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