net/ipv4/ip_tunnel_core.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * Copyright (c) 2013 Nicira, Inc.
   4  */
   5
   6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   7
   8 #include <linux/types.h>
   9 #include <linux/kernel.h>
  10 #include <linux/skbuff.h>
  11 #include <linux/netdevice.h>
  12 #include <linux/in.h>
  13 #include <linux/if_arp.h>
  14 #include <linux/init.h>
  15 #include <linux/in6.h>
  16 #include <linux/inetdevice.h>
  17 #include <linux/netfilter_ipv4.h>
  18 #include <linux/etherdevice.h>
  19 #include <linux/if_ether.h>
  20 #include <linux/if_vlan.h>
  21 #include <linux/static_key.h>
  22
  23 #include <net/ip.h>
  24 #include <net/icmp.h>
  25 #include <net/protocol.h>
  26 #include <net/ip_tunnels.h>
  27 #include <net/ip6_tunnel.h>
  28 #include <net/arp.h>
  29 #include <net/checksum.h>
  30 #include <net/dsfield.h>
  31 #include <net/inet_ecn.h>
  32 #include <net/xfrm.h>
  33 #include <net/net_namespace.h>
  34 #include <net/netns/generic.h>
  35 #include <net/rtnetlink.h>
  36 #include <net/dst_metadata.h>
  37 #include <net/geneve.h>
  38
  39 const struct ip_tunnel_encap_ops __rcu *
  40                 iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
  41 EXPORT_SYMBOL(iptun_encaps);
  42
  43 const struct ip6_tnl_encap_ops __rcu *
  44                 ip6tun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
  45 EXPORT_SYMBOL(ip6tun_encaps);
  46
  47 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
  48                    __be32 src, __be32 dst, __u8 proto,
  49                    __u8 tos, __u8 ttl, __be16 df, bool xnet)
  50 {
  51         int pkt_len = skb->len - skb_inner_network_offset(skb);
  52         struct net *net = dev_net(rt->dst.dev);
  53         struct net_device *dev = skb->dev;
  54         struct iphdr *iph;
  55         int err;
  56
  57         skb_scrub_packet(skb, xnet);
  58
  59         skb_clear_hash_if_not_l4(skb);
  60         skb_dst_set(skb, &rt->dst);
  61         memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
  62
  63         /* Push down and install the IP header. */
  64         skb_push(skb, sizeof(struct iphdr));
  65         skb_reset_network_header(skb);
  66
  67         iph = ip_hdr(skb);
  68
  69         iph->version    =       4;
  70         iph->ihl        =       sizeof(struct iphdr) >> 2;
  71         iph->frag_off   =       ip_mtu_locked(&rt->dst) ? 0 : df;
  72         iph->protocol   =       proto;
  73         iph->tos        =       tos;
  74         iph->daddr      =       dst;
  75         iph->saddr      =       src;
  76         iph->ttl        =       ttl;
  77         __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
  78
  79         err = ip_local_out(net, sk, skb);
  80
  81         if (dev) {
  82                 if (unlikely(net_xmit_eval(err)))
  83                         pkt_len = 0;
  84                 iptunnel_xmit_stats(dev, pkt_len);
  85         }
  86 }
  87 EXPORT_SYMBOL_GPL(iptunnel_xmit);
  88
  89 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
  90                            __be16 inner_proto, bool raw_proto, bool xnet)
  91 {
  92         if (unlikely(!pskb_may_pull(skb, hdr_len)))
  93                 return -ENOMEM;
  94
  95         skb_pull_rcsum(skb, hdr_len);
  96
  97         if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
  98                 struct ethhdr *eh;
  99
 100                 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
 101                         return -ENOMEM;
 102
 103                 eh = (struct ethhdr *)skb->data;
 104                 if (likely(eth_proto_is_802_3(eh->h_proto)))
 105                         skb->protocol = eh->h_proto;
 106                 else
 107                         skb->protocol = htons(ETH_P_802_2);
 108
 109         } else {
 110                 skb->protocol = inner_proto;
 111         }
 112
 113         skb_clear_hash_if_not_l4(skb);
 114         __vlan_hwaccel_clear_tag(skb);
 115         skb_set_queue_mapping(skb, 0);
 116         skb_scrub_packet(skb, xnet);
 117
 118         return iptunnel_pull_offloads(skb);
 119 }
 120 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
 121
 122 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
 123                                              gfp_t flags)
 124 {
 125         struct metadata_dst *res;
 126         struct ip_tunnel_info *dst, *src;
 127
 128         if (!md || md->type != METADATA_IP_TUNNEL ||
 129             md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
 130                 return NULL;
 131
 132         src = &md->u.tun_info;
 133         res = metadata_dst_alloc(src->options_len, METADATA_IP_TUNNEL, flags);
 134         if (!res)
 135                 return NULL;
 136
 137         dst = &res->u.tun_info;
 138         dst->key.tun_id = src->key.tun_id;
 139         if (src->mode & IP_TUNNEL_INFO_IPV6)
 140                 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
 141                        sizeof(struct in6_addr));
 142         else
 143                 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
 144         dst->key.tun_flags = src->key.tun_flags;
 145         dst->mode = src->mode | IP_TUNNEL_INFO_TX;
 146         ip_tunnel_info_opts_set(dst, ip_tunnel_info_opts(src),
 147                                 src->options_len, 0);
 148
 149         return res;
 150 }
 151 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
 152
 153 int iptunnel_handle_offloads(struct sk_buff *skb,
 154                              int gso_type_mask)
 155 {
 156         int err;
 157
 158         if (likely(!skb->encapsulation)) {
 159                 skb_reset_inner_headers(skb);
 160                 skb->encapsulation = 1;
 161         }
 162
 163         if (skb_is_gso(skb)) {
 164                 err = skb_header_unclone(skb, GFP_ATOMIC);
 165                 if (unlikely(err))
 166                         return err;
 167                 skb_shinfo(skb)->gso_type |= gso_type_mask;
 168                 return 0;
 169         }
 170
 171         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 172                 skb->ip_summed = CHECKSUM_NONE;
 173                 /* We clear encapsulation here to prevent badly-written
 174                  * drivers potentially deciding to offload an inner checksum
 175                  * if we set CHECKSUM_PARTIAL on the outer header.
 176                  * This should go away when the drivers are all fixed.
 177                  */
 178                 skb->encapsulation = 0;
 179         }
 180
 181         return 0;
 182 }
 183 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
 184
 185 /* Often modified stats are per cpu, other are shared (netdev->stats) */
 186 void ip_tunnel_get_stats64(struct net_device *dev,
 187                            struct rtnl_link_stats64 *tot)
 188 {
 189         int i;
 190
 191         netdev_stats_to_stats64(tot, &dev->stats);
 192
 193         for_each_possible_cpu(i) {
 194                 const struct pcpu_sw_netstats *tstats =
 195                                                    per_cpu_ptr(dev->tstats, i);
 196                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
 197                 unsigned int start;
 198
 199                 do {
 200                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 201                         rx_packets = tstats->rx_packets;
 202                         tx_packets = tstats->tx_packets;
 203                         rx_bytes = tstats->rx_bytes;
 204                         tx_bytes = tstats->tx_bytes;
 205                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 206
 207                 tot->rx_packets += rx_packets;
 208                 tot->tx_packets += tx_packets;
 209                 tot->rx_bytes   += rx_bytes;
 210                 tot->tx_bytes   += tx_bytes;
 211         }
 212 }
 213 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
 214
 215 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
 216         [LWTUNNEL_IP_ID]        = { .type = NLA_U64 },
 217         [LWTUNNEL_IP_DST]       = { .type = NLA_U32 },
 218         [LWTUNNEL_IP_SRC]       = { .type = NLA_U32 },
 219         [LWTUNNEL_IP_TTL]       = { .type = NLA_U8 },
 220         [LWTUNNEL_IP_TOS]       = { .type = NLA_U8 },
 221         [LWTUNNEL_IP_FLAGS]     = { .type = NLA_U16 },
 222         [LWTUNNEL_IP_OPTS]      = { .type = NLA_NESTED },
 223 };
 224
 225 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
 226         [LWTUNNEL_IP_OPTS_GENEVE]       = { .type = NLA_NESTED },
 227 };
 228
 229 static const struct nla_policy
 230 geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
 231         [LWTUNNEL_IP_OPT_GENEVE_CLASS]  = { .type = NLA_U16 },
 232         [LWTUNNEL_IP_OPT_GENEVE_TYPE]   = { .type = NLA_U8 },
 233         [LWTUNNEL_IP_OPT_GENEVE_DATA]   = { .type = NLA_BINARY, .len = 128 },
 234 };
 235
 236 static int ip_tun_parse_opts_geneve(struct nlattr *attr,
 237                                     struct ip_tunnel_info *info,
 238                                     struct netlink_ext_ack *extack)
 239 {
 240         struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
 241         int data_len, err;
 242
 243         err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_OPT_GENEVE_MAX,
 244                                           attr, geneve_opt_policy, extack);
 245         if (err)
 246                 return err;
 247
 248         if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] ||
 249             !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] ||
 250             !tb[LWTUNNEL_IP_OPT_GENEVE_DATA])
 251                 return -EINVAL;
 252
 253         attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA];
 254         data_len = nla_len(attr);
 255         if (data_len % 4)
 256                 return -EINVAL;
 257
 258         if (info) {
 259                 struct geneve_opt *opt = ip_tunnel_info_opts(info);
 260
 261                 memcpy(opt->opt_data, nla_data(attr), data_len);
 262                 opt->length = data_len / 4;
 263                 attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS];
 264                 opt->opt_class = nla_get_be16(attr);
 265                 attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE];
 266                 opt->type = nla_get_u8(attr);
 267                 info->key.tun_flags |= TUNNEL_GENEVE_OPT;
 268         }
 269
 270         return sizeof(struct geneve_opt) + data_len;
 271 }
 272
 273 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
 274                              struct netlink_ext_ack *extack)
 275 {
 276         struct nlattr *tb[LWTUNNEL_IP_OPTS_MAX + 1];
 277         int err;
 278
 279         if (!attr)
 280                 return 0;
 281
 282         err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_OPTS_MAX, attr,
 283                                           ip_opts_policy, extack);
 284         if (err)
 285                 return err;
 286
 287         if (tb[LWTUNNEL_IP_OPTS_GENEVE])
 288                 err = ip_tun_parse_opts_geneve(tb[LWTUNNEL_IP_OPTS_GENEVE],
 289                                                info, extack);
 290         else
 291                 err = -EINVAL;
 292
 293         return err;
 294 }
 295
 296 static int ip_tun_get_optlen(struct nlattr *attr,
 297                              struct netlink_ext_ack *extack)
 298 {
 299         return ip_tun_parse_opts(attr, NULL, extack);
 300 }
 301
 302 static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info,
 303                            struct netlink_ext_ack *extack)
 304 {
 305         return ip_tun_parse_opts(attr, info, extack);
 306 }
 307
 308 static int ip_tun_build_state(struct nlattr *attr,
 309                               unsigned int family, const void *cfg,
 310                               struct lwtunnel_state **ts,
 311                               struct netlink_ext_ack *extack)
 312 {
 313         struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
 314         struct lwtunnel_state *new_state;
 315         struct ip_tunnel_info *tun_info;
 316         int err, opt_len;
 317
 318         err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr,
 319                                           ip_tun_policy, extack);
 320         if (err < 0)
 321                 return err;
 322
 323         opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack);
 324         if (opt_len < 0)
 325                 return opt_len;
 326
 327         new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
 328         if (!new_state)
 329                 return -ENOMEM;
 330
 331         new_state->type = LWTUNNEL_ENCAP_IP;
 332
 333         tun_info = lwt_tun_info(new_state);
 334
 335         err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack);
 336         if (err < 0) {
 337                 lwtstate_free(new_state);
 338                 return err;
 339         }
 340
 341 #ifdef CONFIG_DST_CACHE
 342         err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
 343         if (err) {
 344                 lwtstate_free(new_state);
 345                 return err;
 346         }
 347 #endif
 348
 349         if (tb[LWTUNNEL_IP_ID])
 350                 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
 351
 352         if (tb[LWTUNNEL_IP_DST])
 353                 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
 354
 355         if (tb[LWTUNNEL_IP_SRC])
 356                 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
 357
 358         if (tb[LWTUNNEL_IP_TTL])
 359                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
 360
 361         if (tb[LWTUNNEL_IP_TOS])
 362                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
 363
 364         if (tb[LWTUNNEL_IP_FLAGS])
 365                 tun_info->key.tun_flags |= nla_get_be16(tb[LWTUNNEL_IP_FLAGS]);
 366
 367         tun_info->mode = IP_TUNNEL_INFO_TX;
 368         tun_info->options_len = opt_len;
 369
 370         *ts = new_state;
 371
 372         return 0;
 373 }
 374
 375 static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
 376 {
 377 #ifdef CONFIG_DST_CACHE
 378         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
 379
 380         dst_cache_destroy(&tun_info->dst_cache);
 381 #endif
 382 }
 383
 384 static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
 385                                          struct ip_tunnel_info *tun_info)
 386 {
 387         struct geneve_opt *opt;
 388         struct nlattr *nest;
 389
 390         nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
 391         if (!nest)
 392                 return -ENOMEM;
 393
 394         opt = ip_tunnel_info_opts(tun_info);
 395         if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS, opt->opt_class) ||
 396             nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
 397             nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
 398                     opt->opt_data)) {
 399                 nla_nest_cancel(skb, nest);
 400                 return -ENOMEM;
 401         }
 402
 403         nla_nest_end(skb, nest);
 404         return 0;
 405 }
 406
 407 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
 408                                   struct ip_tunnel_info *tun_info)
 409 {
 410         struct nlattr *nest;
 411         int err = 0;
 412
 413         if (!(tun_info->key.tun_flags & TUNNEL_GENEVE_OPT))
 414                 return 0;
 415
 416         nest = nla_nest_start_noflag(skb, type);
 417         if (!nest)
 418                 return -ENOMEM;
 419
 420         if (tun_info->key.tun_flags & TUNNEL_GENEVE_OPT)
 421                 err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
 422
 423         if (err) {
 424                 nla_nest_cancel(skb, nest);
 425                 return err;
 426         }
 427
 428         nla_nest_end(skb, nest);
 429         return 0;
 430 }
 431
 432 static int ip_tun_fill_encap_info(struct sk_buff *skb,
 433                                   struct lwtunnel_state *lwtstate)
 434 {
 435         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
 436
 437         if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
 438                          LWTUNNEL_IP_PAD) ||
 439             nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
 440             nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
 441             nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
 442             nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
 443             nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags) ||
 444             ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info))
 445                 return -ENOMEM;
 446
 447         return 0;
 448 }
 449
 450 static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
 451 {
 452         int opt_len;
 453
 454         if (!(info->key.tun_flags & TUNNEL_GENEVE_OPT))
 455                 return 0;
 456
 457         opt_len = nla_total_size(0);            /* LWTUNNEL_IP_OPTS */
 458         if (info->key.tun_flags & TUNNEL_GENEVE_OPT) {
 459                 struct geneve_opt *opt = ip_tunnel_info_opts(info);
 460
 461                 opt_len += nla_total_size(0)    /* LWTUNNEL_IP_OPTS_GENEVE */
 462                            + nla_total_size(2)  /* OPT_GENEVE_CLASS */
 463                            + nla_total_size(1)  /* OPT_GENEVE_TYPE */
 464                            + nla_total_size(opt->length * 4);
 465                                                 /* OPT_GENEVE_DATA */
 466         }
 467
 468         return opt_len;
 469 }
 470
 471 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
 472 {
 473         return nla_total_size_64bit(8)  /* LWTUNNEL_IP_ID */
 474                 + nla_total_size(4)     /* LWTUNNEL_IP_DST */
 475                 + nla_total_size(4)     /* LWTUNNEL_IP_SRC */
 476                 + nla_total_size(1)     /* LWTUNNEL_IP_TOS */
 477                 + nla_total_size(1)     /* LWTUNNEL_IP_TTL */
 478                 + nla_total_size(2)     /* LWTUNNEL_IP_FLAGS */
 479                 + ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
 480                                         /* LWTUNNEL_IP_OPTS */
 481 }
 482
 483 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
 484 {
 485         struct ip_tunnel_info *info_a = lwt_tun_info(a);
 486         struct ip_tunnel_info *info_b = lwt_tun_info(b);
 487
 488         return memcmp(info_a, info_b, sizeof(info_a->key)) ||
 489                info_a->mode != info_b->mode ||
 490                info_a->options_len != info_b->options_len ||
 491                memcmp(ip_tunnel_info_opts(info_a),
 492                       ip_tunnel_info_opts(info_b), info_a->options_len);
 493 }
 494
 495 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
 496         .build_state = ip_tun_build_state,
 497         .destroy_state = ip_tun_destroy_state,
 498         .fill_encap = ip_tun_fill_encap_info,
 499         .get_encap_size = ip_tun_encap_nlsize,
 500         .cmp_encap = ip_tun_cmp_encap,
 501         .owner = THIS_MODULE,
 502 };
 503
 504 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
 505         [LWTUNNEL_IP6_ID]               = { .type = NLA_U64 },
 506         [LWTUNNEL_IP6_DST]              = { .len = sizeof(struct in6_addr) },
 507         [LWTUNNEL_IP6_SRC]              = { .len = sizeof(struct in6_addr) },
 508         [LWTUNNEL_IP6_HOPLIMIT]         = { .type = NLA_U8 },
 509         [LWTUNNEL_IP6_TC]               = { .type = NLA_U8 },
 510         [LWTUNNEL_IP6_FLAGS]            = { .type = NLA_U16 },
 511 };
 512
 513 static int ip6_tun_build_state(struct nlattr *attr,
 514                                unsigned int family, const void *cfg,
 515                                struct lwtunnel_state **ts,
 516                                struct netlink_ext_ack *extack)
 517 {
 518         struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
 519         struct lwtunnel_state *new_state;
 520         struct ip_tunnel_info *tun_info;
 521         int err, opt_len;
 522
 523         err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr,
 524                                           ip6_tun_policy, extack);
 525         if (err < 0)
 526                 return err;
 527
 528         opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack);
 529         if (opt_len < 0)
 530                 return opt_len;
 531
 532         new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
 533         if (!new_state)
 534                 return -ENOMEM;
 535
 536         new_state->type = LWTUNNEL_ENCAP_IP6;
 537
 538         tun_info = lwt_tun_info(new_state);
 539
 540         err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack);
 541         if (err < 0) {
 542                 lwtstate_free(new_state);
 543                 return err;
 544         }
 545
 546         if (tb[LWTUNNEL_IP6_ID])
 547                 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
 548
 549         if (tb[LWTUNNEL_IP6_DST])
 550                 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
 551
 552         if (tb[LWTUNNEL_IP6_SRC])
 553                 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
 554
 555         if (tb[LWTUNNEL_IP6_HOPLIMIT])
 556                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
 557
 558         if (tb[LWTUNNEL_IP6_TC])
 559                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
 560
 561         if (tb[LWTUNNEL_IP6_FLAGS])
 562                 tun_info->key.tun_flags |= nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
 563
 564         tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
 565         tun_info->options_len = opt_len;
 566
 567         *ts = new_state;
 568
 569         return 0;
 570 }
 571
 572 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
 573                                    struct lwtunnel_state *lwtstate)
 574 {
 575         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
 576
 577         if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
 578                          LWTUNNEL_IP6_PAD) ||
 579             nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
 580             nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
 581             nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
 582             nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
 583             nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags) ||
 584             ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info))
 585                 return -ENOMEM;
 586
 587         return 0;
 588 }
 589
 590 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
 591 {
 592         return nla_total_size_64bit(8)  /* LWTUNNEL_IP6_ID */
 593                 + nla_total_size(16)    /* LWTUNNEL_IP6_DST */
 594                 + nla_total_size(16)    /* LWTUNNEL_IP6_SRC */
 595                 + nla_total_size(1)     /* LWTUNNEL_IP6_HOPLIMIT */
 596                 + nla_total_size(1)     /* LWTUNNEL_IP6_TC */
 597                 + nla_total_size(2)     /* LWTUNNEL_IP6_FLAGS */
 598                 + ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
 599                                         /* LWTUNNEL_IP6_OPTS */
 600 }
 601
 602 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
 603         .build_state = ip6_tun_build_state,
 604         .fill_encap = ip6_tun_fill_encap_info,
 605         .get_encap_size = ip6_tun_encap_nlsize,
 606         .cmp_encap = ip_tun_cmp_encap,
 607         .owner = THIS_MODULE,
 608 };
 609
 610 void __init ip_tunnel_core_init(void)
 611 {
 612         /* If you land here, make sure whether increasing ip_tunnel_info's
 613          * options_len is a reasonable choice with its usage in front ends
 614          * (f.e., it's part of flow keys, etc).
 615          */
 616         BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
 617
 618         lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
 619         lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
 620 }
 621
 622 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
 623 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
 624
 625 void ip_tunnel_need_metadata(void)
 626 {
 627         static_branch_inc(&ip_tunnel_metadata_cnt);
 628 }
 629 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
 630
 631 void ip_tunnel_unneed_metadata(void)
 632 {
 633         static_branch_dec(&ip_tunnel_metadata_cnt);
 634 }
 635 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);