lib/netdev-vport.c

   1 /*
   2  * Copyright (c) 2010, 2011, 2012 Nicira, Inc.
   3  *
   4  * Licensed under the Apache License, Version 2.0 (the "License");
   5  * you may not use this file except in compliance with the License.
   6  * You may obtain a copy of the License at:
   7  *
   8  *     http://www.apache.org/licenses/LICENSE-2.0
   9  *
  10  * Unless required by applicable law or agreed to in writing, software
  11  * distributed under the License is distributed on an "AS IS" BASIS,
  12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13  * See the License for the specific language governing permissions and
  14  * limitations under the License.
  15  */
  16
  17 #include <config.h>
  18
  19 #include "netdev-vport.h"
  20
  21 #include <errno.h>
  22 #include <fcntl.h>
  23 #include <sys/socket.h>
  24 #include <linux/openvswitch.h>
  25 #include <linux/rtnetlink.h>
  26 #include <net/if.h>
  27 #include <sys/ioctl.h>
  28
  29 #include "byte-order.h"
  30 #include "daemon.h"
  31 #include "dirs.h"
  32 #include "dpif-linux.h"
  33 #include "hash.h"
  34 #include "hmap.h"
  35 #include "list.h"
  36 #include "netdev-linux.h"
  37 #include "netdev-provider.h"
  38 #include "netlink.h"
  39 #include "netlink-notifier.h"
  40 #include "netlink-socket.h"
  41 #include "ofpbuf.h"
  42 #include "openvswitch/tunnel.h"
  43 #include "packets.h"
  44 #include "route-table.h"
  45 #include "shash.h"
  46 #include "socket-util.h"
  47 #include "unaligned.h"
  48 #include "vlog.h"
  49
  50 VLOG_DEFINE_THIS_MODULE(netdev_vport);
  51
  52 struct netdev_dev_vport {
  53     struct netdev_dev netdev_dev;
  54     struct ofpbuf *options;
  55     int dp_ifindex;             /* -1 if unknown. */
  56     uint32_t port_no;           /* UINT32_MAX if unknown. */
  57     unsigned int change_seq;
  58 };
  59
  60 struct netdev_vport {
  61     struct netdev netdev;
  62 };
  63
  64 struct vport_class {
  65     enum ovs_vport_type type;
  66     struct netdev_class netdev_class;
  67     int (*parse_config)(const char *name, const char *type,
  68                         const struct smap *args, struct ofpbuf *options);
  69     int (*unparse_config)(const char *name, const char *type,
  70                           const struct nlattr *options, size_t options_len,
  71                           struct smap *args);
  72 };
  73
  74 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
  75
  76 static int netdev_vport_create(const struct netdev_class *, const char *,
  77                                struct netdev_dev **);
  78 static void netdev_vport_poll_notify(const struct netdev *);
  79 static int tnl_port_config_from_nlattr(const struct nlattr *options,
  80                                        size_t options_len,
  81                                        struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]);
  82
  83 static const char *netdev_vport_get_tnl_iface(const struct netdev *netdev);
  84
  85 static bool
  86 is_vport_class(const struct netdev_class *class)
  87 {
  88     return class->create == netdev_vport_create;
  89 }
  90
  91 static const struct vport_class *
  92 vport_class_cast(const struct netdev_class *class)
  93 {
  94     assert(is_vport_class(class));
  95     return CONTAINER_OF(class, struct vport_class, netdev_class);
  96 }
  97
  98 static struct netdev_dev_vport *
  99 netdev_dev_vport_cast(const struct netdev_dev *netdev_dev)
 100 {
 101     assert(is_vport_class(netdev_dev_get_class(netdev_dev)));
 102     return CONTAINER_OF(netdev_dev, struct netdev_dev_vport, netdev_dev);
 103 }
 104
 105 static struct netdev_vport *
 106 netdev_vport_cast(const struct netdev *netdev)
 107 {
 108     struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
 109     assert(is_vport_class(netdev_dev_get_class(netdev_dev)));
 110     return CONTAINER_OF(netdev, struct netdev_vport, netdev);
 111 }
 112
 113 /* If 'netdev' is a vport netdev, returns an ofpbuf that contains Netlink
 114  * options to include in OVS_VPORT_ATTR_OPTIONS for configuring that vport.
 115  * Otherwise returns NULL. */
 116 const struct ofpbuf *
 117 netdev_vport_get_options(const struct netdev *netdev)
 118 {
 119     const struct netdev_dev *dev = netdev_get_dev(netdev);
 120
 121     return (is_vport_class(netdev_dev_get_class(dev))
 122             ? netdev_dev_vport_cast(dev)->options
 123             : NULL);
 124 }
 125
 126 enum ovs_vport_type
 127 netdev_vport_get_vport_type(const struct netdev *netdev)
 128 {
 129     const struct netdev_dev *dev = netdev_get_dev(netdev);
 130     const struct netdev_class *class = netdev_dev_get_class(dev);
 131
 132     return (is_vport_class(class) ? vport_class_cast(class)->type
 133             : class == &netdev_internal_class ? OVS_VPORT_TYPE_INTERNAL
 134             : (class == &netdev_linux_class ||
 135                class == &netdev_tap_class) ? OVS_VPORT_TYPE_NETDEV
 136             : OVS_VPORT_TYPE_UNSPEC);
 137 }
 138
 139 const char *
 140 netdev_vport_get_netdev_type(const struct dpif_linux_vport *vport)
 141 {
 142     struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1];
 143
 144     switch (vport->type) {
 145     case OVS_VPORT_TYPE_UNSPEC:
 146         break;
 147
 148     case OVS_VPORT_TYPE_NETDEV:
 149         return "system";
 150
 151     case OVS_VPORT_TYPE_INTERNAL:
 152         return "internal";
 153
 154     case OVS_VPORT_TYPE_PATCH:
 155         return "patch";
 156
 157     case OVS_VPORT_TYPE_GRE:
 158         if (tnl_port_config_from_nlattr(vport->options, vport->options_len,
 159                                         a)) {
 160             break;
 161         }
 162         return (nl_attr_get_u32(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC
 163                 ? "ipsec_gre" : "gre");
 164
 165     case OVS_VPORT_TYPE_CAPWAP:
 166         return "capwap";
 167
 168     case __OVS_VPORT_TYPE_MAX:
 169         break;
 170     }
 171
 172     VLOG_WARN_RL(&rl, "dp%d: port `%s' has unsupported type %u",
 173                  vport->dp_ifindex, vport->name, (unsigned int) vport->type);
 174     return "unknown";
 175 }
 176
 177 static int
 178 netdev_vport_create(const struct netdev_class *netdev_class, const char *name,
 179                     struct netdev_dev **netdev_devp)
 180 {
 181     struct netdev_dev_vport *dev;
 182
 183     dev = xmalloc(sizeof *dev);
 184     netdev_dev_init(&dev->netdev_dev, name, netdev_class);
 185     dev->options = NULL;
 186     dev->dp_ifindex = -1;
 187     dev->port_no = UINT32_MAX;
 188     dev->change_seq = 1;
 189
 190     *netdev_devp = &dev->netdev_dev;
 191     route_table_register();
 192
 193     return 0;
 194 }
 195
 196 static void
 197 netdev_vport_destroy(struct netdev_dev *netdev_dev_)
 198 {
 199     struct netdev_dev_vport *netdev_dev = netdev_dev_vport_cast(netdev_dev_);
 200
 201     ofpbuf_delete(netdev_dev->options);
 202     route_table_unregister();
 203     free(netdev_dev);
 204 }
 205
 206 static int
 207 netdev_vport_open(struct netdev_dev *netdev_dev_, struct netdev **netdevp)
 208 {
 209     struct netdev_vport *netdev;
 210
 211     netdev = xmalloc(sizeof *netdev);
 212     netdev_init(&netdev->netdev, netdev_dev_);
 213
 214     *netdevp = &netdev->netdev;
 215     return 0;
 216 }
 217
 218 static void
 219 netdev_vport_close(struct netdev *netdev_)
 220 {
 221     struct netdev_vport *netdev = netdev_vport_cast(netdev_);
 222     free(netdev);
 223 }
 224
 225 static int
 226 netdev_vport_get_config(struct netdev_dev *dev_, struct smap *args)
 227 {
 228     const struct netdev_class *netdev_class = netdev_dev_get_class(dev_);
 229     const struct vport_class *vport_class = vport_class_cast(netdev_class);
 230     struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
 231     const char *name = netdev_dev_get_name(dev_);
 232     int error;
 233
 234     if (!dev->options) {
 235         struct dpif_linux_vport reply;
 236         struct ofpbuf *buf;
 237
 238         error = dpif_linux_vport_get(name, &reply, &buf);
 239         if (error) {
 240             VLOG_ERR_RL(&rl, "%s: vport query failed (%s)",
 241                         name, strerror(error));
 242             return error;
 243         }
 244
 245         dev->options = ofpbuf_clone_data(reply.options, reply.options_len);
 246         dev->dp_ifindex = reply.dp_ifindex;
 247         dev->port_no = reply.port_no;
 248         ofpbuf_delete(buf);
 249     }
 250
 251     error = vport_class->unparse_config(name, netdev_class->type,
 252                                         dev->options->data,
 253                                         dev->options->size,
 254                                         args);
 255     if (error) {
 256         VLOG_ERR_RL(&rl, "%s: failed to parse kernel config (%s)",
 257                     name, strerror(error));
 258     }
 259     return error;
 260 }
 261
 262 static int
 263 netdev_vport_set_config(struct netdev_dev *dev_, const struct smap *args)
 264 {
 265     const struct netdev_class *netdev_class = netdev_dev_get_class(dev_);
 266     const struct vport_class *vport_class = vport_class_cast(netdev_class);
 267     struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
 268     const char *name = netdev_dev_get_name(dev_);
 269     struct ofpbuf *options;
 270     int error;
 271
 272     options = ofpbuf_new(64);
 273     error = vport_class->parse_config(name, netdev_dev_get_type(dev_),
 274                                       args, options);
 275     if (!error
 276         && (!dev->options
 277             || options->size != dev->options->size
 278             || memcmp(options->data, dev->options->data, options->size))) {
 279         struct dpif_linux_vport vport;
 280
 281         dpif_linux_vport_init(&vport);
 282         vport.cmd = OVS_VPORT_CMD_SET;
 283         vport.name = name;
 284         vport.options = options->data;
 285         vport.options_len = options->size;
 286         error = dpif_linux_vport_transact(&vport, NULL, NULL);
 287         if (!error || error == ENODEV) {
 288             /* Either reconfiguration succeeded or this vport is not installed
 289              * in the kernel (e.g. it hasn't been added to a dpif yet with
 290              * dpif_port_add()). */
 291             ofpbuf_delete(dev->options);
 292             dev->options = options;
 293             options = NULL;
 294             error = 0;
 295         }
 296     }
 297     ofpbuf_delete(options);
 298
 299     return error;
 300 }
 301
 302 static int
 303 netdev_vport_send(struct netdev *netdev, const void *data, size_t size)
 304 {
 305     struct netdev_dev *dev_ = netdev_get_dev(netdev);
 306     struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
 307
 308     if (dev->dp_ifindex == -1) {
 309         const char *name = netdev_get_name(netdev);
 310         struct dpif_linux_vport reply;
 311         struct ofpbuf *buf;
 312         int error;
 313
 314         error = dpif_linux_vport_get(name, &reply, &buf);
 315         if (error) {
 316             VLOG_ERR_RL(&rl, "%s: failed to query vport for send (%s)",
 317                         name, strerror(error));
 318             return error;
 319         }
 320         dev->dp_ifindex = reply.dp_ifindex;
 321         dev->port_no = reply.port_no;
 322         ofpbuf_delete(buf);
 323     }
 324
 325     return dpif_linux_vport_send(dev->dp_ifindex, dev->port_no, data, size);
 326 }
 327
 328 static int
 329 netdev_vport_set_etheraddr(struct netdev *netdev,
 330                            const uint8_t mac[ETH_ADDR_LEN])
 331 {
 332     struct dpif_linux_vport vport;
 333     int error;
 334
 335     dpif_linux_vport_init(&vport);
 336     vport.cmd = OVS_VPORT_CMD_SET;
 337     vport.name = netdev_get_name(netdev);
 338     vport.address = mac;
 339
 340     error = dpif_linux_vport_transact(&vport, NULL, NULL);
 341     if (!error) {
 342         netdev_vport_poll_notify(netdev);
 343     }
 344     return error;
 345 }
 346
 347 static int
 348 netdev_vport_get_etheraddr(const struct netdev *netdev,
 349                            uint8_t mac[ETH_ADDR_LEN])
 350 {
 351     struct dpif_linux_vport reply;
 352     struct ofpbuf *buf;
 353     int error;
 354
 355     error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf);
 356     if (!error) {
 357         if (reply.address) {
 358             memcpy(mac, reply.address, ETH_ADDR_LEN);
 359         } else {
 360             error = EOPNOTSUPP;
 361         }
 362         ofpbuf_delete(buf);
 363     }
 364     return error;
 365 }
 366
 367 /* Copies 'src' into 'dst', performing format conversion in the process.
 368  *
 369  * 'src' is allowed to be misaligned. */
 370 static void
 371 netdev_stats_from_ovs_vport_stats(struct netdev_stats *dst,
 372                                   const struct ovs_vport_stats *src)
 373 {
 374     dst->rx_packets = get_unaligned_u64(&src->rx_packets);
 375     dst->tx_packets = get_unaligned_u64(&src->tx_packets);
 376     dst->rx_bytes = get_unaligned_u64(&src->rx_bytes);
 377     dst->tx_bytes = get_unaligned_u64(&src->tx_bytes);
 378     dst->rx_errors = get_unaligned_u64(&src->rx_errors);
 379     dst->tx_errors = get_unaligned_u64(&src->tx_errors);
 380     dst->rx_dropped = get_unaligned_u64(&src->rx_dropped);
 381     dst->tx_dropped = get_unaligned_u64(&src->tx_dropped);
 382     dst->multicast = 0;
 383     dst->collisions = 0;
 384     dst->rx_length_errors = 0;
 385     dst->rx_over_errors = 0;
 386     dst->rx_crc_errors = 0;
 387     dst->rx_frame_errors = 0;
 388     dst->rx_fifo_errors = 0;
 389     dst->rx_missed_errors = 0;
 390     dst->tx_aborted_errors = 0;
 391     dst->tx_carrier_errors = 0;
 392     dst->tx_fifo_errors = 0;
 393     dst->tx_heartbeat_errors = 0;
 394     dst->tx_window_errors = 0;
 395 }
 396
 397 /* Copies 'src' into 'dst', performing format conversion in the process. */
 398 static void
 399 netdev_stats_to_ovs_vport_stats(struct ovs_vport_stats *dst,
 400                                 const struct netdev_stats *src)
 401 {
 402     dst->rx_packets = src->rx_packets;
 403     dst->tx_packets = src->tx_packets;
 404     dst->rx_bytes = src->rx_bytes;
 405     dst->tx_bytes = src->tx_bytes;
 406     dst->rx_errors = src->rx_errors;
 407     dst->tx_errors = src->tx_errors;
 408     dst->rx_dropped = src->rx_dropped;
 409     dst->tx_dropped = src->tx_dropped;
 410 }
 411
 412 int
 413 netdev_vport_get_stats(const struct netdev *netdev, struct netdev_stats *stats)
 414 {
 415     struct dpif_linux_vport reply;
 416     struct ofpbuf *buf;
 417     int error;
 418
 419     error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf);
 420     if (error) {
 421         return error;
 422     } else if (!reply.stats) {
 423         ofpbuf_delete(buf);
 424         return EOPNOTSUPP;
 425     }
 426
 427     netdev_stats_from_ovs_vport_stats(stats, reply.stats);
 428
 429     ofpbuf_delete(buf);
 430
 431     return 0;
 432 }
 433
 434 int
 435 netdev_vport_set_stats(struct netdev *netdev, const struct netdev_stats *stats)
 436 {
 437     struct ovs_vport_stats rtnl_stats;
 438     struct dpif_linux_vport vport;
 439     int err;
 440
 441     netdev_stats_to_ovs_vport_stats(&rtnl_stats, stats);
 442
 443     dpif_linux_vport_init(&vport);
 444     vport.cmd = OVS_VPORT_CMD_SET;
 445     vport.name = netdev_get_name(netdev);
 446     vport.stats = &rtnl_stats;
 447
 448     err = dpif_linux_vport_transact(&vport, NULL, NULL);
 449
 450     /* If the vport layer doesn't know about the device, that doesn't mean it
 451      * doesn't exist (after all were able to open it when netdev_open() was
 452      * called), it just means that it isn't attached and we'll be getting
 453      * stats a different way. */
 454     if (err == ENODEV) {
 455         err = EOPNOTSUPP;
 456     }
 457
 458     return err;
 459 }
 460
 461 static int
 462 netdev_vport_get_drv_info(const struct netdev *netdev, struct smap *smap)
 463 {
 464     const char *iface = netdev_vport_get_tnl_iface(netdev);
 465
 466     if (iface) {
 467         struct netdev *egress_netdev;
 468
 469         smap_add(smap, "tunnel_egress_iface", iface);
 470
 471         if (!netdev_open(iface, "system", &egress_netdev)) {
 472             smap_add(smap, "tunnel_egress_iface_carrier",
 473                      netdev_get_carrier(egress_netdev) ? "up" : "down");
 474             netdev_close(egress_netdev);
 475         }
 476     }
 477
 478     return 0;
 479 }
 480
 481 static int
 482 netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED,
 483                         enum netdev_flags off, enum netdev_flags on OVS_UNUSED,
 484                         enum netdev_flags *old_flagsp)
 485 {
 486     if (off & (NETDEV_UP | NETDEV_PROMISC)) {
 487         return EOPNOTSUPP;
 488     }
 489
 490     *old_flagsp = NETDEV_UP | NETDEV_PROMISC;
 491     return 0;
 492 }
 493
 494 static unsigned int
 495 netdev_vport_change_seq(const struct netdev *netdev)
 496 {
 497     return netdev_dev_vport_cast(netdev_get_dev(netdev))->change_seq;
 498 }
 499
 500 static void
 501 netdev_vport_run(void)
 502 {
 503     route_table_run();
 504 }
 505
 506 static void
 507 netdev_vport_wait(void)
 508 {
 509     route_table_wait();
 510 }
 511 \f
 512 /* get_tnl_iface() implementation. */
 513 static const char *
 514 netdev_vport_get_tnl_iface(const struct netdev *netdev)
 515 {
 516     struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1];
 517     ovs_be32 route;
 518     struct netdev_dev_vport *ndv;
 519     static char name[IFNAMSIZ];
 520
 521     ndv = netdev_dev_vport_cast(netdev_get_dev(netdev));
 522     if (tnl_port_config_from_nlattr(ndv->options->data, ndv->options->size,
 523                                     a)) {
 524         return NULL;
 525     }
 526     route = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]);
 527
 528     if (route_table_get_name(route, name)) {
 529         return name;
 530     }
 531
 532     return NULL;
 533 }
 534 \f
 535 /* Helper functions. */
 536
 537 static void
 538 netdev_vport_poll_notify(const struct netdev *netdev)
 539 {
 540     struct netdev_dev_vport *ndv;
 541
 542     ndv = netdev_dev_vport_cast(netdev_get_dev(netdev));
 543
 544     ndv->change_seq++;
 545     if (!ndv->change_seq) {
 546         ndv->change_seq++;
 547     }
 548 }
 549 \f
 550 /* Code specific to individual vport types. */
 551
 552 static void
 553 set_key(const struct smap *args, const char *name, uint16_t type,
 554         struct ofpbuf *options)
 555 {
 556     const char *s;
 557
 558     s = smap_get(args, name);
 559     if (!s) {
 560         s = smap_get(args, "key");
 561         if (!s) {
 562             s = "0";
 563         }
 564     }
 565
 566     if (!strcmp(s, "flow")) {
 567         /* This is the default if no attribute is present. */
 568     } else {
 569         nl_msg_put_be64(options, type, htonll(strtoull(s, NULL, 0)));
 570     }
 571 }
 572
 573 static int
 574 parse_tunnel_config(const char *name, const char *type,
 575                     const struct smap *args, struct ofpbuf *options)
 576 {
 577     bool is_gre = false;
 578     bool is_ipsec = false;
 579     struct smap_node *node;
 580     bool ipsec_mech_set = false;
 581     ovs_be32 daddr = htonl(0);
 582     ovs_be32 saddr = htonl(0);
 583     uint32_t flags;
 584
 585     flags = TNL_F_DF_DEFAULT | TNL_F_PMTUD | TNL_F_HDR_CACHE;
 586     if (!strcmp(type, "gre")) {
 587         is_gre = true;
 588     } else if (!strcmp(type, "ipsec_gre")) {
 589         is_gre = true;
 590         is_ipsec = true;
 591         flags |= TNL_F_IPSEC;
 592         flags &= ~TNL_F_HDR_CACHE;
 593     }
 594
 595     SMAP_FOR_EACH (node, args) {
 596         if (!strcmp(node->key, "remote_ip")) {
 597             struct in_addr in_addr;
 598             if (lookup_ip(node->value, &in_addr)) {
 599                 VLOG_WARN("%s: bad %s 'remote_ip'", name, type);
 600             } else {
 601                 daddr = in_addr.s_addr;
 602             }
 603         } else if (!strcmp(node->key, "local_ip")) {
 604             struct in_addr in_addr;
 605             if (lookup_ip(node->value, &in_addr)) {
 606                 VLOG_WARN("%s: bad %s 'local_ip'", name, type);
 607             } else {
 608                 saddr = in_addr.s_addr;
 609             }
 610         } else if (!strcmp(node->key, "tos")) {
 611             if (!strcmp(node->value, "inherit")) {
 612                 flags |= TNL_F_TOS_INHERIT;
 613             } else {
 614                 char *endptr;
 615                 int tos;
 616                 tos = strtol(node->value, &endptr, 0);
 617                 if (*endptr == '\0') {
 618                     nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TOS, tos);
 619                 }
 620             }
 621         } else if (!strcmp(node->key, "ttl")) {
 622             if (!strcmp(node->value, "inherit")) {
 623                 flags |= TNL_F_TTL_INHERIT;
 624             } else {
 625                 nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TTL, atoi(node->value));
 626             }
 627         } else if (!strcmp(node->key, "csum") && is_gre) {
 628             if (!strcmp(node->value, "true")) {
 629                 flags |= TNL_F_CSUM;
 630             }
 631         } else if (!strcmp(node->key, "df_inherit")) {
 632             if (!strcmp(node->value, "true")) {
 633                 flags |= TNL_F_DF_INHERIT;
 634             }
 635         } else if (!strcmp(node->key, "df_default")) {
 636             if (!strcmp(node->value, "false")) {
 637                 flags &= ~TNL_F_DF_DEFAULT;
 638             }
 639         } else if (!strcmp(node->key, "pmtud")) {
 640             if (!strcmp(node->value, "false")) {
 641                 flags &= ~TNL_F_PMTUD;
 642             }
 643         } else if (!strcmp(node->key, "header_cache")) {
 644             if (!strcmp(node->value, "false")) {
 645                 flags &= ~TNL_F_HDR_CACHE;
 646             }
 647         } else if (!strcmp(node->key, "peer_cert") && is_ipsec) {
 648             if (smap_get(args, "certificate")) {
 649                 ipsec_mech_set = true;
 650             } else {
 651                 const char *use_ssl_cert;
 652
 653                 /* If the "use_ssl_cert" is true, then "certificate" and
 654                  * "private_key" will be pulled from the SSL table.  The
 655                  * use of this option is strongly discouraged, since it
 656                  * will like be removed when multiple SSL configurations
 657                  * are supported by OVS.
 658                  */
 659                 use_ssl_cert = smap_get(args, "use_ssl_cert");
 660                 if (!use_ssl_cert || strcmp(use_ssl_cert, "true")) {
 661                     VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument",
 662                              name);
 663                     return EINVAL;
 664                 }
 665                 ipsec_mech_set = true;
 666             }
 667         } else if (!strcmp(node->key, "psk") && is_ipsec) {
 668             ipsec_mech_set = true;
 669         } else if (is_ipsec
 670                 && (!strcmp(node->key, "certificate")
 671                     || !strcmp(node->key, "private_key")
 672                     || !strcmp(node->key, "use_ssl_cert"))) {
 673             /* Ignore options not used by the netdev. */
 674         } else if (!strcmp(node->key, "key") ||
 675                    !strcmp(node->key, "in_key") ||
 676                    !strcmp(node->key, "out_key")) {
 677             /* Handled separately below. */
 678         } else {
 679             VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->key);
 680         }
 681     }
 682
 683     if (is_ipsec) {
 684         static pid_t pid = 0;
 685         if (pid == 0) {
 686             char *file_name = xasprintf("%s/%s", ovs_rundir(),
 687                                         "ovs-monitor-ipsec.pid");
 688             pid = read_pidfile(file_name);
 689             free(file_name);
 690         }
 691
 692         if (pid < 0) {
 693             VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon",
 694                      name);
 695             return EINVAL;
 696         }
 697
 698         if (smap_get(args, "peer_cert") && smap_get(args, "psk")) {
 699             VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name);
 700             return EINVAL;
 701         }
 702
 703         if (!ipsec_mech_set) {
 704             VLOG_ERR("%s: IPsec requires an 'peer_cert' or psk' argument",
 705                      name);
 706             return EINVAL;
 707         }
 708     }
 709
 710     set_key(args, "in_key", OVS_TUNNEL_ATTR_IN_KEY, options);
 711     set_key(args, "out_key", OVS_TUNNEL_ATTR_OUT_KEY, options);
 712
 713     if (!daddr) {
 714         VLOG_ERR("%s: %s type requires valid 'remote_ip' argument",
 715                  name, type);
 716         return EINVAL;
 717     }
 718     nl_msg_put_be32(options, OVS_TUNNEL_ATTR_DST_IPV4, daddr);
 719
 720     if (saddr) {
 721         if (ip_is_multicast(daddr)) {
 722             VLOG_WARN("%s: remote_ip is multicast, ignoring local_ip", name);
 723         } else {
 724             nl_msg_put_be32(options, OVS_TUNNEL_ATTR_SRC_IPV4, saddr);
 725         }
 726     }
 727
 728     nl_msg_put_u32(options, OVS_TUNNEL_ATTR_FLAGS, flags);
 729
 730     return 0;
 731 }
 732
 733 static int
 734 tnl_port_config_from_nlattr(const struct nlattr *options, size_t options_len,
 735                             struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1])
 736 {
 737     static const struct nl_policy ovs_tunnel_policy[] = {
 738         [OVS_TUNNEL_ATTR_FLAGS] = { .type = NL_A_U32 },
 739         [OVS_TUNNEL_ATTR_DST_IPV4] = { .type = NL_A_BE32 },
 740         [OVS_TUNNEL_ATTR_SRC_IPV4] = { .type = NL_A_BE32, .optional = true },
 741         [OVS_TUNNEL_ATTR_IN_KEY] = { .type = NL_A_BE64, .optional = true },
 742         [OVS_TUNNEL_ATTR_OUT_KEY] = { .type = NL_A_BE64, .optional = true },
 743         [OVS_TUNNEL_ATTR_TOS] = { .type = NL_A_U8, .optional = true },
 744         [OVS_TUNNEL_ATTR_TTL] = { .type = NL_A_U8, .optional = true },
 745     };
 746     struct ofpbuf buf;
 747
 748     ofpbuf_use_const(&buf, options, options_len);
 749     if (!nl_policy_parse(&buf, 0, ovs_tunnel_policy,
 750                          a, ARRAY_SIZE(ovs_tunnel_policy))) {
 751         return EINVAL;
 752     }
 753     return 0;
 754 }
 755
 756 static uint64_t
 757 get_be64_or_zero(const struct nlattr *a)
 758 {
 759     return a ? ntohll(nl_attr_get_be64(a)) : 0;
 760 }
 761
 762 static int
 763 unparse_tunnel_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED,
 764                       const struct nlattr *options, size_t options_len,
 765                       struct smap *args)
 766 {
 767     struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1];
 768     ovs_be32 daddr;
 769     uint32_t flags;
 770     int error;
 771
 772     error = tnl_port_config_from_nlattr(options, options_len, a);
 773     if (error) {
 774         return error;
 775     }
 776
 777     flags = nl_attr_get_u32(a[OVS_TUNNEL_ATTR_FLAGS]);
 778     if (!(flags & TNL_F_HDR_CACHE) == !(flags & TNL_F_IPSEC)) {
 779         smap_add(args, "header_cache",
 780                  flags & TNL_F_HDR_CACHE ? "true" : "false");
 781     }
 782
 783     daddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]);
 784     smap_add_format(args, "remote_ip", IP_FMT, IP_ARGS(&daddr));
 785
 786     if (a[OVS_TUNNEL_ATTR_SRC_IPV4]) {
 787         ovs_be32 saddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_SRC_IPV4]);
 788         smap_add_format(args, "local_ip", IP_FMT, IP_ARGS(&saddr));
 789     }
 790
 791     if (!a[OVS_TUNNEL_ATTR_IN_KEY] && !a[OVS_TUNNEL_ATTR_OUT_KEY]) {
 792         smap_add(args, "key", "flow");
 793     } else {
 794         uint64_t in_key = get_be64_or_zero(a[OVS_TUNNEL_ATTR_IN_KEY]);
 795         uint64_t out_key = get_be64_or_zero(a[OVS_TUNNEL_ATTR_OUT_KEY]);
 796
 797         if (in_key && in_key == out_key) {
 798             smap_add_format(args, "key", "%"PRIu64, in_key);
 799         } else {
 800             if (!a[OVS_TUNNEL_ATTR_IN_KEY]) {
 801                 smap_add(args, "in_key", "flow");
 802             } else if (in_key) {
 803                 smap_add_format(args, "in_key", "%"PRIu64, in_key);
 804             }
 805
 806             if (!a[OVS_TUNNEL_ATTR_OUT_KEY]) {
 807                 smap_add(args, "out_key", "flow");
 808             } else if (out_key) {
 809                 smap_add_format(args, "out_key", "%"PRIu64, out_key);
 810             }
 811         }
 812     }
 813
 814     if (flags & TNL_F_TTL_INHERIT) {
 815         smap_add(args, "tos", "inherit");
 816     } else if (a[OVS_TUNNEL_ATTR_TTL]) {
 817         int ttl = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TTL]);
 818         smap_add_format(args, "tos", "%d", ttl);
 819     }
 820
 821     if (flags & TNL_F_TOS_INHERIT) {
 822         smap_add(args, "tos", "inherit");
 823     } else if (a[OVS_TUNNEL_ATTR_TOS]) {
 824         int tos = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TOS]);
 825         smap_add_format(args, "tos", "0x%x", tos);
 826     }
 827
 828     if (flags & TNL_F_CSUM) {
 829         smap_add(args, "csum", "true");
 830     }
 831     if (flags & TNL_F_DF_INHERIT) {
 832         smap_add(args, "df_inherit", "true");
 833     }
 834     if (!(flags & TNL_F_DF_DEFAULT)) {
 835         smap_add(args, "df_default", "false");
 836     }
 837     if (!(flags & TNL_F_PMTUD)) {
 838         smap_add(args, "pmtud", "false");
 839     }
 840
 841     return 0;
 842 }
 843
 844 static int
 845 parse_patch_config(const char *name, const char *type OVS_UNUSED,
 846                    const struct smap *args, struct ofpbuf *options)
 847 {
 848     const char *peer;
 849
 850     peer = smap_get(args, "peer");
 851     if (!peer) {
 852         VLOG_ERR("%s: patch type requires valid 'peer' argument", name);
 853         return EINVAL;
 854     }
 855
 856     if (smap_count(args) > 1) {
 857         VLOG_ERR("%s: patch type takes only a 'peer' argument", name);
 858         return EINVAL;
 859     }
 860
 861     if (strlen(peer) >= IFNAMSIZ) {
 862         VLOG_ERR("%s: patch 'peer' arg too long", name);
 863         return EINVAL;
 864     }
 865
 866     if (!strcmp(name, peer)) {
 867         VLOG_ERR("%s: patch peer must not be self", name);
 868         return EINVAL;
 869     }
 870
 871     nl_msg_put_string(options, OVS_PATCH_ATTR_PEER, peer);
 872
 873     return 0;
 874 }
 875
 876 static int
 877 unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED,
 878                      const struct nlattr *options, size_t options_len,
 879                      struct smap *args)
 880 {
 881     static const struct nl_policy ovs_patch_policy[] = {
 882         [OVS_PATCH_ATTR_PEER] = { .type = NL_A_STRING,
 883                                .max_len = IFNAMSIZ,
 884                                .optional = false }
 885     };
 886
 887     struct nlattr *a[ARRAY_SIZE(ovs_patch_policy)];
 888     struct ofpbuf buf;
 889
 890     ofpbuf_use_const(&buf, options, options_len);
 891     if (!nl_policy_parse(&buf, 0, ovs_patch_policy,
 892                          a, ARRAY_SIZE(ovs_patch_policy))) {
 893         return EINVAL;
 894     }
 895
 896     smap_add(args, "peer", nl_attr_get_string(a[OVS_PATCH_ATTR_PEER]));
 897     return 0;
 898 }
 899 \f
 900 #define VPORT_FUNCTIONS(GET_STATUS)                         \
 901     NULL,                                                   \
 902     netdev_vport_run,                                       \
 903     netdev_vport_wait,                                      \
 904                                                             \
 905     netdev_vport_create,                                    \
 906     netdev_vport_destroy,                                   \
 907     netdev_vport_get_config,                                \
 908     netdev_vport_set_config,                                \
 909                                                             \
 910     netdev_vport_open,                                      \
 911     netdev_vport_close,                                     \
 912                                                             \
 913     NULL,                       /* listen */                \
 914     NULL,                       /* recv */                  \
 915     NULL,                       /* recv_wait */             \
 916     NULL,                       /* drain */                 \
 917                                                             \
 918     netdev_vport_send,          /* send */                  \
 919     NULL,                       /* send_wait */             \
 920                                                             \
 921     netdev_vport_set_etheraddr,                             \
 922     netdev_vport_get_etheraddr,                             \
 923     NULL,                       /* get_mtu */               \
 924     NULL,                       /* set_mtu */               \
 925     NULL,                       /* get_ifindex */           \
 926     NULL,                       /* get_carrier */           \
 927     NULL,                       /* get_carrier_resets */    \
 928     NULL,                       /* get_miimon */            \
 929     netdev_vport_get_stats,                                 \
 930     netdev_vport_set_stats,                                 \
 931                                                             \
 932     NULL,                       /* get_features */          \
 933     NULL,                       /* set_advertisements */    \
 934                                                             \
 935     NULL,                       /* set_policing */          \
 936     NULL,                       /* get_qos_types */         \
 937     NULL,                       /* get_qos_capabilities */  \
 938     NULL,                       /* get_qos */               \
 939     NULL,                       /* set_qos */               \
 940     NULL,                       /* get_queue */             \
 941     NULL,                       /* set_queue */             \
 942     NULL,                       /* delete_queue */          \
 943     NULL,                       /* get_queue_stats */       \
 944     NULL,                       /* dump_queues */           \
 945     NULL,                       /* dump_queue_stats */      \
 946                                                             \
 947     NULL,                       /* get_in4 */               \
 948     NULL,                       /* set_in4 */               \
 949     NULL,                       /* get_in6 */               \
 950     NULL,                       /* add_router */            \
 951     NULL,                       /* get_next_hop */          \
 952     GET_STATUS,                                             \
 953     NULL,                       /* arp_lookup */            \
 954                                                             \
 955     netdev_vport_update_flags,                              \
 956                                                             \
 957     netdev_vport_change_seq
 958
 959 void
 960 netdev_vport_register(void)
 961 {
 962     static const struct vport_class vport_classes[] = {
 963         { OVS_VPORT_TYPE_GRE,
 964           { "gre", VPORT_FUNCTIONS(netdev_vport_get_drv_info) },
 965           parse_tunnel_config, unparse_tunnel_config },
 966
 967         { OVS_VPORT_TYPE_GRE,
 968           { "ipsec_gre", VPORT_FUNCTIONS(netdev_vport_get_drv_info) },
 969           parse_tunnel_config, unparse_tunnel_config },
 970
 971         { OVS_VPORT_TYPE_CAPWAP,
 972           { "capwap", VPORT_FUNCTIONS(netdev_vport_get_drv_info) },
 973           parse_tunnel_config, unparse_tunnel_config },
 974
 975         { OVS_VPORT_TYPE_PATCH,
 976           { "patch", VPORT_FUNCTIONS(NULL) },
 977           parse_patch_config, unparse_patch_config }
 978     };
 979
 980     int i;
 981
 982     for (i = 0; i < ARRAY_SIZE(vport_classes); i++) {
 983         netdev_register_provider(&vport_classes[i].netdev_class);
 984     }
 985 }