zebra/rt_socket.c

   1 /*
   2  * Kernel routing table updates by routing socket.
   3  * Copyright (C) 1997, 98 Kunihiro Ishiguro
   4  *
   5  * This file is part of GNU Zebra.
   6  *
   7  * GNU Zebra is free software; you can redistribute it and/or modify it
   8  * under the terms of the GNU General Public License as published by the
   9  * Free Software Foundation; either version 2, or (at your option) any
  10  * later version.
  11  *
  12  * GNU Zebra is distributed in the hope that it will be useful, but
  13  * WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License along
  18  * with this program; see the file COPYING; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 #include <zebra.h>
  23
  24 #ifndef HAVE_NETLINK
  25
  26 #ifdef __OpenBSD__
  27 #include <netmpls/mpls.h>
  28 #endif
  29
  30 #include "if.h"
  31 #include "prefix.h"
  32 #include "sockunion.h"
  33 #include "log.h"
  34 #include "privs.h"
  35 #include "vxlan.h"
  36 #include "lib_errors.h"
  37
  38 #include "zebra/debug.h"
  39 #include "zebra/rib.h"
  40 #include "zebra/rt.h"
  41 #include "zebra/kernel_socket.h"
  42 #include "zebra/zebra_mpls.h"
  43 #include "zebra/zebra_errors.h"
  44
  45 extern struct zebra_privs_t zserv_privs;
  46
  47 #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
  48 /* Adjust netmask socket length. Return value is a adjusted sin_len
  49    value. */
  50 static int sin_masklen(struct in_addr mask)
  51 {
  52         char *p, *lim;
  53         int len;
  54         struct sockaddr_in sin;
  55
  56         if (mask.s_addr == 0)
  57                 return sizeof(long);
  58
  59         sin.sin_addr = mask;
  60         len = sizeof(struct sockaddr_in);
  61
  62         lim = (char *)&sin.sin_addr;
  63         p = lim + sizeof(sin.sin_addr);
  64
  65         while (*--p == 0 && p >= lim)
  66                 len--;
  67         return len;
  68 }
  69 #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
  70
  71 #ifdef __OpenBSD__
  72 static int kernel_rtm_add_labels(struct mpls_label_stack *nh_label,
  73                                  struct sockaddr_mpls *smpls)
  74 {
  75         if (nh_label->num_labels > 1) {
  76                 flog_warn(EC_ZEBRA_MAX_LABELS_PUSH,
  77                           "%s: can't push %u labels at "
  78                           "once (maximum is 1)",
  79                           __func__, nh_label->num_labels);
  80                 return -1;
  81         }
  82
  83         memset(smpls, 0, sizeof(*smpls));
  84         smpls->smpls_len = sizeof(*smpls);
  85         smpls->smpls_family = AF_MPLS;
  86         smpls->smpls_label = htonl(nh_label->label[0] << MPLS_LABEL_OFFSET);
  87
  88         return 0;
  89 }
  90 #endif
  91
  92 /* Interface between zebra message and rtm message. */
  93 static int kernel_rtm_ipv4(int cmd, const struct prefix *p,
  94                            const struct nexthop_group *ng, uint32_t metric)
  95
  96 {
  97         struct sockaddr_in *mask = NULL;
  98         struct sockaddr_in sin_dest, sin_mask, sin_gate;
  99 #ifdef __OpenBSD__
 100         struct sockaddr_mpls smpls;
 101 #endif
 102         union sockunion *smplsp = NULL;
 103         struct nexthop *nexthop;
 104         int nexthop_num = 0;
 105         ifindex_t ifindex = 0;
 106         int gate = 0;
 107         int error;
 108         char prefix_buf[PREFIX_STRLEN];
 109         enum blackhole_type bh_type = BLACKHOLE_UNSPEC;
 110
 111         if (IS_ZEBRA_DEBUG_RIB)
 112                 prefix2str(p, prefix_buf, sizeof(prefix_buf));
 113         memset(&sin_dest, 0, sizeof(struct sockaddr_in));
 114         sin_dest.sin_family = AF_INET;
 115 #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
 116         sin_dest.sin_len = sizeof(struct sockaddr_in);
 117 #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
 118         sin_dest.sin_addr = p->u.prefix4;
 119
 120         memset(&sin_mask, 0, sizeof(struct sockaddr_in));
 121
 122         memset(&sin_gate, 0, sizeof(struct sockaddr_in));
 123         sin_gate.sin_family = AF_INET;
 124 #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
 125         sin_gate.sin_len = sizeof(struct sockaddr_in);
 126 #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
 127
 128         /* Make gateway. */
 129         for (ALL_NEXTHOPS_PTR(ng, nexthop)) {
 130                 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
 131                         continue;
 132
 133                 gate = 0;
 134                 char gate_buf[INET_ADDRSTRLEN] = "NULL";
 135
 136                 /*
 137                  * XXX We need to refrain from kernel operations in some cases,
 138                  * but this if statement seems overly cautious - what about
 139                  * other than ADD and DELETE?
 140                  */
 141                 if ((cmd == RTM_ADD && NEXTHOP_IS_ACTIVE(nexthop->flags))
 142                     || (cmd == RTM_DELETE)) {
 143                         if (nexthop->type == NEXTHOP_TYPE_IPV4
 144                             || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
 145                                 sin_gate.sin_addr = nexthop->gate.ipv4;
 146                                 gate = 1;
 147                         }
 148                         if (nexthop->type == NEXTHOP_TYPE_IFINDEX
 149                             || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX)
 150                                 ifindex = nexthop->ifindex;
 151                         if (nexthop->type == NEXTHOP_TYPE_BLACKHOLE) {
 152                                 struct in_addr loopback;
 153                                 loopback.s_addr = htonl(INADDR_LOOPBACK);
 154                                 sin_gate.sin_addr = loopback;
 155                                 bh_type = nexthop->bh_type;
 156                                 gate = 1;
 157                         }
 158
 159                         if (gate && p->prefixlen == 32)
 160                                 mask = NULL;
 161                         else {
 162                                 masklen2ip(p->prefixlen, &sin_mask.sin_addr);
 163                                 sin_mask.sin_family = AF_INET;
 164 #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
 165                                 sin_mask.sin_len =
 166                                         sin_masklen(sin_mask.sin_addr);
 167 #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
 168                                 mask = &sin_mask;
 169                         }
 170
 171 #ifdef __OpenBSD__
 172                         if (nexthop->nh_label
 173                             && !kernel_rtm_add_labels(nexthop->nh_label,
 174                                                       &smpls))
 175                                 continue;
 176                         smplsp = (union sockunion *)&smpls;
 177 #endif
 178
 179                         error = rtm_write(cmd, (union sockunion *)&sin_dest,
 180                                           (union sockunion *)mask,
 181                                           gate ? (union sockunion *)&sin_gate
 182                                                : NULL,
 183                                           smplsp, ifindex, bh_type, metric);
 184
 185                         if (IS_ZEBRA_DEBUG_KERNEL) {
 186                                 if (!gate) {
 187                                         zlog_debug(
 188                                                 "%s: %s: attention! gate not found for re",
 189                                                 __func__, prefix_buf);
 190                                 } else
 191                                         inet_ntop(AF_INET, &sin_gate.sin_addr,
 192                                                   gate_buf, INET_ADDRSTRLEN);
 193                         }
 194
 195                         switch (error) {
 196                         /* We only flag nexthops as being in FIB if rtm_write()
 197                          * did its work. */
 198                         case ZEBRA_ERR_NOERROR:
 199                                 nexthop_num++;
 200                                 if (IS_ZEBRA_DEBUG_KERNEL)
 201                                         zlog_debug(
 202                                                 "%s: %s: successfully did NH %s",
 203                                                 __func__, prefix_buf, gate_buf);
 204
 205                                 if (cmd == RTM_ADD)
 206                                         SET_FLAG(nexthop->flags,
 207                                                  NEXTHOP_FLAG_FIB);
 208
 209                                 break;
 210
 211                         /* The only valid case for this error is kernel's
 212                          * failure to install
 213                          * a multipath route, which is common for FreeBSD. This
 214                          * should be
 215                          * ignored silently, but logged as an error otherwise.
 216                          */
 217                         case ZEBRA_ERR_RTEXIST:
 218                                 if (cmd != RTM_ADD)
 219                                         flog_err(
 220                                                 EC_LIB_SYSTEM_CALL,
 221                                                 "%s: rtm_write() returned %d for command %d",
 222                                                 __func__, error, cmd);
 223                                 continue;
 224
 225                         /* Note any unexpected status returns */
 226                         default:
 227                                 flog_err(
 228                                         EC_LIB_SYSTEM_CALL,
 229                                         "%s: %s: rtm_write() unexpectedly returned %d for command %s",
 230                                         __func__,
 231                                         prefix2str(p, prefix_buf,
 232                                                    sizeof(prefix_buf)),
 233                                         error,
 234                                         lookup_msg(rtm_type_str, cmd, NULL));
 235                                 break;
 236                         }
 237                 } /* if (cmd and flags make sense) */
 238                 else if (IS_ZEBRA_DEBUG_KERNEL)
 239                         zlog_debug("%s: odd command %s for flags %d", __func__,
 240                                    lookup_msg(rtm_type_str, cmd, NULL),
 241                                    nexthop->flags);
 242         } /* for (ALL_NEXTHOPS(...))*/
 243
 244         /* If there was no useful nexthop, then complain. */
 245         if (nexthop_num == 0) {
 246                 if (IS_ZEBRA_DEBUG_KERNEL)
 247                         zlog_debug("%s: No useful nexthops were found in RIB prefix %s",
 248                                    __func__, prefix2str(p, prefix_buf,
 249                                                         sizeof(prefix_buf)));
 250                 return 1;
 251         }
 252
 253         return 0; /*XXX*/
 254 }
 255
 256 #ifdef SIN6_LEN
 257 /* Calculate sin6_len value for netmask socket value. */
 258 static int sin6_masklen(struct in6_addr mask)
 259 {
 260         struct sockaddr_in6 sin6;
 261         char *p, *lim;
 262         int len;
 263
 264         if (IN6_IS_ADDR_UNSPECIFIED(&mask))
 265                 return sizeof(long);
 266
 267         sin6.sin6_addr = mask;
 268         len = sizeof(struct sockaddr_in6);
 269
 270         lim = (char *)&sin6.sin6_addr;
 271         p = lim + sizeof(sin6.sin6_addr);
 272
 273         while (*--p == 0 && p >= lim)
 274                 len--;
 275
 276         return len;
 277 }
 278 #endif /* SIN6_LEN */
 279
 280 /* Interface between zebra message and rtm message. */
 281 static int kernel_rtm_ipv6(int cmd, const struct prefix *p,
 282                            const struct nexthop_group *ng, uint32_t metric)
 283 {
 284         struct sockaddr_in6 *mask;
 285         struct sockaddr_in6 sin_dest, sin_mask, sin_gate;
 286 #ifdef __OpenBSD__
 287         struct sockaddr_mpls smpls;
 288 #endif
 289         union sockunion *smplsp = NULL;
 290         struct nexthop *nexthop;
 291         int nexthop_num = 0;
 292         ifindex_t ifindex = 0;
 293         int gate = 0;
 294         int error;
 295         enum blackhole_type bh_type = BLACKHOLE_UNSPEC;
 296
 297         memset(&sin_dest, 0, sizeof(struct sockaddr_in6));
 298         sin_dest.sin6_family = AF_INET6;
 299 #ifdef SIN6_LEN
 300         sin_dest.sin6_len = sizeof(struct sockaddr_in6);
 301 #endif /* SIN6_LEN */
 302         sin_dest.sin6_addr = p->u.prefix6;
 303
 304         memset(&sin_mask, 0, sizeof(struct sockaddr_in6));
 305
 306         memset(&sin_gate, 0, sizeof(struct sockaddr_in6));
 307         sin_gate.sin6_family = AF_INET6;
 308 #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
 309         sin_gate.sin6_len = sizeof(struct sockaddr_in6);
 310 #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
 311
 312         /* Make gateway. */
 313         for (ALL_NEXTHOPS_PTR(ng, nexthop)) {
 314                 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
 315                         continue;
 316
 317                 gate = 0;
 318
 319                 if ((cmd == RTM_ADD && NEXTHOP_IS_ACTIVE(nexthop->flags))
 320                     || (cmd == RTM_DELETE)) {
 321                         if (nexthop->type == NEXTHOP_TYPE_IPV6
 322                             || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
 323                                 sin_gate.sin6_addr = nexthop->gate.ipv6;
 324                                 gate = 1;
 325                         }
 326                         if (nexthop->type == NEXTHOP_TYPE_IFINDEX
 327                             || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)
 328                                 ifindex = nexthop->ifindex;
 329
 330                         if (nexthop->type == NEXTHOP_TYPE_BLACKHOLE)
 331                                 bh_type = nexthop->bh_type;
 332                 }
 333
 334 /* Under kame set interface index to link local address. */
 335 #ifdef KAME
 336
 337 #define SET_IN6_LINKLOCAL_IFINDEX(a, i)                                        \
 338         do {                                                                   \
 339                 (a).s6_addr[2] = ((i) >> 8) & 0xff;                            \
 340                 (a).s6_addr[3] = (i)&0xff;                                     \
 341         } while (0)
 342
 343                 if (gate && IN6_IS_ADDR_LINKLOCAL(&sin_gate.sin6_addr))
 344                         SET_IN6_LINKLOCAL_IFINDEX(sin_gate.sin6_addr, ifindex);
 345 #endif /* KAME */
 346
 347                 if (gate && p->prefixlen == 128)
 348                         mask = NULL;
 349                 else {
 350                         masklen2ip6(p->prefixlen, &sin_mask.sin6_addr);
 351                         sin_mask.sin6_family = AF_INET6;
 352 #ifdef SIN6_LEN
 353                         sin_mask.sin6_len = sin6_masklen(sin_mask.sin6_addr);
 354 #endif /* SIN6_LEN */
 355                         mask = &sin_mask;
 356                 }
 357
 358 #ifdef __OpenBSD__
 359                 if (nexthop->nh_label
 360                     && !kernel_rtm_add_labels(nexthop->nh_label, &smpls))
 361                         continue;
 362                 smplsp = (union sockunion *)&smpls;
 363 #endif
 364
 365                 error = rtm_write(cmd, (union sockunion *)&sin_dest,
 366                                   (union sockunion *)mask,
 367                                   gate ? (union sockunion *)&sin_gate : NULL,
 368                                   smplsp, ifindex, bh_type, metric);
 369
 370                 /* Update installed nexthop info on success */
 371                 if ((cmd == RTM_ADD) && (error == ZEBRA_ERR_NOERROR))
 372                         SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB);
 373
 374                 nexthop_num++;
 375         }
 376
 377         /* If there is no useful nexthop then return. */
 378         if (nexthop_num == 0) {
 379                 if (IS_ZEBRA_DEBUG_KERNEL)
 380                         zlog_debug("kernel_rtm_ipv6(): No useful nexthop.");
 381                 return 1;
 382         }
 383
 384         return 0; /*XXX*/
 385 }
 386
 387 static int kernel_rtm(int cmd, const struct prefix *p,
 388                       const struct nexthop_group *ng, uint32_t metric)
 389 {
 390         switch (PREFIX_FAMILY(p)) {
 391         case AF_INET:
 392                 return kernel_rtm_ipv4(cmd, p, ng, metric);
 393         case AF_INET6:
 394                 return kernel_rtm_ipv6(cmd, p, ng, metric);
 395         }
 396         return 0;
 397 }
 398
 399 /*
 400  * Update or delete a prefix from the kernel,
 401  * using info from a dataplane context struct.
 402  */
 403 enum zebra_dplane_result kernel_route_update(dplane_ctx_h ctx)
 404 {
 405         enum zebra_dplane_result res = ZEBRA_DPLANE_REQUEST_SUCCESS;
 406
 407         if (dplane_ctx_get_src(ctx) != NULL) {
 408                 zlog_err("route add: IPv6 sourcedest routes unsupported!");
 409                 res = ZEBRA_DPLANE_REQUEST_FAILURE;
 410                 goto done;
 411         }
 412
 413         frr_elevate_privs(ZPRIVS_RAISE) {
 414
 415                 if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_DELETE)
 416                         kernel_rtm(RTM_DELETE, dplane_ctx_get_dest(ctx),
 417                                    dplane_ctx_get_ng(ctx),
 418                                    dplane_ctx_get_metric(ctx));
 419                 else if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_INSTALL)
 420                         kernel_rtm(RTM_ADD, dplane_ctx_get_dest(ctx),
 421                                    dplane_ctx_get_ng(ctx),
 422                                    dplane_ctx_get_metric(ctx));
 423                 else if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_UPDATE) {
 424                         /* Must do delete and add separately -
 425                          * no update available
 426                          */
 427                         kernel_rtm(RTM_DELETE, dplane_ctx_get_dest(ctx),
 428                                    dplane_ctx_get_old_ng(ctx),
 429                                    dplane_ctx_get_old_metric(ctx));
 430
 431                         kernel_rtm(RTM_ADD, dplane_ctx_get_dest(ctx),
 432                                    dplane_ctx_get_ng(ctx),
 433                                    dplane_ctx_get_metric(ctx));
 434                 } else {
 435                         zlog_err("Invalid routing socket update op %s (%u)",
 436                                  dplane_op2str(dplane_ctx_get_op(ctx)),
 437                                  dplane_ctx_get_op(ctx));
 438                         res = ZEBRA_DPLANE_REQUEST_FAILURE;
 439                 }
 440         } /* Elevated privs */
 441
 442 done:
 443
 444         return res;
 445 }
 446
 447 enum zebra_dplane_result kernel_route_rib(struct route_node *rn,
 448                                           const struct prefix *p,
 449                                           const struct prefix *src_p,
 450                                           struct route_entry *old,
 451                                           struct route_entry *new)
 452 {
 453         int route = 0;
 454
 455         if (src_p && src_p->prefixlen) {
 456                 flog_warn(EC_ZEBRA_UNSUPPORTED_V6_SRCDEST,
 457                           "%s: IPv6 sourcedest routes unsupported!", __func__);
 458                 return ZEBRA_DPLANE_REQUEST_FAILURE;
 459         }
 460
 461         frr_elevate_privs(&zserv_privs) {
 462                 if (old)
 463                         route |= kernel_rtm(RTM_DELETE, p,
 464                                             &old->ng, old->metric);
 465                 if (new)
 466                         route |= kernel_rtm(RTM_ADD, p, &new->ng, new->metric);
 467         }
 468
 469         if (new) {
 470                 kernel_route_rib_pass_fail(
 471                         rn, p, new,
 472                         (!route) ? ZEBRA_DPLANE_INSTALL_SUCCESS
 473                                  : ZEBRA_DPLANE_INSTALL_FAILURE);
 474         } else {
 475                 kernel_route_rib_pass_fail(rn, p, old,
 476                                            (!route)
 477                                                    ? ZEBRA_DPLANE_DELETE_SUCCESS
 478                                                    : ZEBRA_DPLANE_DELETE_FAILURE);
 479         }
 480
 481         return ZEBRA_DPLANE_REQUEST_SUCCESS;
 482 }
 483
 484 int kernel_neigh_update(int add, int ifindex, uint32_t addr, char *lla,
 485                         int llalen, ns_id_t ns_id)
 486 {
 487         /* TODO */
 488         return 0;
 489 }
 490
 491 extern int kernel_get_ipmr_sg_stats(struct zebra_vrf *zvrf, void *mroute)
 492 {
 493         return 0;
 494 }
 495
 496 int kernel_add_vtep(vni_t vni, struct interface *ifp, struct in_addr *vtep_ip)
 497 {
 498         return 0;
 499 }
 500
 501 int kernel_del_vtep(vni_t vni, struct interface *ifp, struct in_addr *vtep_ip)
 502 {
 503         return 0;
 504 }
 505
 506 int kernel_add_mac(struct interface *ifp, vlanid_t vid, struct ethaddr *mac,
 507                    struct in_addr vtep_ip, bool sticky)
 508 {
 509         return 0;
 510 }
 511
 512 int kernel_del_mac(struct interface *ifp, vlanid_t vid, struct ethaddr *mac,
 513                    struct in_addr vtep_ip)
 514 {
 515         return 0;
 516 }
 517
 518 int kernel_add_neigh(struct interface *ifp, struct ipaddr *ip,
 519                      struct ethaddr *mac, uint8_t flags)
 520 {
 521         return 0;
 522 }
 523
 524 int kernel_del_neigh(struct interface *ifp, struct ipaddr *ip)
 525 {
 526         return 0;
 527 }
 528
 529 extern int kernel_interface_set_master(struct interface *master,
 530                                        struct interface *slave)
 531 {
 532         return 0;
 533 }
 534
 535 uint32_t kernel_get_speed(struct interface *ifp)
 536 {
 537         return ifp->speed;
 538 }
 539
 540 #endif /* !HAVE_NETLINK */