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 #include "bgpd/bgp_rd.h"
  48
  49 extern struct zclient *zclient;
  50
  51 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
  52                                int is_bgp_static_route);
  53 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
  54                                  int is_bgp_static_route);
  55 static void evaluate_paths(struct bgp_nexthop_cache *bnc);
  56 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p);
  57
  58 static int bgp_isvalid_nexthop(struct bgp_nexthop_cache *bnc)
  59 {
  60         return (bgp_zebra_num_connects() == 0
  61                 || (bnc && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID)));
  62 }
  63
  64 static int bgp_isvalid_labeled_nexthop(struct bgp_nexthop_cache *bnc)
  65 {
  66         return (bgp_zebra_num_connects() == 0
  67                 || (bnc && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_LABELED_VALID)));
  68 }
  69
  70 static void bgp_unlink_nexthop_check(struct bgp_nexthop_cache *bnc)
  71 {
  72         if (LIST_EMPTY(&(bnc->paths)) && !bnc->nht_info) {
  73                 if (BGP_DEBUG(nht, NHT)) {
  74                         char buf[PREFIX2STR_BUFFER];
  75                         zlog_debug("bgp_unlink_nexthop: freeing bnc %s(%s)",
  76                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER),
  77                                    bnc->bgp->name_pretty);
  78                 }
  79                 unregister_zebra_rnh(bnc,
  80                                      CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE));
  81                 bgp_dest_set_bgp_nexthop_info(bnc->dest, NULL);
  82                 bgp_dest_unlock_node(bnc->dest);
  83                 bnc->dest = NULL;
  84                 bnc_free(bnc);
  85         }
  86 }
  87
  88 void bgp_unlink_nexthop(struct bgp_path_info *path)
  89 {
  90         struct bgp_nexthop_cache *bnc = path->nexthop;
  91
  92         if (!bnc)
  93                 return;
  94
  95         path_nh_map(path, NULL, false);
  96
  97         bgp_unlink_nexthop_check(bnc);
  98 }
  99
 100 void bgp_unlink_nexthop_by_peer(struct peer *peer)
 101 {
 102         struct prefix p;
 103         struct bgp_dest *dest;
 104         struct bgp_nexthop_cache *bnc;
 105         afi_t afi = family2afi(peer->su.sa.sa_family);
 106
 107         if (!sockunion2hostprefix(&peer->su, &p))
 108                 return;
 109
 110         dest = bgp_node_get(peer->bgp->nexthop_cache_table[afi], &p);
 111
 112         bnc = bgp_dest_get_bgp_nexthop_info(dest);
 113         if (!bnc)
 114                 return;
 115
 116         /* cleanup the peer reference */
 117         bnc->nht_info = NULL;
 118
 119         bgp_unlink_nexthop_check(bnc);
 120 }
 121
 122 /*
 123  * A route and its nexthop might belong to different VRFs. Therefore,
 124  * we need both the bgp_route and bgp_nexthop pointers.
 125  */
 126 int bgp_find_or_add_nexthop(struct bgp *bgp_route, struct bgp *bgp_nexthop,
 127                             afi_t afi, struct bgp_path_info *pi,
 128                             struct peer *peer, int connected)
 129 {
 130         struct bgp_dest *dest;
 131         struct bgp_nexthop_cache *bnc;
 132         struct prefix p;
 133         int is_bgp_static_route = 0;
 134         const struct prefix *bnc_p;
 135
 136         if (pi) {
 137                 is_bgp_static_route = ((pi->type == ZEBRA_ROUTE_BGP)
 138                                        && (pi->sub_type == BGP_ROUTE_STATIC))
 139                                               ? 1
 140                                               : 0;
 141
 142                 /* Since Extended Next-hop Encoding (RFC5549) support, we want
 143                    to derive
 144                    address-family from the next-hop. */
 145                 if (!is_bgp_static_route)
 146                         afi = BGP_ATTR_NEXTHOP_AFI_IP6(pi->attr) ? AFI_IP6
 147                                                                  : AFI_IP;
 148
 149                 /* Validation for the ipv4 mapped ipv6 nexthop. */
 150                 if (IS_MAPPED_IPV6(&pi->attr->mp_nexthop_global)) {
 151                         afi = AFI_IP;
 152                 }
 153
 154                 /* This will return true if the global IPv6 NH is a link local
 155                  * addr */
 156                 if (make_prefix(afi, pi, &p) < 0)
 157                         return 1;
 158         } else if (peer) {
 159                 if (!sockunion2hostprefix(&peer->su, &p)) {
 160                         if (BGP_DEBUG(nht, NHT)) {
 161                                 zlog_debug(
 162                                         "%s: Attempting to register with unknown AFI %d (not %d or %d)",
 163                                         __func__, afi, AFI_IP, AFI_IP6);
 164                         }
 165                         return 0;
 166                 }
 167         } else
 168                 return 0;
 169
 170         if (is_bgp_static_route)
 171                 dest = bgp_node_get(bgp_nexthop->import_check_table[afi], &p);
 172         else
 173                 dest = bgp_node_get(bgp_nexthop->nexthop_cache_table[afi], &p);
 174
 175         bnc = bgp_dest_get_bgp_nexthop_info(dest);
 176         if (!bnc) {
 177                 bnc = bnc_new();
 178                 bgp_dest_set_bgp_nexthop_info(dest, bnc);
 179                 bnc->dest = dest;
 180                 bnc->bgp = bgp_nexthop;
 181                 bgp_dest_lock_node(dest);
 182                 if (BGP_DEBUG(nht, NHT)) {
 183                         char buf[PREFIX2STR_BUFFER];
 184
 185                         zlog_debug("Allocated bnc %s(%s) peer %p",
 186                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER),
 187                                    bnc->bgp->name_pretty, peer);
 188                 }
 189         }
 190
 191         bnc_p = bgp_dest_get_prefix(bnc->dest);
 192
 193         bgp_dest_unlock_node(dest);
 194         if (is_bgp_static_route) {
 195                 SET_FLAG(bnc->flags, BGP_STATIC_ROUTE);
 196
 197                 /* If we're toggling the type, re-register */
 198                 if ((CHECK_FLAG(bgp_route->flags, BGP_FLAG_IMPORT_CHECK))
 199                     && !CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)) {
 200                         SET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 201                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 202                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 203                 } else if ((!CHECK_FLAG(bgp_route->flags,
 204                                         BGP_FLAG_IMPORT_CHECK))
 205                            && CHECK_FLAG(bnc->flags,
 206                                          BGP_STATIC_ROUTE_EXACT_MATCH)) {
 207                         UNSET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 208                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 209                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 210                 }
 211         }
 212         /* When nexthop is already known, but now requires 'connected'
 213          * resolution,
 214          * re-register it. The reverse scenario where the nexthop currently
 215          * requires
 216          * 'connected' resolution does not need a re-register (i.e., we treat
 217          * 'connected-required' as an override) except in the scenario where
 218          * this
 219          * is actually a case of tracking a peer for connectivity (e.g., after
 220          * disable connected-check).
 221          * NOTE: We don't track the number of paths separately for 'connected-
 222          * required' vs 'connected-not-required' as this change is not a common
 223          * scenario.
 224          */
 225         else if (connected && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 226                 SET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 227                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 228                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 229         } else if (peer && !connected
 230                    && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 231                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 232                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 233                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 234         }
 235         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW) {
 236                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 237                 SET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 238         } else if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED)
 239                    && !is_default_host_route(bnc_p))
 240                 register_zebra_rnh(bnc, is_bgp_static_route);
 241
 242         if (pi && pi->nexthop != bnc) {
 243                 /* Unlink from existing nexthop cache, if any. This will also
 244                  * free
 245                  * the nexthop cache entry, if appropriate.
 246                  */
 247                 bgp_unlink_nexthop(pi);
 248
 249                 /* updates NHT pi list reference */
 250                 path_nh_map(pi, bnc, true);
 251
 252                 if (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID) && bnc->metric)
 253                         (bgp_path_info_extra_get(pi))->igpmetric = bnc->metric;
 254                 else if (pi->extra)
 255                         pi->extra->igpmetric = 0;
 256         } else if (peer)
 257                 bnc->nht_info = (void *)peer; /* NHT peer reference */
 258
 259         /*
 260          * We are cheating here.  Views have no associated underlying
 261          * ability to detect nexthops.  So when we have a view
 262          * just tell everyone the nexthop is valid
 263          */
 264         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW)
 265                 return 1;
 266         else
 267                 return (bgp_isvalid_nexthop(bnc));
 268 }
 269
 270 void bgp_delete_connected_nexthop(afi_t afi, struct peer *peer)
 271 {
 272         struct bgp_dest *dest;
 273         struct bgp_nexthop_cache *bnc;
 274         struct prefix p;
 275
 276         if (!peer)
 277                 return;
 278
 279         if (!sockunion2hostprefix(&peer->su, &p))
 280                 return;
 281
 282         dest = bgp_node_lookup(
 283                 peer->bgp->nexthop_cache_table[family2afi(p.family)], &p);
 284         if (!dest) {
 285                 if (BGP_DEBUG(nht, NHT))
 286                         zlog_debug(
 287                                 "Cannot find connected NHT node for peer %s(%s)",
 288                                 peer->host, peer->bgp->name_pretty);
 289                 return;
 290         }
 291
 292         bnc = bgp_dest_get_bgp_nexthop_info(dest);
 293         if (!bnc) {
 294                 if (BGP_DEBUG(nht, NHT))
 295                         zlog_debug(
 296                                 "Cannot find connected NHT node for peer %s(%s) on route_node as expected",
 297                                 peer->host, peer->bgp->name_pretty);
 298                 bgp_dest_unlock_node(dest);
 299                 return;
 300         }
 301         bgp_dest_unlock_node(dest);
 302
 303         if (bnc->nht_info != peer) {
 304                 if (BGP_DEBUG(nht, NHT))
 305                         zlog_debug(
 306                                 "Connected NHT %p node for peer %s(%s) points to %p",
 307                                 bnc, peer->host, bnc->bgp->name_pretty,
 308                                 bnc->nht_info);
 309                 return;
 310         }
 311
 312         bnc->nht_info = NULL;
 313
 314         if (LIST_EMPTY(&(bnc->paths))) {
 315                 if (BGP_DEBUG(nht, NHT))
 316                         zlog_debug(
 317                                 "Freeing connected NHT node %p for peer %s(%s)",
 318                                 bnc, peer->host, bnc->bgp->name_pretty);
 319                 unregister_zebra_rnh(bnc, 0);
 320                 bgp_dest_set_bgp_nexthop_info(bnc->dest, NULL);
 321                 bgp_dest_unlock_node(bnc->dest);
 322                 bnc_free(bnc);
 323         }
 324 }
 325
 326 void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 327 {
 328         struct bgp_dest *dest = NULL;
 329         struct bgp_nexthop_cache *bnc;
 330         struct nexthop *nexthop;
 331         struct nexthop *oldnh;
 332         struct nexthop *nhlist_head = NULL;
 333         struct nexthop *nhlist_tail = NULL;
 334         int i;
 335         struct bgp *bgp;
 336         struct zapi_route nhr;
 337
 338         bgp = bgp_lookup_by_vrf_id(vrf_id);
 339         if (!bgp) {
 340                 flog_err(
 341                         EC_BGP_NH_UPD,
 342                         "parse nexthop update: instance not found for vrf_id %u",
 343                         vrf_id);
 344                 return;
 345         }
 346
 347         if (!zapi_nexthop_update_decode(zclient->ibuf, &nhr)) {
 348                 if (BGP_DEBUG(nht, NHT))
 349                         zlog_debug("%s[%s]: Failure to decode nexthop update",
 350                                    __func__, bgp->name_pretty);
 351                 return;
 352         }
 353
 354         if (command == ZEBRA_NEXTHOP_UPDATE)
 355                 dest = bgp_node_lookup(
 356                         bgp->nexthop_cache_table[family2afi(nhr.prefix.family)],
 357                         &nhr.prefix);
 358         else if (command == ZEBRA_IMPORT_CHECK_UPDATE)
 359                 dest = bgp_node_lookup(
 360                         bgp->import_check_table[family2afi(nhr.prefix.family)],
 361                         &nhr.prefix);
 362
 363         if (!dest) {
 364                 if (BGP_DEBUG(nht, NHT)) {
 365                         char buf[PREFIX2STR_BUFFER];
 366                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 367                         zlog_debug("parse nexthop update(%s(%s)): rn not found",
 368                                    buf, bgp->name_pretty);
 369                 }
 370                 return;
 371         }
 372
 373         bnc = bgp_dest_get_bgp_nexthop_info(dest);
 374         if (!bnc) {
 375                 if (BGP_DEBUG(nht, NHT)) {
 376                         char buf[PREFIX2STR_BUFFER];
 377
 378                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 379                         zlog_debug(
 380                                 "parse nexthop update(%s(%s)): bnc node info not found",
 381                                 buf, bgp->name_pretty);
 382                 }
 383                 bgp_dest_unlock_node(dest);
 384                 return;
 385         }
 386
 387         bgp_dest_unlock_node(dest);
 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                         "%s(%u): Rcvd NH update %s - metric %d/%d #nhops %d/%d flags 0x%x",
 397                         bnc->bgp->name_pretty, vrf_id, buf, nhr.metric,
 398                         bnc->metric, nhr.nexthop_num, bnc->nexthop_num,
 399                         bnc->flags);
 400         }
 401
 402         if (nhr.metric != bnc->metric)
 403                 bnc->change_flags |= BGP_NEXTHOP_METRIC_CHANGED;
 404
 405         if (nhr.nexthop_num != bnc->nexthop_num)
 406                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 407
 408         if (nhr.nexthop_num) {
 409                 struct peer *peer = bnc->nht_info;
 410
 411                 /* notify bgp fsm if nbr ip goes from invalid->valid */
 412                 if (!bnc->nexthop_num)
 413                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 414
 415                 bnc->flags |= BGP_NEXTHOP_VALID;
 416                 bnc->metric = nhr.metric;
 417                 bnc->nexthop_num = nhr.nexthop_num;
 418
 419                 bnc->flags &= ~BGP_NEXTHOP_LABELED_VALID; /* check below */
 420
 421                 for (i = 0; i < nhr.nexthop_num; i++) {
 422                         int num_labels = 0;
 423
 424                         nexthop = nexthop_from_zapi_nexthop(&nhr.nexthops[i]);
 425
 426                         /*
 427                          * Turn on RA for the v6 nexthops
 428                          * we receive from bgp.  This is to allow us
 429                          * to work with v4 routing over v6 nexthops
 430                          */
 431                         if (peer && !peer->ifp
 432                             && CHECK_FLAG(peer->flags,
 433                                           PEER_FLAG_CAPABILITY_ENHE)
 434                             && nhr.prefix.family == AF_INET6
 435                             && nexthop->type != NEXTHOP_TYPE_BLACKHOLE) {
 436                                 struct interface *ifp;
 437
 438                                 ifp = if_lookup_by_index(nexthop->ifindex,
 439                                                          nexthop->vrf_id);
 440                                 if (ifp)
 441                                         zclient_send_interface_radv_req(
 442                                                 zclient, nexthop->vrf_id, ifp,
 443                                                 true,
 444                                                 BGP_UNNUM_DEFAULT_RA_INTERVAL);
 445                         }
 446                         /* There is at least one label-switched path */
 447                         if (nexthop->nh_label &&
 448                                 nexthop->nh_label->num_labels) {
 449
 450                                 bnc->flags |= BGP_NEXTHOP_LABELED_VALID;
 451                                 num_labels = nexthop->nh_label->num_labels;
 452                         }
 453
 454                         if (BGP_DEBUG(nht, NHT)) {
 455                                 char buf[NEXTHOP_STRLEN];
 456                                 zlog_debug(
 457                                         "    nhop via %s (%d labels)",
 458                                         nexthop2str(nexthop, buf, sizeof(buf)),
 459                                         num_labels);
 460                         }
 461
 462                         if (nhlist_tail) {
 463                                 nhlist_tail->next = nexthop;
 464                                 nhlist_tail = nexthop;
 465                         } else {
 466                                 nhlist_tail = nexthop;
 467                                 nhlist_head = nexthop;
 468                         }
 469
 470                         /* No need to evaluate the nexthop if we have already
 471                          * determined
 472                          * that there has been a change.
 473                          */
 474                         if (bnc->change_flags & BGP_NEXTHOP_CHANGED)
 475                                 continue;
 476
 477                         for (oldnh = bnc->nexthop; oldnh; oldnh = oldnh->next)
 478                                 if (nexthop_same(oldnh, nexthop))
 479                                         break;
 480
 481                         if (!oldnh)
 482                                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 483                 }
 484                 bnc_nexthop_free(bnc);
 485                 bnc->nexthop = nhlist_head;
 486         } else {
 487                 bnc->flags &= ~BGP_NEXTHOP_VALID;
 488                 bnc->nexthop_num = nhr.nexthop_num;
 489
 490                 /* notify bgp fsm if nbr ip goes from valid->invalid */
 491                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 492
 493                 bnc_nexthop_free(bnc);
 494                 bnc->nexthop = NULL;
 495         }
 496
 497         evaluate_paths(bnc);
 498 }
 499
 500 /*
 501  * Cleanup nexthop registration and status information for BGP nexthops
 502  * pertaining to this VRF. This is invoked upon VRF deletion.
 503  */
 504 void bgp_cleanup_nexthops(struct bgp *bgp)
 505 {
 506         afi_t afi;
 507         struct bgp_dest *dest;
 508         struct bgp_nexthop_cache *bnc;
 509
 510         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 511                 if (!bgp->nexthop_cache_table[afi])
 512                         continue;
 513
 514                 for (dest = bgp_table_top(bgp->nexthop_cache_table[afi]); dest;
 515                      dest = bgp_route_next(dest)) {
 516                         bnc = bgp_dest_get_bgp_nexthop_info(dest);
 517                         if (!bnc)
 518                                 continue;
 519
 520                         /* Clear relevant flags. */
 521                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 522                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 523                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 524                 }
 525         }
 526 }
 527
 528 /**
 529  * make_prefix - make a prefix structure from the path (essentially
 530  * path's node.
 531  */
 532 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p)
 533 {
 534
 535         int is_bgp_static = ((pi->type == ZEBRA_ROUTE_BGP)
 536                              && (pi->sub_type == BGP_ROUTE_STATIC))
 537                                     ? 1
 538                                     : 0;
 539         struct bgp_dest *net = pi->net;
 540         const struct prefix *p_orig = bgp_dest_get_prefix(net);
 541         struct in_addr ipv4;
 542
 543         if (p_orig->family == AF_FLOWSPEC) {
 544                 if (!pi->peer)
 545                         return -1;
 546                 return bgp_flowspec_get_first_nh(pi->peer->bgp,
 547                                                  pi, p);
 548         }
 549         memset(p, 0, sizeof(struct prefix));
 550         switch (afi) {
 551         case AFI_IP:
 552                 p->family = AF_INET;
 553                 if (is_bgp_static) {
 554                         p->u.prefix4 = p_orig->u.prefix4;
 555                         p->prefixlen = p_orig->prefixlen;
 556                 } else {
 557                         if (IS_MAPPED_IPV6(&pi->attr->mp_nexthop_global)) {
 558                                 ipv4_mapped_ipv6_to_ipv4(
 559                                         &pi->attr->mp_nexthop_global, &ipv4);
 560                                 p->u.prefix4 = ipv4;
 561                                 p->prefixlen = IPV4_MAX_BITLEN;
 562                         } else {
 563                                 p->u.prefix4 = pi->attr->nexthop;
 564                                 p->prefixlen = IPV4_MAX_BITLEN;
 565                         }
 566                 }
 567                 break;
 568         case AFI_IP6:
 569                 p->family = AF_INET6;
 570
 571                 if (is_bgp_static) {
 572                         p->u.prefix6 = p_orig->u.prefix6;
 573                         p->prefixlen = p_orig->prefixlen;
 574                 } else {
 575                         /* If we receive MP_REACH nexthop with ::(LL)
 576                          * or LL(LL), use LL address as nexthop cache.
 577                          */
 578                         if (pi->attr->mp_nexthop_len
 579                                     == BGP_ATTR_NHLEN_IPV6_GLOBAL_AND_LL
 580                             && (IN6_IS_ADDR_UNSPECIFIED(
 581                                         &pi->attr->mp_nexthop_global)
 582                                 || IN6_IS_ADDR_LINKLOCAL(
 583                                         &pi->attr->mp_nexthop_global)))
 584                                 p->u.prefix6 = pi->attr->mp_nexthop_local;
 585                         else
 586                                 p->u.prefix6 = pi->attr->mp_nexthop_global;
 587                         p->prefixlen = IPV6_MAX_BITLEN;
 588                 }
 589                 break;
 590         default:
 591                 if (BGP_DEBUG(nht, NHT)) {
 592                         zlog_debug(
 593                                 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
 594                                 __func__, afi, AFI_IP, AFI_IP6);
 595                 }
 596                 break;
 597         }
 598         return 0;
 599 }
 600
 601 /**
 602  * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
 603  *   command to Zebra.
 604  * ARGUMENTS:
 605  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 606  *   int command -- command to send to zebra
 607  * RETURNS:
 608  *   void.
 609  */
 610 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache *bnc, int command)
 611 {
 612         const struct prefix *p;
 613         bool exact_match = false;
 614         int ret;
 615
 616         if (!zclient)
 617                 return;
 618
 619         /* Don't try to register if Zebra doesn't know of this instance. */
 620         if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc->bgp)) {
 621                 if (BGP_DEBUG(zebra, ZEBRA))
 622                         zlog_debug(
 623                                 "%s: No zebra instance to talk to, not installing NHT entry",
 624                                 __func__);
 625                 return;
 626         }
 627
 628         if (!bgp_zebra_num_connects()) {
 629                 if (BGP_DEBUG(zebra, ZEBRA))
 630                         zlog_debug(
 631                                 "%s: We have not connected yet, cannot send nexthops",
 632                                 __func__);
 633         }
 634         p = bgp_dest_get_prefix(bnc->dest);
 635         if ((command == ZEBRA_NEXTHOP_REGISTER
 636              || command == ZEBRA_IMPORT_ROUTE_REGISTER)
 637             && (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)
 638                 || CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)))
 639                 exact_match = true;
 640
 641         if (BGP_DEBUG(zebra, ZEBRA)) {
 642                 char buf[PREFIX2STR_BUFFER];
 643
 644                 prefix2str(p, buf, PREFIX2STR_BUFFER);
 645                 zlog_debug("%s: sending cmd %s for %s (vrf %s)",
 646                         __func__, zserv_command_string(command), buf,
 647                         bnc->bgp->name_pretty);
 648         }
 649
 650         ret = zclient_send_rnh(zclient, command, p, exact_match,
 651                                bnc->bgp->vrf_id);
 652         /* TBD: handle the failure */
 653         if (ret < 0)
 654                 flog_warn(EC_BGP_ZEBRA_SEND,
 655                           "sendmsg_nexthop: zclient_send_message() failed");
 656
 657         if ((command == ZEBRA_NEXTHOP_REGISTER)
 658             || (command == ZEBRA_IMPORT_ROUTE_REGISTER))
 659                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 660         else if ((command == ZEBRA_NEXTHOP_UNREGISTER)
 661                  || (command == ZEBRA_IMPORT_ROUTE_UNREGISTER))
 662                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 663         return;
 664 }
 665
 666 /**
 667  * register_zebra_rnh - register a NH/route with Zebra for notification
 668  *    when the route or the route to the nexthop changes.
 669  * ARGUMENTS:
 670  *   struct bgp_nexthop_cache *bnc
 671  * RETURNS:
 672  *   void.
 673  */
 674 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
 675                                int is_bgp_import_route)
 676 {
 677         /* Check if we have already registered */
 678         if (bnc->flags & BGP_NEXTHOP_REGISTERED)
 679                 return;
 680         if (is_bgp_import_route)
 681                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_REGISTER);
 682         else
 683                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_REGISTER);
 684 }
 685
 686 /**
 687  * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
 688  * ARGUMENTS:
 689  *   struct bgp_nexthop_cache *bnc
 690  * RETURNS:
 691  *   void.
 692  */
 693 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
 694                                  int is_bgp_import_route)
 695 {
 696         /* Check if we have already registered */
 697         if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 698                 return;
 699
 700         if (is_bgp_import_route)
 701                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_UNREGISTER);
 702         else
 703                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_UNREGISTER);
 704 }
 705
 706 /**
 707  * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
 708  * ARGUMENTS:
 709  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 710  * RETURNS:
 711  *   void.
 712  */
 713 static void evaluate_paths(struct bgp_nexthop_cache *bnc)
 714 {
 715         struct bgp_dest *dest;
 716         struct bgp_path_info *path;
 717         int afi;
 718         struct peer *peer = (struct peer *)bnc->nht_info;
 719         struct bgp_table *table;
 720         safi_t safi;
 721         struct bgp *bgp_path;
 722         const struct prefix *p;
 723
 724         if (BGP_DEBUG(nht, NHT)) {
 725                 char buf[PREFIX2STR_BUFFER];
 726                 bnc_str(bnc, buf, PREFIX2STR_BUFFER);
 727                 zlog_debug(
 728                         "NH update for %s %s flags 0x%x chgflags 0x%x - evaluate paths",
 729                         buf, bnc->bgp->name_pretty, bnc->flags,
 730                         bnc->change_flags);
 731         }
 732
 733         LIST_FOREACH (path, &(bnc->paths), nh_thread) {
 734                 if (!(path->type == ZEBRA_ROUTE_BGP
 735                       && ((path->sub_type == BGP_ROUTE_NORMAL)
 736                           || (path->sub_type == BGP_ROUTE_STATIC)
 737                           || (path->sub_type == BGP_ROUTE_IMPORTED))))
 738                         continue;
 739
 740                 dest = path->net;
 741                 assert(dest && bgp_dest_table(dest));
 742                 p = bgp_dest_get_prefix(dest);
 743                 afi = family2afi(p->family);
 744                 table = bgp_dest_table(dest);
 745                 safi = table->safi;
 746
 747                 /*
 748                  * handle routes from other VRFs (they can have a
 749                  * nexthop in THIS VRF). bgp_path is the bgp instance
 750                  * that owns the route referencing this nexthop.
 751                  */
 752                 bgp_path = table->bgp;
 753
 754                 /*
 755                  * Path becomes valid/invalid depending on whether the nexthop
 756                  * reachable/unreachable.
 757                  *
 758                  * In case of unicast routes that were imported from vpn
 759                  * and that have labels, they are valid only if there are
 760                  * nexthops with labels
 761                  */
 762
 763                 bool bnc_is_valid_nexthop = false;
 764                 bool path_valid = false;
 765
 766                 if (safi == SAFI_UNICAST &&
 767                         path->sub_type == BGP_ROUTE_IMPORTED &&
 768                         path->extra &&
 769                         path->extra->num_labels) {
 770
 771                         bnc_is_valid_nexthop =
 772                                 bgp_isvalid_labeled_nexthop(bnc) ? true : false;
 773                 } else {
 774                         if (bgp_update_martian_nexthop(
 775                                     bnc->bgp, afi, safi, path->type,
 776                                     path->sub_type, path->attr, dest)) {
 777                                 if (BGP_DEBUG(nht, NHT))
 778                                         zlog_debug(
 779                                                 "%s: prefix %pRN (vrf %s), ignoring path due to martian or self-next-hop",
 780                                                 __func__, dest, bgp_path->name);
 781                         } else
 782                                 bnc_is_valid_nexthop =
 783                                         bgp_isvalid_nexthop(bnc) ? true : false;
 784                 }
 785
 786                 if (BGP_DEBUG(nht, NHT)) {
 787                         char buf1[RD_ADDRSTRLEN];
 788
 789                         if (dest->pdest) {
 790                                 prefix_rd2str((struct prefix_rd *)bgp_dest_get_prefix(dest->pdest),
 791                                         buf1, sizeof(buf1));
 792                                 zlog_debug(
 793                                         "... eval path %d/%d %pRN RD %s %s flags 0x%x",
 794                                         afi, safi, dest, buf1,
 795                                         bgp_path->name_pretty, path->flags);
 796                         } else
 797                                 zlog_debug(
 798                                         "... eval path %d/%d %pRN %s flags 0x%x",
 799                                         afi, safi, dest, bgp_path->name_pretty,
 800                                         path->flags);
 801                 }
 802
 803                 /* Skip paths marked for removal or as history. */
 804                 if (CHECK_FLAG(path->flags, BGP_PATH_REMOVED)
 805                     || CHECK_FLAG(path->flags, BGP_PATH_HISTORY))
 806                         continue;
 807
 808                 /* Copy the metric to the path. Will be used for bestpath
 809                  * computation */
 810                 if (bgp_isvalid_nexthop(bnc) && bnc->metric)
 811                         (bgp_path_info_extra_get(path))->igpmetric =
 812                                 bnc->metric;
 813                 else if (path->extra)
 814                         path->extra->igpmetric = 0;
 815
 816                 if (CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_METRIC_CHANGED)
 817                     || CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_CHANGED))
 818                         SET_FLAG(path->flags, BGP_PATH_IGP_CHANGED);
 819
 820                 path_valid = !!CHECK_FLAG(path->flags, BGP_PATH_VALID);
 821                 if (path_valid != bnc_is_valid_nexthop) {
 822                         if (path_valid) {
 823                                 /* No longer valid, clear flag; also for EVPN
 824                                  * routes, unimport from VRFs if needed.
 825                                  */
 826                                 bgp_aggregate_decrement(bgp_path, p, path, afi,
 827                                                         safi);
 828                                 bgp_path_info_unset_flag(dest, path,
 829                                                          BGP_PATH_VALID);
 830                                 if (safi == SAFI_EVPN &&
 831                                     bgp_evpn_is_prefix_nht_supported(bgp_dest_get_prefix(dest)))
 832                                         bgp_evpn_unimport_route(bgp_path,
 833                                                 afi, safi, bgp_dest_get_prefix(dest), path);
 834                         } else {
 835                                 /* Path becomes valid, set flag; also for EVPN
 836                                  * routes, import from VRFs if needed.
 837                                  */
 838                                 bgp_path_info_set_flag(dest, path,
 839                                                        BGP_PATH_VALID);
 840                                 bgp_aggregate_increment(bgp_path, p, path, afi,
 841                                                         safi);
 842                                 if (safi == SAFI_EVPN &&
 843                                     bgp_evpn_is_prefix_nht_supported(bgp_dest_get_prefix(dest)))
 844                                         bgp_evpn_import_route(bgp_path,
 845                                                 afi, safi, bgp_dest_get_prefix(dest), path);
 846                         }
 847                 }
 848
 849                 bgp_process(bgp_path, dest, afi, safi);
 850         }
 851
 852         if (peer) {
 853                 int valid_nexthops = bgp_isvalid_nexthop(bnc);
 854
 855                 if (valid_nexthops)
 856                         peer->last_reset = PEER_DOWN_WAITING_OPEN;
 857                 else
 858                         peer->last_reset = PEER_DOWN_WAITING_NHT;
 859
 860                 if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED)) {
 861                         if (BGP_DEBUG(nht, NHT))
 862                                 zlog_debug(
 863                                         "%s: Updating peer (%s(%s)) status with NHT",
 864                                         __func__, peer->host,
 865                                         peer->bgp->name_pretty);
 866                         bgp_fsm_event_update(peer, valid_nexthops);
 867                         SET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 868                 }
 869         }
 870
 871         RESET_FLAG(bnc->change_flags);
 872 }
 873
 874 /**
 875  * path_nh_map - make or break path-to-nexthop association.
 876  * ARGUMENTS:
 877  *   path - pointer to the path structure
 878  *   bnc - pointer to the nexthop structure
 879  *   make - if set, make the association. if unset, just break the existing
 880  *          association.
 881  */
 882 void path_nh_map(struct bgp_path_info *path, struct bgp_nexthop_cache *bnc,
 883                  bool make)
 884 {
 885         if (path->nexthop) {
 886                 LIST_REMOVE(path, nh_thread);
 887                 path->nexthop->path_count--;
 888                 path->nexthop = NULL;
 889         }
 890         if (make) {
 891                 LIST_INSERT_HEAD(&(bnc->paths), path, nh_thread);
 892                 path->nexthop = bnc;
 893                 path->nexthop->path_count++;
 894         }
 895 }
 896
 897 /*
 898  * This function is called to register nexthops to zebra
 899  * as that we may have tried to install the nexthops
 900  * before we actually have a zebra connection
 901  */
 902 void bgp_nht_register_nexthops(struct bgp *bgp)
 903 {
 904         struct bgp_dest *dest;
 905         struct bgp_nexthop_cache *bnc;
 906         afi_t afi;
 907
 908         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 909                 if (!bgp->nexthop_cache_table[afi])
 910                         continue;
 911
 912                 for (dest = bgp_table_top(bgp->nexthop_cache_table[afi]); dest;
 913                      dest = bgp_route_next(dest)) {
 914                         bnc = bgp_dest_get_bgp_nexthop_info(dest);
 915
 916                         if (!bnc)
 917                                 continue;
 918
 919                         register_zebra_rnh(bnc, 0);
 920                 }
 921         }
 922 }
 923
 924 void bgp_nht_reg_enhe_cap_intfs(struct peer *peer)
 925 {
 926         struct bgp *bgp;
 927         struct bgp_dest *dest;
 928         struct bgp_nexthop_cache *bnc;
 929         struct nexthop *nhop;
 930         struct interface *ifp;
 931         struct prefix p;
 932
 933         if (peer->ifp)
 934                 return;
 935
 936         bgp = peer->bgp;
 937
 938         if (!bgp->nexthop_cache_table[AFI_IP6])
 939                 return;
 940
 941         if (!sockunion2hostprefix(&peer->su, &p)) {
 942                 zlog_warn("%s: Unable to convert sockunion to prefix for %s",
 943                           __func__, peer->host);
 944                 return;
 945         }
 946
 947         if (p.family != AF_INET6)
 948                 return;
 949         dest = bgp_node_lookup(bgp->nexthop_cache_table[AFI_IP6], &p);
 950         if (!dest)
 951                 return;
 952
 953         bnc = bgp_dest_get_bgp_nexthop_info(dest);
 954         if (!bnc)
 955                 return;
 956
 957         if (peer != bnc->nht_info)
 958                 return;
 959
 960         for (nhop = bnc->nexthop; nhop; nhop = nhop->next) {
 961                 ifp = if_lookup_by_index(nhop->ifindex, nhop->vrf_id);
 962
 963                 if (!ifp)
 964                         continue;
 965
 966                 zclient_send_interface_radv_req(zclient,
 967                                                 nhop->vrf_id,
 968                                                 ifp, true,
 969                                                 BGP_UNNUM_DEFAULT_RA_INTERVAL);
 970         }
 971 }
 972
 973 void bgp_nht_dereg_enhe_cap_intfs(struct peer *peer)
 974 {
 975         struct bgp *bgp;
 976         struct bgp_dest *dest;
 977         struct bgp_nexthop_cache *bnc;
 978         struct nexthop *nhop;
 979         struct interface *ifp;
 980         struct prefix p;
 981
 982         if (peer->ifp)
 983                 return;
 984
 985         bgp = peer->bgp;
 986
 987         if (!bgp->nexthop_cache_table[AFI_IP6])
 988                 return;
 989
 990         if (!sockunion2hostprefix(&peer->su, &p)) {
 991                 zlog_warn("%s: Unable to convert sockunion to prefix for %s",
 992                           __func__, peer->host);
 993                 return;
 994         }
 995
 996         if (p.family != AF_INET6)
 997                 return;
 998
 999         dest = bgp_node_lookup(bgp->nexthop_cache_table[AFI_IP6], &p);
1000         if (!dest)
1001                 return;
1002
1003         bnc = bgp_dest_get_bgp_nexthop_info(dest);
1004         if (!bnc)
1005                 return;
1006
1007         if (peer != bnc->nht_info)
1008                 return;
1009
1010         for (nhop = bnc->nexthop; nhop; nhop = nhop->next) {
1011                 ifp = if_lookup_by_index(nhop->ifindex, nhop->vrf_id);
1012
1013                 if (!ifp)
1014                         continue;
1015
1016                 zclient_send_interface_radv_req(zclient, nhop->vrf_id, ifp, 0,
1017                                                 0);
1018         }
1019 }