net/ipv6/ip6_gre.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      GRE over IPv6 protocol decoder.
   4  *
   5  *      Authors: Dmitry Kozlov (xeb@mail.ru)
   6  */
   7
   8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   9
  10 #include <linux/capability.h>
  11 #include <linux/module.h>
  12 #include <linux/types.h>
  13 #include <linux/kernel.h>
  14 #include <linux/slab.h>
  15 #include <linux/uaccess.h>
  16 #include <linux/skbuff.h>
  17 #include <linux/netdevice.h>
  18 #include <linux/in.h>
  19 #include <linux/tcp.h>
  20 #include <linux/udp.h>
  21 #include <linux/if_arp.h>
  22 #include <linux/init.h>
  23 #include <linux/in6.h>
  24 #include <linux/inetdevice.h>
  25 #include <linux/igmp.h>
  26 #include <linux/netfilter_ipv4.h>
  27 #include <linux/etherdevice.h>
  28 #include <linux/if_ether.h>
  29 #include <linux/hash.h>
  30 #include <linux/if_tunnel.h>
  31 #include <linux/ip6_tunnel.h>
  32
  33 #include <net/sock.h>
  34 #include <net/ip.h>
  35 #include <net/ip_tunnels.h>
  36 #include <net/icmp.h>
  37 #include <net/protocol.h>
  38 #include <net/addrconf.h>
  39 #include <net/arp.h>
  40 #include <net/checksum.h>
  41 #include <net/dsfield.h>
  42 #include <net/inet_ecn.h>
  43 #include <net/xfrm.h>
  44 #include <net/net_namespace.h>
  45 #include <net/netns/generic.h>
  46 #include <net/rtnetlink.h>
  47
  48 #include <net/ipv6.h>
  49 #include <net/ip6_fib.h>
  50 #include <net/ip6_route.h>
  51 #include <net/ip6_tunnel.h>
  52 #include <net/gre.h>
  53 #include <net/erspan.h>
  54 #include <net/dst_metadata.h>
  55
  56
  57 static bool log_ecn_error = true;
  58 module_param(log_ecn_error, bool, 0644);
  59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  60
  61 #define IP6_GRE_HASH_SIZE_SHIFT  5
  62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
  63
  64 static unsigned int ip6gre_net_id __read_mostly;
  65 struct ip6gre_net {
  66         struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
  67
  68         struct ip6_tnl __rcu *collect_md_tun;
  69         struct ip6_tnl __rcu *collect_md_tun_erspan;
  70         struct net_device *fb_tunnel_dev;
  71 };
  72
  73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
  75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
  76 static int ip6gre_tunnel_init(struct net_device *dev);
  77 static void ip6gre_tunnel_setup(struct net_device *dev);
  78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  81
  82 /* Tunnel hash table */
  83
  84 /*
  85    4 hash tables:
  86
  87    3: (remote,local)
  88    2: (remote,*)
  89    1: (*,local)
  90    0: (*,*)
  91
  92    We require exact key match i.e. if a key is present in packet
  93    it will match only tunnel with the same key; if it is not present,
  94    it will match only keyless tunnel.
  95
  96    All keysless packets, if not matched configured keyless tunnels
  97    will match fallback tunnel.
  98  */
  99
 100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
 101 static u32 HASH_ADDR(const struct in6_addr *addr)
 102 {
 103         u32 hash = ipv6_addr_hash(addr);
 104
 105         return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
 106 }
 107
 108 #define tunnels_r_l     tunnels[3]
 109 #define tunnels_r       tunnels[2]
 110 #define tunnels_l       tunnels[1]
 111 #define tunnels_wc      tunnels[0]
 112
 113 /* Given src, dst and key, find appropriate for input tunnel. */
 114
 115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 116                 const struct in6_addr *remote, const struct in6_addr *local,
 117                 __be32 key, __be16 gre_proto)
 118 {
 119         struct net *net = dev_net(dev);
 120         int link = dev->ifindex;
 121         unsigned int h0 = HASH_ADDR(remote);
 122         unsigned int h1 = HASH_KEY(key);
 123         struct ip6_tnl *t, *cand = NULL;
 124         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 125         int dev_type = (gre_proto == htons(ETH_P_TEB) ||
 126                         gre_proto == htons(ETH_P_ERSPAN) ||
 127                         gre_proto == htons(ETH_P_ERSPAN2)) ?
 128                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 129         int score, cand_score = 4;
 130         struct net_device *ndev;
 131
 132         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 133                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 134                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 135                     key != t->parms.i_key ||
 136                     !(t->dev->flags & IFF_UP))
 137                         continue;
 138
 139                 if (t->dev->type != ARPHRD_IP6GRE &&
 140                     t->dev->type != dev_type)
 141                         continue;
 142
 143                 score = 0;
 144                 if (t->parms.link != link)
 145                         score |= 1;
 146                 if (t->dev->type != dev_type)
 147                         score |= 2;
 148                 if (score == 0)
 149                         return t;
 150
 151                 if (score < cand_score) {
 152                         cand = t;
 153                         cand_score = score;
 154                 }
 155         }
 156
 157         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 158                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 159                     key != t->parms.i_key ||
 160                     !(t->dev->flags & IFF_UP))
 161                         continue;
 162
 163                 if (t->dev->type != ARPHRD_IP6GRE &&
 164                     t->dev->type != dev_type)
 165                         continue;
 166
 167                 score = 0;
 168                 if (t->parms.link != link)
 169                         score |= 1;
 170                 if (t->dev->type != dev_type)
 171                         score |= 2;
 172                 if (score == 0)
 173                         return t;
 174
 175                 if (score < cand_score) {
 176                         cand = t;
 177                         cand_score = score;
 178                 }
 179         }
 180
 181         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 182                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 183                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 184                                  !ipv6_addr_is_multicast(local))) ||
 185                     key != t->parms.i_key ||
 186                     !(t->dev->flags & IFF_UP))
 187                         continue;
 188
 189                 if (t->dev->type != ARPHRD_IP6GRE &&
 190                     t->dev->type != dev_type)
 191                         continue;
 192
 193                 score = 0;
 194                 if (t->parms.link != link)
 195                         score |= 1;
 196                 if (t->dev->type != dev_type)
 197                         score |= 2;
 198                 if (score == 0)
 199                         return t;
 200
 201                 if (score < cand_score) {
 202                         cand = t;
 203                         cand_score = score;
 204                 }
 205         }
 206
 207         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 208                 if (t->parms.i_key != key ||
 209                     !(t->dev->flags & IFF_UP))
 210                         continue;
 211
 212                 if (t->dev->type != ARPHRD_IP6GRE &&
 213                     t->dev->type != dev_type)
 214                         continue;
 215
 216                 score = 0;
 217                 if (t->parms.link != link)
 218                         score |= 1;
 219                 if (t->dev->type != dev_type)
 220                         score |= 2;
 221                 if (score == 0)
 222                         return t;
 223
 224                 if (score < cand_score) {
 225                         cand = t;
 226                         cand_score = score;
 227                 }
 228         }
 229
 230         if (cand)
 231                 return cand;
 232
 233         if (gre_proto == htons(ETH_P_ERSPAN) ||
 234             gre_proto == htons(ETH_P_ERSPAN2))
 235                 t = rcu_dereference(ign->collect_md_tun_erspan);
 236         else
 237                 t = rcu_dereference(ign->collect_md_tun);
 238
 239         if (t && t->dev->flags & IFF_UP)
 240                 return t;
 241
 242         ndev = READ_ONCE(ign->fb_tunnel_dev);
 243         if (ndev && ndev->flags & IFF_UP)
 244                 return netdev_priv(ndev);
 245
 246         return NULL;
 247 }
 248
 249 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 250                 const struct __ip6_tnl_parm *p)
 251 {
 252         const struct in6_addr *remote = &p->raddr;
 253         const struct in6_addr *local = &p->laddr;
 254         unsigned int h = HASH_KEY(p->i_key);
 255         int prio = 0;
 256
 257         if (!ipv6_addr_any(local))
 258                 prio |= 1;
 259         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 260                 prio |= 2;
 261                 h ^= HASH_ADDR(remote);
 262         }
 263
 264         return &ign->tunnels[prio][h];
 265 }
 266
 267 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
 268 {
 269         if (t->parms.collect_md)
 270                 rcu_assign_pointer(ign->collect_md_tun, t);
 271 }
 272
 273 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
 274 {
 275         if (t->parms.collect_md)
 276                 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
 277 }
 278
 279 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
 280 {
 281         if (t->parms.collect_md)
 282                 rcu_assign_pointer(ign->collect_md_tun, NULL);
 283 }
 284
 285 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
 286                                        struct ip6_tnl *t)
 287 {
 288         if (t->parms.collect_md)
 289                 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
 290 }
 291
 292 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 293                 const struct ip6_tnl *t)
 294 {
 295         return __ip6gre_bucket(ign, &t->parms);
 296 }
 297
 298 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 299 {
 300         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 301
 302         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 303         rcu_assign_pointer(*tp, t);
 304 }
 305
 306 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 307 {
 308         struct ip6_tnl __rcu **tp;
 309         struct ip6_tnl *iter;
 310
 311         for (tp = ip6gre_bucket(ign, t);
 312              (iter = rtnl_dereference(*tp)) != NULL;
 313              tp = &iter->next) {
 314                 if (t == iter) {
 315                         rcu_assign_pointer(*tp, t->next);
 316                         break;
 317                 }
 318         }
 319 }
 320
 321 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 322                                            const struct __ip6_tnl_parm *parms,
 323                                            int type)
 324 {
 325         const struct in6_addr *remote = &parms->raddr;
 326         const struct in6_addr *local = &parms->laddr;
 327         __be32 key = parms->i_key;
 328         int link = parms->link;
 329         struct ip6_tnl *t;
 330         struct ip6_tnl __rcu **tp;
 331         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 332
 333         for (tp = __ip6gre_bucket(ign, parms);
 334              (t = rtnl_dereference(*tp)) != NULL;
 335              tp = &t->next)
 336                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 337                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 338                     key == t->parms.i_key &&
 339                     link == t->parms.link &&
 340                     type == t->dev->type)
 341                         break;
 342
 343         return t;
 344 }
 345
 346 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 347                 const struct __ip6_tnl_parm *parms, int create)
 348 {
 349         struct ip6_tnl *t, *nt;
 350         struct net_device *dev;
 351         char name[IFNAMSIZ];
 352         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 353
 354         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 355         if (t && create)
 356                 return NULL;
 357         if (t || !create)
 358                 return t;
 359
 360         if (parms->name[0]) {
 361                 if (!dev_valid_name(parms->name))
 362                         return NULL;
 363                 strscpy(name, parms->name, IFNAMSIZ);
 364         } else {
 365                 strcpy(name, "ip6gre%d");
 366         }
 367         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 368                            ip6gre_tunnel_setup);
 369         if (!dev)
 370                 return NULL;
 371
 372         dev_net_set(dev, net);
 373
 374         nt = netdev_priv(dev);
 375         nt->parms = *parms;
 376         dev->rtnl_link_ops = &ip6gre_link_ops;
 377
 378         nt->dev = dev;
 379         nt->net = dev_net(dev);
 380
 381         if (register_netdevice(dev) < 0)
 382                 goto failed_free;
 383
 384         ip6gre_tnl_link_config(nt, 1);
 385         ip6gre_tunnel_link(ign, nt);
 386         return nt;
 387
 388 failed_free:
 389         free_netdev(dev);
 390         return NULL;
 391 }
 392
 393 static void ip6erspan_tunnel_uninit(struct net_device *dev)
 394 {
 395         struct ip6_tnl *t = netdev_priv(dev);
 396         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 397
 398         ip6erspan_tunnel_unlink_md(ign, t);
 399         ip6gre_tunnel_unlink(ign, t);
 400         dst_cache_reset(&t->dst_cache);
 401         netdev_put(dev, &t->dev_tracker);
 402 }
 403
 404 static void ip6gre_tunnel_uninit(struct net_device *dev)
 405 {
 406         struct ip6_tnl *t = netdev_priv(dev);
 407         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 408
 409         ip6gre_tunnel_unlink_md(ign, t);
 410         ip6gre_tunnel_unlink(ign, t);
 411         if (ign->fb_tunnel_dev == dev)
 412                 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
 413         dst_cache_reset(&t->dst_cache);
 414         netdev_put(dev, &t->dev_tracker);
 415 }
 416
 417
 418 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 419                        u8 type, u8 code, int offset, __be32 info)
 420 {
 421         struct net *net = dev_net(skb->dev);
 422         const struct ipv6hdr *ipv6h;
 423         struct tnl_ptk_info tpi;
 424         struct ip6_tnl *t;
 425
 426         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
 427                              offset) < 0)
 428                 return -EINVAL;
 429
 430         ipv6h = (const struct ipv6hdr *)skb->data;
 431         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 432                                  tpi.key, tpi.proto);
 433         if (!t)
 434                 return -ENOENT;
 435
 436         switch (type) {
 437         case ICMPV6_DEST_UNREACH:
 438                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 439                                     t->parms.name);
 440                 if (code != ICMPV6_PORT_UNREACH)
 441                         break;
 442                 return 0;
 443         case ICMPV6_TIME_EXCEED:
 444                 if (code == ICMPV6_EXC_HOPLIMIT) {
 445                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 446                                             t->parms.name);
 447                         break;
 448                 }
 449                 return 0;
 450         case ICMPV6_PARAMPROB: {
 451                 struct ipv6_tlv_tnl_enc_lim *tel;
 452                 __u32 teli;
 453
 454                 teli = 0;
 455                 if (code == ICMPV6_HDR_FIELD)
 456                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 457
 458                 if (teli && teli == be32_to_cpu(info) - 2) {
 459                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 460                         if (tel->encap_limit == 0) {
 461                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 462                                                     t->parms.name);
 463                         }
 464                 } else {
 465                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 466                                             t->parms.name);
 467                 }
 468                 return 0;
 469         }
 470         case ICMPV6_PKT_TOOBIG:
 471                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
 472                 return 0;
 473         case NDISC_REDIRECT:
 474                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 475                              sock_net_uid(net, NULL));
 476                 return 0;
 477         }
 478
 479         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 480                 t->err_count++;
 481         else
 482                 t->err_count = 1;
 483         t->err_time = jiffies;
 484
 485         return 0;
 486 }
 487
 488 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 489 {
 490         const struct ipv6hdr *ipv6h;
 491         struct ip6_tnl *tunnel;
 492
 493         ipv6h = ipv6_hdr(skb);
 494         tunnel = ip6gre_tunnel_lookup(skb->dev,
 495                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 496                                       tpi->proto);
 497         if (tunnel) {
 498                 if (tunnel->parms.collect_md) {
 499                         struct metadata_dst *tun_dst;
 500                         __be64 tun_id;
 501                         __be16 flags;
 502
 503                         flags = tpi->flags;
 504                         tun_id = key32_to_tunnel_id(tpi->key);
 505
 506                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
 507                         if (!tun_dst)
 508                                 return PACKET_REJECT;
 509
 510                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
 511                 } else {
 512                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 513                 }
 514
 515                 return PACKET_RCVD;
 516         }
 517
 518         return PACKET_REJECT;
 519 }
 520
 521 static int ip6erspan_rcv(struct sk_buff *skb,
 522                          struct tnl_ptk_info *tpi,
 523                          int gre_hdr_len)
 524 {
 525         struct erspan_base_hdr *ershdr;
 526         const struct ipv6hdr *ipv6h;
 527         struct erspan_md2 *md2;
 528         struct ip6_tnl *tunnel;
 529         u8 ver;
 530
 531         if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
 532                 return PACKET_REJECT;
 533
 534         ipv6h = ipv6_hdr(skb);
 535         ershdr = (struct erspan_base_hdr *)skb->data;
 536         ver = ershdr->ver;
 537
 538         tunnel = ip6gre_tunnel_lookup(skb->dev,
 539                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 540                                       tpi->proto);
 541         if (tunnel) {
 542                 int len = erspan_hdr_len(ver);
 543
 544                 if (unlikely(!pskb_may_pull(skb, len)))
 545                         return PACKET_REJECT;
 546
 547                 if (__iptunnel_pull_header(skb, len,
 548                                            htons(ETH_P_TEB),
 549                                            false, false) < 0)
 550                         return PACKET_REJECT;
 551
 552                 if (tunnel->parms.collect_md) {
 553                         struct erspan_metadata *pkt_md, *md;
 554                         struct metadata_dst *tun_dst;
 555                         struct ip_tunnel_info *info;
 556                         unsigned char *gh;
 557                         __be64 tun_id;
 558                         __be16 flags;
 559
 560                         tpi->flags |= TUNNEL_KEY;
 561                         flags = tpi->flags;
 562                         tun_id = key32_to_tunnel_id(tpi->key);
 563
 564                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
 565                                                   sizeof(*md));
 566                         if (!tun_dst)
 567                                 return PACKET_REJECT;
 568
 569                         /* skb can be uncloned in __iptunnel_pull_header, so
 570                          * old pkt_md is no longer valid and we need to reset
 571                          * it
 572                          */
 573                         gh = skb_network_header(skb) +
 574                              skb_network_header_len(skb);
 575                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
 576                                                             sizeof(*ershdr));
 577                         info = &tun_dst->u.tun_info;
 578                         md = ip_tunnel_info_opts(info);
 579                         md->version = ver;
 580                         md2 = &md->u.md2;
 581                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
 582                                                        ERSPAN_V2_MDSIZE);
 583                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
 584                         info->options_len = sizeof(*md);
 585
 586                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
 587
 588                 } else {
 589                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 590                 }
 591
 592                 return PACKET_RCVD;
 593         }
 594
 595         return PACKET_REJECT;
 596 }
 597
 598 static int gre_rcv(struct sk_buff *skb)
 599 {
 600         struct tnl_ptk_info tpi;
 601         bool csum_err = false;
 602         int hdr_len;
 603
 604         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
 605         if (hdr_len < 0)
 606                 goto drop;
 607
 608         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
 609                 goto drop;
 610
 611         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
 612                      tpi.proto == htons(ETH_P_ERSPAN2))) {
 613                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
 614                         return 0;
 615                 goto out;
 616         }
 617
 618         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
 619                 return 0;
 620
 621 out:
 622         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 623 drop:
 624         kfree_skb(skb);
 625         return 0;
 626 }
 627
 628 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
 629 {
 630         return iptunnel_handle_offloads(skb,
 631                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 632 }
 633
 634 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
 635                                      struct net_device *dev,
 636                                      struct flowi6 *fl6, __u8 *dsfield,
 637                                      int *encap_limit)
 638 {
 639         const struct iphdr *iph = ip_hdr(skb);
 640         struct ip6_tnl *t = netdev_priv(dev);
 641
 642         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 643                 *encap_limit = t->parms.encap_limit;
 644
 645         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 646
 647         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 648                 *dsfield = ipv4_get_dsfield(iph);
 649         else
 650                 *dsfield = ip6_tclass(t->parms.flowinfo);
 651
 652         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 653                 fl6->flowi6_mark = skb->mark;
 654         else
 655                 fl6->flowi6_mark = t->parms.fwmark;
 656
 657         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 658 }
 659
 660 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
 661                                     struct net_device *dev,
 662                                     struct flowi6 *fl6, __u8 *dsfield,
 663                                     int *encap_limit)
 664 {
 665         struct ipv6hdr *ipv6h;
 666         struct ip6_tnl *t = netdev_priv(dev);
 667         __u16 offset;
 668
 669         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 670         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 671         ipv6h = ipv6_hdr(skb);
 672
 673         if (offset > 0) {
 674                 struct ipv6_tlv_tnl_enc_lim *tel;
 675
 676                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 677                 if (tel->encap_limit == 0) {
 678                         icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
 679                                         ICMPV6_HDR_FIELD, offset + 2);
 680                         return -1;
 681                 }
 682                 *encap_limit = tel->encap_limit - 1;
 683         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
 684                 *encap_limit = t->parms.encap_limit;
 685         }
 686
 687         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 688
 689         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 690                 *dsfield = ipv6_get_dsfield(ipv6h);
 691         else
 692                 *dsfield = ip6_tclass(t->parms.flowinfo);
 693
 694         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 695                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
 696
 697         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 698                 fl6->flowi6_mark = skb->mark;
 699         else
 700                 fl6->flowi6_mark = t->parms.fwmark;
 701
 702         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 703
 704         return 0;
 705 }
 706
 707 static int prepare_ip6gre_xmit_other(struct sk_buff *skb,
 708                                      struct net_device *dev,
 709                                      struct flowi6 *fl6, __u8 *dsfield,
 710                                      int *encap_limit)
 711 {
 712         struct ip6_tnl *t = netdev_priv(dev);
 713
 714         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 715                 *encap_limit = t->parms.encap_limit;
 716
 717         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 718
 719         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 720                 *dsfield = 0;
 721         else
 722                 *dsfield = ip6_tclass(t->parms.flowinfo);
 723
 724         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 725                 fl6->flowi6_mark = skb->mark;
 726         else
 727                 fl6->flowi6_mark = t->parms.fwmark;
 728
 729         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 730
 731         return 0;
 732 }
 733
 734 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
 735 {
 736         struct ip_tunnel_info *tun_info;
 737
 738         tun_info = skb_tunnel_info(skb);
 739         if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
 740                 return ERR_PTR(-EINVAL);
 741
 742         return tun_info;
 743 }
 744
 745 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
 746                                struct net_device *dev, __u8 dsfield,
 747                                struct flowi6 *fl6, int encap_limit,
 748                                __u32 *pmtu, __be16 proto)
 749 {
 750         struct ip6_tnl *tunnel = netdev_priv(dev);
 751         __be16 protocol;
 752         __be16 flags;
 753
 754         if (dev->type == ARPHRD_ETHER)
 755                 IPCB(skb)->flags = 0;
 756
 757         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
 758                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
 759         else
 760                 fl6->daddr = tunnel->parms.raddr;
 761
 762         /* Push GRE header. */
 763         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
 764
 765         if (tunnel->parms.collect_md) {
 766                 struct ip_tunnel_info *tun_info;
 767                 const struct ip_tunnel_key *key;
 768                 int tun_hlen;
 769
 770                 tun_info = skb_tunnel_info_txcheck(skb);
 771                 if (IS_ERR(tun_info) ||
 772                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
 773                         return -EINVAL;
 774
 775                 key = &tun_info->key;
 776                 memset(fl6, 0, sizeof(*fl6));
 777                 fl6->flowi6_proto = IPPROTO_GRE;
 778                 fl6->daddr = key->u.ipv6.dst;
 779                 fl6->flowlabel = key->label;
 780                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 781                 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
 782
 783                 dsfield = key->tos;
 784                 flags = key->tun_flags &
 785                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
 786                 tun_hlen = gre_calc_hlen(flags);
 787
 788                 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
 789                         return -ENOMEM;
 790
 791                 gre_build_header(skb, tun_hlen,
 792                                  flags, protocol,
 793                                  tunnel_id_to_key32(tun_info->key.tun_id),
 794                                  (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
 795                                                       : 0);
 796
 797         } else {
 798                 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
 799                         return -ENOMEM;
 800
 801                 flags = tunnel->parms.o_flags;
 802
 803                 gre_build_header(skb, tunnel->tun_hlen, flags,
 804                                  protocol, tunnel->parms.o_key,
 805                                  (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
 806                                                       : 0);
 807         }
 808
 809         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
 810                             NEXTHDR_GRE);
 811 }
 812
 813 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 814 {
 815         struct ip6_tnl *t = netdev_priv(dev);
 816         int encap_limit = -1;
 817         struct flowi6 fl6;
 818         __u8 dsfield = 0;
 819         __u32 mtu;
 820         int err;
 821
 822         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 823
 824         if (!t->parms.collect_md)
 825                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
 826                                          &dsfield, &encap_limit);
 827
 828         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 829         if (err)
 830                 return -1;
 831
 832         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 833                           skb->protocol);
 834         if (err != 0) {
 835                 /* XXX: send ICMP error even if DF is not set. */
 836                 if (err == -EMSGSIZE)
 837                         icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 838                                       htonl(mtu));
 839                 return -1;
 840         }
 841
 842         return 0;
 843 }
 844
 845 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 846 {
 847         struct ip6_tnl *t = netdev_priv(dev);
 848         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 849         int encap_limit = -1;
 850         struct flowi6 fl6;
 851         __u8 dsfield = 0;
 852         __u32 mtu;
 853         int err;
 854
 855         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 856                 return -1;
 857
 858         if (!t->parms.collect_md &&
 859             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
 860                 return -1;
 861
 862         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
 863                 return -1;
 864
 865         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
 866                           &mtu, skb->protocol);
 867         if (err != 0) {
 868                 if (err == -EMSGSIZE)
 869                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 870                 return -1;
 871         }
 872
 873         return 0;
 874 }
 875
 876 /**
 877  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
 878  *   @t: the outgoing tunnel device
 879  *   @hdr: IPv6 header from the incoming packet
 880  *
 881  * Description:
 882  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 883  *   doesn't match source of incoming packet.
 884  *
 885  * Return:
 886  *   1 if conflict,
 887  *   0 else
 888  **/
 889
 890 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 891         const struct ipv6hdr *hdr)
 892 {
 893         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 894 }
 895
 896 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 897 {
 898         struct ip6_tnl *t = netdev_priv(dev);
 899         int encap_limit = -1;
 900         struct flowi6 fl6;
 901         __u8 dsfield = 0;
 902         __u32 mtu;
 903         int err;
 904
 905         if (!t->parms.collect_md &&
 906             prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit))
 907                 return -1;
 908
 909         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 910         if (err)
 911                 return err;
 912         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol);
 913
 914         return err;
 915 }
 916
 917 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 918         struct net_device *dev)
 919 {
 920         struct ip6_tnl *t = netdev_priv(dev);
 921         struct net_device_stats *stats = &t->dev->stats;
 922         __be16 payload_protocol;
 923         int ret;
 924
 925         if (!pskb_inet_may_pull(skb))
 926                 goto tx_err;
 927
 928         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 929                 goto tx_err;
 930
 931         payload_protocol = skb_protocol(skb, true);
 932         switch (payload_protocol) {
 933         case htons(ETH_P_IP):
 934                 ret = ip6gre_xmit_ipv4(skb, dev);
 935                 break;
 936         case htons(ETH_P_IPV6):
 937                 ret = ip6gre_xmit_ipv6(skb, dev);
 938                 break;
 939         default:
 940                 ret = ip6gre_xmit_other(skb, dev);
 941                 break;
 942         }
 943
 944         if (ret < 0)
 945                 goto tx_err;
 946
 947         return NETDEV_TX_OK;
 948
 949 tx_err:
 950         if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
 951                 stats->tx_errors++;
 952         stats->tx_dropped++;
 953         kfree_skb(skb);
 954         return NETDEV_TX_OK;
 955 }
 956
 957 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
 958                                          struct net_device *dev)
 959 {
 960         struct ip_tunnel_info *tun_info = NULL;
 961         struct ip6_tnl *t = netdev_priv(dev);
 962         struct dst_entry *dst = skb_dst(skb);
 963         struct net_device_stats *stats;
 964         bool truncate = false;
 965         int encap_limit = -1;
 966         __u8 dsfield = false;
 967         struct flowi6 fl6;
 968         int err = -EINVAL;
 969         __be16 proto;
 970         __u32 mtu;
 971         int nhoff;
 972
 973         if (!pskb_inet_may_pull(skb))
 974                 goto tx_err;
 975
 976         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 977                 goto tx_err;
 978
 979         if (gre_handle_offloads(skb, false))
 980                 goto tx_err;
 981
 982         if (skb->len > dev->mtu + dev->hard_header_len) {
 983                 if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
 984                         goto tx_err;
 985                 truncate = true;
 986         }
 987
 988         nhoff = skb_network_offset(skb);
 989         if (skb->protocol == htons(ETH_P_IP) &&
 990             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
 991                 truncate = true;
 992
 993         if (skb->protocol == htons(ETH_P_IPV6)) {
 994                 int thoff;
 995
 996                 if (skb_transport_header_was_set(skb))
 997                         thoff = skb_transport_offset(skb);
 998                 else
 999                         thoff = nhoff + sizeof(struct ipv6hdr);
1000                 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
1001                         truncate = true;
1002         }
1003
1004         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
1005                 goto tx_err;
1006
1007         t->parms.o_flags &= ~TUNNEL_KEY;
1008         IPCB(skb)->flags = 0;
1009
1010         /* For collect_md mode, derive fl6 from the tunnel key,
1011          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
1012          */
1013         if (t->parms.collect_md) {
1014                 const struct ip_tunnel_key *key;
1015                 struct erspan_metadata *md;
1016                 __be32 tun_id;
1017
1018                 tun_info = skb_tunnel_info_txcheck(skb);
1019                 if (IS_ERR(tun_info) ||
1020                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
1021                         goto tx_err;
1022
1023                 key = &tun_info->key;
1024                 memset(&fl6, 0, sizeof(fl6));
1025                 fl6.flowi6_proto = IPPROTO_GRE;
1026                 fl6.daddr = key->u.ipv6.dst;
1027                 fl6.flowlabel = key->label;
1028                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1029                 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
1030
1031                 dsfield = key->tos;
1032                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1033                         goto tx_err;
1034                 if (tun_info->options_len < sizeof(*md))
1035                         goto tx_err;
1036                 md = ip_tunnel_info_opts(tun_info);
1037
1038                 tun_id = tunnel_id_to_key32(key->tun_id);
1039                 if (md->version == 1) {
1040                         erspan_build_header(skb,
1041                                             ntohl(tun_id),
1042                                             ntohl(md->u.index), truncate,
1043                                             false);
1044                         proto = htons(ETH_P_ERSPAN);
1045                 } else if (md->version == 2) {
1046                         erspan_build_header_v2(skb,
1047                                                ntohl(tun_id),
1048                                                md->u.md2.dir,
1049                                                get_hwid(&md->u.md2),
1050                                                truncate, false);
1051                         proto = htons(ETH_P_ERSPAN2);
1052                 } else {
1053                         goto tx_err;
1054                 }
1055         } else {
1056                 switch (skb->protocol) {
1057                 case htons(ETH_P_IP):
1058                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1059                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1060                                                  &dsfield, &encap_limit);
1061                         break;
1062                 case htons(ETH_P_IPV6):
1063                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1064                                 goto tx_err;
1065                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1066                                                      &dsfield, &encap_limit))
1067                                 goto tx_err;
1068                         break;
1069                 default:
1070                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1071                         break;
1072                 }
1073
1074                 if (t->parms.erspan_ver == 1) {
1075                         erspan_build_header(skb, ntohl(t->parms.o_key),
1076                                             t->parms.index,
1077                                             truncate, false);
1078                         proto = htons(ETH_P_ERSPAN);
1079                 } else if (t->parms.erspan_ver == 2) {
1080                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1081                                                t->parms.dir,
1082                                                t->parms.hwid,
1083                                                truncate, false);
1084                         proto = htons(ETH_P_ERSPAN2);
1085                 } else {
1086                         goto tx_err;
1087                 }
1088
1089                 fl6.daddr = t->parms.raddr;
1090         }
1091
1092         /* Push GRE header. */
1093         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno)));
1094
1095         /* TooBig packet may have updated dst->dev's mtu */
1096         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1097                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1098
1099         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1100                            NEXTHDR_GRE);
1101         if (err != 0) {
1102                 /* XXX: send ICMP error even if DF is not set. */
1103                 if (err == -EMSGSIZE) {
1104                         if (skb->protocol == htons(ETH_P_IP))
1105                                 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1106                                               ICMP_FRAG_NEEDED, htonl(mtu));
1107                         else
1108                                 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1109                 }
1110
1111                 goto tx_err;
1112         }
1113         return NETDEV_TX_OK;
1114
1115 tx_err:
1116         stats = &t->dev->stats;
1117         if (!IS_ERR(tun_info))
1118                 stats->tx_errors++;
1119         stats->tx_dropped++;
1120         kfree_skb(skb);
1121         return NETDEV_TX_OK;
1122 }
1123
1124 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1125 {
1126         struct net_device *dev = t->dev;
1127         struct __ip6_tnl_parm *p = &t->parms;
1128         struct flowi6 *fl6 = &t->fl.u.ip6;
1129
1130         if (dev->type != ARPHRD_ETHER) {
1131                 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1132                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1133         }
1134
1135         /* Set up flowi template */
1136         fl6->saddr = p->laddr;
1137         fl6->daddr = p->raddr;
1138         fl6->flowi6_oif = p->link;
1139         fl6->flowlabel = 0;
1140         fl6->flowi6_proto = IPPROTO_GRE;
1141         fl6->fl6_gre_key = t->parms.o_key;
1142
1143         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1144                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1145         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1146                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1147
1148         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1149         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1150
1151         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1152                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1153                 dev->flags |= IFF_POINTOPOINT;
1154         else
1155                 dev->flags &= ~IFF_POINTOPOINT;
1156 }
1157
1158 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1159                                          int t_hlen)
1160 {
1161         const struct __ip6_tnl_parm *p = &t->parms;
1162         struct net_device *dev = t->dev;
1163
1164         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1165                 int strict = (ipv6_addr_type(&p->raddr) &
1166                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1167
1168                 struct rt6_info *rt = rt6_lookup(t->net,
1169                                                  &p->raddr, &p->laddr,
1170                                                  p->link, NULL, strict);
1171
1172                 if (!rt)
1173                         return;
1174
1175                 if (rt->dst.dev) {
1176                         unsigned short dst_len = rt->dst.dev->hard_header_len +
1177                                                  t_hlen;
1178
1179                         if (t->dev->header_ops)
1180                                 dev->hard_header_len = dst_len;
1181                         else
1182                                 dev->needed_headroom = dst_len;
1183
1184                         if (set_mtu) {
1185                                 int mtu = rt->dst.dev->mtu - t_hlen;
1186
1187                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1188                                         mtu -= 8;
1189                                 if (dev->type == ARPHRD_ETHER)
1190                                         mtu -= ETH_HLEN;
1191
1192                                 if (mtu < IPV6_MIN_MTU)
1193                                         mtu = IPV6_MIN_MTU;
1194                                 WRITE_ONCE(dev->mtu, mtu);
1195                         }
1196                 }
1197                 ip6_rt_put(rt);
1198         }
1199 }
1200
1201 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1202 {
1203         int t_hlen;
1204
1205         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1206         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1207
1208         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1209
1210         if (tunnel->dev->header_ops)
1211                 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1212         else
1213                 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1214
1215         return t_hlen;
1216 }
1217
1218 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1219 {
1220         ip6gre_tnl_link_config_common(t);
1221         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1222 }
1223
1224 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1225                                      const struct __ip6_tnl_parm *p)
1226 {
1227         t->parms.laddr = p->laddr;
1228         t->parms.raddr = p->raddr;
1229         t->parms.flags = p->flags;
1230         t->parms.hop_limit = p->hop_limit;
1231         t->parms.encap_limit = p->encap_limit;
1232         t->parms.flowinfo = p->flowinfo;
1233         t->parms.link = p->link;
1234         t->parms.proto = p->proto;
1235         t->parms.i_key = p->i_key;
1236         t->parms.o_key = p->o_key;
1237         t->parms.i_flags = p->i_flags;
1238         t->parms.o_flags = p->o_flags;
1239         t->parms.fwmark = p->fwmark;
1240         t->parms.erspan_ver = p->erspan_ver;
1241         t->parms.index = p->index;
1242         t->parms.dir = p->dir;
1243         t->parms.hwid = p->hwid;
1244         dst_cache_reset(&t->dst_cache);
1245 }
1246
1247 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1248                              int set_mtu)
1249 {
1250         ip6gre_tnl_copy_tnl_parm(t, p);
1251         ip6gre_tnl_link_config(t, set_mtu);
1252         return 0;
1253 }
1254
1255 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1256         const struct ip6_tnl_parm2 *u)
1257 {
1258         p->laddr = u->laddr;
1259         p->raddr = u->raddr;
1260         p->flags = u->flags;
1261         p->hop_limit = u->hop_limit;
1262         p->encap_limit = u->encap_limit;
1263         p->flowinfo = u->flowinfo;
1264         p->link = u->link;
1265         p->i_key = u->i_key;
1266         p->o_key = u->o_key;
1267         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1268         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1269         memcpy(p->name, u->name, sizeof(u->name));
1270 }
1271
1272 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1273         const struct __ip6_tnl_parm *p)
1274 {
1275         u->proto = IPPROTO_GRE;
1276         u->laddr = p->laddr;
1277         u->raddr = p->raddr;
1278         u->flags = p->flags;
1279         u->hop_limit = p->hop_limit;
1280         u->encap_limit = p->encap_limit;
1281         u->flowinfo = p->flowinfo;
1282         u->link = p->link;
1283         u->i_key = p->i_key;
1284         u->o_key = p->o_key;
1285         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1286         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1287         memcpy(u->name, p->name, sizeof(u->name));
1288 }
1289
1290 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev,
1291                                         struct ifreq *ifr, void __user *data,
1292                                         int cmd)
1293 {
1294         int err = 0;
1295         struct ip6_tnl_parm2 p;
1296         struct __ip6_tnl_parm p1;
1297         struct ip6_tnl *t = netdev_priv(dev);
1298         struct net *net = t->net;
1299         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1300
1301         memset(&p1, 0, sizeof(p1));
1302
1303         switch (cmd) {
1304         case SIOCGETTUNNEL:
1305                 if (dev == ign->fb_tunnel_dev) {
1306                         if (copy_from_user(&p, data, sizeof(p))) {
1307                                 err = -EFAULT;
1308                                 break;
1309                         }
1310                         ip6gre_tnl_parm_from_user(&p1, &p);
1311                         t = ip6gre_tunnel_locate(net, &p1, 0);
1312                         if (!t)
1313                                 t = netdev_priv(dev);
1314                 }
1315                 memset(&p, 0, sizeof(p));
1316                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1317                 if (copy_to_user(data, &p, sizeof(p)))
1318                         err = -EFAULT;
1319                 break;
1320
1321         case SIOCADDTUNNEL:
1322         case SIOCCHGTUNNEL:
1323                 err = -EPERM;
1324                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1325                         goto done;
1326
1327                 err = -EFAULT;
1328                 if (copy_from_user(&p, data, sizeof(p)))
1329                         goto done;
1330
1331                 err = -EINVAL;
1332                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1333                         goto done;
1334
1335                 if (!(p.i_flags&GRE_KEY))
1336                         p.i_key = 0;
1337                 if (!(p.o_flags&GRE_KEY))
1338                         p.o_key = 0;
1339
1340                 ip6gre_tnl_parm_from_user(&p1, &p);
1341                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1342
1343                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1344                         if (t) {
1345                                 if (t->dev != dev) {
1346                                         err = -EEXIST;
1347                                         break;
1348                                 }
1349                         } else {
1350                                 t = netdev_priv(dev);
1351
1352                                 ip6gre_tunnel_unlink(ign, t);
1353                                 synchronize_net();
1354                                 ip6gre_tnl_change(t, &p1, 1);
1355                                 ip6gre_tunnel_link(ign, t);
1356                                 netdev_state_change(dev);
1357                         }
1358                 }
1359
1360                 if (t) {
1361                         err = 0;
1362
1363                         memset(&p, 0, sizeof(p));
1364                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1365                         if (copy_to_user(data, &p, sizeof(p)))
1366                                 err = -EFAULT;
1367                 } else
1368                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1369                 break;
1370
1371         case SIOCDELTUNNEL:
1372                 err = -EPERM;
1373                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1374                         goto done;
1375
1376                 if (dev == ign->fb_tunnel_dev) {
1377                         err = -EFAULT;
1378                         if (copy_from_user(&p, data, sizeof(p)))
1379                                 goto done;
1380                         err = -ENOENT;
1381                         ip6gre_tnl_parm_from_user(&p1, &p);
1382                         t = ip6gre_tunnel_locate(net, &p1, 0);
1383                         if (!t)
1384                                 goto done;
1385                         err = -EPERM;
1386                         if (t == netdev_priv(ign->fb_tunnel_dev))
1387                                 goto done;
1388                         dev = t->dev;
1389                 }
1390                 unregister_netdevice(dev);
1391                 err = 0;
1392                 break;
1393
1394         default:
1395                 err = -EINVAL;
1396         }
1397
1398 done:
1399         return err;
1400 }
1401
1402 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1403                          unsigned short type, const void *daddr,
1404                          const void *saddr, unsigned int len)
1405 {
1406         struct ip6_tnl *t = netdev_priv(dev);
1407         struct ipv6hdr *ipv6h;
1408         __be16 *p;
1409
1410         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1411         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1412                                                   t->fl.u.ip6.flowlabel,
1413                                                   true, &t->fl.u.ip6));
1414         ipv6h->hop_limit = t->parms.hop_limit;
1415         ipv6h->nexthdr = NEXTHDR_GRE;
1416         ipv6h->saddr = t->parms.laddr;
1417         ipv6h->daddr = t->parms.raddr;
1418
1419         p = (__be16 *)(ipv6h + 1);
1420         p[0] = t->parms.o_flags;
1421         p[1] = htons(type);
1422
1423         /*
1424          *      Set the source hardware address.
1425          */
1426
1427         if (saddr)
1428                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1429         if (daddr)
1430                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1431         if (!ipv6_addr_any(&ipv6h->daddr))
1432                 return t->hlen;
1433
1434         return -t->hlen;
1435 }
1436
1437 static const struct header_ops ip6gre_header_ops = {
1438         .create = ip6gre_header,
1439 };
1440
1441 static const struct net_device_ops ip6gre_netdev_ops = {
1442         .ndo_init               = ip6gre_tunnel_init,
1443         .ndo_uninit             = ip6gre_tunnel_uninit,
1444         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1445         .ndo_siocdevprivate     = ip6gre_tunnel_siocdevprivate,
1446         .ndo_change_mtu         = ip6_tnl_change_mtu,
1447         .ndo_get_stats64        = dev_get_tstats64,
1448         .ndo_get_iflink         = ip6_tnl_get_iflink,
1449 };
1450
1451 static void ip6gre_dev_free(struct net_device *dev)
1452 {
1453         struct ip6_tnl *t = netdev_priv(dev);
1454
1455         gro_cells_destroy(&t->gro_cells);
1456         dst_cache_destroy(&t->dst_cache);
1457         free_percpu(dev->tstats);
1458 }
1459
1460 static void ip6gre_tunnel_setup(struct net_device *dev)
1461 {
1462         dev->netdev_ops = &ip6gre_netdev_ops;
1463         dev->needs_free_netdev = true;
1464         dev->priv_destructor = ip6gre_dev_free;
1465
1466         dev->type = ARPHRD_IP6GRE;
1467
1468         dev->flags |= IFF_NOARP;
1469         dev->addr_len = sizeof(struct in6_addr);
1470         netif_keep_dst(dev);
1471         /* This perm addr will be used as interface identifier by IPv6 */
1472         dev->addr_assign_type = NET_ADDR_RANDOM;
1473         eth_random_addr(dev->perm_addr);
1474 }
1475
1476 #define GRE6_FEATURES (NETIF_F_SG |             \
1477                        NETIF_F_FRAGLIST |       \
1478                        NETIF_F_HIGHDMA |        \
1479                        NETIF_F_HW_CSUM)
1480
1481 static void ip6gre_tnl_init_features(struct net_device *dev)
1482 {
1483         struct ip6_tnl *nt = netdev_priv(dev);
1484         __be16 flags;
1485
1486         dev->features           |= GRE6_FEATURES | NETIF_F_LLTX;
1487         dev->hw_features        |= GRE6_FEATURES;
1488
1489         flags = nt->parms.o_flags;
1490
1491         /* TCP offload with GRE SEQ is not supported, nor can we support 2
1492          * levels of outer headers requiring an update.
1493          */
1494         if (flags & TUNNEL_SEQ)
1495                 return;
1496         if (flags & TUNNEL_CSUM && nt->encap.type != TUNNEL_ENCAP_NONE)
1497                 return;
1498
1499         dev->features |= NETIF_F_GSO_SOFTWARE;
1500         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1501 }
1502
1503 static int ip6gre_tunnel_init_common(struct net_device *dev)
1504 {
1505         struct ip6_tnl *tunnel;
1506         int ret;
1507         int t_hlen;
1508
1509         tunnel = netdev_priv(dev);
1510
1511         tunnel->dev = dev;
1512         tunnel->net = dev_net(dev);
1513         strcpy(tunnel->parms.name, dev->name);
1514
1515         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1516         if (!dev->tstats)
1517                 return -ENOMEM;
1518
1519         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1520         if (ret)
1521                 goto cleanup_alloc_pcpu_stats;
1522
1523         ret = gro_cells_init(&tunnel->gro_cells, dev);
1524         if (ret)
1525                 goto cleanup_dst_cache_init;
1526
1527         t_hlen = ip6gre_calc_hlen(tunnel);
1528         dev->mtu = ETH_DATA_LEN - t_hlen;
1529         if (dev->type == ARPHRD_ETHER)
1530                 dev->mtu -= ETH_HLEN;
1531         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1532                 dev->mtu -= 8;
1533
1534         if (tunnel->parms.collect_md) {
1535                 netif_keep_dst(dev);
1536         }
1537         ip6gre_tnl_init_features(dev);
1538
1539         netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1540         return 0;
1541
1542 cleanup_dst_cache_init:
1543         dst_cache_destroy(&tunnel->dst_cache);
1544 cleanup_alloc_pcpu_stats:
1545         free_percpu(dev->tstats);
1546         dev->tstats = NULL;
1547         return ret;
1548 }
1549
1550 static int ip6gre_tunnel_init(struct net_device *dev)
1551 {
1552         struct ip6_tnl *tunnel;
1553         int ret;
1554
1555         ret = ip6gre_tunnel_init_common(dev);
1556         if (ret)
1557                 return ret;
1558
1559         tunnel = netdev_priv(dev);
1560
1561         if (tunnel->parms.collect_md)
1562                 return 0;
1563
1564         __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr));
1565         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1566
1567         if (ipv6_addr_any(&tunnel->parms.raddr))
1568                 dev->header_ops = &ip6gre_header_ops;
1569
1570         return 0;
1571 }
1572
1573 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1574 {
1575         struct ip6_tnl *tunnel = netdev_priv(dev);
1576
1577         tunnel->dev = dev;
1578         tunnel->net = dev_net(dev);
1579         strcpy(tunnel->parms.name, dev->name);
1580
1581         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1582 }
1583
1584 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1585         .handler     = gre_rcv,
1586         .err_handler = ip6gre_err,
1587         .flags       = INET6_PROTO_FINAL,
1588 };
1589
1590 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1591 {
1592         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1593         struct net_device *dev, *aux;
1594         int prio;
1595
1596         for_each_netdev_safe(net, dev, aux)
1597                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1598                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1599                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1600                         unregister_netdevice_queue(dev, head);
1601
1602         for (prio = 0; prio < 4; prio++) {
1603                 int h;
1604                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1605                         struct ip6_tnl *t;
1606
1607                         t = rtnl_dereference(ign->tunnels[prio][h]);
1608
1609                         while (t) {
1610                                 /* If dev is in the same netns, it has already
1611                                  * been added to the list by the previous loop.
1612                                  */
1613                                 if (!net_eq(dev_net(t->dev), net))
1614                                         unregister_netdevice_queue(t->dev,
1615                                                                    head);
1616                                 t = rtnl_dereference(t->next);
1617                         }
1618                 }
1619         }
1620 }
1621
1622 static int __net_init ip6gre_init_net(struct net *net)
1623 {
1624         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1625         struct net_device *ndev;
1626         int err;
1627
1628         if (!net_has_fallback_tunnels(net))
1629                 return 0;
1630         ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1631                             NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1632         if (!ndev) {
1633                 err = -ENOMEM;
1634                 goto err_alloc_dev;
1635         }
1636         ign->fb_tunnel_dev = ndev;
1637         dev_net_set(ign->fb_tunnel_dev, net);
1638         /* FB netdevice is special: we have one, and only one per netns.
1639          * Allowing to move it to another netns is clearly unsafe.
1640          */
1641         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1642
1643
1644         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1645         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1646
1647         err = register_netdev(ign->fb_tunnel_dev);
1648         if (err)
1649                 goto err_reg_dev;
1650
1651         rcu_assign_pointer(ign->tunnels_wc[0],
1652                            netdev_priv(ign->fb_tunnel_dev));
1653         return 0;
1654
1655 err_reg_dev:
1656         free_netdev(ndev);
1657 err_alloc_dev:
1658         return err;
1659 }
1660
1661 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1662 {
1663         struct net *net;
1664         LIST_HEAD(list);
1665
1666         rtnl_lock();
1667         list_for_each_entry(net, net_list, exit_list)
1668                 ip6gre_destroy_tunnels(net, &list);
1669         unregister_netdevice_many(&list);
1670         rtnl_unlock();
1671 }
1672
1673 static struct pernet_operations ip6gre_net_ops = {
1674         .init = ip6gre_init_net,
1675         .exit_batch = ip6gre_exit_batch_net,
1676         .id   = &ip6gre_net_id,
1677         .size = sizeof(struct ip6gre_net),
1678 };
1679
1680 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1681                                   struct netlink_ext_ack *extack)
1682 {
1683         __be16 flags;
1684
1685         if (!data)
1686                 return 0;
1687
1688         flags = 0;
1689         if (data[IFLA_GRE_IFLAGS])
1690                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1691         if (data[IFLA_GRE_OFLAGS])
1692                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1693         if (flags & (GRE_VERSION|GRE_ROUTING))
1694                 return -EINVAL;
1695
1696         return 0;
1697 }
1698
1699 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1700                                struct netlink_ext_ack *extack)
1701 {
1702         struct in6_addr daddr;
1703
1704         if (tb[IFLA_ADDRESS]) {
1705                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1706                         return -EINVAL;
1707                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1708                         return -EADDRNOTAVAIL;
1709         }
1710
1711         if (!data)
1712                 goto out;
1713
1714         if (data[IFLA_GRE_REMOTE]) {
1715                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1716                 if (ipv6_addr_any(&daddr))
1717                         return -EINVAL;
1718         }
1719
1720 out:
1721         return ip6gre_tunnel_validate(tb, data, extack);
1722 }
1723
1724 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1725                                   struct netlink_ext_ack *extack)
1726 {
1727         __be16 flags = 0;
1728         int ret, ver = 0;
1729
1730         if (!data)
1731                 return 0;
1732
1733         ret = ip6gre_tap_validate(tb, data, extack);
1734         if (ret)
1735                 return ret;
1736
1737         /* ERSPAN should only have GRE sequence and key flag */
1738         if (data[IFLA_GRE_OFLAGS])
1739                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1740         if (data[IFLA_GRE_IFLAGS])
1741                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1742         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1743             flags != (GRE_SEQ | GRE_KEY))
1744                 return -EINVAL;
1745
1746         /* ERSPAN Session ID only has 10-bit. Since we reuse
1747          * 32-bit key field as ID, check it's range.
1748          */
1749         if (data[IFLA_GRE_IKEY] &&
1750             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1751                 return -EINVAL;
1752
1753         if (data[IFLA_GRE_OKEY] &&
1754             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1755                 return -EINVAL;
1756
1757         if (data[IFLA_GRE_ERSPAN_VER]) {
1758                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1759                 if (ver != 1 && ver != 2)
1760                         return -EINVAL;
1761         }
1762
1763         if (ver == 1) {
1764                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1765                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1766
1767                         if (index & ~INDEX_MASK)
1768                                 return -EINVAL;
1769                 }
1770         } else if (ver == 2) {
1771                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1772                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1773
1774                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1775                                 return -EINVAL;
1776                 }
1777
1778                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1779                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1780
1781                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1782                                 return -EINVAL;
1783                 }
1784         }
1785
1786         return 0;
1787 }
1788
1789 static void ip6erspan_set_version(struct nlattr *data[],
1790                                   struct __ip6_tnl_parm *parms)
1791 {
1792         if (!data)
1793                 return;
1794
1795         parms->erspan_ver = 1;
1796         if (data[IFLA_GRE_ERSPAN_VER])
1797                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1798
1799         if (parms->erspan_ver == 1) {
1800                 if (data[IFLA_GRE_ERSPAN_INDEX])
1801                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1802         } else if (parms->erspan_ver == 2) {
1803                 if (data[IFLA_GRE_ERSPAN_DIR])
1804                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1805                 if (data[IFLA_GRE_ERSPAN_HWID])
1806                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1807         }
1808 }
1809
1810 static void ip6gre_netlink_parms(struct nlattr *data[],
1811                                 struct __ip6_tnl_parm *parms)
1812 {
1813         memset(parms, 0, sizeof(*parms));
1814
1815         if (!data)
1816                 return;
1817
1818         if (data[IFLA_GRE_LINK])
1819                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1820
1821         if (data[IFLA_GRE_IFLAGS])
1822                 parms->i_flags = gre_flags_to_tnl_flags(
1823                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1824
1825         if (data[IFLA_GRE_OFLAGS])
1826                 parms->o_flags = gre_flags_to_tnl_flags(
1827                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1828
1829         if (data[IFLA_GRE_IKEY])
1830                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1831
1832         if (data[IFLA_GRE_OKEY])
1833                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1834
1835         if (data[IFLA_GRE_LOCAL])
1836                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1837
1838         if (data[IFLA_GRE_REMOTE])
1839                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1840
1841         if (data[IFLA_GRE_TTL])
1842                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1843
1844         if (data[IFLA_GRE_ENCAP_LIMIT])
1845                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1846
1847         if (data[IFLA_GRE_FLOWINFO])
1848                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1849
1850         if (data[IFLA_GRE_FLAGS])
1851                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1852
1853         if (data[IFLA_GRE_FWMARK])
1854                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1855
1856         if (data[IFLA_GRE_COLLECT_METADATA])
1857                 parms->collect_md = true;
1858 }
1859
1860 static int ip6gre_tap_init(struct net_device *dev)
1861 {
1862         int ret;
1863
1864         ret = ip6gre_tunnel_init_common(dev);
1865         if (ret)
1866                 return ret;
1867
1868         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1869
1870         return 0;
1871 }
1872
1873 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1874         .ndo_init = ip6gre_tap_init,
1875         .ndo_uninit = ip6gre_tunnel_uninit,
1876         .ndo_start_xmit = ip6gre_tunnel_xmit,
1877         .ndo_set_mac_address = eth_mac_addr,
1878         .ndo_validate_addr = eth_validate_addr,
1879         .ndo_change_mtu = ip6_tnl_change_mtu,
1880         .ndo_get_stats64 = dev_get_tstats64,
1881         .ndo_get_iflink = ip6_tnl_get_iflink,
1882 };
1883
1884 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1885 {
1886         int t_hlen;
1887
1888         tunnel->tun_hlen = 8;
1889         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1890                        erspan_hdr_len(tunnel->parms.erspan_ver);
1891
1892         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1893         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1894         return t_hlen;
1895 }
1896
1897 static int ip6erspan_tap_init(struct net_device *dev)
1898 {
1899         struct ip6_tnl *tunnel;
1900         int t_hlen;
1901         int ret;
1902
1903         tunnel = netdev_priv(dev);
1904
1905         tunnel->dev = dev;
1906         tunnel->net = dev_net(dev);
1907         strcpy(tunnel->parms.name, dev->name);
1908
1909         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1910         if (!dev->tstats)
1911                 return -ENOMEM;
1912
1913         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1914         if (ret)
1915                 goto cleanup_alloc_pcpu_stats;
1916
1917         ret = gro_cells_init(&tunnel->gro_cells, dev);
1918         if (ret)
1919                 goto cleanup_dst_cache_init;
1920
1921         t_hlen = ip6erspan_calc_hlen(tunnel);
1922         dev->mtu = ETH_DATA_LEN - t_hlen;
1923         if (dev->type == ARPHRD_ETHER)
1924                 dev->mtu -= ETH_HLEN;
1925         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1926                 dev->mtu -= 8;
1927
1928         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1929         ip6erspan_tnl_link_config(tunnel, 1);
1930
1931         netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1932         return 0;
1933
1934 cleanup_dst_cache_init:
1935         dst_cache_destroy(&tunnel->dst_cache);
1936 cleanup_alloc_pcpu_stats:
1937         free_percpu(dev->tstats);
1938         dev->tstats = NULL;
1939         return ret;
1940 }
1941
1942 static const struct net_device_ops ip6erspan_netdev_ops = {
1943         .ndo_init =             ip6erspan_tap_init,
1944         .ndo_uninit =           ip6erspan_tunnel_uninit,
1945         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1946         .ndo_set_mac_address =  eth_mac_addr,
1947         .ndo_validate_addr =    eth_validate_addr,
1948         .ndo_change_mtu =       ip6_tnl_change_mtu,
1949         .ndo_get_stats64 =      dev_get_tstats64,
1950         .ndo_get_iflink =       ip6_tnl_get_iflink,
1951 };
1952
1953 static void ip6gre_tap_setup(struct net_device *dev)
1954 {
1955
1956         ether_setup(dev);
1957
1958         dev->max_mtu = 0;
1959         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1960         dev->needs_free_netdev = true;
1961         dev->priv_destructor = ip6gre_dev_free;
1962
1963         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1964         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1965         netif_keep_dst(dev);
1966 }
1967
1968 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1969                                        struct ip_tunnel_encap *ipencap)
1970 {
1971         bool ret = false;
1972
1973         memset(ipencap, 0, sizeof(*ipencap));
1974
1975         if (!data)
1976                 return ret;
1977
1978         if (data[IFLA_GRE_ENCAP_TYPE]) {
1979                 ret = true;
1980                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1981         }
1982
1983         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1984                 ret = true;
1985                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1986         }
1987
1988         if (data[IFLA_GRE_ENCAP_SPORT]) {
1989                 ret = true;
1990                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1991         }
1992
1993         if (data[IFLA_GRE_ENCAP_DPORT]) {
1994                 ret = true;
1995                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1996         }
1997
1998         return ret;
1999 }
2000
2001 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
2002                                  struct nlattr *tb[], struct nlattr *data[],
2003                                  struct netlink_ext_ack *extack)
2004 {
2005         struct ip6_tnl *nt;
2006         struct ip_tunnel_encap ipencap;
2007         int err;
2008
2009         nt = netdev_priv(dev);
2010
2011         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2012                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2013
2014                 if (err < 0)
2015                         return err;
2016         }
2017
2018         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
2019                 eth_hw_addr_random(dev);
2020
2021         nt->dev = dev;
2022         nt->net = dev_net(dev);
2023
2024         err = register_netdevice(dev);
2025         if (err)
2026                 goto out;
2027
2028         if (tb[IFLA_MTU])
2029                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2030
2031 out:
2032         return err;
2033 }
2034
2035 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2036                           struct nlattr *tb[], struct nlattr *data[],
2037                           struct netlink_ext_ack *extack)
2038 {
2039         struct ip6_tnl *nt = netdev_priv(dev);
2040         struct net *net = dev_net(dev);
2041         struct ip6gre_net *ign;
2042         int err;
2043
2044         ip6gre_netlink_parms(data, &nt->parms);
2045         ign = net_generic(net, ip6gre_net_id);
2046
2047         if (nt->parms.collect_md) {
2048                 if (rtnl_dereference(ign->collect_md_tun))
2049                         return -EEXIST;
2050         } else {
2051                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2052                         return -EEXIST;
2053         }
2054
2055         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2056         if (!err) {
2057                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2058                 ip6gre_tunnel_link_md(ign, nt);
2059                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2060         }
2061         return err;
2062 }
2063
2064 static struct ip6_tnl *
2065 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2066                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2067                          struct netlink_ext_ack *extack)
2068 {
2069         struct ip6_tnl *t, *nt = netdev_priv(dev);
2070         struct net *net = nt->net;
2071         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2072         struct ip_tunnel_encap ipencap;
2073
2074         if (dev == ign->fb_tunnel_dev)
2075                 return ERR_PTR(-EINVAL);
2076
2077         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2078                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2079
2080                 if (err < 0)
2081                         return ERR_PTR(err);
2082         }
2083
2084         ip6gre_netlink_parms(data, p_p);
2085
2086         t = ip6gre_tunnel_locate(net, p_p, 0);
2087
2088         if (t) {
2089                 if (t->dev != dev)
2090                         return ERR_PTR(-EEXIST);
2091         } else {
2092                 t = nt;
2093         }
2094
2095         return t;
2096 }
2097
2098 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2099                              struct nlattr *data[],
2100                              struct netlink_ext_ack *extack)
2101 {
2102         struct ip6_tnl *t = netdev_priv(dev);
2103         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2104         struct __ip6_tnl_parm p;
2105
2106         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2107         if (IS_ERR(t))
2108                 return PTR_ERR(t);
2109
2110         ip6gre_tunnel_unlink_md(ign, t);
2111         ip6gre_tunnel_unlink(ign, t);
2112         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2113         ip6gre_tunnel_link_md(ign, t);
2114         ip6gre_tunnel_link(ign, t);
2115         return 0;
2116 }
2117
2118 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2119 {
2120         struct net *net = dev_net(dev);
2121         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2122
2123         if (dev != ign->fb_tunnel_dev)
2124                 unregister_netdevice_queue(dev, head);
2125 }
2126
2127 static size_t ip6gre_get_size(const struct net_device *dev)
2128 {
2129         return
2130                 /* IFLA_GRE_LINK */
2131                 nla_total_size(4) +
2132                 /* IFLA_GRE_IFLAGS */
2133                 nla_total_size(2) +
2134                 /* IFLA_GRE_OFLAGS */
2135                 nla_total_size(2) +
2136                 /* IFLA_GRE_IKEY */
2137                 nla_total_size(4) +
2138                 /* IFLA_GRE_OKEY */
2139                 nla_total_size(4) +
2140                 /* IFLA_GRE_LOCAL */
2141                 nla_total_size(sizeof(struct in6_addr)) +
2142                 /* IFLA_GRE_REMOTE */
2143                 nla_total_size(sizeof(struct in6_addr)) +
2144                 /* IFLA_GRE_TTL */
2145                 nla_total_size(1) +
2146                 /* IFLA_GRE_ENCAP_LIMIT */
2147                 nla_total_size(1) +
2148                 /* IFLA_GRE_FLOWINFO */
2149                 nla_total_size(4) +
2150                 /* IFLA_GRE_FLAGS */
2151                 nla_total_size(4) +
2152                 /* IFLA_GRE_ENCAP_TYPE */
2153                 nla_total_size(2) +
2154                 /* IFLA_GRE_ENCAP_FLAGS */
2155                 nla_total_size(2) +
2156                 /* IFLA_GRE_ENCAP_SPORT */
2157                 nla_total_size(2) +
2158                 /* IFLA_GRE_ENCAP_DPORT */
2159                 nla_total_size(2) +
2160                 /* IFLA_GRE_COLLECT_METADATA */
2161                 nla_total_size(0) +
2162                 /* IFLA_GRE_FWMARK */
2163                 nla_total_size(4) +
2164                 /* IFLA_GRE_ERSPAN_INDEX */
2165                 nla_total_size(4) +
2166                 0;
2167 }
2168
2169 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2170 {
2171         struct ip6_tnl *t = netdev_priv(dev);
2172         struct __ip6_tnl_parm *p = &t->parms;
2173         __be16 o_flags = p->o_flags;
2174
2175         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2176                 if (!p->collect_md)
2177                         o_flags |= TUNNEL_KEY;
2178
2179                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2180                         goto nla_put_failure;
2181
2182                 if (p->erspan_ver == 1) {
2183                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2184                                 goto nla_put_failure;
2185                 } else {
2186                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2187                                 goto nla_put_failure;
2188                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2189                                 goto nla_put_failure;
2190                 }
2191         }
2192
2193         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2194             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2195                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2196             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2197                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2198             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2199             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2200             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2201             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2202             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2203             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2204             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2205             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2206             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2207                 goto nla_put_failure;
2208
2209         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2210                         t->encap.type) ||
2211             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2212                          t->encap.sport) ||
2213             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2214                          t->encap.dport) ||
2215             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2216                         t->encap.flags))
2217                 goto nla_put_failure;
2218
2219         if (p->collect_md) {
2220                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2221                         goto nla_put_failure;
2222         }
2223
2224         return 0;
2225
2226 nla_put_failure:
2227         return -EMSGSIZE;
2228 }
2229
2230 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2231         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2232         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2233         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2234         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2235         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2236         [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2237         [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2238         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2239         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2240         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2241         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2242         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2243         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2244         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2245         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2246         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2247         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2248         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2249         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2250         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2251         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2252 };
2253
2254 static void ip6erspan_tap_setup(struct net_device *dev)
2255 {
2256         ether_setup(dev);
2257
2258         dev->max_mtu = 0;
2259         dev->netdev_ops = &ip6erspan_netdev_ops;
2260         dev->needs_free_netdev = true;
2261         dev->priv_destructor = ip6gre_dev_free;
2262
2263         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2264         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2265         netif_keep_dst(dev);
2266 }
2267
2268 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2269                              struct nlattr *tb[], struct nlattr *data[],
2270                              struct netlink_ext_ack *extack)
2271 {
2272         struct ip6_tnl *nt = netdev_priv(dev);
2273         struct net *net = dev_net(dev);
2274         struct ip6gre_net *ign;
2275         int err;
2276
2277         ip6gre_netlink_parms(data, &nt->parms);
2278         ip6erspan_set_version(data, &nt->parms);
2279         ign = net_generic(net, ip6gre_net_id);
2280
2281         if (nt->parms.collect_md) {
2282                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2283                         return -EEXIST;
2284         } else {
2285                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2286                         return -EEXIST;
2287         }
2288
2289         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2290         if (!err) {
2291                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2292                 ip6erspan_tunnel_link_md(ign, nt);
2293                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2294         }
2295         return err;
2296 }
2297
2298 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2299 {
2300         ip6gre_tnl_link_config_common(t);
2301         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2302 }
2303
2304 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2305                                 const struct __ip6_tnl_parm *p, int set_mtu)
2306 {
2307         ip6gre_tnl_copy_tnl_parm(t, p);
2308         ip6erspan_tnl_link_config(t, set_mtu);
2309         return 0;
2310 }
2311
2312 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2313                                 struct nlattr *data[],
2314                                 struct netlink_ext_ack *extack)
2315 {
2316         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2317         struct __ip6_tnl_parm p;
2318         struct ip6_tnl *t;
2319
2320         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2321         if (IS_ERR(t))
2322                 return PTR_ERR(t);
2323
2324         ip6erspan_set_version(data, &p);
2325         ip6gre_tunnel_unlink_md(ign, t);
2326         ip6gre_tunnel_unlink(ign, t);
2327         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2328         ip6erspan_tunnel_link_md(ign, t);
2329         ip6gre_tunnel_link(ign, t);
2330         return 0;
2331 }
2332
2333 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2334         .kind           = "ip6gre",
2335         .maxtype        = IFLA_GRE_MAX,
2336         .policy         = ip6gre_policy,
2337         .priv_size      = sizeof(struct ip6_tnl),
2338         .setup          = ip6gre_tunnel_setup,
2339         .validate       = ip6gre_tunnel_validate,
2340         .newlink        = ip6gre_newlink,
2341         .changelink     = ip6gre_changelink,
2342         .dellink        = ip6gre_dellink,
2343         .get_size       = ip6gre_get_size,
2344         .fill_info      = ip6gre_fill_info,
2345         .get_link_net   = ip6_tnl_get_link_net,
2346 };
2347
2348 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2349         .kind           = "ip6gretap",
2350         .maxtype        = IFLA_GRE_MAX,
2351         .policy         = ip6gre_policy,
2352         .priv_size      = sizeof(struct ip6_tnl),
2353         .setup          = ip6gre_tap_setup,
2354         .validate       = ip6gre_tap_validate,
2355         .newlink        = ip6gre_newlink,
2356         .changelink     = ip6gre_changelink,
2357         .get_size       = ip6gre_get_size,
2358         .fill_info      = ip6gre_fill_info,
2359         .get_link_net   = ip6_tnl_get_link_net,
2360 };
2361
2362 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2363         .kind           = "ip6erspan",
2364         .maxtype        = IFLA_GRE_MAX,
2365         .policy         = ip6gre_policy,
2366         .priv_size      = sizeof(struct ip6_tnl),
2367         .setup          = ip6erspan_tap_setup,
2368         .validate       = ip6erspan_tap_validate,
2369         .newlink        = ip6erspan_newlink,
2370         .changelink     = ip6erspan_changelink,
2371         .get_size       = ip6gre_get_size,
2372         .fill_info      = ip6gre_fill_info,
2373         .get_link_net   = ip6_tnl_get_link_net,
2374 };
2375
2376 /*
2377  *      And now the modules code and kernel interface.
2378  */
2379
2380 static int __init ip6gre_init(void)
2381 {
2382         int err;
2383
2384         pr_info("GRE over IPv6 tunneling driver\n");
2385
2386         err = register_pernet_device(&ip6gre_net_ops);
2387         if (err < 0)
2388                 return err;
2389
2390         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2391         if (err < 0) {
2392                 pr_info("%s: can't add protocol\n", __func__);
2393                 goto add_proto_failed;
2394         }
2395
2396         err = rtnl_link_register(&ip6gre_link_ops);
2397         if (err < 0)
2398                 goto rtnl_link_failed;
2399
2400         err = rtnl_link_register(&ip6gre_tap_ops);
2401         if (err < 0)
2402                 goto tap_ops_failed;
2403
2404         err = rtnl_link_register(&ip6erspan_tap_ops);
2405         if (err < 0)
2406                 goto erspan_link_failed;
2407
2408 out:
2409         return err;
2410
2411 erspan_link_failed:
2412         rtnl_link_unregister(&ip6gre_tap_ops);
2413 tap_ops_failed:
2414         rtnl_link_unregister(&ip6gre_link_ops);
2415 rtnl_link_failed:
2416         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2417 add_proto_failed:
2418         unregister_pernet_device(&ip6gre_net_ops);
2419         goto out;
2420 }
2421
2422 static void __exit ip6gre_fini(void)
2423 {
2424         rtnl_link_unregister(&ip6gre_tap_ops);
2425         rtnl_link_unregister(&ip6gre_link_ops);
2426         rtnl_link_unregister(&ip6erspan_tap_ops);
2427         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2428         unregister_pernet_device(&ip6gre_net_ops);
2429 }
2430
2431 module_init(ip6gre_init);
2432 module_exit(ip6gre_fini);
2433 MODULE_LICENSE("GPL");
2434 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2435 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2436 MODULE_ALIAS_RTNL_LINK("ip6gre");
2437 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2438 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2439 MODULE_ALIAS_NETDEV("ip6gre0");