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 bareudp_dev *bareudp; 58 unsigned short family; 59 unsigned int len; 60 __be16 proto; 61 void *oiph; 62 int err; 63 64 bareudp = rcu_dereference_sk_user_data(sk); 65 if (!bareudp) 66 goto drop; 67 68 if (skb->protocol == htons(ETH_P_IP)) 69 family = AF_INET; 70 else 71 family = AF_INET6; 72 73 if (bareudp->ethertype == htons(ETH_P_IP)) { 74 struct iphdr *iphdr; 75 76 iphdr = (struct iphdr *)(skb->data + BAREUDP_BASE_HLEN); 77 if (iphdr->version == 4) { 78 proto = bareudp->ethertype; 79 } else if (bareudp->multi_proto_mode && (iphdr->version == 6)) { 80 proto = htons(ETH_P_IPV6); 81 } else { 82 bareudp->dev->stats.rx_dropped++; 83 goto drop; 84 } 85 } else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) { 86 struct iphdr *tunnel_hdr; 87 88 tunnel_hdr = (struct iphdr *)skb_network_header(skb); 89 if (tunnel_hdr->version == 4) { 90 if (!ipv4_is_multicast(tunnel_hdr->daddr)) { 91 proto = bareudp->ethertype; 92 } else if (bareudp->multi_proto_mode && 93 ipv4_is_multicast(tunnel_hdr->daddr)) { 94 proto = htons(ETH_P_MPLS_MC); 95 } else { 96 bareudp->dev->stats.rx_dropped++; 97 goto drop; 98 } 99 } else { 100 int addr_type; 101 struct ipv6hdr *tunnel_hdr_v6; 102 103 tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb); 104 addr_type = 105 ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr); 106 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 107 proto = bareudp->ethertype; 108 } else if (bareudp->multi_proto_mode && 109 (addr_type & IPV6_ADDR_MULTICAST)) { 110 proto = htons(ETH_P_MPLS_MC); 111 } else { 112 bareudp->dev->stats.rx_dropped++; 113 goto drop; 114 } 115 } 116 } else { 117 proto = bareudp->ethertype; 118 } 119 120 if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN, 121 proto, 122 !net_eq(bareudp->net, 123 dev_net(bareudp->dev)))) { 124 bareudp->dev->stats.rx_dropped++; 125 goto drop; 126 } 127 tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0); 128 if (!tun_dst) { 129 bareudp->dev->stats.rx_dropped++; 130 goto drop; 131 } 132 skb_dst_set(skb, &tun_dst->dst); 133 skb->dev = bareudp->dev; 134 oiph = skb_network_header(skb); 135 skb_reset_network_header(skb); 136 137 if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET) 138 err = IP_ECN_decapsulate(oiph, skb); 139 else 140 err = IP6_ECN_decapsulate(oiph, skb); 141 142 if (unlikely(err)) { 143 if (log_ecn_error) { 144 if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET) 145 net_info_ratelimited("non-ECT from %pI4 " 146 "with TOS=%#x\n", 147 &((struct iphdr *)oiph)->saddr, 148 ((struct iphdr *)oiph)->tos); 149 else 150 net_info_ratelimited("non-ECT from %pI6\n", 151 &((struct ipv6hdr *)oiph)->saddr); 152 } 153 if (err > 1) { 154 ++bareudp->dev->stats.rx_frame_errors; 155 ++bareudp->dev->stats.rx_errors; 156 goto drop; 157 } 158 } 159 160 len = skb->len; 161 err = gro_cells_receive(&bareudp->gro_cells, skb); 162 if (likely(err == NET_RX_SUCCESS)) 163 dev_sw_netstats_rx_add(bareudp->dev, len); 164 165 return 0; 166 drop: 167 /* Consume bad packet */ 168 kfree_skb(skb); 169 170 return 0; 171 } 172 173 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb) 174 { 175 return 0; 176 } 177 178 static int bareudp_init(struct net_device *dev) 179 { 180 struct bareudp_dev *bareudp = netdev_priv(dev); 181 int err; 182 183 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 184 if (!dev->tstats) 185 return -ENOMEM; 186 187 err = gro_cells_init(&bareudp->gro_cells, dev); 188 if (err) { 189 free_percpu(dev->tstats); 190 return err; 191 } 192 return 0; 193 } 194 195 static void bareudp_uninit(struct net_device *dev) 196 { 197 struct bareudp_dev *bareudp = netdev_priv(dev); 198 199 gro_cells_destroy(&bareudp->gro_cells); 200 free_percpu(dev->tstats); 201 } 202 203 static struct socket *bareudp_create_sock(struct net *net, __be16 port) 204 { 205 struct udp_port_cfg udp_conf; 206 struct socket *sock; 207 int err; 208 209 memset(&udp_conf, 0, sizeof(udp_conf)); 210 #if IS_ENABLED(CONFIG_IPV6) 211 udp_conf.family = AF_INET6; 212 #else 213 udp_conf.family = AF_INET; 214 #endif 215 udp_conf.local_udp_port = port; 216 /* Open UDP socket */ 217 err = udp_sock_create(net, &udp_conf, &sock); 218 if (err < 0) 219 return ERR_PTR(err); 220 221 return sock; 222 } 223 224 /* Create new listen socket if needed */ 225 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port) 226 { 227 struct udp_tunnel_sock_cfg tunnel_cfg; 228 struct socket *sock; 229 230 sock = bareudp_create_sock(bareudp->net, port); 231 if (IS_ERR(sock)) 232 return PTR_ERR(sock); 233 234 /* Mark socket as an encapsulation socket */ 235 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); 236 tunnel_cfg.sk_user_data = bareudp; 237 tunnel_cfg.encap_type = 1; 238 tunnel_cfg.encap_rcv = bareudp_udp_encap_recv; 239 tunnel_cfg.encap_err_lookup = bareudp_err_lookup; 240 tunnel_cfg.encap_destroy = NULL; 241 setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg); 242 243 /* As the setup_udp_tunnel_sock does not call udp_encap_enable if the 244 * socket type is v6 an explicit call to udp_encap_enable is needed. 245 */ 246 if (sock->sk->sk_family == AF_INET6) 247 udp_encap_enable(); 248 249 rcu_assign_pointer(bareudp->sock, sock); 250 return 0; 251 } 252 253 static int bareudp_open(struct net_device *dev) 254 { 255 struct bareudp_dev *bareudp = netdev_priv(dev); 256 int ret = 0; 257 258 ret = bareudp_socket_create(bareudp, bareudp->port); 259 return ret; 260 } 261 262 static void bareudp_sock_release(struct bareudp_dev *bareudp) 263 { 264 struct socket *sock; 265 266 sock = bareudp->sock; 267 rcu_assign_pointer(bareudp->sock, NULL); 268 synchronize_net(); 269 udp_tunnel_sock_release(sock); 270 } 271 272 static int bareudp_stop(struct net_device *dev) 273 { 274 struct bareudp_dev *bareudp = netdev_priv(dev); 275 276 bareudp_sock_release(bareudp); 277 return 0; 278 } 279 280 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev, 281 struct bareudp_dev *bareudp, 282 const struct ip_tunnel_info *info) 283 { 284 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev)); 285 bool use_cache = ip_tunnel_dst_cache_usable(skb, info); 286 struct socket *sock = rcu_dereference(bareudp->sock); 287 bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); 288 const struct ip_tunnel_key *key = &info->key; 289 struct rtable *rt; 290 __be16 sport, df; 291 int min_headroom; 292 __u8 tos, ttl; 293 __be32 saddr; 294 int err; 295 296 if (!sock) 297 return -ESHUTDOWN; 298 299 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info, 300 IPPROTO_UDP, use_cache); 301 302 if (IS_ERR(rt)) 303 return PTR_ERR(rt); 304 305 skb_tunnel_check_pmtu(skb, &rt->dst, 306 BAREUDP_IPV4_HLEN + info->options_len, false); 307 308 sport = udp_flow_src_port(bareudp->net, skb, 309 bareudp->sport_min, USHRT_MAX, 310 true); 311 tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb); 312 ttl = key->ttl; 313 df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0; 314 skb_scrub_packet(skb, xnet); 315 316 err = -ENOSPC; 317 if (!skb_pull(skb, skb_network_offset(skb))) 318 goto free_dst; 319 320 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len + 321 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr); 322 323 err = skb_cow_head(skb, min_headroom); 324 if (unlikely(err)) 325 goto free_dst; 326 327 err = udp_tunnel_handle_offloads(skb, udp_sum); 328 if (err) 329 goto free_dst; 330 331 skb_set_inner_protocol(skb, bareudp->ethertype); 332 udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst, 333 tos, ttl, df, sport, bareudp->port, 334 !net_eq(bareudp->net, dev_net(bareudp->dev)), 335 !(info->key.tun_flags & TUNNEL_CSUM)); 336 return 0; 337 338 free_dst: 339 dst_release(&rt->dst); 340 return err; 341 } 342 343 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev, 344 struct bareudp_dev *bareudp, 345 const struct ip_tunnel_info *info) 346 { 347 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev)); 348 bool use_cache = ip_tunnel_dst_cache_usable(skb, info); 349 struct socket *sock = rcu_dereference(bareudp->sock); 350 bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); 351 const struct ip_tunnel_key *key = &info->key; 352 struct dst_entry *dst = NULL; 353 struct in6_addr saddr, daddr; 354 int min_headroom; 355 __u8 prio, ttl; 356 __be16 sport; 357 int err; 358 359 if (!sock) 360 return -ESHUTDOWN; 361 362 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info, 363 IPPROTO_UDP, use_cache); 364 if (IS_ERR(dst)) 365 return PTR_ERR(dst); 366 367 skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len, 368 false); 369 370 sport = udp_flow_src_port(bareudp->net, skb, 371 bareudp->sport_min, USHRT_MAX, 372 true); 373 prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb); 374 ttl = key->ttl; 375 376 skb_scrub_packet(skb, xnet); 377 378 err = -ENOSPC; 379 if (!skb_pull(skb, skb_network_offset(skb))) 380 goto free_dst; 381 382 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len + 383 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr); 384 385 err = skb_cow_head(skb, min_headroom); 386 if (unlikely(err)) 387 goto free_dst; 388 389 err = udp_tunnel_handle_offloads(skb, udp_sum); 390 if (err) 391 goto free_dst; 392 393 daddr = info->key.u.ipv6.dst; 394 udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev, 395 &saddr, &daddr, prio, ttl, 396 info->key.label, sport, bareudp->port, 397 !(info->key.tun_flags & TUNNEL_CSUM)); 398 return 0; 399 400 free_dst: 401 dst_release(dst); 402 return err; 403 } 404 405 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto) 406 { 407 if (bareudp->ethertype == proto) 408 return true; 409 410 if (!bareudp->multi_proto_mode) 411 return false; 412 413 if (bareudp->ethertype == htons(ETH_P_MPLS_UC) && 414 proto == htons(ETH_P_MPLS_MC)) 415 return true; 416 417 if (bareudp->ethertype == htons(ETH_P_IP) && 418 proto == htons(ETH_P_IPV6)) 419 return true; 420 421 return false; 422 } 423 424 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev) 425 { 426 struct bareudp_dev *bareudp = netdev_priv(dev); 427 struct ip_tunnel_info *info = NULL; 428 int err; 429 430 if (!bareudp_proto_valid(bareudp, skb->protocol)) { 431 err = -EINVAL; 432 goto tx_error; 433 } 434 435 info = skb_tunnel_info(skb); 436 if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) { 437 err = -EINVAL; 438 goto tx_error; 439 } 440 441 rcu_read_lock(); 442 if (IS_ENABLED(CONFIG_IPV6) && info->mode & IP_TUNNEL_INFO_IPV6) 443 err = bareudp6_xmit_skb(skb, dev, bareudp, info); 444 else 445 err = bareudp_xmit_skb(skb, dev, bareudp, info); 446 447 rcu_read_unlock(); 448 449 if (likely(!err)) 450 return NETDEV_TX_OK; 451 tx_error: 452 dev_kfree_skb(skb); 453 454 if (err == -ELOOP) 455 dev->stats.collisions++; 456 else if (err == -ENETUNREACH) 457 dev->stats.tx_carrier_errors++; 458 459 dev->stats.tx_errors++; 460 return NETDEV_TX_OK; 461 } 462 463 static int bareudp_fill_metadata_dst(struct net_device *dev, 464 struct sk_buff *skb) 465 { 466 struct ip_tunnel_info *info = skb_tunnel_info(skb); 467 struct bareudp_dev *bareudp = netdev_priv(dev); 468 bool use_cache; 469 470 use_cache = ip_tunnel_dst_cache_usable(skb, info); 471 472 if (!IS_ENABLED(CONFIG_IPV6) || ip_tunnel_info_af(info) == AF_INET) { 473 struct rtable *rt; 474 __be32 saddr; 475 476 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, 477 info, IPPROTO_UDP, use_cache); 478 if (IS_ERR(rt)) 479 return PTR_ERR(rt); 480 481 ip_rt_put(rt); 482 info->key.u.ipv4.src = saddr; 483 } else if (ip_tunnel_info_af(info) == AF_INET6) { 484 struct dst_entry *dst; 485 struct in6_addr saddr; 486 struct socket *sock = rcu_dereference(bareudp->sock); 487 488 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, 489 &saddr, info, IPPROTO_UDP, 490 use_cache); 491 if (IS_ERR(dst)) 492 return PTR_ERR(dst); 493 494 dst_release(dst); 495 info->key.u.ipv6.src = saddr; 496 } else { 497 return -EINVAL; 498 } 499 500 info->key.tp_src = udp_flow_src_port(bareudp->net, skb, 501 bareudp->sport_min, 502 USHRT_MAX, true); 503 info->key.tp_dst = bareudp->port; 504 return 0; 505 } 506 507 static const struct net_device_ops bareudp_netdev_ops = { 508 .ndo_init = bareudp_init, 509 .ndo_uninit = bareudp_uninit, 510 .ndo_open = bareudp_open, 511 .ndo_stop = bareudp_stop, 512 .ndo_start_xmit = bareudp_xmit, 513 .ndo_get_stats64 = ip_tunnel_get_stats64, 514 .ndo_fill_metadata_dst = bareudp_fill_metadata_dst, 515 }; 516 517 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = { 518 [IFLA_BAREUDP_PORT] = { .type = NLA_U16 }, 519 [IFLA_BAREUDP_ETHERTYPE] = { .type = NLA_U16 }, 520 [IFLA_BAREUDP_SRCPORT_MIN] = { .type = NLA_U16 }, 521 [IFLA_BAREUDP_MULTIPROTO_MODE] = { .type = NLA_FLAG }, 522 }; 523 524 /* Info for udev, that this is a virtual tunnel endpoint */ 525 static struct device_type bareudp_type = { 526 .name = "bareudp", 527 }; 528 529 /* Initialize the device structure. */ 530 static void bareudp_setup(struct net_device *dev) 531 { 532 dev->netdev_ops = &bareudp_netdev_ops; 533 dev->needs_free_netdev = true; 534 SET_NETDEV_DEVTYPE(dev, &bareudp_type); 535 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM; 536 dev->features |= NETIF_F_RXCSUM; 537 dev->features |= NETIF_F_GSO_SOFTWARE; 538 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM; 539 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 540 dev->hard_header_len = 0; 541 dev->addr_len = 0; 542 dev->mtu = ETH_DATA_LEN; 543 dev->min_mtu = IPV4_MIN_MTU; 544 dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN; 545 dev->type = ARPHRD_NONE; 546 netif_keep_dst(dev); 547 dev->priv_flags |= IFF_NO_QUEUE; 548 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; 549 } 550 551 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[], 552 struct netlink_ext_ack *extack) 553 { 554 if (!data) { 555 NL_SET_ERR_MSG(extack, 556 "Not enough attributes provided to perform the operation"); 557 return -EINVAL; 558 } 559 return 0; 560 } 561 562 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf, 563 struct netlink_ext_ack *extack) 564 { 565 memset(conf, 0, sizeof(*conf)); 566 567 if (!data[IFLA_BAREUDP_PORT]) { 568 NL_SET_ERR_MSG(extack, "port not specified"); 569 return -EINVAL; 570 } 571 if (!data[IFLA_BAREUDP_ETHERTYPE]) { 572 NL_SET_ERR_MSG(extack, "ethertype not specified"); 573 return -EINVAL; 574 } 575 576 if (data[IFLA_BAREUDP_PORT]) 577 conf->port = nla_get_u16(data[IFLA_BAREUDP_PORT]); 578 579 if (data[IFLA_BAREUDP_ETHERTYPE]) 580 conf->ethertype = nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]); 581 582 if (data[IFLA_BAREUDP_SRCPORT_MIN]) 583 conf->sport_min = nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]); 584 585 if (data[IFLA_BAREUDP_MULTIPROTO_MODE]) 586 conf->multi_proto_mode = true; 587 588 return 0; 589 } 590 591 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn, 592 const struct bareudp_conf *conf) 593 { 594 struct bareudp_dev *bareudp, *t = NULL; 595 596 list_for_each_entry(bareudp, &bn->bareudp_list, next) { 597 if (conf->port == bareudp->port) 598 t = bareudp; 599 } 600 return t; 601 } 602 603 static int bareudp_configure(struct net *net, struct net_device *dev, 604 struct bareudp_conf *conf) 605 { 606 struct bareudp_net *bn = net_generic(net, bareudp_net_id); 607 struct bareudp_dev *t, *bareudp = netdev_priv(dev); 608 int err; 609 610 bareudp->net = net; 611 bareudp->dev = dev; 612 t = bareudp_find_dev(bn, conf); 613 if (t) 614 return -EBUSY; 615 616 if (conf->multi_proto_mode && 617 (conf->ethertype != htons(ETH_P_MPLS_UC) && 618 conf->ethertype != htons(ETH_P_IP))) 619 return -EINVAL; 620 621 bareudp->port = conf->port; 622 bareudp->ethertype = conf->ethertype; 623 bareudp->sport_min = conf->sport_min; 624 bareudp->multi_proto_mode = conf->multi_proto_mode; 625 626 err = register_netdevice(dev); 627 if (err) 628 return err; 629 630 list_add(&bareudp->next, &bn->bareudp_list); 631 return 0; 632 } 633 634 static int bareudp_link_config(struct net_device *dev, 635 struct nlattr *tb[]) 636 { 637 int err; 638 639 if (tb[IFLA_MTU]) { 640 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 641 if (err) 642 return err; 643 } 644 return 0; 645 } 646 647 static int bareudp_newlink(struct net *net, struct net_device *dev, 648 struct nlattr *tb[], struct nlattr *data[], 649 struct netlink_ext_ack *extack) 650 { 651 struct bareudp_conf conf; 652 int err; 653 654 err = bareudp2info(data, &conf, extack); 655 if (err) 656 return err; 657 658 err = bareudp_configure(net, dev, &conf); 659 if (err) 660 return err; 661 662 err = bareudp_link_config(dev, tb); 663 if (err) 664 return err; 665 666 return 0; 667 } 668 669 static void bareudp_dellink(struct net_device *dev, struct list_head *head) 670 { 671 struct bareudp_dev *bareudp = netdev_priv(dev); 672 673 list_del(&bareudp->next); 674 unregister_netdevice_queue(dev, head); 675 } 676 677 static size_t bareudp_get_size(const struct net_device *dev) 678 { 679 return nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_PORT */ 680 nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_ETHERTYPE */ 681 nla_total_size(sizeof(__u16)) + /* IFLA_BAREUDP_SRCPORT_MIN */ 682 nla_total_size(0) + /* IFLA_BAREUDP_MULTIPROTO_MODE */ 683 0; 684 } 685 686 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev) 687 { 688 struct bareudp_dev *bareudp = netdev_priv(dev); 689 690 if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port)) 691 goto nla_put_failure; 692 if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype)) 693 goto nla_put_failure; 694 if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min)) 695 goto nla_put_failure; 696 if (bareudp->multi_proto_mode && 697 nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE)) 698 goto nla_put_failure; 699 700 return 0; 701 702 nla_put_failure: 703 return -EMSGSIZE; 704 } 705 706 static struct rtnl_link_ops bareudp_link_ops __read_mostly = { 707 .kind = "bareudp", 708 .maxtype = IFLA_BAREUDP_MAX, 709 .policy = bareudp_policy, 710 .priv_size = sizeof(struct bareudp_dev), 711 .setup = bareudp_setup, 712 .validate = bareudp_validate, 713 .newlink = bareudp_newlink, 714 .dellink = bareudp_dellink, 715 .get_size = bareudp_get_size, 716 .fill_info = bareudp_fill_info, 717 }; 718 719 struct net_device *bareudp_dev_create(struct net *net, const char *name, 720 u8 name_assign_type, 721 struct bareudp_conf *conf) 722 { 723 struct nlattr *tb[IFLA_MAX + 1]; 724 struct net_device *dev; 725 LIST_HEAD(list_kill); 726 int err; 727 728 memset(tb, 0, sizeof(tb)); 729 dev = rtnl_create_link(net, name, name_assign_type, 730 &bareudp_link_ops, tb, NULL); 731 if (IS_ERR(dev)) 732 return dev; 733 734 err = bareudp_configure(net, dev, conf); 735 if (err) { 736 free_netdev(dev); 737 return ERR_PTR(err); 738 } 739 err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN); 740 if (err) 741 goto err; 742 743 err = rtnl_configure_link(dev, NULL); 744 if (err < 0) 745 goto err; 746 747 return dev; 748 err: 749 bareudp_dellink(dev, &list_kill); 750 unregister_netdevice_many(&list_kill); 751 return ERR_PTR(err); 752 } 753 EXPORT_SYMBOL_GPL(bareudp_dev_create); 754 755 static __net_init int bareudp_init_net(struct net *net) 756 { 757 struct bareudp_net *bn = net_generic(net, bareudp_net_id); 758 759 INIT_LIST_HEAD(&bn->bareudp_list); 760 return 0; 761 } 762 763 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head) 764 { 765 struct bareudp_net *bn = net_generic(net, bareudp_net_id); 766 struct bareudp_dev *bareudp, *next; 767 768 list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next) 769 unregister_netdevice_queue(bareudp->dev, head); 770 } 771 772 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list) 773 { 774 struct net *net; 775 LIST_HEAD(list); 776 777 rtnl_lock(); 778 list_for_each_entry(net, net_list, exit_list) 779 bareudp_destroy_tunnels(net, &list); 780 781 /* unregister the devices gathered above */ 782 unregister_netdevice_many(&list); 783 rtnl_unlock(); 784 } 785 786 static struct pernet_operations bareudp_net_ops = { 787 .init = bareudp_init_net, 788 .exit_batch = bareudp_exit_batch_net, 789 .id = &bareudp_net_id, 790 .size = sizeof(struct bareudp_net), 791 }; 792 793 static int __init bareudp_init_module(void) 794 { 795 int rc; 796 797 rc = register_pernet_subsys(&bareudp_net_ops); 798 if (rc) 799 goto out1; 800 801 rc = rtnl_link_register(&bareudp_link_ops); 802 if (rc) 803 goto out2; 804 805 return 0; 806 out2: 807 unregister_pernet_subsys(&bareudp_net_ops); 808 out1: 809 return rc; 810 } 811 late_initcall(bareudp_init_module); 812 813 static void __exit bareudp_cleanup_module(void) 814 { 815 rtnl_link_unregister(&bareudp_link_ops); 816 unregister_pernet_subsys(&bareudp_net_ops); 817 } 818 module_exit(bareudp_cleanup_module); 819 820 MODULE_ALIAS_RTNL_LINK("bareudp"); 821 MODULE_LICENSE("GPL"); 822 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>"); 823 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic"); 824