net/ipv6/ip6_gre.c

   1 /*
   2  *      GRE over IPv6 protocol decoder.
   3  *
   4  *      Authors: Dmitry Kozlov (xeb@mail.ru)
   5  *
   6  *      This program is free software; you can redistribute it and/or
   7  *      modify it under the terms of the GNU General Public License
   8  *      as published by the Free Software Foundation; either version
   9  *      2 of the License, or (at your option) any later version.
  10  *
  11  */
  12
  13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  14
  15 #include <linux/capability.h>
  16 #include <linux/module.h>
  17 #include <linux/types.h>
  18 #include <linux/kernel.h>
  19 #include <linux/slab.h>
  20 #include <linux/uaccess.h>
  21 #include <linux/skbuff.h>
  22 #include <linux/netdevice.h>
  23 #include <linux/in.h>
  24 #include <linux/tcp.h>
  25 #include <linux/udp.h>
  26 #include <linux/if_arp.h>
  27 #include <linux/init.h>
  28 #include <linux/in6.h>
  29 #include <linux/inetdevice.h>
  30 #include <linux/igmp.h>
  31 #include <linux/netfilter_ipv4.h>
  32 #include <linux/etherdevice.h>
  33 #include <linux/if_ether.h>
  34 #include <linux/hash.h>
  35 #include <linux/if_tunnel.h>
  36 #include <linux/ip6_tunnel.h>
  37
  38 #include <net/sock.h>
  39 #include <net/ip.h>
  40 #include <net/ip_tunnels.h>
  41 #include <net/icmp.h>
  42 #include <net/protocol.h>
  43 #include <net/addrconf.h>
  44 #include <net/arp.h>
  45 #include <net/checksum.h>
  46 #include <net/dsfield.h>
  47 #include <net/inet_ecn.h>
  48 #include <net/xfrm.h>
  49 #include <net/net_namespace.h>
  50 #include <net/netns/generic.h>
  51 #include <net/rtnetlink.h>
  52
  53 #include <net/ipv6.h>
  54 #include <net/ip6_fib.h>
  55 #include <net/ip6_route.h>
  56 #include <net/ip6_tunnel.h>
  57 #include <net/gre.h>
  58 #include <net/erspan.h>
  59 #include <net/dst_metadata.h>
  60
  61
  62 static bool log_ecn_error = true;
  63 module_param(log_ecn_error, bool, 0644);
  64 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  65
  66 #define IP6_GRE_HASH_SIZE_SHIFT  5
  67 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
  68
  69 static unsigned int ip6gre_net_id __read_mostly;
  70 struct ip6gre_net {
  71         struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
  72
  73         struct ip6_tnl __rcu *collect_md_tun;
  74         struct net_device *fb_tunnel_dev;
  75 };
  76
  77 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  78 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
  79 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
  80 static int ip6gre_tunnel_init(struct net_device *dev);
  81 static void ip6gre_tunnel_setup(struct net_device *dev);
  82 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  83 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  84
  85 /* Tunnel hash table */
  86
  87 /*
  88    4 hash tables:
  89
  90    3: (remote,local)
  91    2: (remote,*)
  92    1: (*,local)
  93    0: (*,*)
  94
  95    We require exact key match i.e. if a key is present in packet
  96    it will match only tunnel with the same key; if it is not present,
  97    it will match only keyless tunnel.
  98
  99    All keysless packets, if not matched configured keyless tunnels
 100    will match fallback tunnel.
 101  */
 102
 103 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
 104 static u32 HASH_ADDR(const struct in6_addr *addr)
 105 {
 106         u32 hash = ipv6_addr_hash(addr);
 107
 108         return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
 109 }
 110
 111 #define tunnels_r_l     tunnels[3]
 112 #define tunnels_r       tunnels[2]
 113 #define tunnels_l       tunnels[1]
 114 #define tunnels_wc      tunnels[0]
 115
 116 /* Given src, dst and key, find appropriate for input tunnel. */
 117
 118 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 119                 const struct in6_addr *remote, const struct in6_addr *local,
 120                 __be32 key, __be16 gre_proto)
 121 {
 122         struct net *net = dev_net(dev);
 123         int link = dev->ifindex;
 124         unsigned int h0 = HASH_ADDR(remote);
 125         unsigned int h1 = HASH_KEY(key);
 126         struct ip6_tnl *t, *cand = NULL;
 127         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 128         int dev_type = (gre_proto == htons(ETH_P_TEB) ||
 129                         gre_proto == htons(ETH_P_ERSPAN) ||
 130                         gre_proto == htons(ETH_P_ERSPAN2)) ?
 131                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 132         int score, cand_score = 4;
 133
 134         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 135                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 136                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 137                     key != t->parms.i_key ||
 138                     !(t->dev->flags & IFF_UP))
 139                         continue;
 140
 141                 if (t->dev->type != ARPHRD_IP6GRE &&
 142                     t->dev->type != dev_type)
 143                         continue;
 144
 145                 score = 0;
 146                 if (t->parms.link != link)
 147                         score |= 1;
 148                 if (t->dev->type != dev_type)
 149                         score |= 2;
 150                 if (score == 0)
 151                         return t;
 152
 153                 if (score < cand_score) {
 154                         cand = t;
 155                         cand_score = score;
 156                 }
 157         }
 158
 159         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 160                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 161                     key != t->parms.i_key ||
 162                     !(t->dev->flags & IFF_UP))
 163                         continue;
 164
 165                 if (t->dev->type != ARPHRD_IP6GRE &&
 166                     t->dev->type != dev_type)
 167                         continue;
 168
 169                 score = 0;
 170                 if (t->parms.link != link)
 171                         score |= 1;
 172                 if (t->dev->type != dev_type)
 173                         score |= 2;
 174                 if (score == 0)
 175                         return t;
 176
 177                 if (score < cand_score) {
 178                         cand = t;
 179                         cand_score = score;
 180                 }
 181         }
 182
 183         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 184                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 185                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 186                                  !ipv6_addr_is_multicast(local))) ||
 187                     key != t->parms.i_key ||
 188                     !(t->dev->flags & IFF_UP))
 189                         continue;
 190
 191                 if (t->dev->type != ARPHRD_IP6GRE &&
 192                     t->dev->type != dev_type)
 193                         continue;
 194
 195                 score = 0;
 196                 if (t->parms.link != link)
 197                         score |= 1;
 198                 if (t->dev->type != dev_type)
 199                         score |= 2;
 200                 if (score == 0)
 201                         return t;
 202
 203                 if (score < cand_score) {
 204                         cand = t;
 205                         cand_score = score;
 206                 }
 207         }
 208
 209         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 210                 if (t->parms.i_key != key ||
 211                     !(t->dev->flags & IFF_UP))
 212                         continue;
 213
 214                 if (t->dev->type != ARPHRD_IP6GRE &&
 215                     t->dev->type != dev_type)
 216                         continue;
 217
 218                 score = 0;
 219                 if (t->parms.link != link)
 220                         score |= 1;
 221                 if (t->dev->type != dev_type)
 222                         score |= 2;
 223                 if (score == 0)
 224                         return t;
 225
 226                 if (score < cand_score) {
 227                         cand = t;
 228                         cand_score = score;
 229                 }
 230         }
 231
 232         if (cand)
 233                 return cand;
 234
 235         t = rcu_dereference(ign->collect_md_tun);
 236         if (t && t->dev->flags & IFF_UP)
 237                 return t;
 238
 239         dev = ign->fb_tunnel_dev;
 240         if (dev && dev->flags & IFF_UP)
 241                 return netdev_priv(dev);
 242
 243         return NULL;
 244 }
 245
 246 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 247                 const struct __ip6_tnl_parm *p)
 248 {
 249         const struct in6_addr *remote = &p->raddr;
 250         const struct in6_addr *local = &p->laddr;
 251         unsigned int h = HASH_KEY(p->i_key);
 252         int prio = 0;
 253
 254         if (!ipv6_addr_any(local))
 255                 prio |= 1;
 256         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 257                 prio |= 2;
 258                 h ^= HASH_ADDR(remote);
 259         }
 260
 261         return &ign->tunnels[prio][h];
 262 }
 263
 264 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 265                 const struct ip6_tnl *t)
 266 {
 267         return __ip6gre_bucket(ign, &t->parms);
 268 }
 269
 270 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 271 {
 272         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 273
 274         if (t->parms.collect_md)
 275                 rcu_assign_pointer(ign->collect_md_tun, t);
 276
 277         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 278         rcu_assign_pointer(*tp, t);
 279 }
 280
 281 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 282 {
 283         struct ip6_tnl __rcu **tp;
 284         struct ip6_tnl *iter;
 285
 286         if (t->parms.collect_md)
 287                 rcu_assign_pointer(ign->collect_md_tun, NULL);
 288
 289         for (tp = ip6gre_bucket(ign, t);
 290              (iter = rtnl_dereference(*tp)) != NULL;
 291              tp = &iter->next) {
 292                 if (t == iter) {
 293                         rcu_assign_pointer(*tp, t->next);
 294                         break;
 295                 }
 296         }
 297 }
 298
 299 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 300                                            const struct __ip6_tnl_parm *parms,
 301                                            int type)
 302 {
 303         const struct in6_addr *remote = &parms->raddr;
 304         const struct in6_addr *local = &parms->laddr;
 305         __be32 key = parms->i_key;
 306         int link = parms->link;
 307         struct ip6_tnl *t;
 308         struct ip6_tnl __rcu **tp;
 309         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 310
 311         for (tp = __ip6gre_bucket(ign, parms);
 312              (t = rtnl_dereference(*tp)) != NULL;
 313              tp = &t->next)
 314                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 315                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 316                     key == t->parms.i_key &&
 317                     link == t->parms.link &&
 318                     type == t->dev->type)
 319                         break;
 320
 321         return t;
 322 }
 323
 324 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 325                 const struct __ip6_tnl_parm *parms, int create)
 326 {
 327         struct ip6_tnl *t, *nt;
 328         struct net_device *dev;
 329         char name[IFNAMSIZ];
 330         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 331
 332         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 333         if (t && create)
 334                 return NULL;
 335         if (t || !create)
 336                 return t;
 337
 338         if (parms->name[0])
 339                 strlcpy(name, parms->name, IFNAMSIZ);
 340         else
 341                 strcpy(name, "ip6gre%d");
 342
 343         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 344                            ip6gre_tunnel_setup);
 345         if (!dev)
 346                 return NULL;
 347
 348         dev_net_set(dev, net);
 349
 350         nt = netdev_priv(dev);
 351         nt->parms = *parms;
 352         dev->rtnl_link_ops = &ip6gre_link_ops;
 353
 354         nt->dev = dev;
 355         nt->net = dev_net(dev);
 356
 357         if (register_netdevice(dev) < 0)
 358                 goto failed_free;
 359
 360         ip6gre_tnl_link_config(nt, 1);
 361
 362         /* Can use a lockless transmit, unless we generate output sequences */
 363         if (!(nt->parms.o_flags & TUNNEL_SEQ))
 364                 dev->features |= NETIF_F_LLTX;
 365
 366         dev_hold(dev);
 367         ip6gre_tunnel_link(ign, nt);
 368         return nt;
 369
 370 failed_free:
 371         free_netdev(dev);
 372         return NULL;
 373 }
 374
 375 static void ip6gre_tunnel_uninit(struct net_device *dev)
 376 {
 377         struct ip6_tnl *t = netdev_priv(dev);
 378         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 379
 380         ip6gre_tunnel_unlink(ign, t);
 381         dst_cache_reset(&t->dst_cache);
 382         dev_put(dev);
 383 }
 384
 385
 386 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 387                        u8 type, u8 code, int offset, __be32 info)
 388 {
 389         struct net *net = dev_net(skb->dev);
 390         const struct gre_base_hdr *greh;
 391         const struct ipv6hdr *ipv6h;
 392         int grehlen = sizeof(*greh);
 393         struct ip6_tnl *t;
 394         int key_off = 0;
 395         __be16 flags;
 396         __be32 key;
 397
 398         if (!pskb_may_pull(skb, offset + grehlen))
 399                 return;
 400         greh = (const struct gre_base_hdr *)(skb->data + offset);
 401         flags = greh->flags;
 402         if (flags & (GRE_VERSION | GRE_ROUTING))
 403                 return;
 404         if (flags & GRE_CSUM)
 405                 grehlen += 4;
 406         if (flags & GRE_KEY) {
 407                 key_off = grehlen + offset;
 408                 grehlen += 4;
 409         }
 410
 411         if (!pskb_may_pull(skb, offset + grehlen))
 412                 return;
 413         ipv6h = (const struct ipv6hdr *)skb->data;
 414         greh = (const struct gre_base_hdr *)(skb->data + offset);
 415         key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
 416
 417         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 418                                  key, greh->protocol);
 419         if (!t)
 420                 return;
 421
 422         switch (type) {
 423                 struct ipv6_tlv_tnl_enc_lim *tel;
 424                 __u32 teli;
 425         case ICMPV6_DEST_UNREACH:
 426                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 427                                     t->parms.name);
 428                 if (code != ICMPV6_PORT_UNREACH)
 429                         break;
 430                 return;
 431         case ICMPV6_TIME_EXCEED:
 432                 if (code == ICMPV6_EXC_HOPLIMIT) {
 433                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 434                                             t->parms.name);
 435                         break;
 436                 }
 437                 return;
 438         case ICMPV6_PARAMPROB:
 439                 teli = 0;
 440                 if (code == ICMPV6_HDR_FIELD)
 441                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 442
 443                 if (teli && teli == be32_to_cpu(info) - 2) {
 444                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 445                         if (tel->encap_limit == 0) {
 446                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 447                                                     t->parms.name);
 448                         }
 449                 } else {
 450                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 451                                             t->parms.name);
 452                 }
 453                 return;
 454         case ICMPV6_PKT_TOOBIG:
 455                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
 456                 return;
 457         case NDISC_REDIRECT:
 458                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 459                              sock_net_uid(net, NULL));
 460                 return;
 461         }
 462
 463         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 464                 t->err_count++;
 465         else
 466                 t->err_count = 1;
 467         t->err_time = jiffies;
 468 }
 469
 470 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 471 {
 472         const struct ipv6hdr *ipv6h;
 473         struct ip6_tnl *tunnel;
 474
 475         ipv6h = ipv6_hdr(skb);
 476         tunnel = ip6gre_tunnel_lookup(skb->dev,
 477                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 478                                       tpi->proto);
 479         if (tunnel) {
 480                 if (tunnel->parms.collect_md) {
 481                         struct metadata_dst *tun_dst;
 482                         __be64 tun_id;
 483                         __be16 flags;
 484
 485                         flags = tpi->flags;
 486                         tun_id = key32_to_tunnel_id(tpi->key);
 487
 488                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
 489                         if (!tun_dst)
 490                                 return PACKET_REJECT;
 491
 492                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
 493                 } else {
 494                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 495                 }
 496
 497                 return PACKET_RCVD;
 498         }
 499
 500         return PACKET_REJECT;
 501 }
 502
 503 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
 504                          struct tnl_ptk_info *tpi)
 505 {
 506         struct erspan_base_hdr *ershdr;
 507         struct erspan_metadata *pkt_md;
 508         const struct ipv6hdr *ipv6h;
 509         struct erspan_md2 *md2;
 510         struct ip6_tnl *tunnel;
 511         u8 ver;
 512
 513         if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
 514                 return PACKET_REJECT;
 515
 516         ipv6h = ipv6_hdr(skb);
 517         ershdr = (struct erspan_base_hdr *)skb->data;
 518         ver = ershdr->ver;
 519         tpi->key = cpu_to_be32(get_session_id(ershdr));
 520
 521         tunnel = ip6gre_tunnel_lookup(skb->dev,
 522                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 523                                       tpi->proto);
 524         if (tunnel) {
 525                 int len = erspan_hdr_len(ver);
 526
 527                 if (unlikely(!pskb_may_pull(skb, len)))
 528                         return PACKET_REJECT;
 529
 530                 ershdr = (struct erspan_base_hdr *)skb->data;
 531                 pkt_md = (struct erspan_metadata *)(ershdr + 1);
 532
 533                 if (__iptunnel_pull_header(skb, len,
 534                                            htons(ETH_P_TEB),
 535                                            false, false) < 0)
 536                         return PACKET_REJECT;
 537
 538                 if (tunnel->parms.collect_md) {
 539                         struct metadata_dst *tun_dst;
 540                         struct ip_tunnel_info *info;
 541                         struct erspan_metadata *md;
 542                         __be64 tun_id;
 543                         __be16 flags;
 544
 545                         tpi->flags |= TUNNEL_KEY;
 546                         flags = tpi->flags;
 547                         tun_id = key32_to_tunnel_id(tpi->key);
 548
 549                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
 550                                                   sizeof(*md));
 551                         if (!tun_dst)
 552                                 return PACKET_REJECT;
 553
 554                         info = &tun_dst->u.tun_info;
 555                         md = ip_tunnel_info_opts(info);
 556                         md->version = ver;
 557                         md2 = &md->u.md2;
 558                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
 559                                                        ERSPAN_V2_MDSIZE);
 560                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
 561                         info->options_len = sizeof(*md);
 562
 563                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
 564
 565                 } else {
 566                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 567                 }
 568
 569                 return PACKET_RCVD;
 570         }
 571
 572         return PACKET_REJECT;
 573 }
 574
 575 static int gre_rcv(struct sk_buff *skb)
 576 {
 577         struct tnl_ptk_info tpi;
 578         bool csum_err = false;
 579         int hdr_len;
 580
 581         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
 582         if (hdr_len < 0)
 583                 goto drop;
 584
 585         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
 586                 goto drop;
 587
 588         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
 589                      tpi.proto == htons(ETH_P_ERSPAN2))) {
 590                 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
 591                         return 0;
 592                 goto out;
 593         }
 594
 595         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
 596                 return 0;
 597
 598 out:
 599         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 600 drop:
 601         kfree_skb(skb);
 602         return 0;
 603 }
 604
 605 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
 606 {
 607         return iptunnel_handle_offloads(skb,
 608                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 609 }
 610
 611 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
 612                                      struct net_device *dev,
 613                                      struct flowi6 *fl6, __u8 *dsfield,
 614                                      int *encap_limit)
 615 {
 616         const struct iphdr *iph = ip_hdr(skb);
 617         struct ip6_tnl *t = netdev_priv(dev);
 618
 619         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 620                 *encap_limit = t->parms.encap_limit;
 621
 622         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 623
 624         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 625                 *dsfield = ipv4_get_dsfield(iph);
 626         else
 627                 *dsfield = ip6_tclass(t->parms.flowinfo);
 628
 629         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 630                 fl6->flowi6_mark = skb->mark;
 631         else
 632                 fl6->flowi6_mark = t->parms.fwmark;
 633
 634         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 635 }
 636
 637 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
 638                                     struct net_device *dev,
 639                                     struct flowi6 *fl6, __u8 *dsfield,
 640                                     int *encap_limit)
 641 {
 642         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 643         struct ip6_tnl *t = netdev_priv(dev);
 644         __u16 offset;
 645
 646         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 647         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 648
 649         if (offset > 0) {
 650                 struct ipv6_tlv_tnl_enc_lim *tel;
 651
 652                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 653                 if (tel->encap_limit == 0) {
 654                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 655                                     ICMPV6_HDR_FIELD, offset + 2);
 656                         return -1;
 657                 }
 658                 *encap_limit = tel->encap_limit - 1;
 659         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
 660                 *encap_limit = t->parms.encap_limit;
 661         }
 662
 663         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 664
 665         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 666                 *dsfield = ipv6_get_dsfield(ipv6h);
 667         else
 668                 *dsfield = ip6_tclass(t->parms.flowinfo);
 669
 670         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 671                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
 672
 673         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 674                 fl6->flowi6_mark = skb->mark;
 675         else
 676                 fl6->flowi6_mark = t->parms.fwmark;
 677
 678         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 679
 680         return 0;
 681 }
 682
 683 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
 684                                struct net_device *dev, __u8 dsfield,
 685                                struct flowi6 *fl6, int encap_limit,
 686                                __u32 *pmtu, __be16 proto)
 687 {
 688         struct ip6_tnl *tunnel = netdev_priv(dev);
 689         __be16 protocol;
 690
 691         if (dev->type == ARPHRD_ETHER)
 692                 IPCB(skb)->flags = 0;
 693
 694         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
 695                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
 696         else
 697                 fl6->daddr = tunnel->parms.raddr;
 698
 699         /* Push GRE header. */
 700         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
 701
 702         if (tunnel->parms.collect_md) {
 703                 struct ip_tunnel_info *tun_info;
 704                 const struct ip_tunnel_key *key;
 705                 __be16 flags;
 706
 707                 tun_info = skb_tunnel_info(skb);
 708                 if (unlikely(!tun_info ||
 709                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
 710                              ip_tunnel_info_af(tun_info) != AF_INET6))
 711                         return -EINVAL;
 712
 713                 key = &tun_info->key;
 714                 memset(fl6, 0, sizeof(*fl6));
 715                 fl6->flowi6_proto = IPPROTO_GRE;
 716                 fl6->daddr = key->u.ipv6.dst;
 717                 fl6->flowlabel = key->label;
 718                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 719
 720                 dsfield = key->tos;
 721                 flags = key->tun_flags &
 722                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
 723                 tunnel->tun_hlen = gre_calc_hlen(flags);
 724
 725                 gre_build_header(skb, tunnel->tun_hlen,
 726                                  flags, protocol,
 727                                  tunnel_id_to_key32(tun_info->key.tun_id),
 728                                  (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
 729                                                       : 0);
 730
 731         } else {
 732                 if (tunnel->parms.o_flags & TUNNEL_SEQ)
 733                         tunnel->o_seqno++;
 734
 735                 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
 736                                  protocol, tunnel->parms.o_key,
 737                                  htonl(tunnel->o_seqno));
 738         }
 739
 740         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
 741                             NEXTHDR_GRE);
 742 }
 743
 744 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 745 {
 746         struct ip6_tnl *t = netdev_priv(dev);
 747         int encap_limit = -1;
 748         struct flowi6 fl6;
 749         __u8 dsfield = 0;
 750         __u32 mtu;
 751         int err;
 752
 753         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 754
 755         if (!t->parms.collect_md)
 756                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
 757                                          &dsfield, &encap_limit);
 758
 759         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 760         if (err)
 761                 return -1;
 762
 763         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 764                           skb->protocol);
 765         if (err != 0) {
 766                 /* XXX: send ICMP error even if DF is not set. */
 767                 if (err == -EMSGSIZE)
 768                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 769                                   htonl(mtu));
 770                 return -1;
 771         }
 772
 773         return 0;
 774 }
 775
 776 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 777 {
 778         struct ip6_tnl *t = netdev_priv(dev);
 779         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 780         int encap_limit = -1;
 781         struct flowi6 fl6;
 782         __u8 dsfield = 0;
 783         __u32 mtu;
 784         int err;
 785
 786         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 787                 return -1;
 788
 789         if (!t->parms.collect_md &&
 790             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
 791                 return -1;
 792
 793         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
 794                 return -1;
 795
 796         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
 797                           &mtu, skb->protocol);
 798         if (err != 0) {
 799                 if (err == -EMSGSIZE)
 800                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 801                 return -1;
 802         }
 803
 804         return 0;
 805 }
 806
 807 /**
 808  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 809  *   @t: the outgoing tunnel device
 810  *   @hdr: IPv6 header from the incoming packet
 811  *
 812  * Description:
 813  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 814  *   doesn't match source of incoming packet.
 815  *
 816  * Return:
 817  *   1 if conflict,
 818  *   0 else
 819  **/
 820
 821 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 822         const struct ipv6hdr *hdr)
 823 {
 824         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 825 }
 826
 827 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 828 {
 829         struct ip6_tnl *t = netdev_priv(dev);
 830         int encap_limit = -1;
 831         struct flowi6 fl6;
 832         __u32 mtu;
 833         int err;
 834
 835         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 836                 encap_limit = t->parms.encap_limit;
 837
 838         if (!t->parms.collect_md)
 839                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 840
 841         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 842         if (err)
 843                 return err;
 844
 845         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
 846
 847         return err;
 848 }
 849
 850 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 851         struct net_device *dev)
 852 {
 853         struct ip6_tnl *t = netdev_priv(dev);
 854         struct net_device_stats *stats = &t->dev->stats;
 855         int ret;
 856
 857         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 858                 goto tx_err;
 859
 860         switch (skb->protocol) {
 861         case htons(ETH_P_IP):
 862                 ret = ip6gre_xmit_ipv4(skb, dev);
 863                 break;
 864         case htons(ETH_P_IPV6):
 865                 ret = ip6gre_xmit_ipv6(skb, dev);
 866                 break;
 867         default:
 868                 ret = ip6gre_xmit_other(skb, dev);
 869                 break;
 870         }
 871
 872         if (ret < 0)
 873                 goto tx_err;
 874
 875         return NETDEV_TX_OK;
 876
 877 tx_err:
 878         stats->tx_errors++;
 879         stats->tx_dropped++;
 880         kfree_skb(skb);
 881         return NETDEV_TX_OK;
 882 }
 883
 884 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
 885                                          struct net_device *dev)
 886 {
 887         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 888         struct ip6_tnl *t = netdev_priv(dev);
 889         struct dst_entry *dst = skb_dst(skb);
 890         struct net_device_stats *stats;
 891         bool truncate = false;
 892         int encap_limit = -1;
 893         __u8 dsfield = false;
 894         struct flowi6 fl6;
 895         int err = -EINVAL;
 896         __u32 mtu;
 897
 898         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 899                 goto tx_err;
 900
 901         if (gre_handle_offloads(skb, false))
 902                 goto tx_err;
 903
 904         if (skb->len > dev->mtu + dev->hard_header_len) {
 905                 pskb_trim(skb, dev->mtu + dev->hard_header_len);
 906                 truncate = true;
 907         }
 908
 909         if (skb_cow_head(skb, dev->needed_headroom))
 910                 goto tx_err;
 911
 912         t->parms.o_flags &= ~TUNNEL_KEY;
 913         IPCB(skb)->flags = 0;
 914
 915         /* For collect_md mode, derive fl6 from the tunnel key,
 916          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
 917          */
 918         if (t->parms.collect_md) {
 919                 struct ip_tunnel_info *tun_info;
 920                 const struct ip_tunnel_key *key;
 921                 struct erspan_metadata *md;
 922                 __be32 tun_id;
 923
 924                 tun_info = skb_tunnel_info(skb);
 925                 if (unlikely(!tun_info ||
 926                              !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
 927                              ip_tunnel_info_af(tun_info) != AF_INET6))
 928                         return -EINVAL;
 929
 930                 key = &tun_info->key;
 931                 memset(&fl6, 0, sizeof(fl6));
 932                 fl6.flowi6_proto = IPPROTO_GRE;
 933                 fl6.daddr = key->u.ipv6.dst;
 934                 fl6.flowlabel = key->label;
 935                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 936
 937                 dsfield = key->tos;
 938                 md = ip_tunnel_info_opts(tun_info);
 939                 if (!md)
 940                         goto tx_err;
 941
 942                 tun_id = tunnel_id_to_key32(key->tun_id);
 943                 if (md->version == 1) {
 944                         erspan_build_header(skb,
 945                                             ntohl(tun_id),
 946                                             ntohl(md->u.index), truncate,
 947                                             false);
 948                 } else if (md->version == 2) {
 949                         erspan_build_header_v2(skb,
 950                                                ntohl(tun_id),
 951                                                md->u.md2.dir,
 952                                                get_hwid(&md->u.md2),
 953                                                truncate, false);
 954                 } else {
 955                         goto tx_err;
 956                 }
 957         } else {
 958                 switch (skb->protocol) {
 959                 case htons(ETH_P_IP):
 960                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 961                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
 962                                                  &dsfield, &encap_limit);
 963                         break;
 964                 case htons(ETH_P_IPV6):
 965                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 966                                 goto tx_err;
 967                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
 968                                                      &dsfield, &encap_limit))
 969                                 goto tx_err;
 970                         break;
 971                 default:
 972                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 973                         break;
 974                 }
 975
 976                 if (t->parms.erspan_ver == 1)
 977                         erspan_build_header(skb, ntohl(t->parms.o_key),
 978                                             t->parms.index,
 979                                             truncate, false);
 980                 else
 981                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
 982                                                t->parms.dir,
 983                                                t->parms.hwid,
 984                                                truncate, false);
 985                 fl6.daddr = t->parms.raddr;
 986         }
 987
 988         /* Push GRE header. */
 989         gre_build_header(skb, 8, TUNNEL_SEQ,
 990                          htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
 991
 992         /* TooBig packet may have updated dst->dev's mtu */
 993         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
 994                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
 995
 996         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 997                            NEXTHDR_GRE);
 998         if (err != 0) {
 999                 /* XXX: send ICMP error even if DF is not set. */
1000                 if (err == -EMSGSIZE) {
1001                         if (skb->protocol == htons(ETH_P_IP))
1002                                 icmp_send(skb, ICMP_DEST_UNREACH,
1003                                           ICMP_FRAG_NEEDED, htonl(mtu));
1004                         else
1005                                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1006                 }
1007
1008                 goto tx_err;
1009         }
1010         return NETDEV_TX_OK;
1011
1012 tx_err:
1013         stats = &t->dev->stats;
1014         stats->tx_errors++;
1015         stats->tx_dropped++;
1016         kfree_skb(skb);
1017         return NETDEV_TX_OK;
1018 }
1019
1020 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1021 {
1022         struct net_device *dev = t->dev;
1023         struct __ip6_tnl_parm *p = &t->parms;
1024         struct flowi6 *fl6 = &t->fl.u.ip6;
1025         int t_hlen;
1026
1027         if (dev->type != ARPHRD_ETHER) {
1028                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1029                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1030         }
1031
1032         /* Set up flowi template */
1033         fl6->saddr = p->laddr;
1034         fl6->daddr = p->raddr;
1035         fl6->flowi6_oif = p->link;
1036         fl6->flowlabel = 0;
1037         fl6->flowi6_proto = IPPROTO_GRE;
1038
1039         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1040                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1041         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1042                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1043
1044         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1045         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1046
1047         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1048                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1049                 dev->flags |= IFF_POINTOPOINT;
1050         else
1051                 dev->flags &= ~IFF_POINTOPOINT;
1052
1053         t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
1054
1055         t->hlen = t->encap_hlen + t->tun_hlen;
1056
1057         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1058
1059         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1060                 int strict = (ipv6_addr_type(&p->raddr) &
1061                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1062
1063                 struct rt6_info *rt = rt6_lookup(t->net,
1064                                                  &p->raddr, &p->laddr,
1065                                                  p->link, NULL, strict);
1066
1067                 if (!rt)
1068                         return;
1069
1070                 if (rt->dst.dev) {
1071                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1072                                                t_hlen;
1073
1074                         if (set_mtu) {
1075                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
1076                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1077                                         dev->mtu -= 8;
1078                                 if (dev->type == ARPHRD_ETHER)
1079                                         dev->mtu -= ETH_HLEN;
1080
1081                                 if (dev->mtu < IPV6_MIN_MTU)
1082                                         dev->mtu = IPV6_MIN_MTU;
1083                         }
1084                 }
1085                 ip6_rt_put(rt);
1086         }
1087 }
1088
1089 static int ip6gre_tnl_change(struct ip6_tnl *t,
1090         const struct __ip6_tnl_parm *p, int set_mtu)
1091 {
1092         t->parms.laddr = p->laddr;
1093         t->parms.raddr = p->raddr;
1094         t->parms.flags = p->flags;
1095         t->parms.hop_limit = p->hop_limit;
1096         t->parms.encap_limit = p->encap_limit;
1097         t->parms.flowinfo = p->flowinfo;
1098         t->parms.link = p->link;
1099         t->parms.proto = p->proto;
1100         t->parms.i_key = p->i_key;
1101         t->parms.o_key = p->o_key;
1102         t->parms.i_flags = p->i_flags;
1103         t->parms.o_flags = p->o_flags;
1104         t->parms.fwmark = p->fwmark;
1105         dst_cache_reset(&t->dst_cache);
1106         ip6gre_tnl_link_config(t, set_mtu);
1107         return 0;
1108 }
1109
1110 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1111         const struct ip6_tnl_parm2 *u)
1112 {
1113         p->laddr = u->laddr;
1114         p->raddr = u->raddr;
1115         p->flags = u->flags;
1116         p->hop_limit = u->hop_limit;
1117         p->encap_limit = u->encap_limit;
1118         p->flowinfo = u->flowinfo;
1119         p->link = u->link;
1120         p->i_key = u->i_key;
1121         p->o_key = u->o_key;
1122         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1123         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1124         memcpy(p->name, u->name, sizeof(u->name));
1125 }
1126
1127 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1128         const struct __ip6_tnl_parm *p)
1129 {
1130         u->proto = IPPROTO_GRE;
1131         u->laddr = p->laddr;
1132         u->raddr = p->raddr;
1133         u->flags = p->flags;
1134         u->hop_limit = p->hop_limit;
1135         u->encap_limit = p->encap_limit;
1136         u->flowinfo = p->flowinfo;
1137         u->link = p->link;
1138         u->i_key = p->i_key;
1139         u->o_key = p->o_key;
1140         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1141         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1142         memcpy(u->name, p->name, sizeof(u->name));
1143 }
1144
1145 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1146         struct ifreq *ifr, int cmd)
1147 {
1148         int err = 0;
1149         struct ip6_tnl_parm2 p;
1150         struct __ip6_tnl_parm p1;
1151         struct ip6_tnl *t = netdev_priv(dev);
1152         struct net *net = t->net;
1153         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1154
1155         memset(&p1, 0, sizeof(p1));
1156
1157         switch (cmd) {
1158         case SIOCGETTUNNEL:
1159                 if (dev == ign->fb_tunnel_dev) {
1160                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1161                                 err = -EFAULT;
1162                                 break;
1163                         }
1164                         ip6gre_tnl_parm_from_user(&p1, &p);
1165                         t = ip6gre_tunnel_locate(net, &p1, 0);
1166                         if (!t)
1167                                 t = netdev_priv(dev);
1168                 }
1169                 memset(&p, 0, sizeof(p));
1170                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1171                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1172                         err = -EFAULT;
1173                 break;
1174
1175         case SIOCADDTUNNEL:
1176         case SIOCCHGTUNNEL:
1177                 err = -EPERM;
1178                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1179                         goto done;
1180
1181                 err = -EFAULT;
1182                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1183                         goto done;
1184
1185                 err = -EINVAL;
1186                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1187                         goto done;
1188
1189                 if (!(p.i_flags&GRE_KEY))
1190                         p.i_key = 0;
1191                 if (!(p.o_flags&GRE_KEY))
1192                         p.o_key = 0;
1193
1194                 ip6gre_tnl_parm_from_user(&p1, &p);
1195                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1196
1197                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1198                         if (t) {
1199                                 if (t->dev != dev) {
1200                                         err = -EEXIST;
1201                                         break;
1202                                 }
1203                         } else {
1204                                 t = netdev_priv(dev);
1205
1206                                 ip6gre_tunnel_unlink(ign, t);
1207                                 synchronize_net();
1208                                 ip6gre_tnl_change(t, &p1, 1);
1209                                 ip6gre_tunnel_link(ign, t);
1210                                 netdev_state_change(dev);
1211                         }
1212                 }
1213
1214                 if (t) {
1215                         err = 0;
1216
1217                         memset(&p, 0, sizeof(p));
1218                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1219                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1220                                 err = -EFAULT;
1221                 } else
1222                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1223                 break;
1224
1225         case SIOCDELTUNNEL:
1226                 err = -EPERM;
1227                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1228                         goto done;
1229
1230                 if (dev == ign->fb_tunnel_dev) {
1231                         err = -EFAULT;
1232                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1233                                 goto done;
1234                         err = -ENOENT;
1235                         ip6gre_tnl_parm_from_user(&p1, &p);
1236                         t = ip6gre_tunnel_locate(net, &p1, 0);
1237                         if (!t)
1238                                 goto done;
1239                         err = -EPERM;
1240                         if (t == netdev_priv(ign->fb_tunnel_dev))
1241                                 goto done;
1242                         dev = t->dev;
1243                 }
1244                 unregister_netdevice(dev);
1245                 err = 0;
1246                 break;
1247
1248         default:
1249                 err = -EINVAL;
1250         }
1251
1252 done:
1253         return err;
1254 }
1255
1256 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1257                          unsigned short type, const void *daddr,
1258                          const void *saddr, unsigned int len)
1259 {
1260         struct ip6_tnl *t = netdev_priv(dev);
1261         struct ipv6hdr *ipv6h;
1262         __be16 *p;
1263
1264         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1265         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1266                                                   t->fl.u.ip6.flowlabel,
1267                                                   true, &t->fl.u.ip6));
1268         ipv6h->hop_limit = t->parms.hop_limit;
1269         ipv6h->nexthdr = NEXTHDR_GRE;
1270         ipv6h->saddr = t->parms.laddr;
1271         ipv6h->daddr = t->parms.raddr;
1272
1273         p = (__be16 *)(ipv6h + 1);
1274         p[0] = t->parms.o_flags;
1275         p[1] = htons(type);
1276
1277         /*
1278          *      Set the source hardware address.
1279          */
1280
1281         if (saddr)
1282                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1283         if (daddr)
1284                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1285         if (!ipv6_addr_any(&ipv6h->daddr))
1286                 return t->hlen;
1287
1288         return -t->hlen;
1289 }
1290
1291 static const struct header_ops ip6gre_header_ops = {
1292         .create = ip6gre_header,
1293 };
1294
1295 static const struct net_device_ops ip6gre_netdev_ops = {
1296         .ndo_init               = ip6gre_tunnel_init,
1297         .ndo_uninit             = ip6gre_tunnel_uninit,
1298         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1299         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1300         .ndo_change_mtu         = ip6_tnl_change_mtu,
1301         .ndo_get_stats64        = ip_tunnel_get_stats64,
1302         .ndo_get_iflink         = ip6_tnl_get_iflink,
1303 };
1304
1305 static void ip6gre_dev_free(struct net_device *dev)
1306 {
1307         struct ip6_tnl *t = netdev_priv(dev);
1308
1309         dst_cache_destroy(&t->dst_cache);
1310         free_percpu(dev->tstats);
1311 }
1312
1313 static void ip6gre_tunnel_setup(struct net_device *dev)
1314 {
1315         dev->netdev_ops = &ip6gre_netdev_ops;
1316         dev->needs_free_netdev = true;
1317         dev->priv_destructor = ip6gre_dev_free;
1318
1319         dev->type = ARPHRD_IP6GRE;
1320
1321         dev->flags |= IFF_NOARP;
1322         dev->addr_len = sizeof(struct in6_addr);
1323         netif_keep_dst(dev);
1324         /* This perm addr will be used as interface identifier by IPv6 */
1325         dev->addr_assign_type = NET_ADDR_RANDOM;
1326         eth_random_addr(dev->perm_addr);
1327 }
1328
1329 #define GRE6_FEATURES (NETIF_F_SG |             \
1330                        NETIF_F_FRAGLIST |       \
1331                        NETIF_F_HIGHDMA |        \
1332                        NETIF_F_HW_CSUM)
1333
1334 static void ip6gre_tnl_init_features(struct net_device *dev)
1335 {
1336         struct ip6_tnl *nt = netdev_priv(dev);
1337
1338         dev->features           |= GRE6_FEATURES;
1339         dev->hw_features        |= GRE6_FEATURES;
1340
1341         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1342                 /* TCP offload with GRE SEQ is not supported, nor
1343                  * can we support 2 levels of outer headers requiring
1344                  * an update.
1345                  */
1346                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1347                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1348                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1349                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1350                 }
1351
1352                 /* Can use a lockless transmit, unless we generate
1353                  * output sequences
1354                  */
1355                 dev->features |= NETIF_F_LLTX;
1356         }
1357 }
1358
1359 static int ip6gre_tunnel_init_common(struct net_device *dev)
1360 {
1361         struct ip6_tnl *tunnel;
1362         int ret;
1363         int t_hlen;
1364
1365         tunnel = netdev_priv(dev);
1366
1367         tunnel->dev = dev;
1368         tunnel->net = dev_net(dev);
1369         strcpy(tunnel->parms.name, dev->name);
1370
1371         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1372         if (!dev->tstats)
1373                 return -ENOMEM;
1374
1375         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1376         if (ret) {
1377                 free_percpu(dev->tstats);
1378                 dev->tstats = NULL;
1379                 return ret;
1380         }
1381
1382         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1383         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1384         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1385
1386         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1387         dev->mtu = ETH_DATA_LEN - t_hlen;
1388         if (dev->type == ARPHRD_ETHER)
1389                 dev->mtu -= ETH_HLEN;
1390         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1391                 dev->mtu -= 8;
1392
1393         if (tunnel->parms.collect_md) {
1394                 dev->features |= NETIF_F_NETNS_LOCAL;
1395                 netif_keep_dst(dev);
1396         }
1397         ip6gre_tnl_init_features(dev);
1398
1399         return 0;
1400 }
1401
1402 static int ip6gre_tunnel_init(struct net_device *dev)
1403 {
1404         struct ip6_tnl *tunnel;
1405         int ret;
1406
1407         ret = ip6gre_tunnel_init_common(dev);
1408         if (ret)
1409                 return ret;
1410
1411         tunnel = netdev_priv(dev);
1412
1413         if (tunnel->parms.collect_md)
1414                 return 0;
1415
1416         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1417         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1418
1419         if (ipv6_addr_any(&tunnel->parms.raddr))
1420                 dev->header_ops = &ip6gre_header_ops;
1421
1422         return 0;
1423 }
1424
1425 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1426 {
1427         struct ip6_tnl *tunnel = netdev_priv(dev);
1428
1429         tunnel->dev = dev;
1430         tunnel->net = dev_net(dev);
1431         strcpy(tunnel->parms.name, dev->name);
1432
1433         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1434
1435         dev_hold(dev);
1436 }
1437
1438 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1439         .handler     = gre_rcv,
1440         .err_handler = ip6gre_err,
1441         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1442 };
1443
1444 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1445 {
1446         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1447         struct net_device *dev, *aux;
1448         int prio;
1449
1450         for_each_netdev_safe(net, dev, aux)
1451                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1452                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1453                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1454                         unregister_netdevice_queue(dev, head);
1455
1456         for (prio = 0; prio < 4; prio++) {
1457                 int h;
1458                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1459                         struct ip6_tnl *t;
1460
1461                         t = rtnl_dereference(ign->tunnels[prio][h]);
1462
1463                         while (t) {
1464                                 /* If dev is in the same netns, it has already
1465                                  * been added to the list by the previous loop.
1466                                  */
1467                                 if (!net_eq(dev_net(t->dev), net))
1468                                         unregister_netdevice_queue(t->dev,
1469                                                                    head);
1470                                 t = rtnl_dereference(t->next);
1471                         }
1472                 }
1473         }
1474 }
1475
1476 static int __net_init ip6gre_init_net(struct net *net)
1477 {
1478         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1479         int err;
1480
1481         if (!net_has_fallback_tunnels(net))
1482                 return 0;
1483         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1484                                           NET_NAME_UNKNOWN,
1485                                           ip6gre_tunnel_setup);
1486         if (!ign->fb_tunnel_dev) {
1487                 err = -ENOMEM;
1488                 goto err_alloc_dev;
1489         }
1490         dev_net_set(ign->fb_tunnel_dev, net);
1491         /* FB netdevice is special: we have one, and only one per netns.
1492          * Allowing to move it to another netns is clearly unsafe.
1493          */
1494         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1495
1496
1497         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1498         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1499
1500         err = register_netdev(ign->fb_tunnel_dev);
1501         if (err)
1502                 goto err_reg_dev;
1503
1504         rcu_assign_pointer(ign->tunnels_wc[0],
1505                            netdev_priv(ign->fb_tunnel_dev));
1506         return 0;
1507
1508 err_reg_dev:
1509         free_netdev(ign->fb_tunnel_dev);
1510 err_alloc_dev:
1511         return err;
1512 }
1513
1514 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1515 {
1516         struct net *net;
1517         LIST_HEAD(list);
1518
1519         rtnl_lock();
1520         list_for_each_entry(net, net_list, exit_list)
1521                 ip6gre_destroy_tunnels(net, &list);
1522         unregister_netdevice_many(&list);
1523         rtnl_unlock();
1524 }
1525
1526 static struct pernet_operations ip6gre_net_ops = {
1527         .init = ip6gre_init_net,
1528         .exit_batch = ip6gre_exit_batch_net,
1529         .id   = &ip6gre_net_id,
1530         .size = sizeof(struct ip6gre_net),
1531 };
1532
1533 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1534                                   struct netlink_ext_ack *extack)
1535 {
1536         __be16 flags;
1537
1538         if (!data)
1539                 return 0;
1540
1541         flags = 0;
1542         if (data[IFLA_GRE_IFLAGS])
1543                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1544         if (data[IFLA_GRE_OFLAGS])
1545                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1546         if (flags & (GRE_VERSION|GRE_ROUTING))
1547                 return -EINVAL;
1548
1549         return 0;
1550 }
1551
1552 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1553                                struct netlink_ext_ack *extack)
1554 {
1555         struct in6_addr daddr;
1556
1557         if (tb[IFLA_ADDRESS]) {
1558                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1559                         return -EINVAL;
1560                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1561                         return -EADDRNOTAVAIL;
1562         }
1563
1564         if (!data)
1565                 goto out;
1566
1567         if (data[IFLA_GRE_REMOTE]) {
1568                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1569                 if (ipv6_addr_any(&daddr))
1570                         return -EINVAL;
1571         }
1572
1573 out:
1574         return ip6gre_tunnel_validate(tb, data, extack);
1575 }
1576
1577 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1578                                   struct netlink_ext_ack *extack)
1579 {
1580         __be16 flags = 0;
1581         int ret, ver = 0;
1582
1583         if (!data)
1584                 return 0;
1585
1586         ret = ip6gre_tap_validate(tb, data, extack);
1587         if (ret)
1588                 return ret;
1589
1590         /* ERSPAN should only have GRE sequence and key flag */
1591         if (data[IFLA_GRE_OFLAGS])
1592                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1593         if (data[IFLA_GRE_IFLAGS])
1594                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1595         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1596             flags != (GRE_SEQ | GRE_KEY))
1597                 return -EINVAL;
1598
1599         /* ERSPAN Session ID only has 10-bit. Since we reuse
1600          * 32-bit key field as ID, check it's range.
1601          */
1602         if (data[IFLA_GRE_IKEY] &&
1603             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1604                 return -EINVAL;
1605
1606         if (data[IFLA_GRE_OKEY] &&
1607             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1608                 return -EINVAL;
1609
1610         if (data[IFLA_GRE_ERSPAN_VER]) {
1611                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1612                 if (ver != 1 && ver != 2)
1613                         return -EINVAL;
1614         }
1615
1616         if (ver == 1) {
1617                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1618                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1619
1620                         if (index & ~INDEX_MASK)
1621                                 return -EINVAL;
1622                 }
1623         } else if (ver == 2) {
1624                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1625                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1626
1627                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1628                                 return -EINVAL;
1629                 }
1630
1631                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1632                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1633
1634                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1635                                 return -EINVAL;
1636                 }
1637         }
1638
1639         return 0;
1640 }
1641
1642 static void ip6gre_netlink_parms(struct nlattr *data[],
1643                                 struct __ip6_tnl_parm *parms)
1644 {
1645         memset(parms, 0, sizeof(*parms));
1646
1647         if (!data)
1648                 return;
1649
1650         if (data[IFLA_GRE_LINK])
1651                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1652
1653         if (data[IFLA_GRE_IFLAGS])
1654                 parms->i_flags = gre_flags_to_tnl_flags(
1655                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1656
1657         if (data[IFLA_GRE_OFLAGS])
1658                 parms->o_flags = gre_flags_to_tnl_flags(
1659                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1660
1661         if (data[IFLA_GRE_IKEY])
1662                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1663
1664         if (data[IFLA_GRE_OKEY])
1665                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1666
1667         if (data[IFLA_GRE_LOCAL])
1668                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1669
1670         if (data[IFLA_GRE_REMOTE])
1671                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1672
1673         if (data[IFLA_GRE_TTL])
1674                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1675
1676         if (data[IFLA_GRE_ENCAP_LIMIT])
1677                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1678
1679         if (data[IFLA_GRE_FLOWINFO])
1680                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1681
1682         if (data[IFLA_GRE_FLAGS])
1683                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1684
1685         if (data[IFLA_GRE_FWMARK])
1686                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1687
1688         if (data[IFLA_GRE_COLLECT_METADATA])
1689                 parms->collect_md = true;
1690
1691         if (data[IFLA_GRE_ERSPAN_VER])
1692                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1693
1694         if (parms->erspan_ver == 1) {
1695                 if (data[IFLA_GRE_ERSPAN_INDEX])
1696                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1697         } else if (parms->erspan_ver == 2) {
1698                 if (data[IFLA_GRE_ERSPAN_DIR])
1699                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1700                 if (data[IFLA_GRE_ERSPAN_HWID])
1701                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1702         }
1703 }
1704
1705 static int ip6gre_tap_init(struct net_device *dev)
1706 {
1707         int ret;
1708
1709         ret = ip6gre_tunnel_init_common(dev);
1710         if (ret)
1711                 return ret;
1712
1713         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1714
1715         return 0;
1716 }
1717
1718 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1719         .ndo_init = ip6gre_tap_init,
1720         .ndo_uninit = ip6gre_tunnel_uninit,
1721         .ndo_start_xmit = ip6gre_tunnel_xmit,
1722         .ndo_set_mac_address = eth_mac_addr,
1723         .ndo_validate_addr = eth_validate_addr,
1724         .ndo_change_mtu = ip6_tnl_change_mtu,
1725         .ndo_get_stats64 = ip_tunnel_get_stats64,
1726         .ndo_get_iflink = ip6_tnl_get_iflink,
1727 };
1728
1729 static int ip6erspan_tap_init(struct net_device *dev)
1730 {
1731         struct ip6_tnl *tunnel;
1732         int t_hlen;
1733         int ret;
1734
1735         tunnel = netdev_priv(dev);
1736
1737         tunnel->dev = dev;
1738         tunnel->net = dev_net(dev);
1739         strcpy(tunnel->parms.name, dev->name);
1740
1741         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1742         if (!dev->tstats)
1743                 return -ENOMEM;
1744
1745         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1746         if (ret) {
1747                 free_percpu(dev->tstats);
1748                 dev->tstats = NULL;
1749                 return ret;
1750         }
1751
1752         tunnel->tun_hlen = 8;
1753         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1754                        erspan_hdr_len(tunnel->parms.erspan_ver);
1755         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1756
1757         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1758         dev->mtu = ETH_DATA_LEN - t_hlen;
1759         if (dev->type == ARPHRD_ETHER)
1760                 dev->mtu -= ETH_HLEN;
1761         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1762                 dev->mtu -= 8;
1763
1764         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1765         ip6gre_tnl_link_config(tunnel, 1);
1766
1767         return 0;
1768 }
1769
1770 static const struct net_device_ops ip6erspan_netdev_ops = {
1771         .ndo_init =             ip6erspan_tap_init,
1772         .ndo_uninit =           ip6gre_tunnel_uninit,
1773         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1774         .ndo_set_mac_address =  eth_mac_addr,
1775         .ndo_validate_addr =    eth_validate_addr,
1776         .ndo_change_mtu =       ip6_tnl_change_mtu,
1777         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1778         .ndo_get_iflink =       ip6_tnl_get_iflink,
1779 };
1780
1781 static void ip6gre_tap_setup(struct net_device *dev)
1782 {
1783
1784         ether_setup(dev);
1785
1786         dev->max_mtu = 0;
1787         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1788         dev->needs_free_netdev = true;
1789         dev->priv_destructor = ip6gre_dev_free;
1790
1791         dev->features |= NETIF_F_NETNS_LOCAL;
1792         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1793         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1794         netif_keep_dst(dev);
1795 }
1796
1797 bool is_ip6gretap_dev(const struct net_device *dev)
1798 {
1799         return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1800 }
1801 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1802
1803 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1804                                        struct ip_tunnel_encap *ipencap)
1805 {
1806         bool ret = false;
1807
1808         memset(ipencap, 0, sizeof(*ipencap));
1809
1810         if (!data)
1811                 return ret;
1812
1813         if (data[IFLA_GRE_ENCAP_TYPE]) {
1814                 ret = true;
1815                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1816         }
1817
1818         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1819                 ret = true;
1820                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1821         }
1822
1823         if (data[IFLA_GRE_ENCAP_SPORT]) {
1824                 ret = true;
1825                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1826         }
1827
1828         if (data[IFLA_GRE_ENCAP_DPORT]) {
1829                 ret = true;
1830                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1831         }
1832
1833         return ret;
1834 }
1835
1836 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1837                           struct nlattr *tb[], struct nlattr *data[],
1838                           struct netlink_ext_ack *extack)
1839 {
1840         struct ip6_tnl *nt;
1841         struct net *net = dev_net(dev);
1842         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1843         struct ip_tunnel_encap ipencap;
1844         int err;
1845
1846         nt = netdev_priv(dev);
1847
1848         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1849                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1850
1851                 if (err < 0)
1852                         return err;
1853         }
1854
1855         ip6gre_netlink_parms(data, &nt->parms);
1856
1857         if (nt->parms.collect_md) {
1858                 if (rtnl_dereference(ign->collect_md_tun))
1859                         return -EEXIST;
1860         } else {
1861                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1862                         return -EEXIST;
1863         }
1864
1865         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1866                 eth_hw_addr_random(dev);
1867
1868         nt->dev = dev;
1869         nt->net = dev_net(dev);
1870
1871         err = register_netdevice(dev);
1872         if (err)
1873                 goto out;
1874
1875         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1876
1877         if (tb[IFLA_MTU])
1878                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1879
1880         dev_hold(dev);
1881         ip6gre_tunnel_link(ign, nt);
1882
1883 out:
1884         return err;
1885 }
1886
1887 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1888                              struct nlattr *data[],
1889                              struct netlink_ext_ack *extack)
1890 {
1891         struct ip6_tnl *t, *nt = netdev_priv(dev);
1892         struct net *net = nt->net;
1893         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1894         struct __ip6_tnl_parm p;
1895         struct ip_tunnel_encap ipencap;
1896
1897         if (dev == ign->fb_tunnel_dev)
1898                 return -EINVAL;
1899
1900         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1901                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1902
1903                 if (err < 0)
1904                         return err;
1905         }
1906
1907         ip6gre_netlink_parms(data, &p);
1908
1909         t = ip6gre_tunnel_locate(net, &p, 0);
1910
1911         if (t) {
1912                 if (t->dev != dev)
1913                         return -EEXIST;
1914         } else {
1915                 t = nt;
1916         }
1917
1918         ip6gre_tunnel_unlink(ign, t);
1919         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1920         ip6gre_tunnel_link(ign, t);
1921         return 0;
1922 }
1923
1924 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1925 {
1926         struct net *net = dev_net(dev);
1927         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1928
1929         if (dev != ign->fb_tunnel_dev)
1930                 unregister_netdevice_queue(dev, head);
1931 }
1932
1933 static size_t ip6gre_get_size(const struct net_device *dev)
1934 {
1935         return
1936                 /* IFLA_GRE_LINK */
1937                 nla_total_size(4) +
1938                 /* IFLA_GRE_IFLAGS */
1939                 nla_total_size(2) +
1940                 /* IFLA_GRE_OFLAGS */
1941                 nla_total_size(2) +
1942                 /* IFLA_GRE_IKEY */
1943                 nla_total_size(4) +
1944                 /* IFLA_GRE_OKEY */
1945                 nla_total_size(4) +
1946                 /* IFLA_GRE_LOCAL */
1947                 nla_total_size(sizeof(struct in6_addr)) +
1948                 /* IFLA_GRE_REMOTE */
1949                 nla_total_size(sizeof(struct in6_addr)) +
1950                 /* IFLA_GRE_TTL */
1951                 nla_total_size(1) +
1952                 /* IFLA_GRE_ENCAP_LIMIT */
1953                 nla_total_size(1) +
1954                 /* IFLA_GRE_FLOWINFO */
1955                 nla_total_size(4) +
1956                 /* IFLA_GRE_FLAGS */
1957                 nla_total_size(4) +
1958                 /* IFLA_GRE_ENCAP_TYPE */
1959                 nla_total_size(2) +
1960                 /* IFLA_GRE_ENCAP_FLAGS */
1961                 nla_total_size(2) +
1962                 /* IFLA_GRE_ENCAP_SPORT */
1963                 nla_total_size(2) +
1964                 /* IFLA_GRE_ENCAP_DPORT */
1965                 nla_total_size(2) +
1966                 /* IFLA_GRE_COLLECT_METADATA */
1967                 nla_total_size(0) +
1968                 /* IFLA_GRE_FWMARK */
1969                 nla_total_size(4) +
1970                 /* IFLA_GRE_ERSPAN_INDEX */
1971                 nla_total_size(4) +
1972                 0;
1973 }
1974
1975 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1976 {
1977         struct ip6_tnl *t = netdev_priv(dev);
1978         struct __ip6_tnl_parm *p = &t->parms;
1979
1980         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1981             nla_put_be16(skb, IFLA_GRE_IFLAGS,
1982                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1983             nla_put_be16(skb, IFLA_GRE_OFLAGS,
1984                          gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1985             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1986             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1987             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1988             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1989             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1990             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1991             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1992             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1993             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
1994             nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
1995                 goto nla_put_failure;
1996
1997         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1998                         t->encap.type) ||
1999             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2000                          t->encap.sport) ||
2001             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2002                          t->encap.dport) ||
2003             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2004                         t->encap.flags))
2005                 goto nla_put_failure;
2006
2007         if (p->collect_md) {
2008                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2009                         goto nla_put_failure;
2010         }
2011
2012         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2013                 goto nla_put_failure;
2014
2015         if (p->erspan_ver == 1) {
2016                 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2017                         goto nla_put_failure;
2018         } else if (p->erspan_ver == 2) {
2019                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2020                         goto nla_put_failure;
2021                 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2022                         goto nla_put_failure;
2023         }
2024
2025         return 0;
2026
2027 nla_put_failure:
2028         return -EMSGSIZE;
2029 }
2030
2031 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2032         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2033         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2034         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2035         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2036         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2037         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2038         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2039         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2040         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2041         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2042         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2043         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2044         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2045         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2046         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2047         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2048         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2049         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2050         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2051         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2052         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2053 };
2054
2055 static void ip6erspan_tap_setup(struct net_device *dev)
2056 {
2057         ether_setup(dev);
2058
2059         dev->netdev_ops = &ip6erspan_netdev_ops;
2060         dev->needs_free_netdev = true;
2061         dev->priv_destructor = ip6gre_dev_free;
2062
2063         dev->features |= NETIF_F_NETNS_LOCAL;
2064         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2065         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2066         netif_keep_dst(dev);
2067 }
2068
2069 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2070         .kind           = "ip6gre",
2071         .maxtype        = IFLA_GRE_MAX,
2072         .policy         = ip6gre_policy,
2073         .priv_size      = sizeof(struct ip6_tnl),
2074         .setup          = ip6gre_tunnel_setup,
2075         .validate       = ip6gre_tunnel_validate,
2076         .newlink        = ip6gre_newlink,
2077         .changelink     = ip6gre_changelink,
2078         .dellink        = ip6gre_dellink,
2079         .get_size       = ip6gre_get_size,
2080         .fill_info      = ip6gre_fill_info,
2081         .get_link_net   = ip6_tnl_get_link_net,
2082 };
2083
2084 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2085         .kind           = "ip6gretap",
2086         .maxtype        = IFLA_GRE_MAX,
2087         .policy         = ip6gre_policy,
2088         .priv_size      = sizeof(struct ip6_tnl),
2089         .setup          = ip6gre_tap_setup,
2090         .validate       = ip6gre_tap_validate,
2091         .newlink        = ip6gre_newlink,
2092         .changelink     = ip6gre_changelink,
2093         .get_size       = ip6gre_get_size,
2094         .fill_info      = ip6gre_fill_info,
2095         .get_link_net   = ip6_tnl_get_link_net,
2096 };
2097
2098 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2099         .kind           = "ip6erspan",
2100         .maxtype        = IFLA_GRE_MAX,
2101         .policy         = ip6gre_policy,
2102         .priv_size      = sizeof(struct ip6_tnl),
2103         .setup          = ip6erspan_tap_setup,
2104         .validate       = ip6erspan_tap_validate,
2105         .newlink        = ip6gre_newlink,
2106         .changelink     = ip6gre_changelink,
2107         .get_size       = ip6gre_get_size,
2108         .fill_info      = ip6gre_fill_info,
2109         .get_link_net   = ip6_tnl_get_link_net,
2110 };
2111
2112 /*
2113  *      And now the modules code and kernel interface.
2114  */
2115
2116 static int __init ip6gre_init(void)
2117 {
2118         int err;
2119
2120         pr_info("GRE over IPv6 tunneling driver\n");
2121
2122         err = register_pernet_device(&ip6gre_net_ops);
2123         if (err < 0)
2124                 return err;
2125
2126         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2127         if (err < 0) {
2128                 pr_info("%s: can't add protocol\n", __func__);
2129                 goto add_proto_failed;
2130         }
2131
2132         err = rtnl_link_register(&ip6gre_link_ops);
2133         if (err < 0)
2134                 goto rtnl_link_failed;
2135
2136         err = rtnl_link_register(&ip6gre_tap_ops);
2137         if (err < 0)
2138                 goto tap_ops_failed;
2139
2140         err = rtnl_link_register(&ip6erspan_tap_ops);
2141         if (err < 0)
2142                 goto erspan_link_failed;
2143
2144 out:
2145         return err;
2146
2147 erspan_link_failed:
2148         rtnl_link_unregister(&ip6gre_tap_ops);
2149 tap_ops_failed:
2150         rtnl_link_unregister(&ip6gre_link_ops);
2151 rtnl_link_failed:
2152         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2153 add_proto_failed:
2154         unregister_pernet_device(&ip6gre_net_ops);
2155         goto out;
2156 }
2157
2158 static void __exit ip6gre_fini(void)
2159 {
2160         rtnl_link_unregister(&ip6gre_tap_ops);
2161         rtnl_link_unregister(&ip6gre_link_ops);
2162         rtnl_link_unregister(&ip6erspan_tap_ops);
2163         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2164         unregister_pernet_device(&ip6gre_net_ops);
2165 }
2166
2167 module_init(ip6gre_init);
2168 module_exit(ip6gre_fini);
2169 MODULE_LICENSE("GPL");
2170 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2171 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2172 MODULE_ALIAS_RTNL_LINK("ip6gre");
2173 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2174 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2175 MODULE_ALIAS_NETDEV("ip6gre0");