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         /* Weight of the nexthop ( for unequal cost ECMP ) */
 120         uint8_t weight;
 121         uint32_t if_index;
 122         union g_addr *gateway;
 123
 124         /*
 125          * Information from the struct nexthop from which this nh was
 126          * derived. For debug purposes only.
 127          */
 128         int recursive;
 129         enum nexthop_types_t type;
 130         struct fpm_nh_encap_info_t encap_info;
 131 };
 132
 133 /*
 134  * netlink_route_info
 135  *
 136  * A structure for holding information for a netlink route message.
 137  */
 138 struct netlink_route_info {
 139         uint32_t nlmsg_pid;
 140         uint16_t nlmsg_type;
 141         uint8_t rtm_type;
 142         uint32_t rtm_table;
 143         uint8_t rtm_protocol;
 144         uint8_t af;
 145         struct prefix *prefix;
 146         uint32_t *metric;
 147         unsigned int num_nhs;
 148
 149         /*
 150          * Nexthop structures
 151          */
 152         struct netlink_nh_info nhs[MULTIPATH_NUM];
 153         union g_addr *pref_src;
 154 };
 155
 156 /*
 157  * netlink_route_info_add_nh
 158  *
 159  * Add information about the given nexthop to the given route info
 160  * structure.
 161  *
 162  * Returns true if a nexthop was added, false otherwise.
 163  */
 164 static int netlink_route_info_add_nh(struct netlink_route_info *ri,
 165                                      struct nexthop *nexthop,
 166                                      struct route_entry *re)
 167 {
 168         struct netlink_nh_info nhi;
 169         union g_addr *src;
 170         struct zebra_vrf *zvrf = NULL;
 171         struct interface *ifp = NULL, *link_if = NULL;
 172         struct zebra_if *zif = NULL;
 173         vni_t vni = 0;
 174
 175         memset(&nhi, 0, sizeof(nhi));
 176         src = NULL;
 177
 178         if (ri->num_nhs >= (int)array_size(ri->nhs))
 179                 return 0;
 180
 181         nhi.recursive = nexthop->rparent ? 1 : 0;
 182         nhi.type = nexthop->type;
 183         nhi.if_index = nexthop->ifindex;
 184         nhi.weight = nexthop->weight;
 185
 186         if (nexthop->type == NEXTHOP_TYPE_IPV4
 187             || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
 188                 nhi.gateway = &nexthop->gate;
 189                 if (nexthop->src.ipv4.s_addr != INADDR_ANY)
 190                         src = &nexthop->src;
 191         }
 192
 193         if (nexthop->type == NEXTHOP_TYPE_IPV6
 194             || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
 195                 /* Special handling for IPv4 route with IPv6 Link Local next hop
 196                  */
 197                 if (ri->af == AF_INET)
 198                         nhi.gateway = &ipv4ll_gateway;
 199                 else
 200                         nhi.gateway = &nexthop->gate;
 201         }
 202
 203         if (nexthop->type == NEXTHOP_TYPE_IFINDEX) {
 204                 if (nexthop->src.ipv4.s_addr != INADDR_ANY)
 205                         src = &nexthop->src;
 206         }
 207
 208         if (!nhi.gateway && nhi.if_index == 0)
 209                 return 0;
 210
 211         if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_EVPN)) {
 212                 nhi.encap_info.encap_type = FPM_NH_ENCAP_VXLAN;
 213
 214                 /* Extract VNI id for the nexthop SVI interface */
 215                 zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id);
 216                 if (zvrf) {
 217                         ifp = if_lookup_by_index_per_ns(zvrf->zns,
 218                                                         nexthop->ifindex);
 219                         if (ifp) {
 220                                 zif = (struct zebra_if *)ifp->info;
 221                                 if (zif) {
 222                                         if (IS_ZEBRA_IF_BRIDGE(ifp))
 223                                                 link_if = ifp;
 224                                         else if (IS_ZEBRA_IF_VLAN(ifp))
 225                                                 link_if =
 226                                                 if_lookup_by_index_per_ns(
 227                                                         zvrf->zns,
 228                                                         zif->link_ifindex);
 229                                         if (link_if)
 230                                                 vni = vni_id_from_svi(ifp,
 231                                                                       link_if);
 232                                 }
 233                         }
 234                 }
 235
 236                 nhi.encap_info.vxlan_encap.vni = vni;
 237         }
 238
 239         /*
 240          * We have a valid nhi. Copy the structure over to the route_info.
 241          */
 242         ri->nhs[ri->num_nhs] = nhi;
 243         ri->num_nhs++;
 244
 245         if (src && !ri->pref_src)
 246                 ri->pref_src = src;
 247
 248         return 1;
 249 }
 250
 251 /*
 252  * netlink_proto_from_route_type
 253  */
 254 static uint8_t netlink_proto_from_route_type(int type)
 255 {
 256         return zebra2proto(type);
 257 }
 258
 259 /*
 260  * netlink_route_info_fill
 261  *
 262  * Fill out the route information object from the given route.
 263  *
 264  * Returns true on success and false on failure.
 265  */
 266 static int netlink_route_info_fill(struct netlink_route_info *ri, int cmd,
 267                                    rib_dest_t *dest, struct route_entry *re)
 268 {
 269         struct nexthop *nexthop;
 270         struct rib_table_info *table_info =
 271                 rib_table_info(rib_dest_table(dest));
 272         struct zebra_vrf *zvrf = table_info->zvrf;
 273
 274         memset(ri, 0, sizeof(*ri));
 275
 276         ri->prefix = rib_dest_prefix(dest);
 277         ri->af = rib_dest_af(dest);
 278
 279         if (zvrf && zvrf->zns)
 280                 ri->nlmsg_pid = zvrf->zns->netlink_dplane_out.snl.nl_pid;
 281
 282         ri->nlmsg_type = cmd;
 283         ri->rtm_table = table_info->table_id;
 284         ri->rtm_protocol = RTPROT_UNSPEC;
 285
 286         /*
 287          * An RTM_DELROUTE need not be accompanied by any nexthops,
 288          * particularly in our communication with the FPM.
 289          */
 290         if (cmd == RTM_DELROUTE && !re)
 291                 return 1;
 292
 293         if (!re) {
 294                 zfpm_debug("%s: Expected non-NULL re pointer", __func__);
 295                 return 0;
 296         }
 297
 298         ri->rtm_protocol = netlink_proto_from_route_type(re->type);
 299         ri->rtm_type = RTN_UNICAST;
 300         ri->metric = &re->metric;
 301
 302         for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) {
 303                 if (ri->num_nhs >= zrouter.multipath_num)
 304                         break;
 305
 306                 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
 307                         continue;
 308
 309                 if (nexthop->type == NEXTHOP_TYPE_BLACKHOLE) {
 310                         switch (nexthop->bh_type) {
 311                         case BLACKHOLE_ADMINPROHIB:
 312                                 ri->rtm_type = RTN_PROHIBIT;
 313                                 break;
 314                         case BLACKHOLE_REJECT:
 315                                 ri->rtm_type = RTN_UNREACHABLE;
 316                                 break;
 317                         case BLACKHOLE_NULL:
 318                         default:
 319                                 ri->rtm_type = RTN_BLACKHOLE;
 320                                 break;
 321                         }
 322                 }
 323
 324                 if ((cmd == RTM_NEWROUTE
 325                      && CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE))
 326                     || (cmd == RTM_DELROUTE
 327                         && CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED))) {
 328                         netlink_route_info_add_nh(ri, nexthop, re);
 329                 }
 330         }
 331
 332         if (ri->num_nhs == 0) {
 333                 switch (ri->rtm_type) {
 334                 case RTN_PROHIBIT:
 335                 case RTN_UNREACHABLE:
 336                 case RTN_BLACKHOLE:
 337                         break;
 338                 default:
 339                         /* If there is no useful nexthop then return. */
 340                         zfpm_debug(
 341                                 "netlink_encode_route(): No useful nexthop.");
 342                         return 0;
 343                 }
 344         }
 345
 346         return 1;
 347 }
 348
 349 /*
 350  * netlink_route_info_encode
 351  *
 352  * Returns the number of bytes written to the buffer. 0 or a negative
 353  * value indicates an error.
 354  */
 355 static int netlink_route_info_encode(struct netlink_route_info *ri,
 356                                      char *in_buf, size_t in_buf_len)
 357 {
 358         size_t bytelen;
 359         unsigned int nexthop_num = 0;
 360         size_t buf_offset;
 361         struct netlink_nh_info *nhi;
 362         enum fpm_nh_encap_type_t encap;
 363         struct rtattr *nest, *inner_nest;
 364         struct rtnexthop *rtnh;
 365         struct vxlan_encap_info_t *vxlan;
 366         struct in6_addr ipv6;
 367
 368         struct {
 369                 struct nlmsghdr n;
 370                 struct rtmsg r;
 371                 char buf[1];
 372         } * req;
 373
 374         req = (void *)in_buf;
 375
 376         buf_offset = ((char *)req->buf) - ((char *)req);
 377
 378         if (in_buf_len < buf_offset) {
 379                 assert(0);
 380                 return 0;
 381         }
 382
 383         memset(req, 0, buf_offset);
 384
 385         bytelen = af_addr_size(ri->af);
 386
 387         req->n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
 388         req->n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
 389         req->n.nlmsg_pid = ri->nlmsg_pid;
 390         req->n.nlmsg_type = ri->nlmsg_type;
 391         req->r.rtm_family = ri->af;
 392
 393         /*
 394          * rtm_table field is a uchar field which can accommodate table_id less
 395          * than 256.
 396          * To support table id greater than 255, if the table_id is greater than
 397          * 255, set rtm_table to RT_TABLE_UNSPEC and add RTA_TABLE attribute
 398          * with 32 bit value as the table_id.
 399          */
 400         if (ri->rtm_table < 256)
 401                 req->r.rtm_table = ri->rtm_table;
 402         else {
 403                 req->r.rtm_table = RT_TABLE_UNSPEC;
 404                 nl_attr_put32(&req->n, in_buf_len, RTA_TABLE, ri->rtm_table);
 405         }
 406
 407         req->r.rtm_dst_len = ri->prefix->prefixlen;
 408         req->r.rtm_protocol = ri->rtm_protocol;
 409         req->r.rtm_scope = RT_SCOPE_UNIVERSE;
 410
 411         nl_attr_put(&req->n, in_buf_len, RTA_DST, &ri->prefix->u.prefix,
 412                     bytelen);
 413
 414         req->r.rtm_type = ri->rtm_type;
 415
 416         /* Metric. */
 417         if (ri->metric)
 418                 nl_attr_put32(&req->n, in_buf_len, RTA_PRIORITY, *ri->metric);
 419
 420         if (ri->num_nhs == 0)
 421                 goto done;
 422
 423         if (ri->num_nhs == 1) {
 424                 nhi = &ri->nhs[0];
 425
 426                 if (nhi->gateway) {
 427                         if (nhi->type == NEXTHOP_TYPE_IPV4_IFINDEX
 428                             && ri->af == AF_INET6) {
 429                                 ipv4_to_ipv4_mapped_ipv6(&ipv6,
 430                                                          nhi->gateway->ipv4);
 431                                 nl_attr_put(&req->n, in_buf_len, RTA_GATEWAY,
 432                                             &ipv6, bytelen);
 433                         } else
 434                                 nl_attr_put(&req->n, in_buf_len, RTA_GATEWAY,
 435                                             nhi->gateway, bytelen);
 436                 }
 437
 438                 if (nhi->if_index) {
 439                         nl_attr_put32(&req->n, in_buf_len, RTA_OIF,
 440                                       nhi->if_index);
 441                 }
 442
 443                 encap = nhi->encap_info.encap_type;
 444                 switch (encap) {
 445                 case FPM_NH_ENCAP_NONE:
 446                 case FPM_NH_ENCAP_MAX:
 447                         break;
 448                 case FPM_NH_ENCAP_VXLAN:
 449                         nl_attr_put16(&req->n, in_buf_len, RTA_ENCAP_TYPE,
 450                                       encap);
 451                         vxlan = &nhi->encap_info.vxlan_encap;
 452                         nest = nl_attr_nest(&req->n, in_buf_len, RTA_ENCAP);
 453                         nl_attr_put32(&req->n, in_buf_len, VXLAN_VNI,
 454                                       vxlan->vni);
 455                         nl_attr_nest_end(&req->n, nest);
 456                         break;
 457                 }
 458
 459                 goto done;
 460         }
 461
 462         /*
 463          * Multipath case.
 464          */
 465         nest = nl_attr_nest(&req->n, in_buf_len, RTA_MULTIPATH);
 466
 467         for (nexthop_num = 0; nexthop_num < ri->num_nhs; nexthop_num++) {
 468                 rtnh = nl_attr_rtnh(&req->n, in_buf_len);
 469                 nhi = &ri->nhs[nexthop_num];
 470
 471                 if (nhi->gateway)
 472                         nl_attr_put(&req->n, in_buf_len, RTA_GATEWAY,
 473                                     nhi->gateway, bytelen);
 474
 475                 if (nhi->if_index) {
 476                         rtnh->rtnh_ifindex = nhi->if_index;
 477                 }
 478
 479                 rtnh->rtnh_hops = nhi->weight;
 480
 481                 encap = nhi->encap_info.encap_type;
 482                 switch (encap) {
 483                 case FPM_NH_ENCAP_NONE:
 484                 case FPM_NH_ENCAP_MAX:
 485                         break;
 486                 case FPM_NH_ENCAP_VXLAN:
 487                         nl_attr_put16(&req->n, in_buf_len, RTA_ENCAP_TYPE,
 488                                       encap);
 489                         vxlan = &nhi->encap_info.vxlan_encap;
 490                         inner_nest =
 491                                 nl_attr_nest(&req->n, in_buf_len, RTA_ENCAP);
 492                         nl_attr_put32(&req->n, in_buf_len, VXLAN_VNI,
 493                                       vxlan->vni);
 494                         nl_attr_nest_end(&req->n, inner_nest);
 495                         break;
 496                 }
 497
 498                 nl_attr_rtnh_end(&req->n, rtnh);
 499         }
 500
 501         nl_attr_nest_end(&req->n, nest);
 502         assert(nest->rta_len > RTA_LENGTH(0));
 503
 504 done:
 505
 506         if (ri->pref_src) {
 507                 nl_attr_put(&req->n, in_buf_len, RTA_PREFSRC, ri->pref_src,
 508                             bytelen);
 509         }
 510
 511         assert(req->n.nlmsg_len < in_buf_len);
 512         return req->n.nlmsg_len;
 513 }
 514
 515 /*
 516  * zfpm_log_route_info
 517  *
 518  * Helper function to log the information in a route_info structure.
 519  */
 520 static void zfpm_log_route_info(struct netlink_route_info *ri,
 521                                 const char *label)
 522 {
 523         struct netlink_nh_info *nhi;
 524         unsigned int i;
 525         char buf[PREFIX_STRLEN];
 526
 527         zfpm_debug("%s : %s %pFX, Proto: %s, Metric: %u", label,
 528                    nl_msg_type_to_str(ri->nlmsg_type), ri->prefix,
 529                    nl_rtproto_to_str(ri->rtm_protocol),
 530                    ri->metric ? *ri->metric : 0);
 531
 532         for (i = 0; i < ri->num_nhs; i++) {
 533                 nhi = &ri->nhs[i];
 534
 535                 if (nhi->gateway) {
 536                         if (ri->af == AF_INET)
 537                                 inet_ntop(AF_INET, nhi->gateway, buf,
 538                                           sizeof(buf));
 539                         else
 540                                 inet_ntop(AF_INET6, nhi->gateway, buf,
 541                                           sizeof(buf));
 542                 } else
 543                         strlcpy(buf, "none", sizeof(buf));
 544
 545                 zfpm_debug("  Intf: %u, Gateway: %s, Recursive: %s, Type: %s, Encap type: %s",
 546                            nhi->if_index, buf, nhi->recursive ? "yes" : "no",
 547                            nexthop_type_to_str(nhi->type),
 548                            fpm_nh_encap_type_to_str(nhi->encap_info.encap_type)
 549                            );
 550         }
 551 }
 552
 553 /*
 554  * zfpm_netlink_encode_route
 555  *
 556  * Create a netlink message corresponding to the given route in the
 557  * given buffer space.
 558  *
 559  * Returns the number of bytes written to the buffer. 0 or a negative
 560  * value indicates an error.
 561  */
 562 int zfpm_netlink_encode_route(int cmd, rib_dest_t *dest, struct route_entry *re,
 563                               char *in_buf, size_t in_buf_len)
 564 {
 565         struct netlink_route_info ri_space, *ri;
 566
 567         ri = &ri_space;
 568
 569         if (!netlink_route_info_fill(ri, cmd, dest, re))
 570                 return 0;
 571
 572         zfpm_log_route_info(ri, __func__);
 573
 574         return netlink_route_info_encode(ri, in_buf, in_buf_len);
 575 }
 576
 577 /*
 578  * zfpm_netlink_encode_mac
 579  *
 580  * Create a netlink message corresponding to the given MAC.
 581  *
 582  * Returns the number of bytes written to the buffer. 0 or a negative
 583  * value indicates an error.
 584  */
 585 int zfpm_netlink_encode_mac(struct fpm_mac_info_t *mac, char *in_buf,
 586                             size_t in_buf_len)
 587 {
 588         size_t buf_offset;
 589
 590         struct macmsg {
 591                 struct nlmsghdr hdr;
 592                 struct ndmsg ndm;
 593                 char buf[0];
 594         } *req;
 595         req = (void *)in_buf;
 596
 597         buf_offset = offsetof(struct macmsg, buf);
 598         if (in_buf_len < buf_offset)
 599                 return 0;
 600         memset(req, 0, buf_offset);
 601
 602         /* Construct nlmsg header */
 603         req->hdr.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
 604         req->hdr.nlmsg_type = CHECK_FLAG(mac->fpm_flags, ZEBRA_MAC_DELETE_FPM) ?
 605                                 RTM_DELNEIGH : RTM_NEWNEIGH;
 606         req->hdr.nlmsg_flags = NLM_F_REQUEST;
 607         if (req->hdr.nlmsg_type == RTM_NEWNEIGH)
 608                 req->hdr.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
 609
 610         /* Construct ndmsg */
 611         req->ndm.ndm_family = AF_BRIDGE;
 612         req->ndm.ndm_ifindex = mac->vxlan_if;
 613
 614         req->ndm.ndm_state = NUD_REACHABLE;
 615         req->ndm.ndm_flags |= NTF_SELF | NTF_MASTER;
 616         if (CHECK_FLAG(mac->zebra_flags,
 617                 (ZEBRA_MAC_STICKY | ZEBRA_MAC_REMOTE_DEF_GW)))
 618                 req->ndm.ndm_state |= NUD_NOARP;
 619         else
 620                 req->ndm.ndm_flags |= NTF_EXT_LEARNED;
 621
 622         /* Add attributes */
 623         nl_attr_put(&req->hdr, in_buf_len, NDA_LLADDR, &mac->macaddr, 6);
 624         nl_attr_put(&req->hdr, in_buf_len, NDA_DST, &mac->r_vtep_ip, 4);
 625         nl_attr_put32(&req->hdr, in_buf_len, NDA_MASTER, mac->svi_if);
 626         nl_attr_put32(&req->hdr, in_buf_len, NDA_VNI, mac->vni);
 627
 628         assert(req->hdr.nlmsg_len < in_buf_len);
 629
 630         zfpm_debug("Tx %s family %s ifindex %u MAC %pEA DEST %pI4",
 631                    nl_msg_type_to_str(req->hdr.nlmsg_type),
 632                    nl_family_to_str(req->ndm.ndm_family), req->ndm.ndm_ifindex,
 633                    &mac->macaddr, &mac->r_vtep_ip);
 634
 635         return req->hdr.nlmsg_len;
 636 }
 637
 638 #endif /* HAVE_NETLINK */