bgpd/bgp_nht.c

   1 /* BGP Nexthop tracking
   2  * Copyright (C) 2013 Cumulus Networks, Inc.
   3  *
   4  * This file is part of GNU Zebra.
   5  *
   6  * GNU Zebra is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the
   8  * Free Software Foundation; either version 2, or (at your option) any
   9  * later version.
  10  *
  11  * GNU Zebra is distributed in the hope that it will be useful, but
  12  * WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License along
  17  * with this program; see the file COPYING; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include <zebra.h>
  22
  23 #include "command.h"
  24 #include "thread.h"
  25 #include "prefix.h"
  26 #include "zclient.h"
  27 #include "stream.h"
  28 #include "network.h"
  29 #include "log.h"
  30 #include "memory.h"
  31 #include "nexthop.h"
  32 #include "vrf.h"
  33 #include "filter.h"
  34
  35 #include "bgpd/bgpd.h"
  36 #include "bgpd/bgp_table.h"
  37 #include "bgpd/bgp_route.h"
  38 #include "bgpd/bgp_attr.h"
  39 #include "bgpd/bgp_nexthop.h"
  40 #include "bgpd/bgp_debug.h"
  41 #include "bgpd/bgp_errors.h"
  42 #include "bgpd/bgp_nht.h"
  43 #include "bgpd/bgp_fsm.h"
  44 #include "bgpd/bgp_zebra.h"
  45 #include "bgpd/bgp_flowspec_util.h"
  46 #include "bgpd/bgp_evpn.h"
  47
  48 extern struct zclient *zclient;
  49
  50 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
  51                                int is_bgp_static_route);
  52 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
  53                                  int is_bgp_static_route);
  54 static void evaluate_paths(struct bgp_nexthop_cache *bnc);
  55 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p);
  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 static void bgp_unlink_nexthop_check(struct bgp_nexthop_cache *bnc)
  70 {
  71         if (LIST_EMPTY(&(bnc->paths)) && !bnc->nht_info) {
  72                 if (BGP_DEBUG(nht, NHT)) {
  73                         char buf[PREFIX2STR_BUFFER];
  74                         zlog_debug("bgp_unlink_nexthop: freeing bnc %s(%s)",
  75                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER),
  76                                    bnc->bgp->name_pretty);
  77                 }
  78                 unregister_zebra_rnh(bnc,
  79                                      CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE));
  80                 bgp_node_set_bgp_nexthop_info(bnc->node, NULL);
  81                 bgp_unlock_node(bnc->node);
  82                 bnc->node = NULL;
  83                 bnc_free(bnc);
  84         }
  85 }
  86
  87 void bgp_unlink_nexthop(struct bgp_path_info *path)
  88 {
  89         struct bgp_nexthop_cache *bnc = path->nexthop;
  90
  91         if (!bnc)
  92                 return;
  93
  94         path_nh_map(path, NULL, false);
  95
  96         bgp_unlink_nexthop_check(bnc);
  97 }
  98
  99 void bgp_unlink_nexthop_by_peer(struct peer *peer)
 100 {
 101         struct prefix p;
 102         struct bgp_node *rn;
 103         struct bgp_nexthop_cache *bnc;
 104         afi_t afi = family2afi(peer->su.sa.sa_family);
 105
 106         if (!sockunion2hostprefix(&peer->su, &p))
 107                 return;
 108
 109         rn = bgp_node_get(peer->bgp->nexthop_cache_table[afi], &p);
 110
 111         bnc = bgp_node_get_bgp_nexthop_info(rn);
 112         if (!bnc)
 113                 return;
 114
 115         /* cleanup the peer reference */
 116         bnc->nht_info = NULL;
 117
 118         bgp_unlink_nexthop_check(bnc);
 119 }
 120
 121 /*
 122  * A route and its nexthop might belong to different VRFs. Therefore,
 123  * we need both the bgp_route and bgp_nexthop pointers.
 124  */
 125 int bgp_find_or_add_nexthop(struct bgp *bgp_route, struct bgp *bgp_nexthop,
 126                             afi_t afi, struct bgp_path_info *pi,
 127                             struct peer *peer, int connected)
 128 {
 129         struct bgp_node *rn;
 130         struct bgp_nexthop_cache *bnc;
 131         struct prefix p;
 132         int is_bgp_static_route = 0;
 133         const struct prefix *bnc_p;
 134
 135         if (pi) {
 136                 is_bgp_static_route = ((pi->type == ZEBRA_ROUTE_BGP)
 137                                        && (pi->sub_type == BGP_ROUTE_STATIC))
 138                                               ? 1
 139                                               : 0;
 140
 141                 /* Since Extended Next-hop Encoding (RFC5549) support, we want
 142                    to derive
 143                    address-family from the next-hop. */
 144                 if (!is_bgp_static_route)
 145                         afi = BGP_ATTR_NEXTHOP_AFI_IP6(pi->attr) ? AFI_IP6
 146                                                                  : AFI_IP;
 147
 148                 /* This will return true if the global IPv6 NH is a link local
 149                  * addr */
 150                 if (make_prefix(afi, pi, &p) < 0)
 151                         return 1;
 152         } else if (peer) {
 153                 if (!sockunion2hostprefix(&peer->su, &p)) {
 154                         if (BGP_DEBUG(nht, NHT)) {
 155                                 zlog_debug(
 156                                         "%s: Attempting to register with unknown AFI %d (not %d or %d)",
 157                                         __func__, afi, AFI_IP, AFI_IP6);
 158                         }
 159                         return 0;
 160                 }
 161         } else
 162                 return 0;
 163
 164         if (is_bgp_static_route)
 165                 rn = bgp_node_get(bgp_nexthop->import_check_table[afi], &p);
 166         else
 167                 rn = bgp_node_get(bgp_nexthop->nexthop_cache_table[afi], &p);
 168
 169         bnc = bgp_node_get_bgp_nexthop_info(rn);
 170         if (!bnc) {
 171                 bnc = bnc_new();
 172                 bgp_node_set_bgp_nexthop_info(rn, bnc);
 173                 bnc->node = rn;
 174                 bnc->bgp = bgp_nexthop;
 175                 bgp_lock_node(rn);
 176                 if (BGP_DEBUG(nht, NHT)) {
 177                         char buf[PREFIX2STR_BUFFER];
 178
 179                         zlog_debug("Allocated bnc %s(%s) peer %p",
 180                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER),
 181                                    bnc->bgp->name_pretty, peer);
 182                 }
 183         }
 184
 185         bnc_p = bgp_node_get_prefix(bnc->node);
 186
 187         bgp_unlock_node(rn);
 188         if (is_bgp_static_route) {
 189                 SET_FLAG(bnc->flags, BGP_STATIC_ROUTE);
 190
 191                 /* If we're toggling the type, re-register */
 192                 if ((CHECK_FLAG(bgp_route->flags, BGP_FLAG_IMPORT_CHECK))
 193                     && !CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)) {
 194                         SET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 195                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 196                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 197                 } else if ((!CHECK_FLAG(bgp_route->flags,
 198                                         BGP_FLAG_IMPORT_CHECK))
 199                            && CHECK_FLAG(bnc->flags,
 200                                          BGP_STATIC_ROUTE_EXACT_MATCH)) {
 201                         UNSET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 202                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 203                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 204                 }
 205         }
 206         /* When nexthop is already known, but now requires 'connected'
 207          * resolution,
 208          * re-register it. The reverse scenario where the nexthop currently
 209          * requires
 210          * 'connected' resolution does not need a re-register (i.e., we treat
 211          * 'connected-required' as an override) except in the scenario where
 212          * this
 213          * is actually a case of tracking a peer for connectivity (e.g., after
 214          * disable connected-check).
 215          * NOTE: We don't track the number of paths separately for 'connected-
 216          * required' vs 'connected-not-required' as this change is not a common
 217          * scenario.
 218          */
 219         else if (connected && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 220                 SET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 221                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 222                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 223         } else if (peer && !connected
 224                    && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 225                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 226                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 227                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 228         }
 229         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW) {
 230                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 231                 SET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 232         } else if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED)
 233                    && !is_default_host_route(bnc_p))
 234                 register_zebra_rnh(bnc, is_bgp_static_route);
 235
 236         if (pi && pi->nexthop != bnc) {
 237                 /* Unlink from existing nexthop cache, if any. This will also
 238                  * free
 239                  * the nexthop cache entry, if appropriate.
 240                  */
 241                 bgp_unlink_nexthop(pi);
 242
 243                 /* updates NHT pi list reference */
 244                 path_nh_map(pi, bnc, true);
 245
 246                 if (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID) && bnc->metric)
 247                         (bgp_path_info_extra_get(pi))->igpmetric = bnc->metric;
 248                 else if (pi->extra)
 249                         pi->extra->igpmetric = 0;
 250         } else if (peer)
 251                 bnc->nht_info = (void *)peer; /* NHT peer reference */
 252
 253         /*
 254          * We are cheating here.  Views have no associated underlying
 255          * ability to detect nexthops.  So when we have a view
 256          * just tell everyone the nexthop is valid
 257          */
 258         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW)
 259                 return 1;
 260         else
 261                 return (bgp_isvalid_nexthop(bnc));
 262 }
 263
 264 void bgp_delete_connected_nexthop(afi_t afi, struct peer *peer)
 265 {
 266         struct bgp_node *rn;
 267         struct bgp_nexthop_cache *bnc;
 268         struct prefix p;
 269
 270         if (!peer)
 271                 return;
 272
 273         if (!sockunion2hostprefix(&peer->su, &p))
 274                 return;
 275
 276         rn = bgp_node_lookup(
 277                 peer->bgp->nexthop_cache_table[family2afi(p.family)], &p);
 278         if (!rn) {
 279                 if (BGP_DEBUG(nht, NHT))
 280                         zlog_debug(
 281                                 "Cannot find connected NHT node for peer %s(%s)",
 282                                 peer->host, peer->bgp->name_pretty);
 283                 return;
 284         }
 285
 286         bnc = bgp_node_get_bgp_nexthop_info(rn);
 287         if (!bnc) {
 288                 if (BGP_DEBUG(nht, NHT))
 289                         zlog_debug(
 290                                 "Cannot find connected NHT node for peer %s(%s) on route_node as expected",
 291                                 peer->host, peer->bgp->name_pretty);
 292                 bgp_unlock_node(rn);
 293                 return;
 294         }
 295         bgp_unlock_node(rn);
 296
 297         if (bnc->nht_info != peer) {
 298                 if (BGP_DEBUG(nht, NHT))
 299                         zlog_debug(
 300                                 "Connected NHT %p node for peer %s(%s) points to %p",
 301                                 bnc, peer->host, bnc->bgp->name_pretty,
 302                                 bnc->nht_info);
 303                 return;
 304         }
 305
 306         bnc->nht_info = NULL;
 307
 308         if (LIST_EMPTY(&(bnc->paths))) {
 309                 if (BGP_DEBUG(nht, NHT))
 310                         zlog_debug(
 311                                 "Freeing connected NHT node %p for peer %s(%s)",
 312                                 bnc, peer->host, bnc->bgp->name_pretty);
 313                 unregister_zebra_rnh(bnc, 0);
 314                 bgp_node_set_bgp_nexthop_info(bnc->node, NULL);
 315                 bgp_unlock_node(bnc->node);
 316                 bnc_free(bnc);
 317         }
 318 }
 319
 320 void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 321 {
 322         struct bgp_node *rn = NULL;
 323         struct bgp_nexthop_cache *bnc;
 324         struct nexthop *nexthop;
 325         struct nexthop *oldnh;
 326         struct nexthop *nhlist_head = NULL;
 327         struct nexthop *nhlist_tail = NULL;
 328         int i;
 329         struct bgp *bgp;
 330         struct zapi_route nhr;
 331
 332         bgp = bgp_lookup_by_vrf_id(vrf_id);
 333         if (!bgp) {
 334                 flog_err(
 335                         EC_BGP_NH_UPD,
 336                         "parse nexthop update: instance not found for vrf_id %u",
 337                         vrf_id);
 338                 return;
 339         }
 340
 341         if (!zapi_nexthop_update_decode(zclient->ibuf, &nhr)) {
 342                 if (BGP_DEBUG(nht, NHT))
 343                         zlog_debug("%s[%s]: Failure to decode nexthop update",
 344                                    __func__, bgp->name_pretty);
 345                 return;
 346         }
 347
 348         if (command == ZEBRA_NEXTHOP_UPDATE)
 349                 rn = bgp_node_lookup(
 350                         bgp->nexthop_cache_table[family2afi(nhr.prefix.family)],
 351                         &nhr.prefix);
 352         else if (command == ZEBRA_IMPORT_CHECK_UPDATE)
 353                 rn = bgp_node_lookup(
 354                         bgp->import_check_table[family2afi(nhr.prefix.family)],
 355                         &nhr.prefix);
 356
 357         if (!rn) {
 358                 if (BGP_DEBUG(nht, NHT)) {
 359                         char buf[PREFIX2STR_BUFFER];
 360                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 361                         zlog_debug("parse nexthop update(%s(%s)): rn not found",
 362                                    buf, bgp->name_pretty);
 363                 }
 364                 return;
 365         }
 366
 367         bnc = bgp_node_get_bgp_nexthop_info(rn);
 368         if (!bnc) {
 369                 if (BGP_DEBUG(nht, NHT)) {
 370                         char buf[PREFIX2STR_BUFFER];
 371
 372                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 373                         zlog_debug(
 374                                 "parse nexthop update(%s(%s)): bnc node info not found",
 375                                 buf, bgp->name_pretty);
 376                 }
 377                 bgp_unlock_node(rn);
 378                 return;
 379         }
 380
 381         bgp_unlock_node(rn);
 382         bnc->last_update = bgp_clock();
 383         bnc->change_flags = 0;
 384
 385         /* debug print the input */
 386         if (BGP_DEBUG(nht, NHT)) {
 387                 char buf[PREFIX2STR_BUFFER];
 388                 prefix2str(&nhr.prefix, buf, sizeof(buf));
 389                 zlog_debug(
 390                         "%s(%u): Rcvd NH update %s - metric %d/%d #nhops %d/%d flags 0x%x",
 391                         bnc->bgp->name_pretty, vrf_id, buf, nhr.metric,
 392                         bnc->metric, nhr.nexthop_num, bnc->nexthop_num,
 393                         bnc->flags);
 394         }
 395
 396         if (nhr.metric != bnc->metric)
 397                 bnc->change_flags |= BGP_NEXTHOP_METRIC_CHANGED;
 398
 399         if (nhr.nexthop_num != bnc->nexthop_num)
 400                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 401
 402         if (nhr.nexthop_num) {
 403                 struct peer *peer = bnc->nht_info;
 404
 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 = nhr.metric;
 411                 bnc->nexthop_num = nhr.nexthop_num;
 412
 413                 bnc->flags &= ~BGP_NEXTHOP_LABELED_VALID; /* check below */
 414
 415                 for (i = 0; i < nhr.nexthop_num; i++) {
 416                         int num_labels = 0;
 417
 418                         nexthop = nexthop_from_zapi_nexthop(&nhr.nexthops[i]);
 419
 420                         /*
 421                          * Turn on RA for the v6 nexthops
 422                          * we receive from bgp.  This is to allow us
 423                          * to work with v4 routing over v6 nexthops
 424                          */
 425                         if (peer && !peer->ifp
 426                             && CHECK_FLAG(peer->flags,
 427                                           PEER_FLAG_CAPABILITY_ENHE)
 428                             && nhr.prefix.family == AF_INET6
 429                             && nexthop->type != NEXTHOP_TYPE_BLACKHOLE) {
 430                                 struct interface *ifp;
 431
 432                                 ifp = if_lookup_by_index(nexthop->ifindex,
 433                                                          nexthop->vrf_id);
 434                                 if (ifp)
 435                                         zclient_send_interface_radv_req(
 436                                                 zclient, nexthop->vrf_id, ifp,
 437                                                 true,
 438                                                 BGP_UNNUM_DEFAULT_RA_INTERVAL);
 439                         }
 440                         /* There is at least one label-switched path */
 441                         if (nexthop->nh_label &&
 442                                 nexthop->nh_label->num_labels) {
 443
 444                                 bnc->flags |= BGP_NEXTHOP_LABELED_VALID;
 445                                 num_labels = nexthop->nh_label->num_labels;
 446                         }
 447
 448                         if (BGP_DEBUG(nht, NHT)) {
 449                                 char buf[NEXTHOP_STRLEN];
 450                                 zlog_debug(
 451                                         "    nhop via %s (%d labels)",
 452                                         nexthop2str(nexthop, buf, sizeof(buf)),
 453                                         num_labels);
 454                         }
 455
 456                         if (nhlist_tail) {
 457                                 nhlist_tail->next = nexthop;
 458                                 nhlist_tail = nexthop;
 459                         } else {
 460                                 nhlist_tail = nexthop;
 461                                 nhlist_head = nexthop;
 462                         }
 463
 464                         /* No need to evaluate the nexthop if we have already
 465                          * determined
 466                          * that there has been a change.
 467                          */
 468                         if (bnc->change_flags & BGP_NEXTHOP_CHANGED)
 469                                 continue;
 470
 471                         for (oldnh = bnc->nexthop; oldnh; oldnh = oldnh->next)
 472                                 if (nexthop_same(oldnh, nexthop))
 473                                         break;
 474
 475                         if (!oldnh)
 476                                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 477                 }
 478                 bnc_nexthop_free(bnc);
 479                 bnc->nexthop = nhlist_head;
 480         } else {
 481                 bnc->flags &= ~BGP_NEXTHOP_VALID;
 482                 bnc->nexthop_num = nhr.nexthop_num;
 483
 484                 /* notify bgp fsm if nbr ip goes from valid->invalid */
 485                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 486
 487                 bnc_nexthop_free(bnc);
 488                 bnc->nexthop = NULL;
 489         }
 490
 491         evaluate_paths(bnc);
 492 }
 493
 494 /*
 495  * Cleanup nexthop registration and status information for BGP nexthops
 496  * pertaining to this VRF. This is invoked upon VRF deletion.
 497  */
 498 void bgp_cleanup_nexthops(struct bgp *bgp)
 499 {
 500         afi_t afi;
 501         struct bgp_node *rn;
 502         struct bgp_nexthop_cache *bnc;
 503
 504         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 505                 if (!bgp->nexthop_cache_table[afi])
 506                         continue;
 507
 508                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 509                      rn = bgp_route_next(rn)) {
 510                         bnc = bgp_node_get_bgp_nexthop_info(rn);
 511                         if (!bnc)
 512                                 continue;
 513
 514                         /* Clear relevant flags. */
 515                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 516                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 517                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 518                 }
 519         }
 520 }
 521
 522 /**
 523  * make_prefix - make a prefix structure from the path (essentially
 524  * path's node.
 525  */
 526 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p)
 527 {
 528
 529         int is_bgp_static = ((pi->type == ZEBRA_ROUTE_BGP)
 530                              && (pi->sub_type == BGP_ROUTE_STATIC))
 531                                     ? 1
 532                                     : 0;
 533         struct bgp_node *net = pi->net;
 534         const struct prefix *p_orig = bgp_node_get_prefix(net);
 535
 536         if (p_orig->family == AF_FLOWSPEC) {
 537                 if (!pi->peer)
 538                         return -1;
 539                 return bgp_flowspec_get_first_nh(pi->peer->bgp,
 540                                                  pi, p);
 541         }
 542         memset(p, 0, sizeof(struct prefix));
 543         switch (afi) {
 544         case AFI_IP:
 545                 p->family = AF_INET;
 546                 if (is_bgp_static) {
 547                         p->u.prefix4 = p_orig->u.prefix4;
 548                         p->prefixlen = p_orig->prefixlen;
 549                 } else {
 550                         p->u.prefix4 = pi->attr->nexthop;
 551                         p->prefixlen = IPV4_MAX_BITLEN;
 552                 }
 553                 break;
 554         case AFI_IP6:
 555                 p->family = AF_INET6;
 556
 557                 if (is_bgp_static) {
 558                         p->u.prefix6 = p_orig->u.prefix6;
 559                         p->prefixlen = p_orig->prefixlen;
 560                 } else {
 561                         p->u.prefix6 = pi->attr->mp_nexthop_global;
 562                         p->prefixlen = IPV6_MAX_BITLEN;
 563                 }
 564                 break;
 565         default:
 566                 if (BGP_DEBUG(nht, NHT)) {
 567                         zlog_debug(
 568                                 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
 569                                 __func__, afi, AFI_IP, AFI_IP6);
 570                 }
 571                 break;
 572         }
 573         return 0;
 574 }
 575
 576 /**
 577  * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
 578  *   command to Zebra.
 579  * ARGUMENTS:
 580  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 581  *   int command -- command to send to zebra
 582  * RETURNS:
 583  *   void.
 584  */
 585 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache *bnc, int command)
 586 {
 587         const struct prefix *p;
 588         bool exact_match = false;
 589         int ret;
 590
 591         if (!zclient)
 592                 return;
 593
 594         /* Don't try to register if Zebra doesn't know of this instance. */
 595         if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc->bgp)) {
 596                 if (BGP_DEBUG(zebra, ZEBRA))
 597                         zlog_debug(
 598                                 "%s: No zebra instance to talk to, not installing NHT entry",
 599                                 __func__);
 600                 return;
 601         }
 602
 603         if (!bgp_zebra_num_connects()) {
 604                 if (BGP_DEBUG(zebra, ZEBRA))
 605                         zlog_debug(
 606                                 "%s: We have not connected yet, cannot send nexthops",
 607                                 __func__);
 608         }
 609         p = bgp_node_get_prefix(bnc->node);
 610         if ((command == ZEBRA_NEXTHOP_REGISTER
 611              || command == ZEBRA_IMPORT_ROUTE_REGISTER)
 612             && (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)
 613                 || CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)))
 614                 exact_match = true;
 615
 616         if (BGP_DEBUG(zebra, ZEBRA)) {
 617                 char buf[PREFIX2STR_BUFFER];
 618
 619                 prefix2str(p, buf, PREFIX2STR_BUFFER);
 620                 zlog_debug("%s: sending cmd %s for %s (vrf %s)",
 621                         __func__, zserv_command_string(command), buf,
 622                         bnc->bgp->name_pretty);
 623         }
 624
 625         ret = zclient_send_rnh(zclient, command, p, exact_match,
 626                                bnc->bgp->vrf_id);
 627         /* TBD: handle the failure */
 628         if (ret < 0)
 629                 flog_warn(EC_BGP_ZEBRA_SEND,
 630                           "sendmsg_nexthop: zclient_send_message() failed");
 631
 632         if ((command == ZEBRA_NEXTHOP_REGISTER)
 633             || (command == ZEBRA_IMPORT_ROUTE_REGISTER))
 634                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 635         else if ((command == ZEBRA_NEXTHOP_UNREGISTER)
 636                  || (command == ZEBRA_IMPORT_ROUTE_UNREGISTER))
 637                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 638         return;
 639 }
 640
 641 /**
 642  * register_zebra_rnh - register a NH/route with Zebra for notification
 643  *    when the route or the route to the nexthop changes.
 644  * ARGUMENTS:
 645  *   struct bgp_nexthop_cache *bnc
 646  * RETURNS:
 647  *   void.
 648  */
 649 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
 650                                int is_bgp_import_route)
 651 {
 652         /* Check if we have already registered */
 653         if (bnc->flags & BGP_NEXTHOP_REGISTERED)
 654                 return;
 655         if (is_bgp_import_route)
 656                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_REGISTER);
 657         else
 658                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_REGISTER);
 659 }
 660
 661 /**
 662  * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
 663  * ARGUMENTS:
 664  *   struct bgp_nexthop_cache *bnc
 665  * RETURNS:
 666  *   void.
 667  */
 668 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
 669                                  int is_bgp_import_route)
 670 {
 671         /* Check if we have already registered */
 672         if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 673                 return;
 674
 675         if (is_bgp_import_route)
 676                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_UNREGISTER);
 677         else
 678                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_UNREGISTER);
 679 }
 680
 681 /**
 682  * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
 683  * ARGUMENTS:
 684  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 685  * RETURNS:
 686  *   void.
 687  */
 688 static void evaluate_paths(struct bgp_nexthop_cache *bnc)
 689 {
 690         struct bgp_node *rn;
 691         struct bgp_path_info *path;
 692         int afi;
 693         struct peer *peer = (struct peer *)bnc->nht_info;
 694         struct bgp_table *table;
 695         safi_t safi;
 696         struct bgp *bgp_path;
 697         const struct prefix *p;
 698
 699         if (BGP_DEBUG(nht, NHT)) {
 700                 char buf[PREFIX2STR_BUFFER];
 701                 bnc_str(bnc, buf, PREFIX2STR_BUFFER);
 702                 zlog_debug(
 703                         "NH update for %s(%s) - flags 0x%x chgflags 0x%x - evaluate paths",
 704                         buf, bnc->bgp->name_pretty, bnc->flags,
 705                         bnc->change_flags);
 706         }
 707
 708         LIST_FOREACH (path, &(bnc->paths), nh_thread) {
 709                 if (!(path->type == ZEBRA_ROUTE_BGP
 710                       && ((path->sub_type == BGP_ROUTE_NORMAL)
 711                           || (path->sub_type == BGP_ROUTE_STATIC)
 712                           || (path->sub_type == BGP_ROUTE_IMPORTED))))
 713                         continue;
 714
 715                 rn = path->net;
 716                 assert(rn && bgp_node_table(rn));
 717                 p = bgp_node_get_prefix(rn);
 718                 afi = family2afi(p->family);
 719                 table = bgp_node_table(rn);
 720                 safi = table->safi;
 721
 722                 /*
 723                  * handle routes from other VRFs (they can have a
 724                  * nexthop in THIS VRF). bgp_path is the bgp instance
 725                  * that owns the route referencing this nexthop.
 726                  */
 727                 bgp_path = table->bgp;
 728
 729                 /*
 730                  * Path becomes valid/invalid depending on whether the nexthop
 731                  * reachable/unreachable.
 732                  *
 733                  * In case of unicast routes that were imported from vpn
 734                  * and that have labels, they are valid only if there are
 735                  * nexthops with labels
 736                  */
 737
 738                 int bnc_is_valid_nexthop = 0;
 739
 740                 if (safi == SAFI_UNICAST &&
 741                         path->sub_type == BGP_ROUTE_IMPORTED &&
 742                         path->extra &&
 743                         path->extra->num_labels) {
 744
 745                         bnc_is_valid_nexthop =
 746                                 bgp_isvalid_labeled_nexthop(bnc) ? 1 : 0;
 747                 } else {
 748                         bnc_is_valid_nexthop =
 749                                 bgp_isvalid_nexthop(bnc) ? 1 : 0;
 750                 }
 751
 752                 if (BGP_DEBUG(nht, NHT))
 753                         zlog_debug("%s: prefix %pRN (vrf %s) %svalid", __func__,
 754                                    rn, bgp_path->name,
 755                                    (bnc_is_valid_nexthop ? "" : "not "));
 756
 757                 if ((CHECK_FLAG(path->flags, BGP_PATH_VALID) ? 1 : 0)
 758                     != bnc_is_valid_nexthop) {
 759                         if (CHECK_FLAG(path->flags, BGP_PATH_VALID)) {
 760                                 bgp_aggregate_decrement(bgp_path, p, path, afi,
 761                                                         safi);
 762                                 bgp_path_info_unset_flag(rn, path,
 763                                                          BGP_PATH_VALID);
 764                         } else {
 765                                 bgp_path_info_set_flag(rn, path,
 766                                                        BGP_PATH_VALID);
 767                                 bgp_aggregate_increment(bgp_path, p, path, afi,
 768                                                         safi);
 769                         }
 770                 }
 771
 772                 /* Copy the metric to the path. Will be used for bestpath
 773                  * computation */
 774                 if (bgp_isvalid_nexthop(bnc) && bnc->metric)
 775                         (bgp_path_info_extra_get(path))->igpmetric =
 776                                 bnc->metric;
 777                 else if (path->extra)
 778                         path->extra->igpmetric = 0;
 779
 780                 if (CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_METRIC_CHANGED)
 781                     || CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_CHANGED))
 782                         SET_FLAG(path->flags, BGP_PATH_IGP_CHANGED);
 783
 784                 if (safi == SAFI_EVPN && bgp_evpn_is_prefix_nht_supported(p)) {
 785                         if (CHECK_FLAG(path->flags, BGP_PATH_VALID))
 786                                 bgp_evpn_import_route(bgp_path, afi, safi, p,
 787                                                       path);
 788                         else
 789                                 bgp_evpn_unimport_route(bgp_path, afi, safi, p,
 790                                                         path);
 791                 }
 792
 793                 bgp_process(bgp_path, rn, afi, safi);
 794         }
 795
 796         if (peer) {
 797                 int valid_nexthops = bgp_isvalid_nexthop(bnc);
 798
 799                 if (valid_nexthops)
 800                         peer->last_reset = PEER_DOWN_WAITING_OPEN;
 801                 else
 802                         peer->last_reset = PEER_DOWN_WAITING_NHT;
 803
 804                 if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED)) {
 805                         if (BGP_DEBUG(nht, NHT))
 806                                 zlog_debug(
 807                                         "%s: Updating peer (%s(%s)) status with NHT",
 808                                         __func__, peer->host,
 809                                         peer->bgp->name_pretty);
 810                         bgp_fsm_event_update(peer, valid_nexthops);
 811                         SET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 812                 }
 813         }
 814
 815         RESET_FLAG(bnc->change_flags);
 816 }
 817
 818 /**
 819  * path_nh_map - make or break path-to-nexthop association.
 820  * ARGUMENTS:
 821  *   path - pointer to the path structure
 822  *   bnc - pointer to the nexthop structure
 823  *   make - if set, make the association. if unset, just break the existing
 824  *          association.
 825  */
 826 void path_nh_map(struct bgp_path_info *path, struct bgp_nexthop_cache *bnc,
 827                  bool make)
 828 {
 829         if (path->nexthop) {
 830                 LIST_REMOVE(path, nh_thread);
 831                 path->nexthop->path_count--;
 832                 path->nexthop = NULL;
 833         }
 834         if (make) {
 835                 LIST_INSERT_HEAD(&(bnc->paths), path, nh_thread);
 836                 path->nexthop = bnc;
 837                 path->nexthop->path_count++;
 838         }
 839 }
 840
 841 /*
 842  * This function is called to register nexthops to zebra
 843  * as that we may have tried to install the nexthops
 844  * before we actually have a zebra connection
 845  */
 846 void bgp_nht_register_nexthops(struct bgp *bgp)
 847 {
 848         struct bgp_node *rn;
 849         struct bgp_nexthop_cache *bnc;
 850         afi_t afi;
 851
 852         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 853                 if (!bgp->nexthop_cache_table[afi])
 854                         continue;
 855
 856                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 857                      rn = bgp_route_next(rn)) {
 858                         bnc = bgp_node_get_bgp_nexthop_info(rn);
 859
 860                         if (!bnc)
 861                                 continue;
 862
 863                         register_zebra_rnh(bnc, 0);
 864                 }
 865         }
 866 }
 867
 868 void bgp_nht_register_enhe_capability_interfaces(struct peer *peer)
 869 {
 870         struct bgp *bgp;
 871         struct bgp_node *rn;
 872         struct bgp_nexthop_cache *bnc;
 873         struct nexthop *nhop;
 874         struct interface *ifp;
 875         struct prefix p;
 876
 877         if (peer->ifp)
 878                 return;
 879
 880         bgp = peer->bgp;
 881
 882         if (!bgp->nexthop_cache_table[AFI_IP6])
 883                 return;
 884
 885         if (!sockunion2hostprefix(&peer->su, &p)) {
 886                 if (BGP_DEBUG(nht, NHT))
 887                         zlog_debug("%s: Unable to convert prefix to sockunion",
 888                                    __func__);
 889                 return;
 890         }
 891
 892         if (p.family != AF_INET6)
 893                 return;
 894         rn = bgp_node_lookup(bgp->nexthop_cache_table[AFI_IP6], &p);
 895         if (!rn)
 896                 return;
 897
 898         bnc = bgp_node_get_bgp_nexthop_info(rn);
 899         if (!bnc)
 900                 return;
 901
 902         if (peer != bnc->nht_info)
 903                 return;
 904
 905         for (nhop = bnc->nexthop; nhop; nhop = nhop->next) {
 906                 ifp = if_lookup_by_index(nhop->ifindex, nhop->vrf_id);
 907
 908                 if (!ifp)
 909                         continue;
 910
 911                 zclient_send_interface_radv_req(zclient,
 912                                                 nhop->vrf_id,
 913                                                 ifp, true,
 914                                                 BGP_UNNUM_DEFAULT_RA_INTERVAL);
 915         }
 916 }