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