zebra/zebra_fpm_netlink.c

   1 /*
   2  * Code for encoding/decoding FPM messages that are in netlink format.
   3  *
   4  * Copyright (C) 1997, 98, 99 Kunihiro Ishiguro
   5  * Copyright (C) 2012 by Open Source Routing.
   6  * Copyright (C) 2012 by Internet Systems Consortium, Inc. ("ISC")
   7  *
   8  * This file is part of GNU Zebra.
   9  *
  10  * GNU Zebra is free software; you can redistribute it and/or modify it
  11  * under the terms of the GNU General Public License as published by the
  12  * Free Software Foundation; either version 2, or (at your option) any
  13  * later version.
  14  *
  15  * GNU Zebra is distributed in the hope that it will be useful, but
  16  * WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public License along
  21  * with this program; see the file COPYING; if not, write to the Free Software
  22  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  23  */
  24
  25 #include <zebra.h>
  26
  27 #ifdef HAVE_NETLINK
  28
  29 #include "log.h"
  30 #include "rib.h"
  31 #include "vty.h"
  32 #include "prefix.h"
  33
  34 #include "zebra/zserv.h"
  35 #include "zebra/zebra_router.h"
  36 #include "zebra/zebra_dplane.h"
  37 #include "zebra/zebra_ns.h"
  38 #include "zebra/zebra_vrf.h"
  39 #include "zebra/kernel_netlink.h"
  40 #include "zebra/rt_netlink.h"
  41 #include "nexthop.h"
  42
  43 #include "zebra/zebra_fpm_private.h"
  44 #include "zebra/zebra_vxlan_private.h"
  45 #include "zebra/interface.h"
  46
  47 /*
  48  * af_addr_size
  49  *
  50  * The size of an address in a given address family.
  51  */
  52 static size_t af_addr_size(uint8_t af)
  53 {
  54         switch (af) {
  55
  56         case AF_INET:
  57                 return 4;
  58         case AF_INET6:
  59                 return 16;
  60         default:
  61                 assert(0);
  62                 return 16;
  63         }
  64 }
  65
  66 /*
  67  * We plan to use RTA_ENCAP_TYPE attribute for VxLAN encap as well.
  68  * Currently, values 0 to 8 for this attribute are used by lwtunnel_encap_types
  69  * So, we cannot use these values for VxLAN encap.
  70  */
  71 enum fpm_nh_encap_type_t {
  72         FPM_NH_ENCAP_NONE = 0,
  73         FPM_NH_ENCAP_VXLAN = 100,
  74         FPM_NH_ENCAP_MAX,
  75 };
  76
  77 /*
  78  * fpm_nh_encap_type_to_str
  79  */
  80 static const char *fpm_nh_encap_type_to_str(enum fpm_nh_encap_type_t encap_type)
  81 {
  82         switch (encap_type) {
  83         case FPM_NH_ENCAP_NONE:
  84                 return "none";
  85
  86         case FPM_NH_ENCAP_VXLAN:
  87                 return "VxLAN";
  88
  89         case FPM_NH_ENCAP_MAX:
  90                 return "invalid";
  91         }
  92
  93         return "invalid";
  94 }
  95
  96 struct vxlan_encap_info_t {
  97         vni_t vni;
  98 };
  99
 100 enum vxlan_encap_info_type_t {
 101         VXLAN_VNI = 0,
 102 };
 103
 104 struct fpm_nh_encap_info_t {
 105         enum fpm_nh_encap_type_t encap_type;
 106         union {
 107                 struct vxlan_encap_info_t vxlan_encap;
 108         };
 109 };
 110
 111 /*
 112  * netlink_nh_info
 113  *
 114  * Holds information about a single nexthop for netlink. These info
 115  * structures are transient and may contain pointers into rib
 116  * data structures for convenience.
 117  */
 118 struct netlink_nh_info {
 119         uint32_t if_index;
 120         union g_addr *gateway;
 121
 122         /*
 123          * Information from the struct nexthop from which this nh was
 124          * derived. For debug purposes only.
 125          */
 126         int recursive;
 127         enum nexthop_types_t type;
 128         struct fpm_nh_encap_info_t encap_info;
 129 };
 130
 131 /*
 132  * netlink_route_info
 133  *
 134  * A structure for holding information for a netlink route message.
 135  */
 136 struct netlink_route_info {
 137         uint32_t nlmsg_pid;
 138         uint16_t nlmsg_type;
 139         uint8_t rtm_type;
 140         uint32_t rtm_table;
 141         uint8_t rtm_protocol;
 142         uint8_t af;
 143         struct prefix *prefix;
 144         uint32_t *metric;
 145         unsigned int num_nhs;
 146
 147         /*
 148          * Nexthop structures
 149          */
 150         struct netlink_nh_info nhs[MULTIPATH_NUM];
 151         union g_addr *pref_src;
 152 };
 153
 154 /*
 155  * netlink_route_info_add_nh
 156  *
 157  * Add information about the given nexthop to the given route info
 158  * structure.
 159  *
 160  * Returns true if a nexthop was added, false otherwise.
 161  */
 162 static int netlink_route_info_add_nh(struct netlink_route_info *ri,
 163                                      struct nexthop *nexthop,
 164                                      struct route_entry *re)
 165 {
 166         struct netlink_nh_info nhi;
 167         union g_addr *src;
 168         struct zebra_vrf *zvrf = NULL;
 169         struct interface *ifp = NULL, *link_if = NULL;
 170         struct zebra_if *zif = NULL;
 171         vni_t vni = 0;
 172
 173         memset(&nhi, 0, sizeof(nhi));
 174         src = NULL;
 175
 176         if (ri->num_nhs >= (int)array_size(ri->nhs))
 177                 return 0;
 178
 179         nhi.recursive = nexthop->rparent ? 1 : 0;
 180         nhi.type = nexthop->type;
 181         nhi.if_index = nexthop->ifindex;
 182
 183         if (nexthop->type == NEXTHOP_TYPE_IPV4
 184             || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
 185                 nhi.gateway = &nexthop->gate;
 186                 if (nexthop->src.ipv4.s_addr != INADDR_ANY)
 187                         src = &nexthop->src;
 188         }
 189
 190         if (nexthop->type == NEXTHOP_TYPE_IPV6
 191             || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
 192                 nhi.gateway = &nexthop->gate;
 193         }
 194
 195         if (nexthop->type == NEXTHOP_TYPE_IFINDEX) {
 196                 if (nexthop->src.ipv4.s_addr != INADDR_ANY)
 197                         src = &nexthop->src;
 198         }
 199
 200         if (!nhi.gateway && nhi.if_index == 0)
 201                 return 0;
 202
 203         if (re && CHECK_FLAG(re->flags, ZEBRA_FLAG_EVPN_ROUTE)) {
 204                 nhi.encap_info.encap_type = FPM_NH_ENCAP_VXLAN;
 205
 206                 /* Extract VNI id for the nexthop SVI interface */
 207                 zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id);
 208                 if (zvrf) {
 209                         ifp = if_lookup_by_index_per_ns(zvrf->zns,
 210                                                         nexthop->ifindex);
 211                         if (ifp) {
 212                                 zif = (struct zebra_if *)ifp->info;
 213                                 if (zif) {
 214                                         if (IS_ZEBRA_IF_BRIDGE(ifp))
 215                                                 link_if = ifp;
 216                                         else if (IS_ZEBRA_IF_VLAN(ifp))
 217                                                 link_if =
 218                                                 if_lookup_by_index_per_ns(
 219                                                         zvrf->zns,
 220                                                         zif->link_ifindex);
 221                                         if (link_if)
 222                                                 vni = vni_id_from_svi(ifp,
 223                                                                       link_if);
 224                                 }
 225                         }
 226                 }
 227
 228                 nhi.encap_info.vxlan_encap.vni = vni;
 229         }
 230
 231         /*
 232          * We have a valid nhi. Copy the structure over to the route_info.
 233          */
 234         ri->nhs[ri->num_nhs] = nhi;
 235         ri->num_nhs++;
 236
 237         if (src && !ri->pref_src)
 238                 ri->pref_src = src;
 239
 240         return 1;
 241 }
 242
 243 /*
 244  * netlink_proto_from_route_type
 245  */
 246 static uint8_t netlink_proto_from_route_type(int type)
 247 {
 248         switch (type) {
 249         case ZEBRA_ROUTE_KERNEL:
 250         case ZEBRA_ROUTE_CONNECT:
 251                 return RTPROT_KERNEL;
 252
 253         default:
 254                 return RTPROT_ZEBRA;
 255         }
 256 }
 257
 258 /*
 259  * netlink_route_info_fill
 260  *
 261  * Fill out the route information object from the given route.
 262  *
 263  * Returns true on success and false on failure.
 264  */
 265 static int netlink_route_info_fill(struct netlink_route_info *ri, int cmd,
 266                                    rib_dest_t *dest, struct route_entry *re)
 267 {
 268         struct nexthop *nexthop;
 269         struct rib_table_info *table_info =
 270                 rib_table_info(rib_dest_table(dest));
 271         struct zebra_vrf *zvrf = table_info->zvrf;
 272
 273         memset(ri, 0, sizeof(*ri));
 274
 275         ri->prefix = rib_dest_prefix(dest);
 276         ri->af = rib_dest_af(dest);
 277
 278         if (zvrf && zvrf->zns)
 279                 ri->nlmsg_pid = zvrf->zns->netlink_dplane.snl.nl_pid;
 280
 281         ri->nlmsg_type = cmd;
 282         ri->rtm_table = table_info->table_id;
 283         ri->rtm_protocol = RTPROT_UNSPEC;
 284
 285         /*
 286          * An RTM_DELROUTE need not be accompanied by any nexthops,
 287          * particularly in our communication with the FPM.
 288          */
 289         if (cmd == RTM_DELROUTE && !re)
 290                 return 1;
 291
 292         if (!re) {
 293                 zfpm_debug("%s: Expected non-NULL re pointer", __func__);
 294                 return 0;
 295         }
 296
 297         ri->rtm_protocol = netlink_proto_from_route_type(re->type);
 298         ri->rtm_type = RTN_UNICAST;
 299         ri->metric = &re->metric;
 300
 301         for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) {
 302                 if (ri->num_nhs >= zrouter.multipath_num)
 303                         break;
 304
 305                 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
 306                         continue;
 307
 308                 if (nexthop->type == NEXTHOP_TYPE_BLACKHOLE) {
 309                         switch (nexthop->bh_type) {
 310                         case BLACKHOLE_ADMINPROHIB:
 311                                 ri->rtm_type = RTN_PROHIBIT;
 312                                 break;
 313                         case BLACKHOLE_REJECT:
 314                                 ri->rtm_type = RTN_UNREACHABLE;
 315                                 break;
 316                         case BLACKHOLE_NULL:
 317                         default:
 318                                 ri->rtm_type = RTN_BLACKHOLE;
 319                                 break;
 320                         }
 321                 }
 322
 323                 if ((cmd == RTM_NEWROUTE
 324                      && CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE))
 325                     || (cmd == RTM_DELROUTE
 326                         && CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED))) {
 327                         netlink_route_info_add_nh(ri, nexthop, re);
 328                 }
 329         }
 330
 331         if (ri->num_nhs == 0) {
 332                 switch (ri->rtm_type) {
 333                 case RTN_PROHIBIT:
 334                 case RTN_UNREACHABLE:
 335                 case RTN_BLACKHOLE:
 336                         break;
 337                 default:
 338                         /* If there is no useful nexthop then return. */
 339                         zfpm_debug(
 340                                 "netlink_encode_route(): No useful nexthop.");
 341                         return 0;
 342                 }
 343         }
 344
 345         return 1;
 346 }
 347
 348 /*
 349  * netlink_route_info_encode
 350  *
 351  * Returns the number of bytes written to the buffer. 0 or a negative
 352  * value indicates an error.
 353  */
 354 static int netlink_route_info_encode(struct netlink_route_info *ri,
 355                                      char *in_buf, size_t in_buf_len)
 356 {
 357         size_t bytelen;
 358         unsigned int nexthop_num = 0;
 359         size_t buf_offset;
 360         struct netlink_nh_info *nhi;
 361         enum fpm_nh_encap_type_t encap;
 362         struct rtattr *nest, *inner_nest;
 363         struct rtnexthop *rtnh;
 364         struct vxlan_encap_info_t *vxlan;
 365         struct in6_addr ipv6;
 366
 367         struct {
 368                 struct nlmsghdr n;
 369                 struct rtmsg r;
 370                 char buf[1];
 371         } * req;
 372
 373         req = (void *)in_buf;
 374
 375         buf_offset = ((char *)req->buf) - ((char *)req);
 376
 377         if (in_buf_len < buf_offset) {
 378                 assert(0);
 379                 return 0;
 380         }
 381
 382         memset(req, 0, buf_offset);
 383
 384         bytelen = af_addr_size(ri->af);
 385
 386         req->n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
 387         req->n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
 388         req->n.nlmsg_pid = ri->nlmsg_pid;
 389         req->n.nlmsg_type = ri->nlmsg_type;
 390         req->r.rtm_family = ri->af;
 391
 392         /*
 393          * rtm_table field is a uchar field which can accomodate table_id less
 394          * than 256.
 395          * To support table id greater than 255, if the table_id is greater than
 396          * 255, set rtm_table to RT_TABLE_UNSPEC and add RTA_TABLE attribute
 397          * with 32 bit value as the table_id.
 398          */
 399         if (ri->rtm_table < 256)
 400                 req->r.rtm_table = ri->rtm_table;
 401         else {
 402                 req->r.rtm_table = RT_TABLE_UNSPEC;
 403                 nl_attr_put32(&req->n, in_buf_len, RTA_TABLE, ri->rtm_table);
 404         }
 405
 406         req->r.rtm_dst_len = ri->prefix->prefixlen;
 407         req->r.rtm_protocol = ri->rtm_protocol;
 408         req->r.rtm_scope = RT_SCOPE_UNIVERSE;
 409
 410         nl_attr_put(&req->n, in_buf_len, RTA_DST, &ri->prefix->u.prefix,
 411                     bytelen);
 412
 413         req->r.rtm_type = ri->rtm_type;
 414
 415         /* Metric. */
 416         if (ri->metric)
 417                 nl_attr_put32(&req->n, in_buf_len, RTA_PRIORITY, *ri->metric);
 418
 419         if (ri->num_nhs == 0)
 420                 goto done;
 421
 422         if (ri->num_nhs == 1) {
 423                 nhi = &ri->nhs[0];
 424
 425                 if (nhi->gateway) {
 426                         if (nhi->type == NEXTHOP_TYPE_IPV4_IFINDEX
 427                             && ri->af == AF_INET6) {
 428                                 ipv4_to_ipv4_mapped_ipv6(&ipv6,
 429                                                          nhi->gateway->ipv4);
 430                                 nl_attr_put(&req->n, in_buf_len, RTA_GATEWAY,
 431                                             &ipv6, bytelen);
 432                         } else
 433                                 nl_attr_put(&req->n, in_buf_len, RTA_GATEWAY,
 434                                             nhi->gateway, bytelen);
 435                 }
 436
 437                 if (nhi->if_index) {
 438                         nl_attr_put32(&req->n, in_buf_len, RTA_OIF,
 439                                       nhi->if_index);
 440                 }
 441
 442                 encap = nhi->encap_info.encap_type;
 443                 switch (encap) {
 444                 case FPM_NH_ENCAP_NONE:
 445                 case FPM_NH_ENCAP_MAX:
 446                         break;
 447                 case FPM_NH_ENCAP_VXLAN:
 448                         nl_attr_put16(&req->n, in_buf_len, RTA_ENCAP_TYPE,
 449                                       encap);
 450                         vxlan = &nhi->encap_info.vxlan_encap;
 451                         nest = nl_attr_nest(&req->n, in_buf_len, RTA_ENCAP);
 452                         nl_attr_put32(&req->n, in_buf_len, VXLAN_VNI,
 453                                       vxlan->vni);
 454                         nl_attr_nest_end(&req->n, nest);
 455                         break;
 456                 }
 457
 458                 goto done;
 459         }
 460
 461         /*
 462          * Multipath case.
 463          */
 464         nest = nl_attr_nest(&req->n, in_buf_len, RTA_MULTIPATH);
 465
 466         for (nexthop_num = 0; nexthop_num < ri->num_nhs; nexthop_num++) {
 467                 rtnh = nl_attr_rtnh(&req->n, in_buf_len);
 468                 nhi = &ri->nhs[nexthop_num];
 469
 470                 if (nhi->gateway)
 471                         nl_attr_put(&req->n, in_buf_len, RTA_GATEWAY,
 472                                     nhi->gateway, bytelen);
 473
 474                 if (nhi->if_index) {
 475                         rtnh->rtnh_ifindex = nhi->if_index;
 476                 }
 477
 478                 encap = nhi->encap_info.encap_type;
 479                 switch (encap) {
 480                 case FPM_NH_ENCAP_NONE:
 481                 case FPM_NH_ENCAP_MAX:
 482                         break;
 483                 case FPM_NH_ENCAP_VXLAN:
 484                         nl_attr_put16(&req->n, in_buf_len, RTA_ENCAP_TYPE,
 485                                       encap);
 486                         vxlan = &nhi->encap_info.vxlan_encap;
 487                         inner_nest =
 488                                 nl_attr_nest(&req->n, in_buf_len, RTA_ENCAP);
 489                         nl_attr_put32(&req->n, in_buf_len, VXLAN_VNI,
 490                                       vxlan->vni);
 491                         nl_attr_nest_end(&req->n, inner_nest);
 492                         break;
 493                 }
 494
 495                 nl_attr_rtnh_end(&req->n, rtnh);
 496         }
 497
 498         nl_attr_nest_end(&req->n, nest);
 499         assert(nest->rta_len > RTA_LENGTH(0));
 500
 501 done:
 502
 503         if (ri->pref_src) {
 504                 nl_attr_put(&req->n, in_buf_len, RTA_PREFSRC, &ri->pref_src,
 505                             bytelen);
 506         }
 507
 508         assert(req->n.nlmsg_len < in_buf_len);
 509         return req->n.nlmsg_len;
 510 }
 511
 512 /*
 513  * zfpm_log_route_info
 514  *
 515  * Helper function to log the information in a route_info structure.
 516  */
 517 static void zfpm_log_route_info(struct netlink_route_info *ri,
 518                                 const char *label)
 519 {
 520         struct netlink_nh_info *nhi;
 521         unsigned int i;
 522         char buf[PREFIX_STRLEN];
 523
 524         zfpm_debug("%s : %s %pFX, Proto: %s, Metric: %u", label,
 525                    nl_msg_type_to_str(ri->nlmsg_type), ri->prefix,
 526                    nl_rtproto_to_str(ri->rtm_protocol),
 527                    ri->metric ? *ri->metric : 0);
 528
 529         for (i = 0; i < ri->num_nhs; i++) {
 530                 nhi = &ri->nhs[i];
 531
 532                 if (ri->af == AF_INET)
 533                         inet_ntop(AF_INET, &nhi->gateway, buf, sizeof(buf));
 534                 else
 535                         inet_ntop(AF_INET6, &nhi->gateway, buf, sizeof(buf));
 536
 537                 zfpm_debug("  Intf: %u, Gateway: %s, Recursive: %s, Type: %s, Encap type: %s",
 538                            nhi->if_index, buf, nhi->recursive ? "yes" : "no",
 539                            nexthop_type_to_str(nhi->type),
 540                            fpm_nh_encap_type_to_str(nhi->encap_info.encap_type)
 541                            );
 542         }
 543 }
 544
 545 /*
 546  * zfpm_netlink_encode_route
 547  *
 548  * Create a netlink message corresponding to the given route in the
 549  * given buffer space.
 550  *
 551  * Returns the number of bytes written to the buffer. 0 or a negative
 552  * value indicates an error.
 553  */
 554 int zfpm_netlink_encode_route(int cmd, rib_dest_t *dest, struct route_entry *re,
 555                               char *in_buf, size_t in_buf_len)
 556 {
 557         struct netlink_route_info ri_space, *ri;
 558
 559         ri = &ri_space;
 560
 561         if (!netlink_route_info_fill(ri, cmd, dest, re))
 562                 return 0;
 563
 564         zfpm_log_route_info(ri, __func__);
 565
 566         return netlink_route_info_encode(ri, in_buf, in_buf_len);
 567 }
 568
 569 /*
 570  * zfpm_netlink_encode_mac
 571  *
 572  * Create a netlink message corresponding to the given MAC.
 573  *
 574  * Returns the number of bytes written to the buffer. 0 or a negative
 575  * value indicates an error.
 576  */
 577 int zfpm_netlink_encode_mac(struct fpm_mac_info_t *mac, char *in_buf,
 578                             size_t in_buf_len)
 579 {
 580         size_t buf_offset;
 581
 582         struct macmsg {
 583                 struct nlmsghdr hdr;
 584                 struct ndmsg ndm;
 585                 char buf[0];
 586         } *req;
 587         req = (void *)in_buf;
 588
 589         buf_offset = offsetof(struct macmsg, buf);
 590         if (in_buf_len < buf_offset)
 591                 return 0;
 592         memset(req, 0, buf_offset);
 593
 594         /* Construct nlmsg header */
 595         req->hdr.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
 596         req->hdr.nlmsg_type = CHECK_FLAG(mac->fpm_flags, ZEBRA_MAC_DELETE_FPM) ?
 597                                 RTM_DELNEIGH : RTM_NEWNEIGH;
 598         req->hdr.nlmsg_flags = NLM_F_REQUEST;
 599         if (req->hdr.nlmsg_type == RTM_NEWNEIGH)
 600                 req->hdr.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
 601
 602         /* Construct ndmsg */
 603         req->ndm.ndm_family = AF_BRIDGE;
 604         req->ndm.ndm_ifindex = mac->vxlan_if;
 605
 606         req->ndm.ndm_state = NUD_REACHABLE;
 607         req->ndm.ndm_flags |= NTF_SELF | NTF_MASTER;
 608         if (CHECK_FLAG(mac->zebra_flags,
 609                 (ZEBRA_MAC_STICKY | ZEBRA_MAC_REMOTE_DEF_GW)))
 610                 req->ndm.ndm_state |= NUD_NOARP;
 611         else
 612                 req->ndm.ndm_flags |= NTF_EXT_LEARNED;
 613
 614         /* Add attributes */
 615         nl_attr_put(&req->hdr, in_buf_len, NDA_LLADDR, &mac->macaddr, 6);
 616         nl_attr_put(&req->hdr, in_buf_len, NDA_DST, &mac->r_vtep_ip, 4);
 617         nl_attr_put32(&req->hdr, in_buf_len, NDA_MASTER, mac->svi_if);
 618         nl_attr_put32(&req->hdr, in_buf_len, NDA_VNI, mac->vni);
 619
 620         assert(req->hdr.nlmsg_len < in_buf_len);
 621
 622         zfpm_debug("Tx %s family %s ifindex %u MAC %pEA DEST %pI4",
 623                    nl_msg_type_to_str(req->hdr.nlmsg_type),
 624                    nl_family_to_str(req->ndm.ndm_family), req->ndm.ndm_ifindex,
 625                    &mac->macaddr, &mac->r_vtep_ip);
 626
 627         return req->hdr.nlmsg_len;
 628 }
 629
 630 #endif /* HAVE_NETLINK */