lib/netdev-vport.c

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