1 // SPDX-License-Identifier: GPL-2.0 2 /* Bareudp: UDP tunnel encasulation for different Payload types like 3 * MPLS, NSH, IP, etc. 4 * Copyright (c) 2019 Nokia, Inc. 5 * Authors: Martin Varghese, <martin.varghese@nokia.com> 6 */ 7 8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 9 10 #include <linux/kernel.h> 11 #include <linux/module.h> 12 #include <linux/etherdevice.h> 13 #include <linux/hash.h> 14 #include <net/dst_metadata.h> 15 #include <net/gro_cells.h> 16 #include <net/rtnetlink.h> 17 #include <net/protocol.h> 18 #include <net/ip6_tunnel.h> 19 #include <net/ip_tunnels.h> 20 #include <net/udp_tunnel.h> 21 #include <net/bareudp.h> 22 23 #define BAREUDP_BASE_HLEN sizeof(struct udphdr) 24 #define BAREUDP_IPV4_HLEN (sizeof(struct iphdr) + \ 25 sizeof(struct udphdr)) 26 #define BAREUDP_IPV6_HLEN (sizeof(struct ipv6hdr) + \ 27 sizeof(struct udphdr)) 28 29 static bool log_ecn_error = true; 30 module_param(log_ecn_error, bool, 0644); 31 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 32 33 /* per-network namespace private data for this module */ 34 35 static unsigned int bareudp_net_id; 36 37 struct bareudp_net { 38 struct list_head bareudp_list; 39 }; 40 41 /* Pseudo network device */ 42 struct bareudp_dev { 43 struct net *net; /* netns for packet i/o */ 44 struct net_device *dev; /* netdev for bareudp tunnel */ 45 __be16 ethertype; 46 __be16 port; 47 u16 sport_min; 48 bool multi_proto_mode; 49 struct socket __rcu *sock; 50 struct list_head next; /* bareudp node on namespace list */ 51 struct gro_cells gro_cells; 52 }; 53 54 static int bareudp_udp_encap_recv(struct sock *sk, struct sk_buff *skb) 55 { 56 struct metadata_dst *tun_dst = NULL; 57 struct pcpu_sw_netstats *stats; 58 struct bareudp_dev *bareudp; 59 unsigned short family; 60 unsigned int len; 61 __be16 proto; 62 void *oiph; 63 int err; 64 65 bareudp = rcu_dereference_sk_user_data(sk); 66 if (!bareudp) 67 goto drop; 68 69 if (skb->protocol == htons(ETH_P_IP)) 70 family = AF_INET; 71 else 72 family = AF_INET6; 73 74 if (bareudp->ethertype == htons(ETH_P_IP)) { 75 struct iphdr *iphdr; 76 77 iphdr = (struct iphdr *)(skb->data + BAREUDP_BASE_HLEN); 78 if (iphdr->version == 4) { 79 proto = bareudp->ethertype; 80 } else if (bareudp->multi_proto_mode && (iphdr->version == 6)) { 81 proto = htons(ETH_P_IPV6); 82 } else { 83 bareudp->dev->stats.rx_dropped++; 84 goto drop; 85 } 86 } else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) { 87 struct iphdr *tunnel_hdr; 88 89 tunnel_hdr = (struct iphdr *)skb_network_header(skb); 90 if (tunnel_hdr->version == 4) { 91 if (!ipv4_is_multicast(tunnel_hdr->daddr)) { 92 proto = bareudp->ethertype; 93 } else if (bareudp->multi_proto_mode && 94 ipv4_is_multicast(tunnel_hdr->daddr)) { 95 proto = htons(ETH_P_MPLS_MC); 96 } else { 97 bareudp->dev->stats.rx_dropped++; 98 goto drop; 99 } 100 } else { 101 int addr_type; 102 struct ipv6hdr *tunnel_hdr_v6; 103 104 tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb); 105 addr_type = 106 ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr); 107 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 108 proto = bareudp->ethertype; 109 } else if (bareudp->multi_proto_mode && 110 (addr_type & IPV6_ADDR_MULTICAST)) { 111 proto = htons(ETH_P_MPLS_MC); 112 } else { 113 bareudp->dev->stats.rx_dropped++; 114 goto drop; 115 } 116 } 117 } else { 118 proto = bareudp->ethertype; 119 } 120 121 if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN, 122 proto, 123 !net_eq(bareudp->net, 124 dev_net(bareudp->dev)))) { 125 bareudp->dev->stats.rx_dropped++; 126 goto drop; 127 } 128 tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0); 129 if (!tun_dst) { 130 bareudp->dev->stats.rx_dropped++; 131 goto drop; 132 } 133 skb_dst_set(skb, &tun_dst->dst); 134 skb->dev = bareudp->dev; 135 oiph = skb_network_header(skb); 136 skb_reset_network_header(skb); 137 138 if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET) 139 err = IP_ECN_decapsulate(oiph, skb); 140 else 141 err = IP6_ECN_decapsulate(oiph, skb); 142 143 if (unlikely(err)) { 144 if (log_ecn_error) { 145 if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET) 146 net_info_ratelimited("non-ECT from %pI4 " 147 "with TOS=%#x\n", 148 &((struct iphdr *)oiph)->saddr, 149 ((struct iphdr *)oiph)->tos); 150 else 151 net_info_ratelimited("non-ECT from %pI6\n", 152 &((struct ipv6hdr *)oiph)->saddr); 153 } 154 if (err > 1) { 155 ++bareudp->dev->stats.rx_frame_errors; 156 ++bareudp->dev->stats.rx_errors; 157 goto drop; 158 } 159 } 160 161 len = skb->len; 162 err = gro_cells_receive(&bareudp->gro_cells, skb); 163 if (likely(err == NET_RX_SUCCESS)) { 164 stats = this_cpu_ptr(bareudp->dev->tstats); 165 u64_stats_update_begin(&stats->syncp); 166 stats->rx_packets++; 167 stats->rx_bytes += len; 168 u64_stats_update_end(&stats->syncp); 169 } 170 return 0; 171 drop: 172 /* Consume bad packet */ 173 kfree_skb(skb); 174 175 return 0; 176 } 177 178 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb) 179 { 180 return 0; 181 } 182 183 static int bareudp_init(struct net_device *dev) 184 { 185 struct bareudp_dev *bareudp = netdev_priv(dev); 186 int err; 187 188 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 189 if (!dev->tstats) 190 return -ENOMEM; 191 192 err = gro_cells_init(&bareudp->gro_cells, dev); 193 if (err) { 194 free_percpu(dev->tstats); 195 return err; 196 } 197 return 0; 198 } 199 200 static void bareudp_uninit(struct net_device *dev) 201 { 202 struct bareudp_dev *bareudp = netdev_priv(dev); 203 204 gro_cells_destroy(&bareudp->gro_cells); 205 free_percpu(dev->tstats); 206 } 207 208 static struct socket *bareudp_create_sock(struct net *net, __be16 port) 209 { 210 struct udp_port_cfg udp_conf; 211 struct socket *sock; 212 int err; 213 214 memset(&udp_conf, 0, sizeof(udp_conf)); 215 #if IS_ENABLED(CONFIG_IPV6) 216 udp_conf.family = AF_INET6; 217 #else 218 udp_conf.family = AF_INET; 219 #endif 220 udp_conf.local_udp_port = port; 221 /* Open UDP socket */ 222 err = udp_sock_create(net, &udp_conf, &sock); 223 if (err < 0) 224 return ERR_PTR(err); 225 226 return sock; 227 } 228 229 /* Create new listen socket if needed */ 230 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port) 231 { 232 struct udp_tunnel_sock_cfg tunnel_cfg; 233 struct socket *sock; 234 235 sock = bareudp_create_sock(bareudp->net, port); 236 if (IS_ERR(sock)) 237 return PTR_ERR(sock); 238 239 /* Mark socket as an encapsulation socket */ 240 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 241 tunnel_cfg.sk_user_data = bareudp; 242 tunnel_cfg.encap_type = 1; 243 tunnel_cfg.encap_rcv = bareudp_udp_encap_recv; 244 tunnel_cfg.encap_err_lookup = bareudp_err_lookup; 245 tunnel_cfg.encap_destroy = NULL; 246 setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg); 247 248 /* As the setup_udp_tunnel_sock does not call udp_encap_enable if the 249 * socket type is v6 an explicit call to udp_encap_enable is needed. 250 */ 251 if (sock->sk->sk_family == AF_INET6) 252 udp_encap_enable(); 253 254 rcu_assign_pointer(bareudp->sock, sock); 255 return 0; 256 } 257 258 static int bareudp_open(struct net_device *dev) 259 { 260 struct bareudp_dev *bareudp = netdev_priv(dev); 261 int ret = 0; 262 263 ret = bareudp_socket_create(bareudp, bareudp->port); 264 return ret; 265 } 266 267 static void bareudp_sock_release(struct bareudp_dev *bareudp) 268 { 269 struct socket *sock; 270 271 sock = bareudp->sock; 272 rcu_assign_pointer(bareudp->sock, NULL); 273 synchronize_net(); 274 udp_tunnel_sock_release(sock); 275 } 276 277 static int bareudp_stop(struct net_device *dev) 278 { 279 struct bareudp_dev *bareudp = netdev_priv(dev); 280 281 bareudp_sock_release(bareudp); 282 return 0; 283 } 284 285 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev, 286 struct bareudp_dev *bareudp, 287 const struct ip_tunnel_info *info) 288 { 289 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev)); 290 bool use_cache = ip_tunnel_dst_cache_usable(skb, info); 291 struct socket *sock = rcu_dereference(bareudp->sock); 292 bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); 293 const struct ip_tunnel_key *key = &info->key; 294 struct rtable *rt; 295 __be16 sport, df; 296 int min_headroom; 297 __u8 tos, ttl; 298 __be32 saddr; 299 int err; 300 301 if (!sock) 302 return -ESHUTDOWN; 303 304 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info, 305 IPPROTO_UDP, use_cache); 306 307 if (IS_ERR(rt)) 308 return PTR_ERR(rt); 309 310 skb_tunnel_check_pmtu(skb, &rt->dst, 311 BAREUDP_IPV4_HLEN + info->options_len); 312 313 sport = udp_flow_src_port(bareudp->net, skb, 314 bareudp->sport_min, USHRT_MAX, 315 true); 316 tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb); 317 ttl = key->ttl; 318 df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0; 319 skb_scrub_packet(skb, xnet); 320 321 err = -ENOSPC; 322 if (!skb_pull(skb, skb_network_offset(skb))) 323 goto free_dst; 324 325 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len + 326 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr); 327 328 err = skb_cow_head(skb, min_headroom); 329 if (unlikely(err)) 330 goto free_dst; 331 332 err = udp_tunnel_handle_offloads(skb, udp_sum); 333 if (err) 334 goto free_dst; 335 336 skb_set_inner_protocol(skb, bareudp->ethertype); 337 udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst, 338 tos, ttl, df, sport, bareudp->port, 339 !net_eq(bareudp->net, dev_net(bareudp->dev)), 340 !(info->key.tun_flags & TUNNEL_CSUM)); 341 return 0; 342 343 free_dst: 344 dst_release(&rt->dst); 345 return err; 346 } 347 348 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev, 349 struct bareudp_dev *bareudp, 350 const struct ip_tunnel_info *info) 351 { 352 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev)); 353 bool use_cache = ip_tunnel_dst_cache_usable(skb, info); 354 struct socket *sock = rcu_dereference(bareudp->sock); 355 bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); 356 const struct ip_tunnel_key *key = &info->key; 357 struct dst_entry *dst = NULL; 358 struct in6_addr saddr, daddr; 359 int min_headroom; 360 __u8 prio, ttl; 361 __be16 sport; 362 int err; 363 364 if (!sock) 365 return -ESHUTDOWN; 366 367 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info, 368 IPPROTO_UDP, use_cache); 369 if (IS_ERR(dst)) 370 return PTR_ERR(dst); 371 372 skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len); 373 374 sport = udp_flow_src_port(bareudp->net, skb, 375 bareudp->sport_min, USHRT_MAX, 376 true); 377 prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb); 378 ttl = key->ttl; 379 380 skb_scrub_packet(skb, xnet); 381 382 err = -ENOSPC; 383 if (!skb_pull(skb, skb_network_offset(skb))) 384 goto free_dst; 385 386 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len + 387 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr); 388 389 err = skb_cow_head(skb, min_headroom); 390 if (unlikely(err)) 391 goto free_dst; 392 393 err = udp_tunnel_handle_offloads(skb, udp_sum); 394 if (err) 395 goto free_dst; 396 397 daddr = info->key.u.ipv6.dst; 398 udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev, 399 &saddr, &daddr, prio, ttl, 400 info->key.label, sport, bareudp->port, 401 !(info->key.tun_flags & TUNNEL_CSUM)); 402 return 0; 403 404 free_dst: 405 dst_release(dst); 406 return err; 407 } 408 409 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev) 410 { 411 struct bareudp_dev *bareudp = netdev_priv(dev); 412 struct ip_tunnel_info *info = NULL; 413 int err; 414 415 if (skb->protocol != bareudp->ethertype) { 416 if (!bareudp->multi_proto_mode || 417 (skb->protocol != htons(ETH_P_MPLS_MC) && 418 skb->protocol != htons(ETH_P_IPV6))) { 419 err = -EINVAL; 420 goto tx_error; 421 } 422 } 423 424 info = skb_tunnel_info(skb); 425 if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) { 426 err = -EINVAL; 427 goto tx_error; 428 } 429 430 rcu_read_lock(); 431 if (IS_ENABLED(CONFIG_IPV6) && info->mode & IP_TUNNEL_INFO_IPV6) 432 err = bareudp6_xmit_skb(skb, dev, bareudp, info); 433 else 434 err = bareudp_xmit_skb(skb, dev, bareudp, info); 435 436 rcu_read_unlock(); 437 438 if (likely(!err)) 439 return NETDEV_TX_OK; 440 tx_error: 441 dev_kfree_skb(skb); 442 443 if (err == -ELOOP) 444 dev->stats.collisions++; 445 else if (err == -ENETUNREACH) 446 dev->stats.tx_carrier_errors++; 447 448 dev->stats.tx_errors++; 449 return NETDEV_TX_OK; 450 } 451 452 static int bareudp_fill_metadata_dst(struct net_device *dev, 453 struct sk_buff *skb) 454 { 455 struct ip_tunnel_info *info = skb_tunnel_info(skb); 456 struct bareudp_dev *bareudp = netdev_priv(dev); 457 bool use_cache; 458 459 use_cache = ip_tunnel_dst_cache_usable(skb, info); 460 461 if (!IS_ENABLED(CONFIG_IPV6) || ip_tunnel_info_af(info) == AF_INET) { 462 struct rtable *rt; 463 __be32 saddr; 464 465 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, 466 info, IPPROTO_UDP, use_cache); 467 if (IS_ERR(rt)) 468 return PTR_ERR(rt); 469 470 ip_rt_put(rt); 471 info->key.u.ipv4.src = saddr; 472 } else if (ip_tunnel_info_af(info) == AF_INET6) { 473 struct dst_entry *dst; 474 struct in6_addr saddr; 475 struct socket *sock = rcu_dereference(bareudp->sock); 476 477 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, 478 &saddr, info, IPPROTO_UDP, 479 use_cache); 480 if (IS_ERR(dst)) 481 return PTR_ERR(dst); 482 483 dst_release(dst); 484 info->key.u.ipv6.src = saddr; 485 } else { 486 return -EINVAL; 487 } 488 489 info->key.tp_src = udp_flow_src_port(bareudp->net, skb, 490 bareudp->sport_min, 491 USHRT_MAX, true); 492 info->key.tp_dst = bareudp->port; 493 return 0; 494 } 495 496 static const struct net_device_ops bareudp_netdev_ops = { 497 .ndo_init = bareudp_init, 498 .ndo_uninit = bareudp_uninit, 499 .ndo_open = bareudp_open, 500 .ndo_stop = bareudp_stop, 501 .ndo_start_xmit = bareudp_xmit, 502 .ndo_get_stats64 = ip_tunnel_get_stats64, 503 .ndo_fill_metadata_dst = bareudp_fill_metadata_dst, 504 }; 505 506 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = { 507 [IFLA_BAREUDP_PORT] = { .type = NLA_U16 }, 508 [IFLA_BAREUDP_ETHERTYPE] = { .type = NLA_U16 }, 509 [IFLA_BAREUDP_SRCPORT_MIN] = { .type = NLA_U16 }, 510 [IFLA_BAREUDP_MULTIPROTO_MODE] = { .type = NLA_FLAG }, 511 }; 512 513 /* Info for udev, that this is a virtual tunnel endpoint */ 514 static struct device_type bareudp_type = { 515 .name = "bareudp", 516 }; 517 518 /* Initialize the device structure. */ 519 static void bareudp_setup(struct net_device *dev) 520 { 521 dev->netdev_ops = &bareudp_netdev_ops; 522 dev->needs_free_netdev = true; 523 SET_NETDEV_DEVTYPE(dev, &bareudp_type); 524 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM; 525 dev->features |= NETIF_F_RXCSUM; 526 dev->features |= NETIF_F_GSO_SOFTWARE; 527 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM; 528 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 529 dev->hard_header_len = 0; 530 dev->addr_len = 0; 531 dev->mtu = ETH_DATA_LEN; 532 dev->min_mtu = IPV4_MIN_MTU; 533 dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN; 534 dev->type = ARPHRD_NONE; 535 netif_keep_dst(dev); 536 dev->priv_flags |= IFF_NO_QUEUE; 537 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 538 } 539 540 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[], 541 struct netlink_ext_ack *extack) 542 { 543 if (!data) { 544 NL_SET_ERR_MSG(extack, 545 "Not enough attributes provided to perform the operation"); 546 return -EINVAL; 547 } 548 return 0; 549 } 550 551 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf, 552 struct netlink_ext_ack *extack) 553 { 554 memset(conf, 0, sizeof(*conf)); 555 556 if (!data[IFLA_BAREUDP_PORT]) { 557 NL_SET_ERR_MSG(extack, "port not specified"); 558 return -EINVAL; 559 } 560 if (!data[IFLA_BAREUDP_ETHERTYPE]) { 561 NL_SET_ERR_MSG(extack, "ethertype not specified"); 562 return -EINVAL; 563 } 564 565 if (data[IFLA_BAREUDP_PORT]) 566 conf->port = nla_get_u16(data[IFLA_BAREUDP_PORT]); 567 568 if (data[IFLA_BAREUDP_ETHERTYPE]) 569 conf->ethertype = nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]); 570 571 if (data[IFLA_BAREUDP_SRCPORT_MIN]) 572 conf->sport_min = nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]); 573 574 if (data[IFLA_BAREUDP_MULTIPROTO_MODE]) 575 conf->multi_proto_mode = true; 576 577 return 0; 578 } 579 580 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn, 581 const struct bareudp_conf *conf) 582 { 583 struct bareudp_dev *bareudp, *t = NULL; 584 585 list_for_each_entry(bareudp, &bn->bareudp_list, next) { 586 if (conf->port == bareudp->port) 587 t = bareudp; 588 } 589 return t; 590 } 591 592 static int bareudp_configure(struct net *net, struct net_device *dev, 593 struct bareudp_conf *conf) 594 { 595 struct bareudp_net *bn = net_generic(net, bareudp_net_id); 596 struct bareudp_dev *t, *bareudp = netdev_priv(dev); 597 int err; 598 599 bareudp->net = net; 600 bareudp->dev = dev; 601 t = bareudp_find_dev(bn, conf); 602 if (t) 603 return -EBUSY; 604 605 if (conf->multi_proto_mode && 606 (conf->ethertype != htons(ETH_P_MPLS_UC) && 607 conf->ethertype != htons(ETH_P_IP))) 608 return -EINVAL; 609 610 bareudp->port = conf->port; 611 bareudp->ethertype = conf->ethertype; 612 bareudp->sport_min = conf->sport_min; 613 bareudp->multi_proto_mode = conf->multi_proto_mode; 614 615 err = register_netdevice(dev); 616 if (err) 617 return err; 618 619 list_add(&bareudp->next, &bn->bareudp_list); 620 return 0; 621 } 622 623 static int bareudp_link_config(struct net_device *dev, 624 struct nlattr *tb[]) 625 { 626 int err; 627 628 if (tb[IFLA_MTU]) { 629 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 630 if (err) 631 return err; 632 } 633 return 0; 634 } 635 636 static int bareudp_newlink(struct net *net, struct net_device *dev, 637 struct nlattr *tb[], struct nlattr *data[], 638 struct netlink_ext_ack *extack) 639 { 640 struct bareudp_conf conf; 641 int err; 642 643 err = bareudp2info(data, &conf, extack); 644 if (err) 645 return err; 646 647 err = bareudp_configure(net, dev, &conf); 648 if (err) 649 return err; 650 651 err = bareudp_link_config(dev, tb); 652 if (err) 653 return err; 654 655 return 0; 656 } 657 658 static void bareudp_dellink(struct net_device *dev, struct list_head *head) 659 { 660 struct bareudp_dev *bareudp = netdev_priv(dev); 661 662 list_del(&bareudp->next); 663 unregister_netdevice_queue(dev, head); 664 } 665 666 static size_t bareudp_get_size(const struct net_device *dev) 667 { 668 return nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_PORT */ 669 nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_ETHERTYPE */ 670 nla_total_size(sizeof(__u16)) + /* IFLA_BAREUDP_SRCPORT_MIN */ 671 nla_total_size(0) + /* IFLA_BAREUDP_MULTIPROTO_MODE */ 672 0; 673 } 674 675 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev) 676 { 677 struct bareudp_dev *bareudp = netdev_priv(dev); 678 679 if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port)) 680 goto nla_put_failure; 681 if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype)) 682 goto nla_put_failure; 683 if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min)) 684 goto nla_put_failure; 685 if (bareudp->multi_proto_mode && 686 nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE)) 687 goto nla_put_failure; 688 689 return 0; 690 691 nla_put_failure: 692 return -EMSGSIZE; 693 } 694 695 static struct rtnl_link_ops bareudp_link_ops __read_mostly = { 696 .kind = "bareudp", 697 .maxtype = IFLA_BAREUDP_MAX, 698 .policy = bareudp_policy, 699 .priv_size = sizeof(struct bareudp_dev), 700 .setup = bareudp_setup, 701 .validate = bareudp_validate, 702 .newlink = bareudp_newlink, 703 .dellink = bareudp_dellink, 704 .get_size = bareudp_get_size, 705 .fill_info = bareudp_fill_info, 706 }; 707 708 struct net_device *bareudp_dev_create(struct net *net, const char *name, 709 u8 name_assign_type, 710 struct bareudp_conf *conf) 711 { 712 struct nlattr *tb[IFLA_MAX + 1]; 713 struct net_device *dev; 714 LIST_HEAD(list_kill); 715 int err; 716 717 memset(tb, 0, sizeof(tb)); 718 dev = rtnl_create_link(net, name, name_assign_type, 719 &bareudp_link_ops, tb, NULL); 720 if (IS_ERR(dev)) 721 return dev; 722 723 err = bareudp_configure(net, dev, conf); 724 if (err) { 725 free_netdev(dev); 726 return ERR_PTR(err); 727 } 728 err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN); 729 if (err) 730 goto err; 731 732 err = rtnl_configure_link(dev, NULL); 733 if (err < 0) 734 goto err; 735 736 return dev; 737 err: 738 bareudp_dellink(dev, &list_kill); 739 unregister_netdevice_many(&list_kill); 740 return ERR_PTR(err); 741 } 742 EXPORT_SYMBOL_GPL(bareudp_dev_create); 743 744 static __net_init int bareudp_init_net(struct net *net) 745 { 746 struct bareudp_net *bn = net_generic(net, bareudp_net_id); 747 748 INIT_LIST_HEAD(&bn->bareudp_list); 749 return 0; 750 } 751 752 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head) 753 { 754 struct bareudp_net *bn = net_generic(net, bareudp_net_id); 755 struct bareudp_dev *bareudp, *next; 756 757 list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next) 758 unregister_netdevice_queue(bareudp->dev, head); 759 } 760 761 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list) 762 { 763 struct net *net; 764 LIST_HEAD(list); 765 766 rtnl_lock(); 767 list_for_each_entry(net, net_list, exit_list) 768 bareudp_destroy_tunnels(net, &list); 769 770 /* unregister the devices gathered above */ 771 unregister_netdevice_many(&list); 772 rtnl_unlock(); 773 } 774 775 static struct pernet_operations bareudp_net_ops = { 776 .init = bareudp_init_net, 777 .exit_batch = bareudp_exit_batch_net, 778 .id = &bareudp_net_id, 779 .size = sizeof(struct bareudp_net), 780 }; 781 782 static int __init bareudp_init_module(void) 783 { 784 int rc; 785 786 rc = register_pernet_subsys(&bareudp_net_ops); 787 if (rc) 788 goto out1; 789 790 rc = rtnl_link_register(&bareudp_link_ops); 791 if (rc) 792 goto out2; 793 794 return 0; 795 out2: 796 unregister_pernet_subsys(&bareudp_net_ops); 797 out1: 798 return rc; 799 } 800 late_initcall(bareudp_init_module); 801 802 static void __exit bareudp_cleanup_module(void) 803 { 804 rtnl_link_unregister(&bareudp_link_ops); 805 unregister_pernet_subsys(&bareudp_net_ops); 806 } 807 module_exit(bareudp_cleanup_module); 808 809 MODULE_ALIAS_RTNL_LINK("bareudp"); 810 MODULE_LICENSE("GPL"); 811 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>"); 812 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic"); 813