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 ((CHECK_FLAG(bgp_route->flags, 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 ((!CHECK_FLAG(bgp_route->flags,
 195                                         BGP_FLAG_IMPORT_CHECK))
 196                            && CHECK_FLAG(bnc->flags,
 197                                          BGP_STATIC_ROUTE_EXACT_MATCH)) {
 198                         UNSET_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH);
 199                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 200                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 201                 }
 202         }
 203         /* When nexthop is already known, but now requires 'connected'
 204          * resolution,
 205          * re-register it. The reverse scenario where the nexthop currently
 206          * requires
 207          * 'connected' resolution does not need a re-register (i.e., we treat
 208          * 'connected-required' as an override) except in the scenario where
 209          * this
 210          * is actually a case of tracking a peer for connectivity (e.g., after
 211          * disable connected-check).
 212          * NOTE: We don't track the number of paths separately for 'connected-
 213          * required' vs 'connected-not-required' as this change is not a common
 214          * scenario.
 215          */
 216         else if (connected && !CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 217                 SET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 218                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 219                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 220         } else if (peer && !connected
 221                    && CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)) {
 222                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED);
 223                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 224                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 225         }
 226         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW) {
 227                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 228                 SET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 229         } else if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED) &&
 230                    !is_default_host_route(&bnc->node->p))
 231                 register_zebra_rnh(bnc, is_bgp_static_route);
 232
 233         if (pi && pi->nexthop != bnc) {
 234                 /* Unlink from existing nexthop cache, if any. This will also
 235                  * free
 236                  * the nexthop cache entry, if appropriate.
 237                  */
 238                 bgp_unlink_nexthop(pi);
 239
 240                 /* updates NHT pi list reference */
 241                 path_nh_map(pi, bnc, true);
 242
 243                 if (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_VALID) && bnc->metric)
 244                         (bgp_path_info_extra_get(pi))->igpmetric = bnc->metric;
 245                 else if (pi->extra)
 246                         pi->extra->igpmetric = 0;
 247         } else if (peer)
 248                 bnc->nht_info = (void *)peer; /* NHT peer reference */
 249
 250         /*
 251          * We are cheating here.  Views have no associated underlying
 252          * ability to detect nexthops.  So when we have a view
 253          * just tell everyone the nexthop is valid
 254          */
 255         if (bgp_route->inst_type == BGP_INSTANCE_TYPE_VIEW)
 256                 return 1;
 257         else
 258                 return (bgp_isvalid_nexthop(bnc));
 259 }
 260
 261 void bgp_delete_connected_nexthop(afi_t afi, struct peer *peer)
 262 {
 263         struct bgp_node *rn;
 264         struct bgp_nexthop_cache *bnc;
 265         struct prefix p;
 266
 267         if (!peer)
 268                 return;
 269
 270         if (!sockunion2hostprefix(&peer->su, &p))
 271                 return;
 272
 273         rn = bgp_node_lookup(
 274                 peer->bgp->nexthop_cache_table[family2afi(p.family)], &p);
 275         if (!rn) {
 276                 if (BGP_DEBUG(nht, NHT))
 277                         zlog_debug(
 278                                 "Cannot find connected NHT node for peer %s(%s)",
 279                                 peer->host, peer->bgp->name_pretty);
 280                 return;
 281         }
 282
 283         bnc = bgp_node_get_bgp_nexthop_info(rn);
 284         if (!bnc) {
 285                 if (BGP_DEBUG(nht, NHT))
 286                         zlog_debug(
 287                                 "Cannot find connected NHT node for peer %s(%s) on route_node as expected",
 288                                 peer->host, peer->bgp->name_pretty);
 289                 bgp_unlock_node(rn);
 290                 return;
 291         }
 292         bgp_unlock_node(rn);
 293
 294         if (bnc->nht_info != peer) {
 295                 if (BGP_DEBUG(nht, NHT))
 296                         zlog_debug(
 297                                 "Connected NHT %p node for peer %s(%s) points to %p",
 298                                 bnc, peer->host, bnc->bgp->name_pretty,
 299                                 bnc->nht_info);
 300                 return;
 301         }
 302
 303         bnc->nht_info = NULL;
 304
 305         if (LIST_EMPTY(&(bnc->paths))) {
 306                 if (BGP_DEBUG(nht, NHT))
 307                         zlog_debug(
 308                                 "Freeing connected NHT node %p for peer %s(%s)",
 309                                 bnc, peer->host, bnc->bgp->name_pretty);
 310                 unregister_zebra_rnh(bnc, 0);
 311                 bgp_node_set_bgp_nexthop_info(bnc->node, NULL);
 312                 bgp_unlock_node(bnc->node);
 313                 bnc_free(bnc);
 314         }
 315 }
 316
 317 void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 318 {
 319         struct bgp_node *rn = NULL;
 320         struct bgp_nexthop_cache *bnc;
 321         struct nexthop *nexthop;
 322         struct nexthop *oldnh;
 323         struct nexthop *nhlist_head = NULL;
 324         struct nexthop *nhlist_tail = NULL;
 325         int i;
 326         struct bgp *bgp;
 327         struct zapi_route nhr;
 328
 329         bgp = bgp_lookup_by_vrf_id(vrf_id);
 330         if (!bgp) {
 331                 flog_err(
 332                         EC_BGP_NH_UPD,
 333                         "parse nexthop update: instance not found for vrf_id %u",
 334                         vrf_id);
 335                 return;
 336         }
 337
 338         if (!zapi_nexthop_update_decode(zclient->ibuf, &nhr)) {
 339                 if (BGP_DEBUG(nht, NHT))
 340                         zlog_debug("%s[%s]: Failure to decode nexthop update",
 341                                    __PRETTY_FUNCTION__, bgp->name_pretty);
 342                 return;
 343         }
 344
 345         if (command == ZEBRA_NEXTHOP_UPDATE)
 346                 rn = bgp_node_lookup(
 347                         bgp->nexthop_cache_table[family2afi(nhr.prefix.family)],
 348                         &nhr.prefix);
 349         else if (command == ZEBRA_IMPORT_CHECK_UPDATE)
 350                 rn = bgp_node_lookup(
 351                         bgp->import_check_table[family2afi(nhr.prefix.family)],
 352                         &nhr.prefix);
 353
 354         if (!rn) {
 355                 if (BGP_DEBUG(nht, NHT)) {
 356                         char buf[PREFIX2STR_BUFFER];
 357                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 358                         zlog_debug("parse nexthop update(%s(%s)): rn not found",
 359                                    buf, bgp->name_pretty);
 360                 }
 361                 return;
 362         }
 363
 364         bnc = bgp_node_get_bgp_nexthop_info(rn);
 365         if (!bnc) {
 366                 if (BGP_DEBUG(nht, NHT)) {
 367                         char buf[PREFIX2STR_BUFFER];
 368
 369                         prefix2str(&nhr.prefix, buf, sizeof(buf));
 370                         zlog_debug(
 371                                 "parse nexthop update(%s(%s)): bnc node info not found",
 372                                 buf, bgp->name_pretty);
 373                 }
 374                 bgp_unlock_node(rn);
 375                 return;
 376         }
 377
 378         bgp_unlock_node(rn);
 379         bnc->last_update = bgp_clock();
 380         bnc->change_flags = 0;
 381
 382         /* debug print the input */
 383         if (BGP_DEBUG(nht, NHT)) {
 384                 char buf[PREFIX2STR_BUFFER];
 385                 prefix2str(&nhr.prefix, buf, sizeof(buf));
 386                 zlog_debug(
 387                         "%s(%u): Rcvd NH update %s - metric %d/%d #nhops %d/%d flags 0x%x",
 388                         bnc->bgp->name_pretty, vrf_id, buf, nhr.metric,
 389                         bnc->metric, nhr.nexthop_num, bnc->nexthop_num,
 390                         bnc->flags);
 391         }
 392
 393         if (nhr.metric != bnc->metric)
 394                 bnc->change_flags |= BGP_NEXTHOP_METRIC_CHANGED;
 395
 396         if (nhr.nexthop_num != bnc->nexthop_num)
 397                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 398
 399         if (nhr.nexthop_num) {
 400                 struct peer *peer = bnc->nht_info;
 401
 402                 /* notify bgp fsm if nbr ip goes from invalid->valid */
 403                 if (!bnc->nexthop_num)
 404                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 405
 406                 bnc->flags |= BGP_NEXTHOP_VALID;
 407                 bnc->metric = nhr.metric;
 408                 bnc->nexthop_num = nhr.nexthop_num;
 409
 410                 bnc->flags &= ~BGP_NEXTHOP_LABELED_VALID; /* check below */
 411
 412                 for (i = 0; i < nhr.nexthop_num; i++) {
 413                         int num_labels = 0;
 414
 415                         nexthop = nexthop_from_zapi_nexthop(&nhr.nexthops[i]);
 416
 417                         /*
 418                          * Turn on RA for the v6 nexthops
 419                          * we receive from bgp.  This is to allow us
 420                          * to work with v4 routing over v6 nexthops
 421                          */
 422                         if (peer && !peer->ifp
 423                             && CHECK_FLAG(peer->flags,
 424                                           PEER_FLAG_CAPABILITY_ENHE)
 425                             && nhr.prefix.family == AF_INET6
 426                             && nexthop->type != NEXTHOP_TYPE_BLACKHOLE) {
 427                                 struct interface *ifp;
 428
 429                                 ifp = if_lookup_by_index(nexthop->ifindex,
 430                                                          nexthop->vrf_id);
 431                                 zclient_send_interface_radv_req(
 432                                         zclient, nexthop->vrf_id, ifp, true,
 433                                         BGP_UNNUM_DEFAULT_RA_INTERVAL);
 434                         }
 435                         /* There is at least one label-switched path */
 436                         if (nexthop->nh_label &&
 437                                 nexthop->nh_label->num_labels) {
 438
 439                                 bnc->flags |= BGP_NEXTHOP_LABELED_VALID;
 440                                 num_labels = nexthop->nh_label->num_labels;
 441                         }
 442
 443                         if (BGP_DEBUG(nht, NHT)) {
 444                                 char buf[NEXTHOP_STRLEN];
 445                                 zlog_debug(
 446                                         "    nhop via %s (%d labels)",
 447                                         nexthop2str(nexthop, buf, sizeof(buf)),
 448                                         num_labels);
 449                         }
 450
 451                         if (nhlist_tail) {
 452                                 nhlist_tail->next = nexthop;
 453                                 nhlist_tail = nexthop;
 454                         } else {
 455                                 nhlist_tail = nexthop;
 456                                 nhlist_head = nexthop;
 457                         }
 458
 459                         /* No need to evaluate the nexthop if we have already
 460                          * determined
 461                          * that there has been a change.
 462                          */
 463                         if (bnc->change_flags & BGP_NEXTHOP_CHANGED)
 464                                 continue;
 465
 466                         for (oldnh = bnc->nexthop; oldnh; oldnh = oldnh->next)
 467                                 if (nexthop_same(oldnh, nexthop))
 468                                         break;
 469
 470                         if (!oldnh)
 471                                 bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 472                 }
 473                 bnc_nexthop_free(bnc);
 474                 bnc->nexthop = nhlist_head;
 475         } else {
 476                 bnc->flags &= ~BGP_NEXTHOP_VALID;
 477                 bnc->nexthop_num = nhr.nexthop_num;
 478
 479                 /* notify bgp fsm if nbr ip goes from valid->invalid */
 480                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 481
 482                 bnc_nexthop_free(bnc);
 483                 bnc->nexthop = NULL;
 484         }
 485
 486         evaluate_paths(bnc);
 487 }
 488
 489 /*
 490  * Cleanup nexthop registration and status information for BGP nexthops
 491  * pertaining to this VRF. This is invoked upon VRF deletion.
 492  */
 493 void bgp_cleanup_nexthops(struct bgp *bgp)
 494 {
 495         afi_t afi;
 496         struct bgp_node *rn;
 497         struct bgp_nexthop_cache *bnc;
 498
 499         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 500                 if (!bgp->nexthop_cache_table[afi])
 501                         continue;
 502
 503                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 504                      rn = bgp_route_next(rn)) {
 505                         bnc = bgp_node_get_bgp_nexthop_info(rn);
 506                         if (!bnc)
 507                                 continue;
 508
 509                         /* Clear relevant flags. */
 510                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_VALID);
 511                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 512                         UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 513                 }
 514         }
 515 }
 516
 517 /**
 518  * make_prefix - make a prefix structure from the path (essentially
 519  * path's node.
 520  */
 521 static int make_prefix(int afi, struct bgp_path_info *pi, struct prefix *p)
 522 {
 523
 524         int is_bgp_static = ((pi->type == ZEBRA_ROUTE_BGP)
 525                              && (pi->sub_type == BGP_ROUTE_STATIC))
 526                                     ? 1
 527                                     : 0;
 528         struct bgp_node *net = pi->net;
 529         struct prefix *p_orig = &net->p;
 530
 531         if (p_orig->family == AF_FLOWSPEC) {
 532                 if (!pi->peer)
 533                         return -1;
 534                 return bgp_flowspec_get_first_nh(pi->peer->bgp,
 535                                                  pi, p);
 536         }
 537         memset(p, 0, sizeof(struct prefix));
 538         switch (afi) {
 539         case AFI_IP:
 540                 p->family = AF_INET;
 541                 if (is_bgp_static) {
 542                         p->u.prefix4 = pi->net->p.u.prefix4;
 543                         p->prefixlen = pi->net->p.prefixlen;
 544                 } else {
 545                         p->u.prefix4 = pi->attr->nexthop;
 546                         p->prefixlen = IPV4_MAX_BITLEN;
 547                 }
 548                 break;
 549         case AFI_IP6:
 550                 p->family = AF_INET6;
 551
 552                 if (is_bgp_static) {
 553                         p->u.prefix6 = pi->net->p.u.prefix6;
 554                         p->prefixlen = pi->net->p.prefixlen;
 555                 } else {
 556                         p->u.prefix6 = pi->attr->mp_nexthop_global;
 557                         p->prefixlen = IPV6_MAX_BITLEN;
 558                 }
 559                 break;
 560         default:
 561                 if (BGP_DEBUG(nht, NHT)) {
 562                         zlog_debug(
 563                                 "%s: Attempting to make prefix with unknown AFI %d (not %d or %d)",
 564                                 __FUNCTION__, afi, AFI_IP, AFI_IP6);
 565                 }
 566                 break;
 567         }
 568         return 0;
 569 }
 570
 571 /**
 572  * sendmsg_zebra_rnh -- Format and send a nexthop register/Unregister
 573  *   command to Zebra.
 574  * ARGUMENTS:
 575  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 576  *   int command -- command to send to zebra
 577  * RETURNS:
 578  *   void.
 579  */
 580 static void sendmsg_zebra_rnh(struct bgp_nexthop_cache *bnc, int command)
 581 {
 582         struct prefix *p;
 583         bool exact_match = false;
 584         int ret;
 585
 586         if (!zclient)
 587                 return;
 588
 589         /* Don't try to register if Zebra doesn't know of this instance. */
 590         if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bnc->bgp)) {
 591                 if (BGP_DEBUG(zebra, ZEBRA))
 592                         zlog_debug("%s: No zebra instance to talk to, not installing NHT entry",
 593                                    __PRETTY_FUNCTION__);
 594                 return;
 595         }
 596
 597         if (!bgp_zebra_num_connects()) {
 598                 if (BGP_DEBUG(zebra, ZEBRA))
 599                         zlog_debug("%s: We have not connected yet, cannot send nexthops",
 600                                    __PRETTY_FUNCTION__);
 601         }
 602         p = &(bnc->node->p);
 603         if ((command == ZEBRA_NEXTHOP_REGISTER
 604              || command == ZEBRA_IMPORT_ROUTE_REGISTER)
 605             && (CHECK_FLAG(bnc->flags, BGP_NEXTHOP_CONNECTED)
 606                 || CHECK_FLAG(bnc->flags, BGP_STATIC_ROUTE_EXACT_MATCH)))
 607                 exact_match = true;
 608
 609         if (BGP_DEBUG(zebra, ZEBRA)) {
 610                 char buf[PREFIX2STR_BUFFER];
 611
 612                 prefix2str(p, buf, PREFIX2STR_BUFFER);
 613                 zlog_debug("%s: sending cmd %s for %s (vrf %s)",
 614                         __func__, zserv_command_string(command), buf,
 615                         bnc->bgp->name_pretty);
 616         }
 617
 618         ret = zclient_send_rnh(zclient, command, p, exact_match,
 619                                bnc->bgp->vrf_id);
 620         /* TBD: handle the failure */
 621         if (ret < 0)
 622                 flog_warn(EC_BGP_ZEBRA_SEND,
 623                           "sendmsg_nexthop: zclient_send_message() failed");
 624
 625         if ((command == ZEBRA_NEXTHOP_REGISTER)
 626             || (command == ZEBRA_IMPORT_ROUTE_REGISTER))
 627                 SET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 628         else if ((command == ZEBRA_NEXTHOP_UNREGISTER)
 629                  || (command == ZEBRA_IMPORT_ROUTE_UNREGISTER))
 630                 UNSET_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED);
 631         return;
 632 }
 633
 634 /**
 635  * register_zebra_rnh - register a NH/route with Zebra for notification
 636  *    when the route or the route to the nexthop changes.
 637  * ARGUMENTS:
 638  *   struct bgp_nexthop_cache *bnc
 639  * RETURNS:
 640  *   void.
 641  */
 642 static void register_zebra_rnh(struct bgp_nexthop_cache *bnc,
 643                                int is_bgp_import_route)
 644 {
 645         /* Check if we have already registered */
 646         if (bnc->flags & BGP_NEXTHOP_REGISTERED)
 647                 return;
 648         if (is_bgp_import_route)
 649                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_REGISTER);
 650         else
 651                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_REGISTER);
 652 }
 653
 654 /**
 655  * unregister_zebra_rnh -- Unregister the route/nexthop from Zebra.
 656  * ARGUMENTS:
 657  *   struct bgp_nexthop_cache *bnc
 658  * RETURNS:
 659  *   void.
 660  */
 661 static void unregister_zebra_rnh(struct bgp_nexthop_cache *bnc,
 662                                  int is_bgp_import_route)
 663 {
 664         /* Check if we have already registered */
 665         if (!CHECK_FLAG(bnc->flags, BGP_NEXTHOP_REGISTERED))
 666                 return;
 667
 668         if (is_bgp_import_route)
 669                 sendmsg_zebra_rnh(bnc, ZEBRA_IMPORT_ROUTE_UNREGISTER);
 670         else
 671                 sendmsg_zebra_rnh(bnc, ZEBRA_NEXTHOP_UNREGISTER);
 672 }
 673
 674 /**
 675  * evaluate_paths - Evaluate the paths/nets associated with a nexthop.
 676  * ARGUMENTS:
 677  *   struct bgp_nexthop_cache *bnc -- the nexthop structure.
 678  * RETURNS:
 679  *   void.
 680  */
 681 static void evaluate_paths(struct bgp_nexthop_cache *bnc)
 682 {
 683         struct bgp_node *rn;
 684         struct bgp_path_info *path;
 685         int afi;
 686         struct peer *peer = (struct peer *)bnc->nht_info;
 687         struct bgp_table *table;
 688         safi_t safi;
 689         struct bgp *bgp_path;
 690
 691         if (BGP_DEBUG(nht, NHT)) {
 692                 char buf[PREFIX2STR_BUFFER];
 693                 bnc_str(bnc, buf, PREFIX2STR_BUFFER);
 694                 zlog_debug(
 695                         "NH update for %s(%s) - flags 0x%x chgflags 0x%x - evaluate paths",
 696                         buf, bnc->bgp->name_pretty, bnc->flags,
 697                         bnc->change_flags);
 698         }
 699
 700         LIST_FOREACH (path, &(bnc->paths), nh_thread) {
 701                 if (!(path->type == ZEBRA_ROUTE_BGP
 702                       && ((path->sub_type == BGP_ROUTE_NORMAL)
 703                           || (path->sub_type == BGP_ROUTE_STATIC)
 704                           || (path->sub_type == BGP_ROUTE_IMPORTED))))
 705                         continue;
 706
 707                 rn = path->net;
 708                 assert(rn && bgp_node_table(rn));
 709                 afi = family2afi(rn->p.family);
 710                 table = bgp_node_table(rn);
 711                 safi = table->safi;
 712
 713                 /*
 714                  * handle routes from other VRFs (they can have a
 715                  * nexthop in THIS VRF). bgp_path is the bgp instance
 716                  * that owns the route referencing this nexthop.
 717                  */
 718                 bgp_path = table->bgp;
 719
 720                 /*
 721                  * Path becomes valid/invalid depending on whether the nexthop
 722                  * reachable/unreachable.
 723                  *
 724                  * In case of unicast routes that were imported from vpn
 725                  * and that have labels, they are valid only if there are
 726                  * nexthops with labels
 727                  */
 728
 729                 int bnc_is_valid_nexthop = 0;
 730
 731                 if (safi == SAFI_UNICAST &&
 732                         path->sub_type == BGP_ROUTE_IMPORTED &&
 733                         path->extra &&
 734                         path->extra->num_labels) {
 735
 736                         bnc_is_valid_nexthop =
 737                                 bgp_isvalid_labeled_nexthop(bnc) ? 1 : 0;
 738                 } else {
 739                         bnc_is_valid_nexthop =
 740                                 bgp_isvalid_nexthop(bnc) ? 1 : 0;
 741                 }
 742
 743                 if (BGP_DEBUG(nht, NHT)) {
 744                         char buf[PREFIX_STRLEN];
 745
 746                         prefix2str(&rn->p, buf, PREFIX_STRLEN);
 747                         zlog_debug("%s: prefix %s (vrf %s) %svalid",
 748                                 __func__, buf, bgp_path->name,
 749                                 (bnc_is_valid_nexthop ? "" : "not "));
 750                 }
 751
 752                 if ((CHECK_FLAG(path->flags, BGP_PATH_VALID) ? 1 : 0)
 753                     != bnc_is_valid_nexthop) {
 754                         if (CHECK_FLAG(path->flags, BGP_PATH_VALID)) {
 755                                 bgp_aggregate_decrement(bgp_path, &rn->p,
 756                                                         path, afi, safi);
 757                                 bgp_path_info_unset_flag(rn, path,
 758                                                          BGP_PATH_VALID);
 759                         } else {
 760                                 bgp_path_info_set_flag(rn, path,
 761                                                        BGP_PATH_VALID);
 762                                 bgp_aggregate_increment(bgp_path, &rn->p,
 763                                                         path, afi, safi);
 764                         }
 765                 }
 766
 767                 /* Copy the metric to the path. Will be used for bestpath
 768                  * computation */
 769                 if (bgp_isvalid_nexthop(bnc) && bnc->metric)
 770                         (bgp_path_info_extra_get(path))->igpmetric =
 771                                 bnc->metric;
 772                 else if (path->extra)
 773                         path->extra->igpmetric = 0;
 774
 775                 if (CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_METRIC_CHANGED)
 776                     || CHECK_FLAG(bnc->change_flags, BGP_NEXTHOP_CHANGED))
 777                         SET_FLAG(path->flags, BGP_PATH_IGP_CHANGED);
 778
 779                 if (safi == SAFI_EVPN &&
 780                     bgp_evpn_is_prefix_nht_supported(&rn->p)) {
 781                         if (CHECK_FLAG(path->flags, BGP_PATH_VALID))
 782                                 bgp_evpn_import_route(bgp_path, afi, safi,
 783                                                       &rn->p, path);
 784                         else
 785                                 bgp_evpn_unimport_route(bgp_path, afi, safi,
 786                                                         &rn->p, path);
 787                 }
 788
 789                 bgp_process(bgp_path, rn, afi, safi);
 790         }
 791
 792         if (peer) {
 793                 int valid_nexthops = bgp_isvalid_nexthop(bnc);
 794
 795                 if (valid_nexthops)
 796                         peer->last_reset = PEER_DOWN_WAITING_OPEN;
 797                 else
 798                         peer->last_reset = PEER_DOWN_WAITING_NHT;
 799
 800                 if (!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, valid_nexthops);
 806                         SET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
 807                 }
 808         }
 809
 810         RESET_FLAG(bnc->change_flags);
 811 }
 812
 813 /**
 814  * path_nh_map - make or break path-to-nexthop association.
 815  * ARGUMENTS:
 816  *   path - pointer to the path structure
 817  *   bnc - pointer to the nexthop structure
 818  *   make - if set, make the association. if unset, just break the existing
 819  *          association.
 820  */
 821 void path_nh_map(struct bgp_path_info *path, struct bgp_nexthop_cache *bnc,
 822                  bool make)
 823 {
 824         if (path->nexthop) {
 825                 LIST_REMOVE(path, nh_thread);
 826                 path->nexthop->path_count--;
 827                 path->nexthop = NULL;
 828         }
 829         if (make) {
 830                 LIST_INSERT_HEAD(&(bnc->paths), path, nh_thread);
 831                 path->nexthop = bnc;
 832                 path->nexthop->path_count++;
 833         }
 834 }
 835
 836 /*
 837  * This function is called to register nexthops to zebra
 838  * as that we may have tried to install the nexthops
 839  * before we actually have a zebra connection
 840  */
 841 void bgp_nht_register_nexthops(struct bgp *bgp)
 842 {
 843         struct bgp_node *rn;
 844         struct bgp_nexthop_cache *bnc;
 845         afi_t afi;
 846
 847         for (afi = AFI_IP; afi < AFI_MAX; afi++) {
 848                 if (!bgp->nexthop_cache_table[afi])
 849                         continue;
 850
 851                 for (rn = bgp_table_top(bgp->nexthop_cache_table[afi]); rn;
 852                      rn = bgp_route_next(rn)) {
 853                         bnc = bgp_node_get_bgp_nexthop_info(rn);
 854
 855                         if (!bnc)
 856                                 continue;
 857
 858                         register_zebra_rnh(bnc, 0);
 859                 }
 860         }
 861 }
 862
 863 void bgp_nht_register_enhe_capability_interfaces(struct peer *peer)
 864 {
 865         struct bgp *bgp;
 866         struct bgp_node *rn;
 867         struct bgp_nexthop_cache *bnc;
 868         struct nexthop *nhop;
 869         struct interface *ifp;
 870         struct prefix p;
 871
 872         if (peer->ifp)
 873                 return;
 874
 875         bgp = peer->bgp;
 876
 877         if (!bgp->nexthop_cache_table[AFI_IP6])
 878                 return;
 879
 880         if (!sockunion2hostprefix(&peer->su, &p)) {
 881                 if (BGP_DEBUG(nht, NHT))
 882                         zlog_debug("%s: Unable to convert prefix to sockunion",
 883                                    __PRETTY_FUNCTION__);
 884                 return;
 885         }
 886
 887         if (p.family != AF_INET6)
 888                 return;
 889         rn = bgp_node_lookup(bgp->nexthop_cache_table[AFI_IP6], &p);
 890         if (!rn)
 891                 return;
 892
 893         bnc = bgp_node_get_bgp_nexthop_info(rn);
 894         if (!bnc)
 895                 return;
 896
 897         if (peer != bnc->nht_info)
 898                 return;
 899
 900         for (nhop = bnc->nexthop; nhop; nhop = nhop->next) {
 901                 ifp = if_lookup_by_index(nhop->ifindex,
 902                                          nhop->vrf_id);
 903                 zclient_send_interface_radv_req(zclient,
 904                                                 nhop->vrf_id,
 905                                                 ifp, true,
 906                                                 BGP_UNNUM_DEFAULT_RA_INTERVAL);
 907         }
 908 }