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
  47 extern struct zclient *zclient;
  48
  49 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
  50                                int is_bgp_static_route);
  51 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
  52                                  int is_bgp_static_route);
  53 static void evaluate_paths(struct bgp_nexthop_cache *bnc);
  54 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p);
  55
  56 static int bgp_isvalid_nexthop(struct bgp_nexthop_cache *bnc)
  57 {
  58         return (bgp_zebra_num_connects() == 0
  59                 || (bnc && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID)));
  60 }
  61
  62 static int bgp_isvalid_labeled_nexthop(struct bgp_nexthop_cache *bnc)
  63 {
  64         return (bgp_zebra_num_connects() == 0
  65                 || (bnc && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_LABELED_VALID)));
  66 }
  67
  68 int bgp_find_nexthop(struct bgp_path_info *path, int connected)
  69 {
  70         struct bgp_nexthop_cache *bnc = path->nexthop;
  71
  72         if (!bnc)
  73                 return 0;
  74
  75         /*
  76          * We are cheating here.  Views have no associated underlying
  77          * ability to detect nexthops.  So when we have a view
  78          * just tell everyone the nexthop is valid
  79          */
  80         if (path->peer && path->peer->bgp->inst_type == BGP_INSTANCE_TYPE_VIEW)
  81                 return 1;
  82
  83         if (connected && !(CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)))
  84                 return 0;
  85
  86         return (bgp_isvalid_nexthop(bnc));
  87 }
  88
  89 static void bgp_unlink_nexthop_check(struct bgp_nexthop_cache *bnc)
  90 {
  91         if (LIST_EMPTY(&(bnc->paths)) && !bnc->nht_info) {
  92                 if (BGP_DEBUG(nht, NHT)) {
  93                         char buf[PREFIX2STR_BUFFER];
  94                         zlog_debug("bgp_unlink_nexthop: freeing bnc %s(%s)",
  95                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER),
  96                                    bnc->bgp->name_pretty);
  97                 }
  98                 unregister_zebra_rnh(bnc,
  99                                      CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE));
 100                 bgp_node_set_bgp_nexthop_info(bnc->node, NULL);
 101                 bgp_unlock_node(bnc->node);
 102                 bnc->node = NULL;
 103                 bnc_free(bnc);
 104         }
 105 }
 106
 107 void bgp_unlink_nexthop(struct bgp_path_info *path)
 108 {
 109         struct bgp_nexthop_cache *bnc = path->nexthop;
 110
 111         if (!bnc)
 112                 return;
 113
 114         path_nh_map(path, NULL, false);
 115
 116         bgp_unlink_nexthop_check(bnc);
 117 }
 118
 119 void bgp_unlink_nexthop_by_peer(struct peer *peer)
 120 {
 121         struct prefix p;
 122         struct bgp_node *rn;
 123         struct bgp_nexthop_cache *bnc;
 124         afi_t afi = family2afi(peer->su.sa.sa_family);
 125
 126         if (!sockunion2hostprefix(&peer->su, &p))
 127                 return;
 128
 129         rn = bgp_node_get(peer->bgp->nexthop_cache_table[afi], &p);
 130
 131         bnc = bgp_node_get_bgp_nexthop_info(rn);
 132         if (!bnc)
 133                 return;
 134
 135         /* cleanup the peer reference */
 136         bnc->nht_info = NULL;
 137
 138         bgp_unlink_nexthop_check(bnc);
 139 }
 140
 141 /*
 142  * A route and its nexthop might belong to different VRFs. Therefore,
 143  * we need both the bgp_route and bgp_nexthop pointers.
 144  */
 145 int bgp_find_or_add_nexthop(struct bgp *bgp_route, struct bgp *bgp_nexthop,
 146                             afi_t afi, struct bgp_path_info *pi,
 147                             struct peer *peer, int connected)
 148 {
 149         struct bgp_node *rn;
 150         struct bgp_nexthop_cache *bnc;
 151         struct prefix p;
 152         int is_bgp_static_route = 0;
 153
 154         if (pi) {
 155                 is_bgp_static_route = ((pi->type == ZEBRA_ROUTE_BGP)
 156                                        && (pi->sub_type == BGP_ROUTE_STATIC))
 157                                               ? 1
 158                                               : 0;
 159
 160                 /* Since Extended Next-hop Encoding (RFC5549) support, we want
 161                    to derive
 162                    address-family from the next-hop. */
 163                 if (!is_bgp_static_route)
 164                         afi = BGP_ATTR_NEXTHOP_AFI_IP6(pi->attr) ? AFI_IP6
 165                                                                  : AFI_IP;
 166
 167                 /* This will return true if the global IPv6 NH is a link local
 168                  * addr */
 169                 if (make_prefix(afi, pi, &p) < 0)
 170                         return 1;
 171         } else if (peer) {
 172                 if (!sockunion2hostprefix(&peer->su, &p)) {
 173                         if (BGP_DEBUG(nht, NHT)) {
 174                                 zlog_debug(
 175                                         "%s: Attempting to register with unknown AFI %d (not %d or %d)",
 176                                         __FUNCTION__, afi, AFI_IP, AFI_IP6);
 177                         }
 178                         return 0;
 179                 }
 180         } else
 181                 return 0;
 182
 183         if (is_bgp_static_route)
 184                 rn = bgp_node_get(bgp_nexthop->import_check_table[afi], &p);
 185         else
 186                 rn = bgp_node_get(bgp_nexthop->nexthop_cache_table[afi], &p);
 187
 188         bnc = bgp_node_get_bgp_nexthop_info(rn);
 189         if (!bnc) {
 190                 bnc = bnc_new();
 191                 bgp_node_set_bgp_nexthop_info(rn, bnc);
 192                 bnc->node = rn;
 193                 bnc->bgp = bgp_nexthop;
 194                 bgp_lock_node(rn);
 195                 if (BGP_DEBUG(nht, NHT)) {
 196                         char buf[PREFIX2STR_BUFFER];
 197
 198                         zlog_debug("Allocated bnc %s(%s) peer %p",
 199                                    bnc_str(bnc, buf, PREFIX2STR_BUFFER),
 200                                    bnc->bgp->name_pretty, peer);
 201                 }
 202         }
 203
 204         bgp_unlock_node(rn);
 205         if (is_bgp_static_route) {
 206                 SET_FLAG(bnc->flags, BGP_STATIC_ROUTE);
 207
 208                 /* If we're toggling the type, re-register */
 209                 if ((bgp_flag_check(bgp_route, BGP_FLAG_IMPORT_CHECK))
 210                     && !CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)) {
 211                         SET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 212                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 213                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 214                 } else if ((!bgp_flag_check(bgp_route, BGP_FLAG_IMPORT_CHECK))
 215                            && CHECK_FLAG(bnc->flags,
 216                                          BGP_STATIC_ROUTE_EXACT_MATCH)) {
 217                         UNSET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 218                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 219                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 220                 }
 221         }
 222         /* When nexthop is already known, but now requires 'connected'
 223          * resolution,
 224          * re-register it. The reverse scenario where the nexthop currently
 225          * requires
 226          * 'connected' resolution does not need a re-register (i.e., we treat
 227          * 'connected-required' as an override) except in the scenario where
 228          * this
 229          * is actually a case of tracking a peer for connectivity (e.g., after
 230          * disable connected-check).
 231          * NOTE: We don't track the number of paths separately for 'connected-
 232          * required' vs 'connected-not-required' as this change is not a common
 233          * scenario.
 234          */
 235         else if (connected && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 236                 SET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 237                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 238                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 239         } else if (peer && !connected
 240                    && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 241                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 242                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 243                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 244         }
 245         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW) {
 246                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 247                 SET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 248         } else if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED) &&
 249                    !is_default_host_route(&bnc->node->p))
 250                 register_zebra_rnh(bnc, is_bgp_static_route);
 251
 252         if (pi && pi->nexthop != bnc) {
 253                 /* Unlink from existing nexthop cache, if any. This will also
 254                  * free
 255                  * the nexthop cache entry, if appropriate.
 256                  */
 257                 bgp_unlink_nexthop(pi);
 258
 259                 /* updates NHT pi list reference */
 260                 path_nh_map(pi, bnc, true);
 261
 262                 if (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID) && bnc->metric)
 263                         (bgp_path_info_extra_get(pi))->igpmetric = bnc->metric;
 264                 else if (pi->extra)
 265                         pi->extra->igpmetric = 0;
 266         } else if (peer)
 267                 bnc->nht_info = (void *)peer; /* NHT peer reference */
 268
 269         /*
 270          * We are cheating here.  Views have no associated underlying
 271          * ability to detect nexthops.  So when we have a view
 272          * just tell everyone the nexthop is valid
 273          */
 274         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW)
 275                 return 1;
 276         else
 277                 return (bgp_isvalid_nexthop(bnc));
 278 }
 279
 280 void bgp_delete_connected_nexthop(afi_t afi, struct peer *peer)
 281 {
 282         struct bgp_node *rn;
 283         struct bgp_nexthop_cache *bnc;
 284         struct prefix p;
 285
 286         if (!peer)
 287                 return;
 288
 289         if (!sockunion2hostprefix(&peer->su, &p))
 290                 return;
 291
 292         rn = bgp_node_lookup(
 293                 peer->bgp->nexthop_cache_table[family2afi(p.family)], &p);
 294         if (!rn) {
 295                 if (BGP_DEBUG(nht, NHT))
 296                         zlog_debug(
 297                                 "Cannot find connected NHT node for peer %s(%s)",
 298                                 peer->host, peer->bgp->name_pretty);
 299                 return;
 300         }
 301
 302         bnc = bgp_node_get_bgp_nexthop_info(rn);
 303         if (!bnc) {
 304                 if (BGP_DEBUG(nht, NHT))
 305                         zlog_debug(
 306                                 "Cannot find connected NHT node for peer %s(%s) on route_node as expected",
 307                                 peer->host, peer->bgp->name_pretty);
 308                 bgp_unlock_node(rn);
 309                 return;
 310         }
 311         bgp_unlock_node(rn);
 312
 313         if (bnc->nht_info != peer) {
 314                 if (BGP_DEBUG(nht, NHT))
 315                         zlog_debug(
 316                                 "Connected NHT %p node for peer %s(%s) points to %p",
 317                                 bnc, peer->host, bnc->bgp->name_pretty,
 318                                 bnc->nht_info);
 319                 return;
 320         }
 321
 322         bnc->nht_info = NULL;
 323
 324         if (LIST_EMPTY(&(bnc->paths))) {
 325                 if (BGP_DEBUG(nht, NHT))
 326                         zlog_debug(
 327                                 "Freeing connected NHT node %p for peer %s(%s)",
 328                                 bnc, peer->host, bnc->bgp->name_pretty);
 329                 unregister_zebra_rnh(bnc, 0);
 330                 bgp_node_set_bgp_nexthop_info(bnc->node, NULL);
 331                 bgp_unlock_node(bnc->node);
 332                 bnc_free(bnc);
 333         }
 334 }
 335
 336 void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 337 {
 338         struct bgp_node *rn = NULL;
 339         struct bgp_nexthop_cache *bnc;
 340         struct nexthop *nexthop;
 341         struct nexthop *oldnh;
 342         struct nexthop *nhlist_head = NULL;
 343         struct nexthop *nhlist_tail = NULL;
 344         int i;
 345         struct bgp *bgp;
 346         struct zapi_route nhr;
 347
 348         bgp = bgp_lookup_by_vrf_id(vrf_id);
 349         if (!bgp) {
 350                 flog_err(
 351                         EC_BGP_NH_UPD,
 352                         "parse nexthop update: instance not found for vrf_id %u",
 353                         vrf_id);
 354                 return;
 355         }
 356
 357         if (!zapi_nexthop_update_decode(zclient->ibuf, &nhr)) {
 358                 if (BGP_DEBUG(nht, NHT))
 359                         zlog_debug("%s[%s]: Failure to decode nexthop update",
 360                                    __PRETTY_FUNCTION__, bgp->name_pretty);
 361                 return;
 362         }
 363
 364         if (command == ZEBRA_NEXTHOP_UPDATE)
 365                 rn = bgp_node_lookup(
 366                         bgp->nexthop_cache_table[family2afi(nhr.prefix.family)],
 367                         &nhr.prefix);
 368         else if (command == ZEBRA_IMPORT_CHECK_UPDATE)
 369                 rn = bgp_node_lookup(
 370                         bgp->import_check_table[family2afi(nhr.prefix.family)],
 371                         &nhr.prefix);
 372
 373         if (!rn) {
 374                 if (BGP_DEBUG(nht, NHT)) {
 375                         char buf[PREFIX2STR_BUFFER];
 376                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 377                         zlog_debug("parse nexthop update(%s(%s)): rn not found",
 378                                    buf, bgp->name_pretty);
 379                 }
 380                 return;
 381         }
 382
 383         bnc = bgp_node_get_bgp_nexthop_info(rn);
 384         if (!bnc) {
 385                 if (BGP_DEBUG(nht, NHT)) {
 386                         char buf[PREFIX2STR_BUFFER];
 387
 388                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 389                         zlog_debug(
 390                                 "parse nexthop update(%s(%s)): bnc node info not found",
 391                                 buf, bgp->name_pretty);
 392                 }
 393                 bgp_unlock_node(rn);
 394                 return;
 395         }
 396
 397         bgp_unlock_node(rn);
 398         bnc->last_update = bgp_clock();
 399         bnc->change_flags = 0;
 400
 401         /* debug print the input */
 402         if (BGP_DEBUG(nht, NHT)) {
 403                 char buf[PREFIX2STR_BUFFER];
 404                 prefix2str(&nhr.prefix, buf, sizeof(buf));
 405                 zlog_debug(
 406                         "%s(%u): Rcvd NH update %s - metric %d/%d #nhops %d/%d flags 0x%x",
 407                         bnc->bgp->name_pretty, vrf_id, buf, nhr.metric,
 408                         bnc->metric, nhr.nexthop_num, bnc->nexthop_num,
 409                         bnc->flags);
 410         }
 411
 412         if (nhr.metric != bnc->metric)
 413                 bnc->change_flags |= BGP_NEXTHOP_METRIC_CHANGED;
 414
 415         if (nhr.nexthop_num != bnc->nexthop_num)
 416                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 417
 418         if (nhr.nexthop_num) {
 419                 struct peer *peer = bnc->nht_info;
 420
 421                 /* notify bgp fsm if nbr ip goes from invalid->valid */
 422                 if (!bnc->nexthop_num)
 423                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 424
 425                 bnc->flags |= BGP_NEXTHOP_VALID;
 426                 bnc->metric = nhr.metric;
 427                 bnc->nexthop_num = nhr.nexthop_num;
 428
 429                 bnc->flags &= ~BGP_NEXTHOP_LABELED_VALID; /* check below */
 430
 431                 for (i = 0; i < nhr.nexthop_num; i++) {
 432                         int num_labels = 0;
 433
 434                         nexthop = nexthop_from_zapi_nexthop(&nhr.nexthops[i]);
 435
 436                         /*
 437                          * Turn on RA for the v6 nexthops
 438                          * we receive from bgp.  This is to allow us
 439                          * to work with v4 routing over v6 nexthops
 440                          */
 441                         if (peer && !peer->ifp
 442                             && CHECK_FLAG(peer->flags,
 443                                           PEER_FLAG_CAPABILITY_ENHE)
 444                             && nhr.prefix.family == AF_INET6) {
 445                                 struct interface *ifp;
 446
 447                                 ifp = if_lookup_by_index(nexthop->ifindex,
 448                                                          nexthop->vrf_id);
 449                                 zclient_send_interface_radv_req(
 450                                         zclient, nexthop->vrf_id, ifp, true,
 451                                         BGP_UNNUM_DEFAULT_RA_INTERVAL);
 452                         }
 453                         /* There is at least one label-switched path */
 454                         if (nexthop->nh_label &&
 455                                 nexthop->nh_label->num_labels) {
 456
 457                                 bnc->flags |= BGP_NEXTHOP_LABELED_VALID;
 458                                 num_labels = nexthop->nh_label->num_labels;
 459                         }
 460
 461                         if (BGP_DEBUG(nht, NHT)) {
 462                                 char buf[NEXTHOP_STRLEN];
 463                                 zlog_debug(
 464                                         "    nhop via %s (%d labels)",
 465                                         nexthop2str(nexthop, buf, sizeof(buf)),
 466                                         num_labels);
 467                         }
 468
 469                         if (nhlist_tail) {
 470                                 nhlist_tail->next = nexthop;
 471                                 nhlist_tail = nexthop;
 472                         } else {
 473                                 nhlist_tail = nexthop;
 474                                 nhlist_head = nexthop;
 475                         }
 476
 477                         /* No need to evaluate the nexthop if we have already
 478                          * determined
 479                          * that there has been a change.
 480                          */
 481                         if (bnc->change_flags & BGP_NEXTHOP_CHANGED)
 482                                 continue;
 483
 484                         for (oldnh = bnc->nexthop; oldnh; oldnh = oldnh->next)
 485                                 if (nexthop_same(oldnh, nexthop))
 486                                         break;
 487
 488                         if (!oldnh)
 489                                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 490                 }
 491                 bnc_nexthop_free(bnc);
 492                 bnc->nexthop = nhlist_head;
 493         } else {
 494                 bnc->flags &= ~BGP_NEXTHOP_VALID;
 495                 bnc->nexthop_num = nhr.nexthop_num;
 496
 497                 /* notify bgp fsm if nbr ip goes from valid->invalid */
 498                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 499
 500                 bnc_nexthop_free(bnc);
 501                 bnc->nexthop = NULL;
 502         }
 503
 504         evaluate_paths(bnc);
 505 }
 506
 507 /*
 508  * Cleanup nexthop registration and status information for BGP nexthops
 509  * pertaining to this VRF. This is invoked upon VRF deletion.
 510  */
 511 void bgp_cleanup_nexthops(struct bgp *bgp)
 512 {
 513         afi_t afi;
 514         struct bgp_node *rn;
 515         struct bgp_nexthop_cache *bnc;
 516
 517         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 518                 if (!bgp->nexthop_cache_table[afi])
 519                         continue;
 520
 521                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 522                      rn = bgp_route_next(rn)) {
 523                         bnc = bgp_node_get_bgp_nexthop_info(rn);
 524                         if (!bnc)
 525                                 continue;
 526
 527                         /* Clear relevant flags. */
 528                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 529                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 530                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 531                 }
 532         }
 533 }
 534
 535 /**
 536  * make_prefix - make a prefix structure from the path (essentially
 537  * path's node.
 538  */
 539 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p)
 540 {
 541
 542         int is_bgp_static = ((pi->type == ZEBRA_ROUTE_BGP)
 543                              && (pi->sub_type == BGP_ROUTE_STATIC))
 544                                     ? 1
 545                                     : 0;
 546         struct bgp_node *net = pi->net;
 547         struct prefix *p_orig = &net->p;
 548
 549         if (p_orig->family == AF_FLOWSPEC) {
 550                 if (!pi->peer)
 551                         return -1;
 552                 return bgp_flowspec_get_first_nh(pi->peer->bgp,
 553                                                  pi, p);
 554         }
 555         memset(p, 0, sizeof(struct prefix));
 556         switch (afi) {
 557         case AFI_IP:
 558                 p->family = AF_INET;
 559                 if (is_bgp_static) {
 560                         p->u.prefix4 = pi->net->p.u.prefix4;
 561                         p->prefixlen = pi->net->p.prefixlen;
 562                 } else {
 563                         p->u.prefix4 = pi->attr->nexthop;
 564                         p->prefixlen = IPV4_MAX_BITLEN;
 565                 }
 566                 break;
 567         case AFI_IP6:
 568                 p->family = AF_INET6;
 569
 570                 if (is_bgp_static) {
 571                         p->u.prefix6 = pi->net->p.u.prefix6;
 572                         p->prefixlen = pi->net->p.prefixlen;
 573                 } else {
 574                         p->u.prefix6 = pi->attr->mp_nexthop_global;
 575                         p->prefixlen = IPV6_MAX_BITLEN;
 576                 }
 577                 break;
 578         default:
 579                 if (BGP_DEBUG(nht, NHT)) {
 580                         zlog_debug(
 581                                 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
 582                                 __FUNCTION__, afi, AFI_IP, AFI_IP6);
 583                 }
 584                 break;
 585         }
 586         return 0;
 587 }
 588
 589 /**
 590  * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
 591  *   command to Zebra.
 592  * ARGUMENTS:
 593  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 594  *   int command -- command to send to zebra
 595  * RETURNS:
 596  *   void.
 597  */
 598 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache *bnc, int command)
 599 {
 600         struct prefix *p;
 601         bool exact_match = false;
 602         int ret;
 603
 604         if (!zclient)
 605                 return;
 606
 607         /* Don't try to register if Zebra doesn't know of this instance. */
 608         if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc->bgp)) {
 609                 if (BGP_DEBUG(zebra, ZEBRA))
 610                         zlog_debug("%s: No zebra instance to talk to, not installing NHT entry",
 611                                    __PRETTY_FUNCTION__);
 612                 return;
 613         }
 614
 615         if (!bgp_zebra_num_connects()) {
 616                 if (BGP_DEBUG(zebra, ZEBRA))
 617                         zlog_debug("%s: We have not connected yet, cannot send nexthops",
 618                                    __PRETTY_FUNCTION__);
 619         }
 620         p = &(bnc->node->p);
 621         if ((command == ZEBRA_NEXTHOP_REGISTER
 622              || command == ZEBRA_IMPORT_ROUTE_REGISTER)
 623             && (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)
 624                 || CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)))
 625                 exact_match = true;
 626
 627         if (BGP_DEBUG(zebra, ZEBRA)) {
 628                 char buf[PREFIX2STR_BUFFER];
 629
 630                 prefix2str(p, buf, PREFIX2STR_BUFFER);
 631                 zlog_debug("%s: sending cmd %s for %s (vrf %s)",
 632                         __func__, zserv_command_string(command), buf,
 633                         bnc->bgp->name_pretty);
 634         }
 635
 636         ret = zclient_send_rnh(zclient, command, p, exact_match,
 637                                bnc->bgp->vrf_id);
 638         /* TBD: handle the failure */
 639         if (ret < 0)
 640                 flog_warn(EC_BGP_ZEBRA_SEND,
 641                           "sendmsg_nexthop: zclient_send_message() failed");
 642
 643         if ((command == ZEBRA_NEXTHOP_REGISTER)
 644             || (command == ZEBRA_IMPORT_ROUTE_REGISTER))
 645                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 646         else if ((command == ZEBRA_NEXTHOP_UNREGISTER)
 647                  || (command == ZEBRA_IMPORT_ROUTE_UNREGISTER))
 648                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 649         return;
 650 }
 651
 652 /**
 653  * register_zebra_rnh - register a NH/route with Zebra for notification
 654  *    when the route or the route to the nexthop changes.
 655  * ARGUMENTS:
 656  *   struct bgp_nexthop_cache *bnc
 657  * RETURNS:
 658  *   void.
 659  */
 660 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
 661                                int is_bgp_import_route)
 662 {
 663         /* Check if we have already registered */
 664         if (bnc->flags & BGP_NEXTHOP_REGISTERED)
 665                 return;
 666         if (is_bgp_import_route)
 667                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_REGISTER);
 668         else
 669                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_REGISTER);
 670 }
 671
 672 /**
 673  * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
 674  * ARGUMENTS:
 675  *   struct bgp_nexthop_cache *bnc
 676  * RETURNS:
 677  *   void.
 678  */
 679 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
 680                                  int is_bgp_import_route)
 681 {
 682         /* Check if we have already registered */
 683         if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 684                 return;
 685
 686         if (is_bgp_import_route)
 687                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_UNREGISTER);
 688         else
 689                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_UNREGISTER);
 690 }
 691
 692 /**
 693  * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
 694  * ARGUMENTS:
 695  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 696  * RETURNS:
 697  *   void.
 698  */
 699 static void evaluate_paths(struct bgp_nexthop_cache *bnc)
 700 {
 701         struct bgp_node *rn;
 702         struct bgp_path_info *path;
 703         int afi;
 704         struct peer *peer = (struct peer *)bnc->nht_info;
 705         struct bgp_table *table;
 706         safi_t safi;
 707         struct bgp *bgp_path;
 708
 709         if (BGP_DEBUG(nht, NHT)) {
 710                 char buf[PREFIX2STR_BUFFER];
 711                 bnc_str(bnc, buf, PREFIX2STR_BUFFER);
 712                 zlog_debug(
 713                         "NH update for %s(%s) - flags 0x%x chgflags 0x%x - evaluate paths",
 714                         buf, bnc->bgp->name_pretty, bnc->flags,
 715                         bnc->change_flags);
 716         }
 717
 718         LIST_FOREACH (path, &(bnc->paths), nh_thread) {
 719                 if (!(path->type == ZEBRA_ROUTE_BGP
 720                       && ((path->sub_type == BGP_ROUTE_NORMAL)
 721                           || (path->sub_type == BGP_ROUTE_STATIC)
 722                           || (path->sub_type == BGP_ROUTE_IMPORTED))))
 723                         continue;
 724
 725                 rn = path->net;
 726                 assert(rn && bgp_node_table(rn));
 727                 afi = family2afi(rn->p.family);
 728                 table = bgp_node_table(rn);
 729                 safi = table->safi;
 730
 731                 /*
 732                  * handle routes from other VRFs (they can have a
 733                  * nexthop in THIS VRF). bgp_path is the bgp instance
 734                  * that owns the route referencing this nexthop.
 735                  */
 736                 bgp_path = table->bgp;
 737
 738                 /*
 739                  * Path becomes valid/invalid depending on whether the nexthop
 740                  * reachable/unreachable.
 741                  *
 742                  * In case of unicast routes that were imported from vpn
 743                  * and that have labels, they are valid only if there are
 744                  * nexthops with labels
 745                  */
 746
 747                 int bnc_is_valid_nexthop = 0;
 748
 749                 if (safi == SAFI_UNICAST &&
 750                         path->sub_type == BGP_ROUTE_IMPORTED &&
 751                         path->extra &&
 752                         path->extra->num_labels) {
 753
 754                         bnc_is_valid_nexthop =
 755                                 bgp_isvalid_labeled_nexthop(bnc) ? 1 : 0;
 756                 } else {
 757                         bnc_is_valid_nexthop =
 758                                 bgp_isvalid_nexthop(bnc) ? 1 : 0;
 759                 }
 760
 761                 if (BGP_DEBUG(nht, NHT)) {
 762                         char buf[PREFIX_STRLEN];
 763
 764                         prefix2str(&rn->p, buf, PREFIX_STRLEN);
 765                         zlog_debug("%s: prefix %s (vrf %s) %svalid",
 766                                 __func__, buf, bgp_path->name,
 767                                 (bnc_is_valid_nexthop ? "" : "not "));
 768                 }
 769
 770                 if ((CHECK_FLAG(path->flags, BGP_PATH_VALID) ? 1 : 0)
 771                     != bnc_is_valid_nexthop) {
 772                         if (CHECK_FLAG(path->flags, BGP_PATH_VALID)) {
 773                                 bgp_aggregate_decrement(bgp_path, &rn->p,
 774                                                         path, afi, safi);
 775                                 bgp_path_info_unset_flag(rn, path,
 776                                                          BGP_PATH_VALID);
 777                         } else {
 778                                 bgp_path_info_set_flag(rn, path,
 779                                                        BGP_PATH_VALID);
 780                                 bgp_aggregate_increment(bgp_path, &rn->p,
 781                                                         path, afi, safi);
 782                         }
 783                 }
 784
 785                 /* Copy the metric to the path. Will be used for bestpath
 786                  * computation */
 787                 if (bgp_isvalid_nexthop(bnc) && bnc->metric)
 788                         (bgp_path_info_extra_get(path))->igpmetric =
 789                                 bnc->metric;
 790                 else if (path->extra)
 791                         path->extra->igpmetric = 0;
 792
 793                 if (CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_METRIC_CHANGED)
 794                     || CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_CHANGED))
 795                         SET_FLAG(path->flags, BGP_PATH_IGP_CHANGED);
 796
 797                 bgp_process(bgp_path, rn, afi, safi);
 798         }
 799
 800         if (peer && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED)) {
 801                 if (BGP_DEBUG(nht, NHT))
 802                         zlog_debug("%s: Updating peer (%s(%s)) status with NHT",
 803                                    __FUNCTION__, peer->host,
 804                                    peer->bgp->name_pretty);
 805                 bgp_fsm_event_update(peer, bgp_isvalid_nexthop(bnc));
 806                 SET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 807         }
 808
 809         RESET_FLAG(bnc->change_flags);
 810 }
 811
 812 /**
 813  * path_nh_map - make or break path-to-nexthop association.
 814  * ARGUMENTS:
 815  *   path - pointer to the path structure
 816  *   bnc - pointer to the nexthop structure
 817  *   make - if set, make the association. if unset, just break the existing
 818  *          association.
 819  */
 820 void path_nh_map(struct bgp_path_info *path, struct bgp_nexthop_cache *bnc,
 821                  bool make)
 822 {
 823         if (path->nexthop) {
 824                 LIST_REMOVE(path, nh_thread);
 825                 path->nexthop->path_count--;
 826                 path->nexthop = NULL;
 827         }
 828         if (make) {
 829                 LIST_INSERT_HEAD(&(bnc->paths), path, nh_thread);
 830                 path->nexthop = bnc;
 831                 path->nexthop->path_count++;
 832         }
 833 }
 834
 835 /*
 836  * This function is called to register nexthops to zebra
 837  * as that we may have tried to install the nexthops
 838  * before we actually have a zebra connection
 839  */
 840 void bgp_nht_register_nexthops(struct bgp *bgp)
 841 {
 842         struct bgp_node *rn;
 843         struct bgp_nexthop_cache *bnc;
 844         afi_t afi;
 845
 846         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 847                 if (!bgp->nexthop_cache_table[afi])
 848                         continue;
 849
 850                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 851                      rn = bgp_route_next(rn)) {
 852                         bnc = bgp_node_get_bgp_nexthop_info(rn);
 853
 854                         if (!bnc)
 855                                 continue;
 856
 857                         register_zebra_rnh(bnc, 0);
 858                 }
 859         }
 860 }
 861
 862 void bgp_nht_register_enhe_capability_interfaces(struct peer *peer)
 863 {
 864         struct bgp *bgp;
 865         struct bgp_node *rn;
 866         struct bgp_nexthop_cache *bnc;
 867         struct nexthop *nhop;
 868         struct interface *ifp;
 869         struct prefix p;
 870
 871         if (peer->ifp)
 872                 return;
 873
 874         bgp = peer->bgp;
 875
 876         if (!bgp->nexthop_cache_table[AFI_IP6])
 877                 return;
 878
 879         if (!sockunion2hostprefix(&peer->su, &p)) {
 880                 if (BGP_DEBUG(nht, NHT))
 881                         zlog_debug("%s: Unable to convert prefix to sockunion",
 882                                    __PRETTY_FUNCTION__);
 883                 return;
 884         }
 885
 886         if (p.family != AF_INET6)
 887                 return;
 888         rn = bgp_node_lookup(bgp->nexthop_cache_table[AFI_IP6], &p);
 889         if (!rn)
 890                 return;
 891
 892         bnc = bgp_node_get_bgp_nexthop_info(rn);
 893         if (!bnc)
 894                 return;
 895
 896         if (peer != bnc->nht_info)
 897                 return;
 898
 899         for (nhop = bnc->nexthop; nhop; nhop = nhop->next) {
 900                 ifp = if_lookup_by_index(nhop->ifindex,
 901                                          nhop->vrf_id);
 902                 zclient_send_interface_radv_req(zclient,
 903                                                 nhop->vrf_id,
 904                                                 ifp, true,
 905                                                 BGP_UNNUM_DEFAULT_RA_INTERVAL);
 906         }
 907 }