"
  39 #include "bgpd/bgp_attr.h"
  40 #include "bgpd/bgp_nexthop.h"
  41 #include "bgpd/bgp_debug.h"
  42 #include "bgpd/bgp_nht.h"
  43 #include "bgpd/bgp_fsm.h"
  44 #include "bgpd/bgp_zebra.h"
  45
  46 extern struct zclient *zclient;
  47
  48 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
  49                                int is_bgp_static_route);
  50 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
  51                                  int is_bgp_static_route);
  52 static void evaluate_paths(struct bgp_nexthop_cache *bnc);
  53 static int make_prefix(int afi, struct bgp_info *ri, struct prefix *p);
  54 static void path_nh_map(struct bgp_info *path, struct bgp_nexthop_cache *bnc,
  55                         int keep);
  56
  57 static int bgp_isvalid_nexthop(struct bgp_nexthop_cache *bnc)
  58 {
  59         return (bgp_zebra_num_connects() == 0
  60                 || (bnc && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID)));
  61 }
  62
  63 int bgp_find_nexthop(struct bgp_info *path, int connected)
  64 {
  65         struct bgp_nexthop_cache *bnc = path->nexthop;
  66
  67         if (!bnc)
  68                 return 0;
  69
  70         /*
  71          * We are cheating here.  Views have no associated underlying
  72          * ability to detect nexthops.  So when we have a view
  73          * just tell everyone the nexthop is valid
  74          */
  75         if (path->peer && path->peer->bgp->inst_type == BGP_INSTANCE_TYPE_VIEW)
  76                 return 1;
  77
  78         if (connected && !(CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)))
  79                 return 0;
  80
  81         return (bgp_isvalid_nexthop(bnc));
  82 }
  83
  84 static void bgp_unlink_nexthop_check(struct bgp_nexthop_cache *bnc)
  85 {
  86         if (LIST_EMPTY(&(bnc->paths)) && !bnc->nht_info) {
  87                 if (BGP_DEBUG(nht, NHT)) {
  88                         char buf[PREFIX2STR_BUFFER];
  89                         zlog_debug("bgp_unlink_nexthop: freeing bnc %s",
  90                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER));
  91                 }
  92                 unregister_zebra_rnh(bnc,
  93                                      CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE));
  94                 bnc->node->info = NULL;
  95                 bgp_unlock_node(bnc->node);
  96                 bnc->node = NULL;
  97                 bnc_free(bnc);
  98         }
  99 }
 100
 101 void bgp_unlink_nexthop(struct bgp_info *path)
 102 {
 103         struct bgp_nexthop_cache *bnc = path->nexthop;
 104
 105         if (!bnc)
 106                 return;
 107
 108         path_nh_map(path, NULL, 0);
 109
 110         bgp_unlink_nexthop_check(bnc);
 111 }
 112
 113 void bgp_unlink_nexthop_by_peer(struct peer *peer)
 114 {
 115         struct prefix p;
 116         struct bgp_node *rn;
 117         struct bgp_nexthop_cache *bnc;
 118         afi_t afi = family2afi(peer->su.sa.sa_family);
 119
 120         if (!sockunion2hostprefix(&peer->su, &p))
 121                 return;
 122
 123         rn = bgp_node_get(peer->bgp->nexthop_cache_table[afi], &p);
 124
 125         if (!rn->info)
 126                 return;
 127
 128         bnc = rn->info;
 129
 130         /* cleanup the peer reference */
 131         bnc->nht_info = NULL;
 132
 133         bgp_unlink_nexthop_check(bnc);
 134 }
 135
 136 int bgp_find_or_add_nexthop(struct bgp *bgp, afi_t afi, struct bgp_info *ri,
 137                             struct peer *peer, int connected)
 138 {
 139         struct bgp_node *rn;
 140         struct bgp_nexthop_cache *bnc;
 141         struct prefix p;
 142         int is_bgp_static_route = 0;
 143
 144         if (ri) {
 145                 is_bgp_static_route = ((ri->type == ZEBRA_ROUTE_BGP)
 146                                        && (ri->sub_type == BGP_ROUTE_STATIC))
 147                                               ? 1
 148                                               : 0;
 149
 150                 /* Since Extended Next-hop Encoding (RFC5549) support, we want
 151                    to derive
 152                    address-family from the next-hop. */
 153                 if (!is_bgp_static_route)
 154                         afi = BGP_ATTR_NEXTHOP_AFI_IP6(ri->attr) ? AFI_IP6
 155                                                                  : AFI_IP;
 156
 157                 /* This will return TRUE if the global IPv6 NH is a link local
 158                  * addr */
 159                 if (make_prefix(afi, ri, &p) < 0)
 160                         return 1;
 161         } else if (peer) {
 162                 /* Don't register link local NH */
 163                 if (afi == AFI_IP6
 164                     && IN6_IS_ADDR_LINKLOCAL(&peer->su.sin6.sin6_addr))
 165                         return 1;
 166
 167                 if (!sockunion2hostprefix(&peer->su, &p)) {
 168                         if (BGP_DEBUG(nht, NHT)) {
 169                                 zlog_debug(
 170                                         "%s: Attempting to register with unknown AFI %d (not %d or %d)",
 171                                         __FUNCTION__, afi, AFI_IP, AFI_IP6);
 172                         }
 173                         return 0;
 174                 }
 175         } else
 176                 return 0;
 177
 178         if (is_bgp_static_route)
 179                 rn = bgp_node_get(bgp->import_check_table[afi], &p);
 180         else
 181                 rn = bgp_node_get(bgp->nexthop_cache_table[afi], &p);
 182
 183         if (!rn->info) {
 184                 bnc = bnc_new();
 185                 rn->info = bnc;
 186                 bnc->node = rn;
 187                 bnc->bgp = bgp;
 188                 bgp_lock_node(rn);
 189                 if (BGP_DEBUG(nht, NHT)) {
 190                         char buf[PREFIX2STR_BUFFER];
 191
 192                         zlog_debug("Allocated bnc %s peer %p",
 193                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER), peer);
 194                 }
 195         }
 196
 197         bnc = rn->info;
 198         bgp_unlock_node(rn);
 199         if (is_bgp_static_route) {
 200                 SET_FLAG(bnc->flags, BGP_STATIC_ROUTE);
 201
 202                 /* If we're toggling the type, re-register */
 203                 if ((bgp_flag_check(bgp, BGP_FLAG_IMPORT_CHECK))
 204                     && !CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)) {
 205                         SET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 206                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 207                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 208                 } else if ((!bgp_flag_check(bgp, BGP_FLAG_IMPORT_CHECK))
 209                            && CHECK_FLAG(bnc->flags,
 210                                          BGP_STATIC_ROUTE_EXACT_MATCH)) {
 211                         UNSET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 212                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 213                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 214                 }
 215         }
 216         /* When nexthop is already known, but now requires 'connected'
 217          * resolution,
 218          * re-register it. The reverse scenario where the nexthop currently
 219          * requires
 220          * 'connected' resolution does not need a re-register (i.e., we treat
 221          * 'connected-required' as an override) except in the scenario where
 222          * this
 223          * is actually a case of tracking a peer for connectivity (e.g., after
 224          * disable connected-check).
 225          * NOTE: We don't track the number of paths separately for 'connected-
 226          * required' vs 'connected-not-required' as this change is not a common
 227          * scenario.
 228          */
 229         else if (connected && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 230                 SET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 231                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 232                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 233         } else if (peer && !connected
 234                    && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 235                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 236                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 237                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 238         }
 239         if (bgp->inst_type == BGP_INSTANCE_TYPE_VIEW) {
 240                 bnc->flags |= BGP_NEXTHOP_REGISTERED;
 241                 bnc->flags |= BGP_NEXTHOP_VALID;
 242         } else if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 243                 register_zebra_rnh(bnc, is_bgp_static_route);
 244         if (ri && ri->nexthop != bnc) {
 245                 /* Unlink from existing nexthop cache, if any. This will also
 246                  * free
 247                  * the nexthop cache entry, if appropriate.
 248                  */
 249                 bgp_unlink_nexthop(ri);
 250
 251                 path_nh_map(ri, bnc, 1); /* updates NHT ri list reference */
 252
 253                 if (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID) && bnc->metric)
 254                         (bgp_info_extra_get(ri))->igpmetric = bnc->metric;
 255                 else if (ri->extra)
 256                         ri->extra->igpmetric = 0;
 257         } else if (peer)
 258                 bnc->nht_info = (void *)peer; /* NHT peer reference */
 259
 260         /*
 261          * We are cheating here.  Views have no associated underlying
 262          * ability to detect nexthops.  So when we have a view
 263          * just tell everyone the nexthop is valid
 264          */
 265         if (bgp->inst_type == BGP_INSTANCE_TYPE_VIEW)
 266                 return 1;
 267         else
 268                 return (bgp_isvalid_nexthop(bnc));
 269 }
 270
 271 void bgp_delete_connected_nexthop(afi_t afi, struct peer *peer)
 272 {
 273         struct bgp_node *rn;
 274         struct bgp_nexthop_cache *bnc;
 275         struct prefix p;
 276
 277         if (!peer)
 278                 return;
 279
 280         /* We don't register link local address for NHT */
 281         if (afi == AFI_IP6 && IN6_IS_ADDR_LINKLOCAL(&peer->su.sin6.sin6_addr))
 282                 return;
 283
 284         if (!sockunion2hostprefix(&peer->su, &p))
 285                 return;
 286
 287         rn = bgp_node_lookup(
 288                 peer->bgp->nexthop_cache_table[family2afi(p.family)], &p);
 289         if (!rn || !rn->info) {
 290                 if (BGP_DEBUG(nht, NHT))
 291                         zlog_debug("Cannot find connected NHT node for peer %s",
 292                                    peer->host);
 293                 if (rn)
 294                         bgp_unlock_node(rn);
 295                 return;
 296         }
 297
 298         bnc = rn->info;
 299         bgp_unlock_node(rn);
 300
 301         if (bnc->nht_info != peer) {
 302                 if (BGP_DEBUG(nht, NHT))
 303                         zlog_debug(
 304                                 "Connected NHT %p node for peer %s points to %p",
 305                                 bnc, peer->host, bnc->nht_info);
 306                 return;
 307         }
 308
 309         bnc->nht_info = NULL;
 310
 311         if (LIST_EMPTY(&(bnc->paths))) {
 312                 if (BGP_DEBUG(nht, NHT))
 313                         zlog_debug("Freeing connected NHT node %p for peer %s",
 314                                    bnc, peer->host);
 315                 unregister_zebra_rnh(bnc, 0);
 316                 bnc->node->info = NULL;
 317                 bgp_unlock_node(bnc->node);
 318                 bnc_free(bnc);
 319         }
 320 }
 321
 322 void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 323 {
 324         struct stream *s;
 325         struct bgp_node *rn = NULL;
 326         struct bgp_nexthop_cache *bnc;
 327         struct nexthop *nexthop;
 328         struct nexthop *oldnh;
 329         struct nexthop *nhlist_head = NULL;
 330         struct nexthop *nhlist_tail = NULL;
 331         uint32_t metric;
 332         u_char nexthop_num;
 333         struct prefix p;
 334         int i;
 335         struct bgp *bgp;
 336
 337         bgp = bgp_lookup_by_vrf_id(vrf_id);
 338         if (!bgp) {
 339                 zlog_err(
 340                         "parse nexthop update: instance not found for vrf_id %d",
 341                         vrf_id);
 342                 return;
 343         }
 344
 345         s = zclient->ibuf;
 346
 347         memset(&p, 0, sizeof(struct prefix));
 348         p.family = stream_getw(s);
 349         p.prefixlen = stream_getc(s);
 350         switch (p.family) {
 351         case AF_INET:
 352                 p.u.prefix4.s_addr = stream_get_ipv4(s);
 353                 break;
 354         case AF_INET6:
 355                 stream_get(&p.u.prefix6, s, 16);
 356                 break;
 357         default:
 358                 break;
 359         }
 360
 361         if (command == ZEBRA_NEXTHOP_UPDATE)
 362                 rn = bgp_node_lookup(
 363                         bgp->nexthop_cache_table[family2afi(p.family)], &p);
 364         else if (command == ZEBRA_IMPORT_CHECK_UPDATE)
 365                 rn = bgp_node_lookup(
 366                         bgp->import_check_table[family2afi(p.family)], &p);
 367
 368         if (!rn || !rn->info) {
 369                 if (BGP_DEBUG(nht, NHT)) {
 370                         char buf[PREFIX2STR_BUFFER];
 371                         prefix2str(&p, buf, sizeof(buf));
 372                         zlog_debug("parse nexthop update(%s): rn not found",
 373                                    buf);
 374                 }
 375                 if (rn)
 376                         bgp_unlock_node(rn);
 377                 return;
 378         }
 379
 380         bnc = rn->info;
 381         bgp_unlock_node(rn);
 382         bnc->last_update = bgp_clock();
 383         bnc->change_flags = 0;
 384         stream_getc(s); // Distance but not currently used
 385         metric = stream_getl(s);
 386         nexthop_num = stream_getc(s);
 387
 388         /* debug print the input */
 389         if (BGP_DEBUG(nht, NHT)) {
 390                 char buf[PREFIX2STR_BUFFER];
 391                 prefix2str(&p, buf, sizeof(buf));
 392                 zlog_debug(
 393                         "%d: NH update for %s - metric %d (cur %d) #nhops %d (cur %d)",
 394                         vrf_id, buf, metric, bnc->metric, nexthop_num,
 395                         bnc->nexthop_num);
 396         }
 397
 398         if (metric != bnc->metric)
 399                 bnc->change_flags |= BGP_NEXTHOP_METRIC_CHANGED;
 400
 401         if (nexthop_num != bnc->nexthop_num)
 402                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 403
 404         if (nexthop_num) {
 405                 /* notify bgp fsm if nbr ip goes from invalid->valid */
 406                 if (!bnc->nexthop_num)
 407                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 408
 409                 bnc->flags |= BGP_NEXTHOP_VALID;
 410                 bnc->metric = metric;
 411                 bnc->nexthop_num = nexthop_num;
 412
 413                 for (i = 0; i < nexthop_num; i++) {
 414                         nexthop = nexthop_new();
 415                         nexthop->type = stream_getc(s);
 416                         switch (nexthop->type) {
 417                         case NEXTHOP_TYPE_IPV4:
 418                                 nexthop->gate.ipv4.s_addr = stream_get_ipv4(s);
 419                                 nexthop->ifindex = stream_getl(s);
 420                                 break;
 421                         case NEXTHOP_TYPE_IFINDEX:
 422                                 nexthop->ifindex = stream_getl(s);
 423                                 break;
 424                         case NEXTHOP_TYPE_IPV4_IFINDEX:
 425                                 nexthop->gate.ipv4.s_addr = stream_get_ipv4(s);
 426                                 nexthop->ifindex = stream_getl(s);
 427                                 break;
 428                         case NEXTHOP_TYPE_IPV6:
 429                                 stream_get(&nexthop->gate.ipv6, s, 16);
 430                                 nexthop->ifindex = stream_getl(s);
 431                                 break;
 432                         case NEXTHOP_TYPE_IPV6_IFINDEX:
 433                                 stream_get(&nexthop->gate.ipv6, s, 16);
 434                                 nexthop->ifindex = stream_getl(s);
 435                                 break;
 436                         default:
 437                                 /* do nothing */
 438                                 break;
 439                         }
 440
 441                         if (BGP_DEBUG(nht, NHT)) {
 442                                 char buf[NEXTHOP_STRLEN];
 443                                 zlog_debug(
 444                                         "    nhop via %s",
 445                                         nexthop2str(nexthop, buf, sizeof(buf)));
 446                         }
 447
 448                         if (nhlist_tail) {
 449                                 nhlist_tail->next = nexthop;
 450                                 nhlist_tail = nexthop;
 451                         } else {
 452                                 nhlist_tail = nexthop;
 453                                 nhlist_head = nexthop;
 454                         }
 455
 456                         /* No need to evaluate the nexthop if we have already
 457                          * determined
 458                          * that there has been a change.
 459                          */
 460                         if (bnc->change_flags & BGP_NEXTHOP_CHANGED)
 461                                 continue;
 462
 463                         for (oldnh = bnc->nexthop; oldnh; oldnh = oldnh->next)
 464                                 if (nexthop_same_no_recurse(oldnh, nexthop))
 465                                         break;
 466
 467                         if (!oldnh)
 468                                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 469                 }
 470                 bnc_nexthop_free(bnc);
 471                 bnc->nexthop = nhlist_head;
 472         } else {
 473                 bnc->flags &= ~BGP_NEXTHOP_VALID;
 474                 bnc->nexthop_num = nexthop_num;
 475
 476                 /* notify bgp fsm if nbr ip goes from valid->invalid */
 477                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 478
 479                 bnc_nexthop_free(bnc);
 480                 bnc->nexthop = NULL;
 481         }
 482
 483         evaluate_paths(bnc);
 484 }
 485
 486 /**
 487  * make_prefix - make a prefix structure from the path (essentially
 488  * path's node.
 489  */
 490 static int make_prefix(int afi, struct bgp_info *ri, struct prefix *p)
 491 {
 492
 493         int is_bgp_static = ((ri->type == ZEBRA_ROUTE_BGP)
 494                              && (ri->sub_type == BGP_ROUTE_STATIC))
 495                                     ? 1
 496                                     : 0;
 497
 498         memset(p, 0, sizeof(struct prefix));
 499         switch (afi) {
 500         case AFI_IP:
 501                 p->family = AF_INET;
 502                 if (is_bgp_static) {
 503                         p->u.prefix4 = ri->net->p.u.prefix4;
 504                         p->prefixlen = ri->net->p.prefixlen;
 505                 } else {
 506                         p->u.prefix4 = ri->attr->nexthop;
 507                         p->prefixlen = IPV4_MAX_BITLEN;
 508                 }
 509                 break;
 510         case AFI_IP6:
 511                 /* We don't register link local NH */
 512                 if (ri->attr->extra->mp_nexthop_len
 513                             != BGP_ATTR_NHLEN_IPV6_GLOBAL
 514                     || IN6_IS_ADDR_LINKLOCAL(
 515                                &ri->attr->extra->mp_nexthop_global))
 516                         return -1;
 517
 518                 p->family = AF_INET6;
 519
 520                 if (is_bgp_static) {
 521                         p->u.prefix6 = ri->net->p.u.prefix6;
 522                         p->prefixlen = ri->net->p.prefixlen;
 523                 } else {
 524                         p->u.prefix6 = ri->attr->extra->mp_nexthop_global;
 525                         p->prefixlen = IPV6_MAX_BITLEN;
 526                 }
 527                 break;
 528         default:
 529                 if (BGP_DEBUG(nht, NHT)) {
 530                         zlog_debug(
 531                                 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
 532                                 __FUNCTION__, afi, AFI_IP, AFI_IP6);
 533                 }
 534                 break;
 535         }
 536         return 0;
 537 }
 538
 539 /**
 540  * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
 541  *   command to Zebra.
 542  * ARGUMENTS:
 543  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 544  *   int command -- command to send to zebra
 545  * RETURNS:
 546  *   void.
 547  */
 548 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache *bnc, int command)
 549 {
 550         struct stream *s;
 551         struct prefix *p;
 552         int ret;
 553
 554         /* Check socket. */
 555         if (!zclient || zclient->sock < 0)
 556                 return;
 557
 558         /* Don't try to register if Zebra doesn't know of this instance. */
 559         if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc->bgp))
 560                 return;
 561
 562         p = &(bnc->node->p);
 563         s = zclient->obuf;
 564         stream_reset(s);
 565         zclient_create_header(s, command, bnc->bgp->vrf_id);
 566         if (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)
 567             || CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH))
 568                 stream_putc(s, 1);
 569         else
 570                 stream_putc(s, 0);
 571
 572         stream_putw(s, PREFIX_FAMILY(p));
 573         stream_putc(s, p->prefixlen);
 574         switch (PREFIX_FAMILY(p)) {
 575         case AF_INET:
 576                 stream_put_in_addr(s, &p->u.prefix4);
 577                 break;
 578         case AF_INET6:
 579                 stream_put(s, &(p->u.prefix6), 16);
 580                 break;
 581         default:
 582                 break;
 583         }
 584         stream_putw_at(s, 0, stream_get_endp(s));
 585
 586         ret = zclient_send_message(zclient);
 587         /* TBD: handle the failure */
 588         if (ret < 0)
 589                 zlog_warn("sendmsg_nexthop: zclient_send_message() failed");
 590
 591         if ((command == ZEBRA_NEXTHOP_REGISTER)
 592             || (command == ZEBRA_IMPORT_ROUTE_REGISTER))
 593                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 594         else if ((command == ZEBRA_NEXTHOP_UNREGISTER)
 595                  || (command == ZEBRA_IMPORT_ROUTE_UNREGISTER))
 596                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 597         return;
 598 }
 599
 600 /**
 601  * register_zebra_rnh - register a NH/route with Zebra for notification
 602  *    when the route or the route to the nexthop changes.
 603  * ARGUMENTS:
 604  *   struct bgp_nexthop_cache *bnc
 605  * RETURNS:
 606  *   void.
 607  */
 608 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
 609                                int is_bgp_import_route)
 610 {
 611         /* Check if we have already registered */
 612         if (bnc->flags & BGP_NEXTHOP_REGISTERED)
 613                 return;
 614         if (is_bgp_import_route)
 615                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_REGISTER);
 616         else
 617                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_REGISTER);
 618 }
 619
 620 /**
 621  * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
 622  * ARGUMENTS:
 623  *   struct bgp_nexthop_cache *bnc
 624  * RETURNS:
 625  *   void.
 626  */
 627 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
 628                                  int is_bgp_import_route)
 629 {
 630         /* Check if we have already registered */
 631         if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 632                 return;
 633
 634         if (is_bgp_import_route)
 635                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_UNREGISTER);
 636         else
 637                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_UNREGISTER);
 638 }
 639
 640 /**
 641  * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
 642  * ARGUMENTS:
 643  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 644  * RETURNS:
 645  *   void.
 646  */
 647 static void evaluate_paths(struct bgp_nexthop_cache *bnc)
 648 {
 649         struct bgp_node *rn;
 650         struct bgp_info *path;
 651         struct bgp *bgp = bnc->bgp;
 652         int afi;
 653         struct peer *peer = (struct peer *)bnc->nht_info;
 654
 655         if (BGP_DEBUG(nht, NHT)) {
 656                 char buf[PREFIX2STR_BUFFER];
 657                 bnc_str(bnc, buf, PREFIX2STR_BUFFER);
 658                 zlog_debug(
 659                         "NH update for %s - flags 0x%x chgflags 0x%x - evaluate paths",
 660                         buf, bnc->flags, bnc->change_flags);
 661         }
 662
 663         LIST_FOREACH(path, &(bnc->paths), nh_thread)
 664         {
 665                 if (!(path->type == ZEBRA_ROUTE_BGP
 666                       && ((path->sub_type == BGP_ROUTE_NORMAL)
 667                           || (path->sub_type == BGP_ROUTE_STATIC))))
 668                         continue;
 669
 670                 rn = path->net;
 671                 afi = family2afi(rn->p.family);
 672
 673                 /* Path becomes valid/invalid depending on whether the nexthop
 674                  * reachable/unreachable.
 675                  */
 676                 if ((CHECK_FLAG(path->flags, BGP_INFO_VALID) ? 1 : 0)
 677                     != (bgp_isvalid_nexthop(bnc) ? 1 : 0)) {
 678                         if (CHECK_FLAG(path->flags, BGP_INFO_VALID)) {
 679                                 bgp_aggregate_decrement(bgp, &rn->p, path, afi,
 680                                                         SAFI_UNICAST);
 681                                 bgp_info_unset_flag(rn, path, BGP_INFO_VALID);
 682                         } else {
 683                                 bgp_info_set_flag(rn, path, BGP_INFO_VALID);
 684                                 bgp_aggregate_increment(bgp, &rn->p, path, afi,
 685                                                         SAFI_UNICAST);
 686                         }
 687                 }
 688
 689                 /* Copy the metric to the path. Will be used for bestpath
 690                  * computation */
 691                 if (bgp_isvalid_nexthop(bnc) && bnc->metric)
 692                         (bgp_info_extra_get(path))->igpmetric = bnc->metric;
 693                 else if (path->extra)
 694                         path->extra->igpmetric = 0;
 695
 696                 if (CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_METRIC_CHANGED)
 697                     || CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_CHANGED))
 698                         SET_FLAG(path->flags, BGP_INFO_IGP_CHANGED);
 699
 700                 bgp_process(bgp, rn, afi, SAFI_UNICAST);
 701         }
 702
 703         if (peer && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED)) {
 704                 if (BGP_DEBUG(nht, NHT))
 705                         zlog_debug("%s: Updating peer (%s) status with NHT",
 706                                    __FUNCTION__, peer->host);
 707                 bgp_fsm_nht_update(peer, bgp_isvalid_nexthop(bnc));
 708                 SET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 709         }
 710
 711         RESET_FLAG(bnc->change_flags);
 712 }
 713
 714 /**
 715  * path_nh_map - make or break path-to-nexthop association.
 716  * ARGUMENTS:
 717  *   path - pointer to the path structure
 718  *   bnc - pointer to the nexthop structure
 719  *   make - if set, make the association. if unset, just break the existing
 720  *          association.
 721  */
 722 static void path_nh_map(struct bgp_info *path, struct bgp_nexthop_cache *bnc,
 723                         int make)
 724 {
 725         if (path->nexthop) {
 726                 LIST_REMOVE(path, nh_thread);
 727                 path->nexthop->path_count--;
 728                 path->nexthop = NULL;
 729         }
 730         if (make) {
 731                 LIST_INSERT_HEAD(&(bnc->paths), path, nh_thread);
 732                 path->nexthop = bnc;
 733                 path->nexthop->path_count++;
 734         }
 735 }