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
  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 static int bgp_isvalid_labeled_nexthop(struct bgp_nexthop_cache *bnc)
  64 {
  65         return (bgp_zebra_num_connects() == 0
  66                 || (bnc && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_LABELED_VALID)));
  67 }
  68
  69 int bgp_find_nexthop(struct bgp_info *path, int connected)
  70 {
  71         struct bgp_nexthop_cache *bnc = path->nexthop;
  72
  73         if (!bnc)
  74                 return 0;
  75
  76         /*
  77          * We are cheating here.  Views have no associated underlying
  78          * ability to detect nexthops.  So when we have a view
  79          * just tell everyone the nexthop is valid
  80          */
  81         if (path->peer && path->peer->bgp->inst_type == BGP_INSTANCE_TYPE_VIEW)
  82                 return 1;
  83
  84         if (connected && !(CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)))
  85                 return 0;
  86
  87         return (bgp_isvalid_nexthop(bnc));
  88 }
  89
  90 static void bgp_unlink_nexthop_check(struct bgp_nexthop_cache *bnc)
  91 {
  92         if (LIST_EMPTY(&(bnc->paths)) && !bnc->nht_info) {
  93                 if (BGP_DEBUG(nht, NHT)) {
  94                         char buf[PREFIX2STR_BUFFER];
  95                         zlog_debug("bgp_unlink_nexthop: freeing bnc %s",
  96                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER));
  97                 }
  98                 unregister_zebra_rnh(bnc,
  99                                      CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE));
 100                 bnc->node->info = NULL;
 101                 bgp_unlock_node(bnc->node);
 102                 bnc->node = NULL;
 103                 bnc_free(bnc);
 104         }
 105 }
 106
 107 void bgp_unlink_nexthop(struct bgp_info *path)
 108 {
 109         struct bgp_nexthop_cache *bnc = path->nexthop;
 110
 111         if (!bnc)
 112                 return;
 113
 114         path_nh_map(path, NULL, 0);
 115
 116         bgp_unlink_nexthop_check(bnc);
 117 }
 118
 119 void bgp_unlink_nexthop_by_peer(struct peer *peer)
 120 {
 121         struct prefix p;
 122         struct bgp_node *rn;
 123         struct bgp_nexthop_cache *bnc;
 124         afi_t afi = family2afi(peer->su.sa.sa_family);
 125
 126         if (!sockunion2hostprefix(&peer->su, &p))
 127                 return;
 128
 129         rn = bgp_node_get(peer->bgp->nexthop_cache_table[afi], &p);
 130
 131         if (!rn->info)
 132                 return;
 133
 134         bnc = rn->info;
 135
 136         /* cleanup the peer reference */
 137         bnc->nht_info = NULL;
 138
 139         bgp_unlink_nexthop_check(bnc);
 140 }
 141
 142 /*
 143  * A route and its nexthop might belong to different VRFs. Therefore,
 144  * we need both the bgp_route and bgp_nexthop pointers.
 145  */
 146 int bgp_find_or_add_nexthop(struct bgp *bgp_route, struct bgp *bgp_nexthop,
 147                             afi_t afi, struct bgp_info *ri,
 148                             struct peer *peer, int connected)
 149 {
 150         struct bgp_node *rn;
 151         struct bgp_nexthop_cache *bnc;
 152         struct prefix p;
 153         int is_bgp_static_route = 0;
 154
 155         if (ri) {
 156                 is_bgp_static_route = ((ri->type == ZEBRA_ROUTE_BGP)
 157                                        && (ri->sub_type == BGP_ROUTE_STATIC))
 158                                               ? 1
 159                                               : 0;
 160
 161                 /* Since Extended Next-hop Encoding (RFC5549) support, we want
 162                    to derive
 163                    address-family from the next-hop. */
 164                 if (!is_bgp_static_route)
 165                         afi = BGP_ATTR_NEXTHOP_AFI_IP6(ri->attr) ? AFI_IP6
 166                                                                  : AFI_IP;
 167
 168                 /* This will return TRUE if the global IPv6 NH is a link local
 169                  * addr */
 170                 if (make_prefix(afi, ri, &p) < 0)
 171                         return 1;
 172         } else if (peer) {
 173                 /* Don't register link local NH */
 174                 if (afi == AFI_IP6
 175                     && IN6_IS_ADDR_LINKLOCAL(&peer->su.sin6.sin6_addr))
 176                         return 1;
 177
 178                 if (!sockunion2hostprefix(&peer->su, &p)) {
 179                         if (BGP_DEBUG(nht, NHT)) {
 180                                 zlog_debug(
 181                                         "%s: Attempting to register with unknown AFI %d (not %d or %d)",
 182                                         __FUNCTION__, afi, AFI_IP, AFI_IP6);
 183                         }
 184                         return 0;
 185                 }
 186         } else
 187                 return 0;
 188
 189         if (is_bgp_static_route)
 190                 rn = bgp_node_get(bgp_nexthop->import_check_table[afi], &p);
 191         else
 192                 rn = bgp_node_get(bgp_nexthop->nexthop_cache_table[afi], &p);
 193
 194         if (!rn->info) {
 195                 bnc = bnc_new();
 196                 rn->info = bnc;
 197                 bnc->node = rn;
 198                 bnc->bgp = bgp_nexthop;
 199                 bgp_lock_node(rn);
 200                 if (BGP_DEBUG(nht, NHT)) {
 201                         char buf[PREFIX2STR_BUFFER];
 202
 203                         zlog_debug("Allocated bnc %s peer %p",
 204                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER), peer);
 205                 }
 206         }
 207
 208         bnc = rn->info;
 209         bgp_unlock_node(rn);
 210         if (is_bgp_static_route) {
 211                 SET_FLAG(bnc->flags, BGP_STATIC_ROUTE);
 212
 213                 /* If we're toggling the type, re-register */
 214                 if ((bgp_flag_check(bgp_route, BGP_FLAG_IMPORT_CHECK))
 215                     && !CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)) {
 216                         SET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 217                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 218                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 219                 } else if ((!bgp_flag_check(bgp_route, BGP_FLAG_IMPORT_CHECK))
 220                            && CHECK_FLAG(bnc->flags,
 221                                          BGP_STATIC_ROUTE_EXACT_MATCH)) {
 222                         UNSET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 223                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 224                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 225                 }
 226         }
 227         /* When nexthop is already known, but now requires 'connected'
 228          * resolution,
 229          * re-register it. The reverse scenario where the nexthop currently
 230          * requires
 231          * 'connected' resolution does not need a re-register (i.e., we treat
 232          * 'connected-required' as an override) except in the scenario where
 233          * this
 234          * is actually a case of tracking a peer for connectivity (e.g., after
 235          * disable connected-check).
 236          * NOTE: We don't track the number of paths separately for 'connected-
 237          * required' vs 'connected-not-required' as this change is not a common
 238          * scenario.
 239          */
 240         else if (connected && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 241                 SET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 242                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 243                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 244         } else if (peer && !connected
 245                    && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 246                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 247                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 248                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 249         }
 250         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW) {
 251                 bnc->flags |= BGP_NEXTHOP_REGISTERED;
 252                 bnc->flags |= BGP_NEXTHOP_VALID;
 253         } else if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 254                 register_zebra_rnh(bnc, is_bgp_static_route);
 255         if (ri && ri->nexthop != bnc) {
 256                 /* Unlink from existing nexthop cache, if any. This will also
 257                  * free
 258                  * the nexthop cache entry, if appropriate.
 259                  */
 260                 bgp_unlink_nexthop(ri);
 261
 262                 path_nh_map(ri, bnc, 1); /* updates NHT ri list reference */
 263
 264                 if (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID) && bnc->metric)
 265                         (bgp_info_extra_get(ri))->igpmetric = bnc->metric;
 266                 else if (ri->extra)
 267                         ri->extra->igpmetric = 0;
 268         } else if (peer)
 269                 bnc->nht_info = (void *)peer; /* NHT peer reference */
 270
 271         /*
 272          * We are cheating here.  Views have no associated underlying
 273          * ability to detect nexthops.  So when we have a view
 274          * just tell everyone the nexthop is valid
 275          */
 276         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW)
 277                 return 1;
 278         else
 279                 return (bgp_isvalid_nexthop(bnc));
 280 }
 281
 282 void bgp_delete_connected_nexthop(afi_t afi, struct peer *peer)
 283 {
 284         struct bgp_node *rn;
 285         struct bgp_nexthop_cache *bnc;
 286         struct prefix p;
 287
 288         if (!peer)
 289                 return;
 290
 291         /* We don't register link local address for NHT */
 292         if (afi == AFI_IP6 && IN6_IS_ADDR_LINKLOCAL(&peer->su.sin6.sin6_addr))
 293                 return;
 294
 295         if (!sockunion2hostprefix(&peer->su, &p))
 296                 return;
 297
 298         rn = bgp_node_lookup(
 299                 peer->bgp->nexthop_cache_table[family2afi(p.family)], &p);
 300         if (!rn || !rn->info) {
 301                 if (BGP_DEBUG(nht, NHT))
 302                         zlog_debug("Cannot find connected NHT node for peer %s",
 303                                    peer->host);
 304                 if (rn)
 305                         bgp_unlock_node(rn);
 306                 return;
 307         }
 308
 309         bnc = rn->info;
 310         bgp_unlock_node(rn);
 311
 312         if (bnc->nht_info != peer) {
 313                 if (BGP_DEBUG(nht, NHT))
 314                         zlog_debug(
 315                                 "Connected NHT %p node for peer %s points to %p",
 316                                 bnc, peer->host, bnc->nht_info);
 317                 return;
 318         }
 319
 320         bnc->nht_info = NULL;
 321
 322         if (LIST_EMPTY(&(bnc->paths))) {
 323                 if (BGP_DEBUG(nht, NHT))
 324                         zlog_debug("Freeing connected NHT node %p for peer %s",
 325                                    bnc, peer->host);
 326                 unregister_zebra_rnh(bnc, 0);
 327                 bnc->node->info = NULL;
 328                 bgp_unlock_node(bnc->node);
 329                 bnc_free(bnc);
 330         }
 331 }
 332
 333 void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 334 {
 335         struct bgp_node *rn = NULL;
 336         struct bgp_nexthop_cache *bnc;
 337         struct nexthop *nexthop;
 338         struct nexthop *oldnh;
 339         struct nexthop *nhlist_head = NULL;
 340         struct nexthop *nhlist_tail = NULL;
 341         int i;
 342         struct bgp *bgp;
 343         struct zapi_route nhr;
 344
 345         bgp = bgp_lookup_by_vrf_id(vrf_id);
 346         if (!bgp) {
 347                 flog_err(
 348                         BGP_ERR_NH_UPD,
 349                         "parse nexthop update: instance not found for vrf_id %u",
 350                         vrf_id);
 351                 return;
 352         }
 353
 354         if (!zapi_nexthop_update_decode(zclient->ibuf, &nhr)) {
 355                 if (BGP_DEBUG(nht, NHT))
 356                         zlog_debug("%s: Failure to decode nexthop update",
 357                                    __PRETTY_FUNCTION__);
 358                 return;
 359         }
 360
 361         if (command == ZEBRA_NEXTHOP_UPDATE)
 362                 rn = bgp_node_lookup(
 363                         bgp->nexthop_cache_table[family2afi(nhr.prefix.family)],
 364                         &nhr.prefix);
 365         else if (command == ZEBRA_IMPORT_CHECK_UPDATE)
 366                 rn = bgp_node_lookup(
 367                         bgp->import_check_table[family2afi(nhr.prefix.family)],
 368                         &nhr.prefix);
 369
 370         if (!rn || !rn->info) {
 371                 if (BGP_DEBUG(nht, NHT)) {
 372                         char buf[PREFIX2STR_BUFFER];
 373                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 374                         zlog_debug("parse nexthop update(%s): rn not found",
 375                                    buf);
 376                 }
 377                 if (rn)
 378                         bgp_unlock_node(rn);
 379                 return;
 380         }
 381
 382         bnc = rn->info;
 383         bgp_unlock_node(rn);
 384         bnc->last_update = bgp_clock();
 385         bnc->change_flags = 0;
 386
 387         /* debug print the input */
 388         if (BGP_DEBUG(nht, NHT)) {
 389                 char buf[PREFIX2STR_BUFFER];
 390                 prefix2str(&nhr.prefix, buf, sizeof(buf));
 391                 zlog_debug(
 392                         "%u: Rcvd NH update %s - metric %d/%d #nhops %d/%d flags 0x%x",
 393                         vrf_id, buf, nhr.metric, bnc->metric, nhr.nexthop_num,
 394                         bnc->nexthop_num, bnc->flags);
 395         }
 396
 397         if (nhr.metric != bnc->metric)
 398                 bnc->change_flags |= BGP_NEXTHOP_METRIC_CHANGED;
 399
 400         if (nhr.nexthop_num != bnc->nexthop_num)
 401                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 402
 403         if (nhr.nexthop_num) {
 404                 /* notify bgp fsm if nbr ip goes from invalid->valid */
 405                 if (!bnc->nexthop_num)
 406                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 407
 408                 bnc->flags |= BGP_NEXTHOP_VALID;
 409                 bnc->metric = nhr.metric;
 410                 bnc->nexthop_num = nhr.nexthop_num;
 411
 412                 bnc->flags &= ~BGP_NEXTHOP_LABELED_VALID; /* check below */
 413
 414                 for (i = 0; i < nhr.nexthop_num; i++) {
 415                         int num_labels = 0;
 416
 417                         nexthop = nexthop_from_zapi_nexthop(&nhr.nexthops[i]);
 418
 419                         /* There is at least one label-switched path */
 420                         if (nexthop->nh_label &&
 421                                 nexthop->nh_label->num_labels) {
 422
 423                                 bnc->flags |= BGP_NEXTHOP_LABELED_VALID;
 424                                 num_labels = nexthop->nh_label->num_labels;
 425                         }
 426
 427                         if (BGP_DEBUG(nht, NHT)) {
 428                                 char buf[NEXTHOP_STRLEN];
 429                                 zlog_debug(
 430                                         "    nhop via %s (%d labels)",
 431                                         nexthop2str(nexthop, buf, sizeof(buf)),
 432                                         num_labels);
 433                         }
 434
 435                         if (nhlist_tail) {
 436                                 nhlist_tail->next = nexthop;
 437                                 nhlist_tail = nexthop;
 438                         } else {
 439                                 nhlist_tail = nexthop;
 440                                 nhlist_head = nexthop;
 441                         }
 442
 443                         /* No need to evaluate the nexthop if we have already
 444                          * determined
 445                          * that there has been a change.
 446                          */
 447                         if (bnc->change_flags & BGP_NEXTHOP_CHANGED)
 448                                 continue;
 449
 450                         for (oldnh = bnc->nexthop; oldnh; oldnh = oldnh->next)
 451                                 if (nexthop_same_no_recurse(oldnh, nexthop) &&
 452                                     nexthop_labels_match(oldnh, nexthop))
 453                                         break;
 454
 455                         if (!oldnh)
 456                                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 457                 }
 458                 bnc_nexthop_free(bnc);
 459                 bnc->nexthop = nhlist_head;
 460         } else {
 461                 bnc->flags &= ~BGP_NEXTHOP_VALID;
 462                 bnc->nexthop_num = nhr.nexthop_num;
 463
 464                 /* notify bgp fsm if nbr ip goes from valid->invalid */
 465                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 466
 467                 bnc_nexthop_free(bnc);
 468                 bnc->nexthop = NULL;
 469         }
 470
 471         evaluate_paths(bnc);
 472 }
 473
 474 /*
 475  * Cleanup nexthop registration and status information for BGP nexthops
 476  * pertaining to this VRF. This is invoked upon VRF deletion.
 477  */
 478 void bgp_cleanup_nexthops(struct bgp *bgp)
 479 {
 480         afi_t afi;
 481         struct bgp_node *rn;
 482         struct bgp_nexthop_cache *bnc;
 483
 484         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 485                 if (!bgp->nexthop_cache_table[afi])
 486                         continue;
 487
 488                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 489                      rn = bgp_route_next(rn)) {
 490                         bnc = rn->info;
 491                         if (!bnc)
 492                                 continue;
 493
 494                         /* Clear relevant flags. */
 495                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 496                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 497                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 498                 }
 499         }
 500 }
 501
 502 /**
 503  * make_prefix - make a prefix structure from the path (essentially
 504  * path's node.
 505  */
 506 static int make_prefix(int afi, struct bgp_info *ri, struct prefix *p)
 507 {
 508
 509         int is_bgp_static = ((ri->type == ZEBRA_ROUTE_BGP)
 510                              && (ri->sub_type == BGP_ROUTE_STATIC))
 511                                     ? 1
 512                                     : 0;
 513
 514         memset(p, 0, sizeof(struct prefix));
 515         switch (afi) {
 516         case AFI_IP:
 517                 p->family = AF_INET;
 518                 if (is_bgp_static) {
 519                         p->u.prefix4 = ri->net->p.u.prefix4;
 520                         p->prefixlen = ri->net->p.prefixlen;
 521                 } else {
 522                         p->u.prefix4 = ri->attr->nexthop;
 523                         p->prefixlen = IPV4_MAX_BITLEN;
 524                 }
 525                 break;
 526         case AFI_IP6:
 527                 /* We don't register link local NH */
 528                 if (ri->attr->mp_nexthop_len != BGP_ATTR_NHLEN_IPV6_GLOBAL
 529                     || IN6_IS_ADDR_LINKLOCAL(&ri->attr->mp_nexthop_global))
 530                         return -1;
 531
 532                 p->family = AF_INET6;
 533
 534                 if (is_bgp_static) {
 535                         p->u.prefix6 = ri->net->p.u.prefix6;
 536                         p->prefixlen = ri->net->p.prefixlen;
 537                 } else {
 538                         p->u.prefix6 = ri->attr->mp_nexthop_global;
 539                         p->prefixlen = IPV6_MAX_BITLEN;
 540                 }
 541                 break;
 542         default:
 543                 if (BGP_DEBUG(nht, NHT)) {
 544                         zlog_debug(
 545                                 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
 546                                 __FUNCTION__, afi, AFI_IP, AFI_IP6);
 547                 }
 548                 break;
 549         }
 550         return 0;
 551 }
 552
 553 /**
 554  * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
 555  *   command to Zebra.
 556  * ARGUMENTS:
 557  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 558  *   int command -- command to send to zebra
 559  * RETURNS:
 560  *   void.
 561  */
 562 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache *bnc, int command)
 563 {
 564         struct prefix *p;
 565         bool exact_match = false;
 566         int ret;
 567
 568         if (!zclient)
 569                 return;
 570
 571         /* Don't try to register if Zebra doesn't know of this instance. */
 572         if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc->bgp))
 573                 return;
 574
 575         p = &(bnc->node->p);
 576         if ((command == ZEBRA_NEXTHOP_REGISTER
 577              || command == ZEBRA_IMPORT_ROUTE_REGISTER)
 578             && (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)
 579                 || CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)))
 580                 exact_match = true;
 581
 582         if (BGP_DEBUG(zebra, ZEBRA)) {
 583                 char buf[PREFIX2STR_BUFFER];
 584
 585                 prefix2str(p, buf, PREFIX2STR_BUFFER);
 586                 zlog_debug("%s: sending cmd %s for %s (vrf %s)",
 587                         __func__, zserv_command_string(command), buf,
 588                         bnc->bgp->name);
 589         }
 590
 591         ret = zclient_send_rnh(zclient, command, p, exact_match,
 592                                bnc->bgp->vrf_id);
 593         /* TBD: handle the failure */
 594         if (ret < 0)
 595                 zlog_warn("sendmsg_nexthop: zclient_send_message() failed");
 596
 597         if ((command == ZEBRA_NEXTHOP_REGISTER)
 598             || (command == ZEBRA_IMPORT_ROUTE_REGISTER))
 599                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 600         else if ((command == ZEBRA_NEXTHOP_UNREGISTER)
 601                  || (command == ZEBRA_IMPORT_ROUTE_UNREGISTER))
 602                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 603         return;
 604 }
 605
 606 /**
 607  * register_zebra_rnh - register a NH/route with Zebra for notification
 608  *    when the route or the route to the nexthop changes.
 609  * ARGUMENTS:
 610  *   struct bgp_nexthop_cache *bnc
 611  * RETURNS:
 612  *   void.
 613  */
 614 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
 615                                int is_bgp_import_route)
 616 {
 617         /* Check if we have already registered */
 618         if (bnc->flags & BGP_NEXTHOP_REGISTERED)
 619                 return;
 620         if (is_bgp_import_route)
 621                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_REGISTER);
 622         else
 623                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_REGISTER);
 624 }
 625
 626 /**
 627  * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
 628  * ARGUMENTS:
 629  *   struct bgp_nexthop_cache *bnc
 630  * RETURNS:
 631  *   void.
 632  */
 633 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
 634                                  int is_bgp_import_route)
 635 {
 636         /* Check if we have already registered */
 637         if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 638                 return;
 639
 640         if (is_bgp_import_route)
 641                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_UNREGISTER);
 642         else
 643                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_UNREGISTER);
 644 }
 645
 646 /**
 647  * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
 648  * ARGUMENTS:
 649  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 650  * RETURNS:
 651  *   void.
 652  */
 653 static void evaluate_paths(struct bgp_nexthop_cache *bnc)
 654 {
 655         struct bgp_node *rn;
 656         struct bgp_info *path;
 657         int afi;
 658         struct peer *peer = (struct peer *)bnc->nht_info;
 659         struct bgp_table *table;
 660         safi_t safi;
 661         struct bgp *bgp_path;
 662
 663         if (BGP_DEBUG(nht, NHT)) {
 664                 char buf[PREFIX2STR_BUFFER];
 665                 bnc_str(bnc, buf, PREFIX2STR_BUFFER);
 666                 zlog_debug(
 667                         "NH update for %s - flags 0x%x chgflags 0x%x - evaluate paths",
 668                         buf, bnc->flags, bnc->change_flags);
 669         }
 670
 671         LIST_FOREACH (path, &(bnc->paths), nh_thread) {
 672                 if (!(path->type == ZEBRA_ROUTE_BGP
 673                       && ((path->sub_type == BGP_ROUTE_NORMAL)
 674                           || (path->sub_type == BGP_ROUTE_STATIC)
 675                           || (path->sub_type == BGP_ROUTE_IMPORTED))))
 676                         continue;
 677
 678                 rn = path->net;
 679                 assert(rn && bgp_node_table(rn));
 680                 afi = family2afi(rn->p.family);
 681                 table = bgp_node_table(rn);
 682                 safi = table->safi;
 683
 684                 /*
 685                  * handle routes from other VRFs (they can have a
 686                  * nexthop in THIS VRF). bgp_path is the bgp instance
 687                  * that owns the route referencing this nexthop.
 688                  */
 689                 bgp_path = table->bgp;
 690
 691                 /*
 692                  * Path becomes valid/invalid depending on whether the nexthop
 693                  * reachable/unreachable.
 694                  *
 695                  * In case of unicast routes that were imported from vpn
 696                  * and that have labels, they are valid only if there are
 697                  * nexthops with labels
 698                  */
 699
 700                 int bnc_is_valid_nexthop = 0;
 701
 702                 if (safi == SAFI_UNICAST &&
 703                         path->sub_type == BGP_ROUTE_IMPORTED &&
 704                         path->extra &&
 705                         path->extra->num_labels) {
 706
 707                         bnc_is_valid_nexthop =
 708                                 bgp_isvalid_labeled_nexthop(bnc) ? 1 : 0;
 709                 } else {
 710                         bnc_is_valid_nexthop =
 711                                 bgp_isvalid_nexthop(bnc) ? 1 : 0;
 712                 }
 713
 714                 if (BGP_DEBUG(nht, NHT)) {
 715                         char buf[PREFIX_STRLEN];
 716
 717                         prefix2str(&rn->p, buf, PREFIX_STRLEN);
 718                         zlog_debug("%s: prefix %s (vrf %s) %svalid",
 719                                 __func__, buf, bgp_path->name,
 720                                 (bnc_is_valid_nexthop ? "" : "not "));
 721                 }
 722
 723                 if ((CHECK_FLAG(path->flags, BGP_INFO_VALID) ? 1 : 0)
 724                     != bnc_is_valid_nexthop) {
 725                         if (CHECK_FLAG(path->flags, BGP_INFO_VALID)) {
 726                                 bgp_aggregate_decrement(bgp_path, &rn->p,
 727                                                         path, afi, safi);
 728                                 bgp_info_unset_flag(rn, path, BGP_INFO_VALID);
 729                         } else {
 730                                 bgp_info_set_flag(rn, path, BGP_INFO_VALID);
 731                                 bgp_aggregate_increment(bgp_path, &rn->p,
 732                                                         path, afi, safi);
 733                         }
 734                 }
 735
 736                 /* Copy the metric to the path. Will be used for bestpath
 737                  * computation */
 738                 if (bgp_isvalid_nexthop(bnc) && bnc->metric)
 739                         (bgp_info_extra_get(path))->igpmetric = bnc->metric;
 740                 else if (path->extra)
 741                         path->extra->igpmetric = 0;
 742
 743                 if (CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_METRIC_CHANGED)
 744                     || CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_CHANGED))
 745                         SET_FLAG(path->flags, BGP_INFO_IGP_CHANGED);
 746
 747                 bgp_process(bgp_path, rn, afi, safi);
 748         }
 749
 750         if (peer && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED)) {
 751                 if (BGP_DEBUG(nht, NHT))
 752                         zlog_debug("%s: Updating peer (%s) status with NHT",
 753                                    __FUNCTION__, peer->host);
 754                 bgp_fsm_nht_update(peer, bgp_isvalid_nexthop(bnc));
 755                 SET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 756         }
 757
 758         RESET_FLAG(bnc->change_flags);
 759 }
 760
 761 /**
 762  * path_nh_map - make or break path-to-nexthop association.
 763  * ARGUMENTS:
 764  *   path - pointer to the path structure
 765  *   bnc - pointer to the nexthop structure
 766  *   make - if set, make the association. if unset, just break the existing
 767  *          association.
 768  */
 769 static void path_nh_map(struct bgp_info *path, struct bgp_nexthop_cache *bnc,
 770                         int make)
 771 {
 772         if (path->nexthop) {
 773                 LIST_REMOVE(path, nh_thread);
 774                 path->nexthop->path_count--;
 775                 path->nexthop = NULL;
 776         }
 777         if (make) {
 778                 LIST_INSERT_HEAD(&(bnc->paths), path, nh_thread);
 779                 path->nexthop = bnc;
 780                 path->nexthop->path_count++;
 781         }
 782 }