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