drivers/net/bareudp.c

   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, <[email protected]>
   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         udp_allow_gso(sock->sk);
 222         return sock;
 223 }
 224
 225 /* Create new listen socket if needed */
 226 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
 227 {
 228         struct udp_tunnel_sock_cfg tunnel_cfg;
 229         struct socket *sock;
 230
 231         sock = bareudp_create_sock(bareudp->net, port);
 232         if (IS_ERR(sock))
 233                 return PTR_ERR(sock);
 234
 235         /* Mark socket as an encapsulation socket */
 236         memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
 237         tunnel_cfg.sk_user_data = bareudp;
 238         tunnel_cfg.encap_type = 1;
 239         tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
 240         tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
 241         tunnel_cfg.encap_destroy = NULL;
 242         setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
 243
 244         rcu_assign_pointer(bareudp->sock, sock);
 245         return 0;
 246 }
 247
 248 static int bareudp_open(struct net_device *dev)
 249 {
 250         struct bareudp_dev *bareudp = netdev_priv(dev);
 251         int ret = 0;
 252
 253         ret =  bareudp_socket_create(bareudp, bareudp->port);
 254         return ret;
 255 }
 256
 257 static void bareudp_sock_release(struct bareudp_dev *bareudp)
 258 {
 259         struct socket *sock;
 260
 261         sock = bareudp->sock;
 262         rcu_assign_pointer(bareudp->sock, NULL);
 263         synchronize_net();
 264         udp_tunnel_sock_release(sock);
 265 }
 266
 267 static int bareudp_stop(struct net_device *dev)
 268 {
 269         struct bareudp_dev *bareudp = netdev_priv(dev);
 270
 271         bareudp_sock_release(bareudp);
 272         return 0;
 273 }
 274
 275 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
 276                             struct bareudp_dev *bareudp,
 277                             const struct ip_tunnel_info *info)
 278 {
 279         bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
 280         bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
 281         struct socket *sock = rcu_dereference(bareudp->sock);
 282         bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
 283         const struct ip_tunnel_key *key = &info->key;
 284         struct rtable *rt;
 285         __be16 sport, df;
 286         int min_headroom;
 287         __u8 tos, ttl;
 288         __be32 saddr;
 289         int err;
 290
 291         if (!sock)
 292                 return -ESHUTDOWN;
 293
 294         rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
 295                                     IPPROTO_UDP, use_cache);
 296
 297         if (IS_ERR(rt))
 298                 return PTR_ERR(rt);
 299
 300         skb_tunnel_check_pmtu(skb, &rt->dst,
 301                               BAREUDP_IPV4_HLEN + info->options_len, false);
 302
 303         sport = udp_flow_src_port(bareudp->net, skb,
 304                                   bareudp->sport_min, USHRT_MAX,
 305                                   true);
 306         tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
 307         ttl = key->ttl;
 308         df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
 309         skb_scrub_packet(skb, xnet);
 310
 311         err = -ENOSPC;
 312         if (!skb_pull(skb, skb_network_offset(skb)))
 313                 goto free_dst;
 314
 315         min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
 316                 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
 317
 318         err = skb_cow_head(skb, min_headroom);
 319         if (unlikely(err))
 320                 goto free_dst;
 321
 322         err = udp_tunnel_handle_offloads(skb, udp_sum);
 323         if (err)
 324                 goto free_dst;
 325
 326         skb_set_inner_protocol(skb, bareudp->ethertype);
 327         udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
 328                             tos, ttl, df, sport, bareudp->port,
 329                             !net_eq(bareudp->net, dev_net(bareudp->dev)),
 330                             !(info->key.tun_flags & TUNNEL_CSUM));
 331         return 0;
 332
 333 free_dst:
 334         dst_release(&rt->dst);
 335         return err;
 336 }
 337
 338 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
 339                              struct bareudp_dev *bareudp,
 340                              const struct ip_tunnel_info *info)
 341 {
 342         bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
 343         bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
 344         struct socket *sock  = rcu_dereference(bareudp->sock);
 345         bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
 346         const struct ip_tunnel_key *key = &info->key;
 347         struct dst_entry *dst = NULL;
 348         struct in6_addr saddr, daddr;
 349         int min_headroom;
 350         __u8 prio, ttl;
 351         __be16 sport;
 352         int err;
 353
 354         if (!sock)
 355                 return -ESHUTDOWN;
 356
 357         dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
 358                                     IPPROTO_UDP, use_cache);
 359         if (IS_ERR(dst))
 360                 return PTR_ERR(dst);
 361
 362         skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
 363                               false);
 364
 365         sport = udp_flow_src_port(bareudp->net, skb,
 366                                   bareudp->sport_min, USHRT_MAX,
 367                                   true);
 368         prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
 369         ttl = key->ttl;
 370
 371         skb_scrub_packet(skb, xnet);
 372
 373         err = -ENOSPC;
 374         if (!skb_pull(skb, skb_network_offset(skb)))
 375                 goto free_dst;
 376
 377         min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
 378                 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct ipv6hdr);
 379
 380         err = skb_cow_head(skb, min_headroom);
 381         if (unlikely(err))
 382                 goto free_dst;
 383
 384         err = udp_tunnel_handle_offloads(skb, udp_sum);
 385         if (err)
 386                 goto free_dst;
 387
 388         daddr = info->key.u.ipv6.dst;
 389         udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
 390                              &saddr, &daddr, prio, ttl,
 391                              info->key.label, sport, bareudp->port,
 392                              !(info->key.tun_flags & TUNNEL_CSUM));
 393         return 0;
 394
 395 free_dst:
 396         dst_release(dst);
 397         return err;
 398 }
 399
 400 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
 401 {
 402         if (bareudp->ethertype == proto)
 403                 return true;
 404
 405         if (!bareudp->multi_proto_mode)
 406                 return false;
 407
 408         if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
 409             proto == htons(ETH_P_MPLS_MC))
 410                 return true;
 411
 412         if (bareudp->ethertype == htons(ETH_P_IP) &&
 413             proto == htons(ETH_P_IPV6))
 414                 return true;
 415
 416         return false;
 417 }
 418
 419 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
 420 {
 421         struct bareudp_dev *bareudp = netdev_priv(dev);
 422         struct ip_tunnel_info *info = NULL;
 423         int err;
 424
 425         if (!bareudp_proto_valid(bareudp, skb->protocol)) {
 426                 err = -EINVAL;
 427                 goto tx_error;
 428         }
 429
 430         info = skb_tunnel_info(skb);
 431         if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
 432                 err = -EINVAL;
 433                 goto tx_error;
 434         }
 435
 436         rcu_read_lock();
 437         if (IS_ENABLED(CONFIG_IPV6) && info->mode & IP_TUNNEL_INFO_IPV6)
 438                 err = bareudp6_xmit_skb(skb, dev, bareudp, info);
 439         else
 440                 err = bareudp_xmit_skb(skb, dev, bareudp, info);
 441
 442         rcu_read_unlock();
 443
 444         if (likely(!err))
 445                 return NETDEV_TX_OK;
 446 tx_error:
 447         dev_kfree_skb(skb);
 448
 449         if (err == -ELOOP)
 450                 dev->stats.collisions++;
 451         else if (err == -ENETUNREACH)
 452                 dev->stats.tx_carrier_errors++;
 453
 454         dev->stats.tx_errors++;
 455         return NETDEV_TX_OK;
 456 }
 457
 458 static int bareudp_fill_metadata_dst(struct net_device *dev,
 459                                      struct sk_buff *skb)
 460 {
 461         struct ip_tunnel_info *info = skb_tunnel_info(skb);
 462         struct bareudp_dev *bareudp = netdev_priv(dev);
 463         bool use_cache;
 464
 465         use_cache = ip_tunnel_dst_cache_usable(skb, info);
 466
 467         if (!IS_ENABLED(CONFIG_IPV6) || ip_tunnel_info_af(info) == AF_INET) {
 468                 struct rtable *rt;
 469                 __be32 saddr;
 470
 471                 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
 472                                             info, IPPROTO_UDP, use_cache);
 473                 if (IS_ERR(rt))
 474                         return PTR_ERR(rt);
 475
 476                 ip_rt_put(rt);
 477                 info->key.u.ipv4.src = saddr;
 478         } else if (ip_tunnel_info_af(info) == AF_INET6) {
 479                 struct dst_entry *dst;
 480                 struct in6_addr saddr;
 481                 struct socket *sock = rcu_dereference(bareudp->sock);
 482
 483                 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
 484                                             &saddr, info, IPPROTO_UDP,
 485                                             use_cache);
 486                 if (IS_ERR(dst))
 487                         return PTR_ERR(dst);
 488
 489                 dst_release(dst);
 490                 info->key.u.ipv6.src = saddr;
 491         } else {
 492                 return -EINVAL;
 493         }
 494
 495         info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
 496                                              bareudp->sport_min,
 497                         USHRT_MAX, true);
 498         info->key.tp_dst = bareudp->port;
 499         return 0;
 500 }
 501
 502 static const struct net_device_ops bareudp_netdev_ops = {
 503         .ndo_init               = bareudp_init,
 504         .ndo_uninit             = bareudp_uninit,
 505         .ndo_open               = bareudp_open,
 506         .ndo_stop               = bareudp_stop,
 507         .ndo_start_xmit         = bareudp_xmit,
 508         .ndo_get_stats64        = dev_get_tstats64,
 509         .ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
 510 };
 511
 512 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
 513         [IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
 514         [IFLA_BAREUDP_ETHERTYPE]           = { .type = NLA_U16 },
 515         [IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
 516         [IFLA_BAREUDP_MULTIPROTO_MODE]     = { .type = NLA_FLAG },
 517 };
 518
 519 /* Info for udev, that this is a virtual tunnel endpoint */
 520 static const struct device_type bareudp_type = {
 521         .name = "bareudp",
 522 };
 523
 524 /* Initialize the device structure. */
 525 static void bareudp_setup(struct net_device *dev)
 526 {
 527         dev->netdev_ops = &bareudp_netdev_ops;
 528         dev->needs_free_netdev = true;
 529         SET_NETDEV_DEVTYPE(dev, &bareudp_type);
 530         dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
 531         dev->features    |= NETIF_F_RXCSUM;
 532         dev->features    |= NETIF_F_LLTX;
 533         dev->features    |= NETIF_F_GSO_SOFTWARE;
 534         dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
 535         dev->hw_features |= NETIF_F_RXCSUM;
 536         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
 537         dev->hard_header_len = 0;
 538         dev->addr_len = 0;
 539         dev->mtu = ETH_DATA_LEN;
 540         dev->min_mtu = IPV4_MIN_MTU;
 541         dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
 542         dev->type = ARPHRD_NONE;
 543         netif_keep_dst(dev);
 544         dev->priv_flags |= IFF_NO_QUEUE;
 545         dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
 546 }
 547
 548 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
 549                             struct netlink_ext_ack *extack)
 550 {
 551         if (!data) {
 552                 NL_SET_ERR_MSG(extack,
 553                                "Not enough attributes provided to perform the operation");
 554                 return -EINVAL;
 555         }
 556         return 0;
 557 }
 558
 559 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
 560                         struct netlink_ext_ack *extack)
 561 {
 562         memset(conf, 0, sizeof(*conf));
 563
 564         if (!data[IFLA_BAREUDP_PORT]) {
 565                 NL_SET_ERR_MSG(extack, "port not specified");
 566                 return -EINVAL;
 567         }
 568         if (!data[IFLA_BAREUDP_ETHERTYPE]) {
 569                 NL_SET_ERR_MSG(extack, "ethertype not specified");
 570                 return -EINVAL;
 571         }
 572
 573         if (data[IFLA_BAREUDP_PORT])
 574                 conf->port =  nla_get_u16(data[IFLA_BAREUDP_PORT]);
 575
 576         if (data[IFLA_BAREUDP_ETHERTYPE])
 577                 conf->ethertype =  nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
 578
 579         if (data[IFLA_BAREUDP_SRCPORT_MIN])
 580                 conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
 581
 582         if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
 583                 conf->multi_proto_mode = true;
 584
 585         return 0;
 586 }
 587
 588 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
 589                                             const struct bareudp_conf *conf)
 590 {
 591         struct bareudp_dev *bareudp, *t = NULL;
 592
 593         list_for_each_entry(bareudp, &bn->bareudp_list, next) {
 594                 if (conf->port == bareudp->port)
 595                         t = bareudp;
 596         }
 597         return t;
 598 }
 599
 600 static int bareudp_configure(struct net *net, struct net_device *dev,
 601                              struct bareudp_conf *conf)
 602 {
 603         struct bareudp_net *bn = net_generic(net, bareudp_net_id);
 604         struct bareudp_dev *t, *bareudp = netdev_priv(dev);
 605         int err;
 606
 607         bareudp->net = net;
 608         bareudp->dev = dev;
 609         t = bareudp_find_dev(bn, conf);
 610         if (t)
 611                 return -EBUSY;
 612
 613         if (conf->multi_proto_mode &&
 614             (conf->ethertype != htons(ETH_P_MPLS_UC) &&
 615              conf->ethertype != htons(ETH_P_IP)))
 616                 return -EINVAL;
 617
 618         bareudp->port = conf->port;
 619         bareudp->ethertype = conf->ethertype;
 620         bareudp->sport_min = conf->sport_min;
 621         bareudp->multi_proto_mode = conf->multi_proto_mode;
 622
 623         err = register_netdevice(dev);
 624         if (err)
 625                 return err;
 626
 627         list_add(&bareudp->next, &bn->bareudp_list);
 628         return 0;
 629 }
 630
 631 static int bareudp_link_config(struct net_device *dev,
 632                                struct nlattr *tb[])
 633 {
 634         int err;
 635
 636         if (tb[IFLA_MTU]) {
 637                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
 638                 if (err)
 639                         return err;
 640         }
 641         return 0;
 642 }
 643
 644 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
 645 {
 646         struct bareudp_dev *bareudp = netdev_priv(dev);
 647
 648         list_del(&bareudp->next);
 649         unregister_netdevice_queue(dev, head);
 650 }
 651
 652 static int bareudp_newlink(struct net *net, struct net_device *dev,
 653                            struct nlattr *tb[], struct nlattr *data[],
 654                            struct netlink_ext_ack *extack)
 655 {
 656         struct bareudp_conf conf;
 657         int err;
 658
 659         err = bareudp2info(data, &conf, extack);
 660         if (err)
 661                 return err;
 662
 663         err = bareudp_configure(net, dev, &conf);
 664         if (err)
 665                 return err;
 666
 667         err = bareudp_link_config(dev, tb);
 668         if (err)
 669                 goto err_unconfig;
 670
 671         return 0;
 672
 673 err_unconfig:
 674         bareudp_dellink(dev, NULL);
 675         return err;
 676 }
 677
 678 static size_t bareudp_get_size(const struct net_device *dev)
 679 {
 680         return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
 681                 nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
 682                 nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
 683                 nla_total_size(0)              +  /* IFLA_BAREUDP_MULTIPROTO_MODE */
 684                 0;
 685 }
 686
 687 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
 688 {
 689         struct bareudp_dev *bareudp = netdev_priv(dev);
 690
 691         if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
 692                 goto nla_put_failure;
 693         if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
 694                 goto nla_put_failure;
 695         if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
 696                 goto nla_put_failure;
 697         if (bareudp->multi_proto_mode &&
 698             nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
 699                 goto nla_put_failure;
 700
 701         return 0;
 702
 703 nla_put_failure:
 704         return -EMSGSIZE;
 705 }
 706
 707 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
 708         .kind           = "bareudp",
 709         .maxtype        = IFLA_BAREUDP_MAX,
 710         .policy         = bareudp_policy,
 711         .priv_size      = sizeof(struct bareudp_dev),
 712         .setup          = bareudp_setup,
 713         .validate       = bareudp_validate,
 714         .newlink        = bareudp_newlink,
 715         .dellink        = bareudp_dellink,
 716         .get_size       = bareudp_get_size,
 717         .fill_info      = bareudp_fill_info,
 718 };
 719
 720 struct net_device *bareudp_dev_create(struct net *net, const char *name,
 721                                       u8 name_assign_type,
 722                                       struct bareudp_conf *conf)
 723 {
 724         struct nlattr *tb[IFLA_MAX + 1];
 725         struct net_device *dev;
 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, NULL);
 750         return ERR_PTR(err);
 751 }
 752 EXPORT_SYMBOL_GPL(bareudp_dev_create);
 753
 754 static __net_init int bareudp_init_net(struct net *net)
 755 {
 756         struct bareudp_net *bn = net_generic(net, bareudp_net_id);
 757
 758         INIT_LIST_HEAD(&bn->bareudp_list);
 759         return 0;
 760 }
 761
 762 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
 763 {
 764         struct bareudp_net *bn = net_generic(net, bareudp_net_id);
 765         struct bareudp_dev *bareudp, *next;
 766
 767         list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
 768                 unregister_netdevice_queue(bareudp->dev, head);
 769 }
 770
 771 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
 772 {
 773         struct net *net;
 774         LIST_HEAD(list);
 775
 776         rtnl_lock();
 777         list_for_each_entry(net, net_list, exit_list)
 778                 bareudp_destroy_tunnels(net, &list);
 779
 780         /* unregister the devices gathered above */
 781         unregister_netdevice_many(&list);
 782         rtnl_unlock();
 783 }
 784
 785 static struct pernet_operations bareudp_net_ops = {
 786         .init = bareudp_init_net,
 787         .exit_batch = bareudp_exit_batch_net,
 788         .id   = &bareudp_net_id,
 789         .size = sizeof(struct bareudp_net),
 790 };
 791
 792 static int __init bareudp_init_module(void)
 793 {
 794         int rc;
 795
 796         rc = register_pernet_subsys(&bareudp_net_ops);
 797         if (rc)
 798                 goto out1;
 799
 800         rc = rtnl_link_register(&bareudp_link_ops);
 801         if (rc)
 802                 goto out2;
 803
 804         return 0;
 805 out2:
 806         unregister_pernet_subsys(&bareudp_net_ops);
 807 out1:
 808         return rc;
 809 }
 810 late_initcall(bareudp_init_module);
 811
 812 static void __exit bareudp_cleanup_module(void)
 813 {
 814         rtnl_link_unregister(&bareudp_link_ops);
 815         unregister_pernet_subsys(&bareudp_net_ops);
 816 }
 817 module_exit(bareudp_cleanup_module);
 818
 819 MODULE_ALIAS_RTNL_LINK("bareudp");
 820 MODULE_LICENSE("GPL");
 821 MODULE_AUTHOR("Martin Varghese <[email protected]>");
 822 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");