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                 strlcpy(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
 386         /* Can use a lockless transmit, unless we generate output sequences */
 387         if (!(nt->parms.o_flags & TUNNEL_SEQ))
 388                 dev->features |= NETIF_F_LLTX;
 389
 390         ip6gre_tunnel_link(ign, nt);
 391         return nt;
 392
 393 failed_free:
 394         free_netdev(dev);
 395         return NULL;
 396 }
 397
 398 static void ip6erspan_tunnel_uninit(struct net_device *dev)
 399 {
 400         struct ip6_tnl *t = netdev_priv(dev);
 401         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 402
 403         ip6erspan_tunnel_unlink_md(ign, t);
 404         ip6gre_tunnel_unlink(ign, t);
 405         dst_cache_reset(&t->dst_cache);
 406         dev_put(dev);
 407 }
 408
 409 static void ip6gre_tunnel_uninit(struct net_device *dev)
 410 {
 411         struct ip6_tnl *t = netdev_priv(dev);
 412         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 413
 414         ip6gre_tunnel_unlink_md(ign, t);
 415         ip6gre_tunnel_unlink(ign, t);
 416         if (ign->fb_tunnel_dev == dev)
 417                 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
 418         dst_cache_reset(&t->dst_cache);
 419         dev_put(dev);
 420 }
 421
 422
 423 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 424                        u8 type, u8 code, int offset, __be32 info)
 425 {
 426         struct net *net = dev_net(skb->dev);
 427         const struct ipv6hdr *ipv6h;
 428         struct tnl_ptk_info tpi;
 429         struct ip6_tnl *t;
 430
 431         if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
 432                              offset) < 0)
 433                 return -EINVAL;
 434
 435         ipv6h = (const struct ipv6hdr *)skb->data;
 436         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 437                                  tpi.key, tpi.proto);
 438         if (!t)
 439                 return -ENOENT;
 440
 441         switch (type) {
 442         case ICMPV6_DEST_UNREACH:
 443                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 444                                     t->parms.name);
 445                 if (code != ICMPV6_PORT_UNREACH)
 446                         break;
 447                 return 0;
 448         case ICMPV6_TIME_EXCEED:
 449                 if (code == ICMPV6_EXC_HOPLIMIT) {
 450                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 451                                             t->parms.name);
 452                         break;
 453                 }
 454                 return 0;
 455         case ICMPV6_PARAMPROB: {
 456                 struct ipv6_tlv_tnl_enc_lim *tel;
 457                 __u32 teli;
 458
 459                 teli = 0;
 460                 if (code == ICMPV6_HDR_FIELD)
 461                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 462
 463                 if (teli && teli == be32_to_cpu(info) - 2) {
 464                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 465                         if (tel->encap_limit == 0) {
 466                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 467                                                     t->parms.name);
 468                         }
 469                 } else {
 470                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 471                                             t->parms.name);
 472                 }
 473                 return 0;
 474         }
 475         case ICMPV6_PKT_TOOBIG:
 476                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
 477                 return 0;
 478         case NDISC_REDIRECT:
 479                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 480                              sock_net_uid(net, NULL));
 481                 return 0;
 482         }
 483
 484         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 485                 t->err_count++;
 486         else
 487                 t->err_count = 1;
 488         t->err_time = jiffies;
 489
 490         return 0;
 491 }
 492
 493 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 494 {
 495         const struct ipv6hdr *ipv6h;
 496         struct ip6_tnl *tunnel;
 497
 498         ipv6h = ipv6_hdr(skb);
 499         tunnel = ip6gre_tunnel_lookup(skb->dev,
 500                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 501                                       tpi->proto);
 502         if (tunnel) {
 503                 if (tunnel->parms.collect_md) {
 504                         struct metadata_dst *tun_dst;
 505                         __be64 tun_id;
 506                         __be16 flags;
 507
 508                         flags = tpi->flags;
 509                         tun_id = key32_to_tunnel_id(tpi->key);
 510
 511                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
 512                         if (!tun_dst)
 513                                 return PACKET_REJECT;
 514
 515                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
 516                 } else {
 517                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 518                 }
 519
 520                 return PACKET_RCVD;
 521         }
 522
 523         return PACKET_REJECT;
 524 }
 525
 526 static int ip6erspan_rcv(struct sk_buff *skb,
 527                          struct tnl_ptk_info *tpi,
 528                          int gre_hdr_len)
 529 {
 530         struct erspan_base_hdr *ershdr;
 531         const struct ipv6hdr *ipv6h;
 532         struct erspan_md2 *md2;
 533         struct ip6_tnl *tunnel;
 534         u8 ver;
 535
 536         if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
 537                 return PACKET_REJECT;
 538
 539         ipv6h = ipv6_hdr(skb);
 540         ershdr = (struct erspan_base_hdr *)skb->data;
 541         ver = ershdr->ver;
 542
 543         tunnel = ip6gre_tunnel_lookup(skb->dev,
 544                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 545                                       tpi->proto);
 546         if (tunnel) {
 547                 int len = erspan_hdr_len(ver);
 548
 549                 if (unlikely(!pskb_may_pull(skb, len)))
 550                         return PACKET_REJECT;
 551
 552                 if (__iptunnel_pull_header(skb, len,
 553                                            htons(ETH_P_TEB),
 554                                            false, false) < 0)
 555                         return PACKET_REJECT;
 556
 557                 if (tunnel->parms.collect_md) {
 558                         struct erspan_metadata *pkt_md, *md;
 559                         struct metadata_dst *tun_dst;
 560                         struct ip_tunnel_info *info;
 561                         unsigned char *gh;
 562                         __be64 tun_id;
 563                         __be16 flags;
 564
 565                         tpi->flags |= TUNNEL_KEY;
 566                         flags = tpi->flags;
 567                         tun_id = key32_to_tunnel_id(tpi->key);
 568
 569                         tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
 570                                                   sizeof(*md));
 571                         if (!tun_dst)
 572                                 return PACKET_REJECT;
 573
 574                         /* skb can be uncloned in __iptunnel_pull_header, so
 575                          * old pkt_md is no longer valid and we need to reset
 576                          * it
 577                          */
 578                         gh = skb_network_header(skb) +
 579                              skb_network_header_len(skb);
 580                         pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
 581                                                             sizeof(*ershdr));
 582                         info = &tun_dst->u.tun_info;
 583                         md = ip_tunnel_info_opts(info);
 584                         md->version = ver;
 585                         md2 = &md->u.md2;
 586                         memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
 587                                                        ERSPAN_V2_MDSIZE);
 588                         info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
 589                         info->options_len = sizeof(*md);
 590
 591                         ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
 592
 593                 } else {
 594                         ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 595                 }
 596
 597                 return PACKET_RCVD;
 598         }
 599
 600         return PACKET_REJECT;
 601 }
 602
 603 static int gre_rcv(struct sk_buff *skb)
 604 {
 605         struct tnl_ptk_info tpi;
 606         bool csum_err = false;
 607         int hdr_len;
 608
 609         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
 610         if (hdr_len < 0)
 611                 goto drop;
 612
 613         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
 614                 goto drop;
 615
 616         if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
 617                      tpi.proto == htons(ETH_P_ERSPAN2))) {
 618                 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
 619                         return 0;
 620                 goto out;
 621         }
 622
 623         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
 624                 return 0;
 625
 626 out:
 627         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 628 drop:
 629         kfree_skb(skb);
 630         return 0;
 631 }
 632
 633 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
 634 {
 635         return iptunnel_handle_offloads(skb,
 636                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 637 }
 638
 639 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
 640                                      struct net_device *dev,
 641                                      struct flowi6 *fl6, __u8 *dsfield,
 642                                      int *encap_limit)
 643 {
 644         const struct iphdr *iph = ip_hdr(skb);
 645         struct ip6_tnl *t = netdev_priv(dev);
 646
 647         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 648                 *encap_limit = t->parms.encap_limit;
 649
 650         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 651
 652         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 653                 *dsfield = ipv4_get_dsfield(iph);
 654         else
 655                 *dsfield = ip6_tclass(t->parms.flowinfo);
 656
 657         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 658                 fl6->flowi6_mark = skb->mark;
 659         else
 660                 fl6->flowi6_mark = t->parms.fwmark;
 661
 662         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 663 }
 664
 665 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
 666                                     struct net_device *dev,
 667                                     struct flowi6 *fl6, __u8 *dsfield,
 668                                     int *encap_limit)
 669 {
 670         struct ipv6hdr *ipv6h;
 671         struct ip6_tnl *t = netdev_priv(dev);
 672         __u16 offset;
 673
 674         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 675         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 676         ipv6h = ipv6_hdr(skb);
 677
 678         if (offset > 0) {
 679                 struct ipv6_tlv_tnl_enc_lim *tel;
 680
 681                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 682                 if (tel->encap_limit == 0) {
 683                         icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
 684                                         ICMPV6_HDR_FIELD, offset + 2);
 685                         return -1;
 686                 }
 687                 *encap_limit = tel->encap_limit - 1;
 688         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
 689                 *encap_limit = t->parms.encap_limit;
 690         }
 691
 692         memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 693
 694         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 695                 *dsfield = ipv6_get_dsfield(ipv6h);
 696         else
 697                 *dsfield = ip6_tclass(t->parms.flowinfo);
 698
 699         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 700                 fl6->flowlabel |= ip6_flowlabel(ipv6h);
 701
 702         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 703                 fl6->flowi6_mark = skb->mark;
 704         else
 705                 fl6->flowi6_mark = t->parms.fwmark;
 706
 707         fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 708
 709         return 0;
 710 }
 711
 712 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
 713 {
 714         struct ip_tunnel_info *tun_info;
 715
 716         tun_info = skb_tunnel_info(skb);
 717         if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
 718                 return ERR_PTR(-EINVAL);
 719
 720         return tun_info;
 721 }
 722
 723 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
 724                                struct net_device *dev, __u8 dsfield,
 725                                struct flowi6 *fl6, int encap_limit,
 726                                __u32 *pmtu, __be16 proto)
 727 {
 728         struct ip6_tnl *tunnel = netdev_priv(dev);
 729         __be16 protocol;
 730         __be16 flags;
 731
 732         if (dev->type == ARPHRD_ETHER)
 733                 IPCB(skb)->flags = 0;
 734
 735         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
 736                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
 737         else
 738                 fl6->daddr = tunnel->parms.raddr;
 739
 740         /* Push GRE header. */
 741         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
 742
 743         if (tunnel->parms.collect_md) {
 744                 struct ip_tunnel_info *tun_info;
 745                 const struct ip_tunnel_key *key;
 746                 int tun_hlen;
 747
 748                 tun_info = skb_tunnel_info_txcheck(skb);
 749                 if (IS_ERR(tun_info) ||
 750                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
 751                         return -EINVAL;
 752
 753                 key = &tun_info->key;
 754                 memset(fl6, 0, sizeof(*fl6));
 755                 fl6->flowi6_proto = IPPROTO_GRE;
 756                 fl6->daddr = key->u.ipv6.dst;
 757                 fl6->flowlabel = key->label;
 758                 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 759                 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
 760
 761                 dsfield = key->tos;
 762                 flags = key->tun_flags &
 763                         (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
 764                 tun_hlen = gre_calc_hlen(flags);
 765
 766                 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
 767                         return -ENOMEM;
 768
 769                 gre_build_header(skb, tun_hlen,
 770                                  flags, protocol,
 771                                  tunnel_id_to_key32(tun_info->key.tun_id),
 772                                  (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
 773                                                       : 0);
 774
 775         } else {
 776                 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
 777                         return -ENOMEM;
 778
 779                 flags = tunnel->parms.o_flags;
 780
 781                 gre_build_header(skb, tunnel->tun_hlen, flags,
 782                                  protocol, tunnel->parms.o_key,
 783                                  (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
 784                                                       : 0);
 785         }
 786
 787         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
 788                             NEXTHDR_GRE);
 789 }
 790
 791 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 792 {
 793         struct ip6_tnl *t = netdev_priv(dev);
 794         int encap_limit = -1;
 795         struct flowi6 fl6;
 796         __u8 dsfield = 0;
 797         __u32 mtu;
 798         int err;
 799
 800         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 801
 802         if (!t->parms.collect_md)
 803                 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
 804                                          &dsfield, &encap_limit);
 805
 806         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 807         if (err)
 808                 return -1;
 809
 810         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 811                           skb->protocol);
 812         if (err != 0) {
 813                 /* XXX: send ICMP error even if DF is not set. */
 814                 if (err == -EMSGSIZE)
 815                         icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 816                                       htonl(mtu));
 817                 return -1;
 818         }
 819
 820         return 0;
 821 }
 822
 823 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 824 {
 825         struct ip6_tnl *t = netdev_priv(dev);
 826         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 827         int encap_limit = -1;
 828         struct flowi6 fl6;
 829         __u8 dsfield = 0;
 830         __u32 mtu;
 831         int err;
 832
 833         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 834                 return -1;
 835
 836         if (!t->parms.collect_md &&
 837             prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
 838                 return -1;
 839
 840         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
 841                 return -1;
 842
 843         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
 844                           &mtu, skb->protocol);
 845         if (err != 0) {
 846                 if (err == -EMSGSIZE)
 847                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 848                 return -1;
 849         }
 850
 851         return 0;
 852 }
 853
 854 /**
 855  * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
 856  *   @t: the outgoing tunnel device
 857  *   @hdr: IPv6 header from the incoming packet
 858  *
 859  * Description:
 860  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 861  *   doesn't match source of incoming packet.
 862  *
 863  * Return:
 864  *   1 if conflict,
 865  *   0 else
 866  **/
 867
 868 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 869         const struct ipv6hdr *hdr)
 870 {
 871         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 872 }
 873
 874 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 875 {
 876         struct ip6_tnl *t = netdev_priv(dev);
 877         int encap_limit = -1;
 878         struct flowi6 fl6;
 879         __u32 mtu;
 880         int err;
 881
 882         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 883                 encap_limit = t->parms.encap_limit;
 884
 885         if (!t->parms.collect_md)
 886                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 887
 888         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 889         if (err)
 890                 return err;
 891
 892         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
 893
 894         return err;
 895 }
 896
 897 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 898         struct net_device *dev)
 899 {
 900         struct ip6_tnl *t = netdev_priv(dev);
 901         struct net_device_stats *stats = &t->dev->stats;
 902         int ret;
 903
 904         if (!pskb_inet_may_pull(skb))
 905                 goto tx_err;
 906
 907         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 908                 goto tx_err;
 909
 910         switch (skb->protocol) {
 911         case htons(ETH_P_IP):
 912                 ret = ip6gre_xmit_ipv4(skb, dev);
 913                 break;
 914         case htons(ETH_P_IPV6):
 915                 ret = ip6gre_xmit_ipv6(skb, dev);
 916                 break;
 917         default:
 918                 ret = ip6gre_xmit_other(skb, dev);
 919                 break;
 920         }
 921
 922         if (ret < 0)
 923                 goto tx_err;
 924
 925         return NETDEV_TX_OK;
 926
 927 tx_err:
 928         if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
 929                 stats->tx_errors++;
 930         stats->tx_dropped++;
 931         kfree_skb(skb);
 932         return NETDEV_TX_OK;
 933 }
 934
 935 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
 936                                          struct net_device *dev)
 937 {
 938         struct ip_tunnel_info *tun_info = NULL;
 939         struct ip6_tnl *t = netdev_priv(dev);
 940         struct dst_entry *dst = skb_dst(skb);
 941         struct net_device_stats *stats;
 942         bool truncate = false;
 943         int encap_limit = -1;
 944         __u8 dsfield = false;
 945         struct flowi6 fl6;
 946         int err = -EINVAL;
 947         __be16 proto;
 948         __u32 mtu;
 949         int nhoff;
 950
 951         if (!pskb_inet_may_pull(skb))
 952                 goto tx_err;
 953
 954         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 955                 goto tx_err;
 956
 957         if (gre_handle_offloads(skb, false))
 958                 goto tx_err;
 959
 960         if (skb->len > dev->mtu + dev->hard_header_len) {
 961                 if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
 962                         goto tx_err;
 963                 truncate = true;
 964         }
 965
 966         nhoff = skb_network_offset(skb);
 967         if (skb->protocol == htons(ETH_P_IP) &&
 968             (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
 969                 truncate = true;
 970
 971         if (skb->protocol == htons(ETH_P_IPV6)) {
 972                 int thoff;
 973
 974                 if (skb_transport_header_was_set(skb))
 975                         thoff = skb_transport_offset(skb);
 976                 else
 977                         thoff = nhoff + sizeof(struct ipv6hdr);
 978                 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
 979                         truncate = true;
 980         }
 981
 982         if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
 983                 goto tx_err;
 984
 985         t->parms.o_flags &= ~TUNNEL_KEY;
 986         IPCB(skb)->flags = 0;
 987
 988         /* For collect_md mode, derive fl6 from the tunnel key,
 989          * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
 990          */
 991         if (t->parms.collect_md) {
 992                 const struct ip_tunnel_key *key;
 993                 struct erspan_metadata *md;
 994                 __be32 tun_id;
 995
 996                 tun_info = skb_tunnel_info_txcheck(skb);
 997                 if (IS_ERR(tun_info) ||
 998                     unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
 999                         goto tx_err;
1000
1001                 key = &tun_info->key;
1002                 memset(&fl6, 0, sizeof(fl6));
1003                 fl6.flowi6_proto = IPPROTO_GRE;
1004                 fl6.daddr = key->u.ipv6.dst;
1005                 fl6.flowlabel = key->label;
1006                 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1007                 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
1008
1009                 dsfield = key->tos;
1010                 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1011                         goto tx_err;
1012                 if (tun_info->options_len < sizeof(*md))
1013                         goto tx_err;
1014                 md = ip_tunnel_info_opts(tun_info);
1015
1016                 tun_id = tunnel_id_to_key32(key->tun_id);
1017                 if (md->version == 1) {
1018                         erspan_build_header(skb,
1019                                             ntohl(tun_id),
1020                                             ntohl(md->u.index), truncate,
1021                                             false);
1022                         proto = htons(ETH_P_ERSPAN);
1023                 } else if (md->version == 2) {
1024                         erspan_build_header_v2(skb,
1025                                                ntohl(tun_id),
1026                                                md->u.md2.dir,
1027                                                get_hwid(&md->u.md2),
1028                                                truncate, false);
1029                         proto = htons(ETH_P_ERSPAN2);
1030                 } else {
1031                         goto tx_err;
1032                 }
1033         } else {
1034                 switch (skb->protocol) {
1035                 case htons(ETH_P_IP):
1036                         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1037                         prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1038                                                  &dsfield, &encap_limit);
1039                         break;
1040                 case htons(ETH_P_IPV6):
1041                         if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1042                                 goto tx_err;
1043                         if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1044                                                      &dsfield, &encap_limit))
1045                                 goto tx_err;
1046                         break;
1047                 default:
1048                         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1049                         break;
1050                 }
1051
1052                 if (t->parms.erspan_ver == 1) {
1053                         erspan_build_header(skb, ntohl(t->parms.o_key),
1054                                             t->parms.index,
1055                                             truncate, false);
1056                         proto = htons(ETH_P_ERSPAN);
1057                 } else if (t->parms.erspan_ver == 2) {
1058                         erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1059                                                t->parms.dir,
1060                                                t->parms.hwid,
1061                                                truncate, false);
1062                         proto = htons(ETH_P_ERSPAN2);
1063                 } else {
1064                         goto tx_err;
1065                 }
1066
1067                 fl6.daddr = t->parms.raddr;
1068         }
1069
1070         /* Push GRE header. */
1071         gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno)));
1072
1073         /* TooBig packet may have updated dst->dev's mtu */
1074         if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1075                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1076
1077         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1078                            NEXTHDR_GRE);
1079         if (err != 0) {
1080                 /* XXX: send ICMP error even if DF is not set. */
1081                 if (err == -EMSGSIZE) {
1082                         if (skb->protocol == htons(ETH_P_IP))
1083                                 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1084                                               ICMP_FRAG_NEEDED, htonl(mtu));
1085                         else
1086                                 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1087                 }
1088
1089                 goto tx_err;
1090         }
1091         return NETDEV_TX_OK;
1092
1093 tx_err:
1094         stats = &t->dev->stats;
1095         if (!IS_ERR(tun_info))
1096                 stats->tx_errors++;
1097         stats->tx_dropped++;
1098         kfree_skb(skb);
1099         return NETDEV_TX_OK;
1100 }
1101
1102 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1103 {
1104         struct net_device *dev = t->dev;
1105         struct __ip6_tnl_parm *p = &t->parms;
1106         struct flowi6 *fl6 = &t->fl.u.ip6;
1107
1108         if (dev->type != ARPHRD_ETHER) {
1109                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1110                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1111         }
1112
1113         /* Set up flowi template */
1114         fl6->saddr = p->laddr;
1115         fl6->daddr = p->raddr;
1116         fl6->flowi6_oif = p->link;
1117         fl6->flowlabel = 0;
1118         fl6->flowi6_proto = IPPROTO_GRE;
1119         fl6->fl6_gre_key = t->parms.o_key;
1120
1121         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1122                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1123         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1124                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1125
1126         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1127         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1128
1129         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1130                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1131                 dev->flags |= IFF_POINTOPOINT;
1132         else
1133                 dev->flags &= ~IFF_POINTOPOINT;
1134 }
1135
1136 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1137                                          int t_hlen)
1138 {
1139         const struct __ip6_tnl_parm *p = &t->parms;
1140         struct net_device *dev = t->dev;
1141
1142         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1143                 int strict = (ipv6_addr_type(&p->raddr) &
1144                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1145
1146                 struct rt6_info *rt = rt6_lookup(t->net,
1147                                                  &p->raddr, &p->laddr,
1148                                                  p->link, NULL, strict);
1149
1150                 if (!rt)
1151                         return;
1152
1153                 if (rt->dst.dev) {
1154                         unsigned short dst_len = rt->dst.dev->hard_header_len +
1155                                                  t_hlen;
1156
1157                         if (t->dev->header_ops)
1158                                 dev->hard_header_len = dst_len;
1159                         else
1160                                 dev->needed_headroom = dst_len;
1161
1162                         if (set_mtu) {
1163                                 int mtu = rt->dst.dev->mtu - t_hlen;
1164
1165                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1166                                         mtu -= 8;
1167                                 if (dev->type == ARPHRD_ETHER)
1168                                         mtu -= ETH_HLEN;
1169
1170                                 if (mtu < IPV6_MIN_MTU)
1171                                         mtu = IPV6_MIN_MTU;
1172                                 WRITE_ONCE(dev->mtu, mtu);
1173                         }
1174                 }
1175                 ip6_rt_put(rt);
1176         }
1177 }
1178
1179 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1180 {
1181         int t_hlen;
1182
1183         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1184         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1185
1186         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1187
1188         if (tunnel->dev->header_ops)
1189                 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1190         else
1191                 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1192
1193         return t_hlen;
1194 }
1195
1196 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1197 {
1198         ip6gre_tnl_link_config_common(t);
1199         ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1200 }
1201
1202 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1203                                      const struct __ip6_tnl_parm *p)
1204 {
1205         t->parms.laddr = p->laddr;
1206         t->parms.raddr = p->raddr;
1207         t->parms.flags = p->flags;
1208         t->parms.hop_limit = p->hop_limit;
1209         t->parms.encap_limit = p->encap_limit;
1210         t->parms.flowinfo = p->flowinfo;
1211         t->parms.link = p->link;
1212         t->parms.proto = p->proto;
1213         t->parms.i_key = p->i_key;
1214         t->parms.o_key = p->o_key;
1215         t->parms.i_flags = p->i_flags;
1216         t->parms.o_flags = p->o_flags;
1217         t->parms.fwmark = p->fwmark;
1218         t->parms.erspan_ver = p->erspan_ver;
1219         t->parms.index = p->index;
1220         t->parms.dir = p->dir;
1221         t->parms.hwid = p->hwid;
1222         dst_cache_reset(&t->dst_cache);
1223 }
1224
1225 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1226                              int set_mtu)
1227 {
1228         ip6gre_tnl_copy_tnl_parm(t, p);
1229         ip6gre_tnl_link_config(t, set_mtu);
1230         return 0;
1231 }
1232
1233 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1234         const struct ip6_tnl_parm2 *u)
1235 {
1236         p->laddr = u->laddr;
1237         p->raddr = u->raddr;
1238         p->flags = u->flags;
1239         p->hop_limit = u->hop_limit;
1240         p->encap_limit = u->encap_limit;
1241         p->flowinfo = u->flowinfo;
1242         p->link = u->link;
1243         p->i_key = u->i_key;
1244         p->o_key = u->o_key;
1245         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1246         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1247         memcpy(p->name, u->name, sizeof(u->name));
1248 }
1249
1250 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1251         const struct __ip6_tnl_parm *p)
1252 {
1253         u->proto = IPPROTO_GRE;
1254         u->laddr = p->laddr;
1255         u->raddr = p->raddr;
1256         u->flags = p->flags;
1257         u->hop_limit = p->hop_limit;
1258         u->encap_limit = p->encap_limit;
1259         u->flowinfo = p->flowinfo;
1260         u->link = p->link;
1261         u->i_key = p->i_key;
1262         u->o_key = p->o_key;
1263         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1264         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1265         memcpy(u->name, p->name, sizeof(u->name));
1266 }
1267
1268 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1269         struct ifreq *ifr, int cmd)
1270 {
1271         int err = 0;
1272         struct ip6_tnl_parm2 p;
1273         struct __ip6_tnl_parm p1;
1274         struct ip6_tnl *t = netdev_priv(dev);
1275         struct net *net = t->net;
1276         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1277
1278         memset(&p1, 0, sizeof(p1));
1279
1280         switch (cmd) {
1281         case SIOCGETTUNNEL:
1282                 if (dev == ign->fb_tunnel_dev) {
1283                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1284                                 err = -EFAULT;
1285                                 break;
1286                         }
1287                         ip6gre_tnl_parm_from_user(&p1, &p);
1288                         t = ip6gre_tunnel_locate(net, &p1, 0);
1289                         if (!t)
1290                                 t = netdev_priv(dev);
1291                 }
1292                 memset(&p, 0, sizeof(p));
1293                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1294                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1295                         err = -EFAULT;
1296                 break;
1297
1298         case SIOCADDTUNNEL:
1299         case SIOCCHGTUNNEL:
1300                 err = -EPERM;
1301                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1302                         goto done;
1303
1304                 err = -EFAULT;
1305                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1306                         goto done;
1307
1308                 err = -EINVAL;
1309                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1310                         goto done;
1311
1312                 if (!(p.i_flags&GRE_KEY))
1313                         p.i_key = 0;
1314                 if (!(p.o_flags&GRE_KEY))
1315                         p.o_key = 0;
1316
1317                 ip6gre_tnl_parm_from_user(&p1, &p);
1318                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1319
1320                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1321                         if (t) {
1322                                 if (t->dev != dev) {
1323                                         err = -EEXIST;
1324                                         break;
1325                                 }
1326                         } else {
1327                                 t = netdev_priv(dev);
1328
1329                                 ip6gre_tunnel_unlink(ign, t);
1330                                 synchronize_net();
1331                                 ip6gre_tnl_change(t, &p1, 1);
1332                                 ip6gre_tunnel_link(ign, t);
1333                                 netdev_state_change(dev);
1334                         }
1335                 }
1336
1337                 if (t) {
1338                         err = 0;
1339
1340                         memset(&p, 0, sizeof(p));
1341                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1342                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1343                                 err = -EFAULT;
1344                 } else
1345                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1346                 break;
1347
1348         case SIOCDELTUNNEL:
1349                 err = -EPERM;
1350                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1351                         goto done;
1352
1353                 if (dev == ign->fb_tunnel_dev) {
1354                         err = -EFAULT;
1355                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1356                                 goto done;
1357                         err = -ENOENT;
1358                         ip6gre_tnl_parm_from_user(&p1, &p);
1359                         t = ip6gre_tunnel_locate(net, &p1, 0);
1360                         if (!t)
1361                                 goto done;
1362                         err = -EPERM;
1363                         if (t == netdev_priv(ign->fb_tunnel_dev))
1364                                 goto done;
1365                         dev = t->dev;
1366                 }
1367                 unregister_netdevice(dev);
1368                 err = 0;
1369                 break;
1370
1371         default:
1372                 err = -EINVAL;
1373         }
1374
1375 done:
1376         return err;
1377 }
1378
1379 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1380                          unsigned short type, const void *daddr,
1381                          const void *saddr, unsigned int len)
1382 {
1383         struct ip6_tnl *t = netdev_priv(dev);
1384         struct ipv6hdr *ipv6h;
1385         __be16 *p;
1386
1387         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1388         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1389                                                   t->fl.u.ip6.flowlabel,
1390                                                   true, &t->fl.u.ip6));
1391         ipv6h->hop_limit = t->parms.hop_limit;
1392         ipv6h->nexthdr = NEXTHDR_GRE;
1393         ipv6h->saddr = t->parms.laddr;
1394         ipv6h->daddr = t->parms.raddr;
1395
1396         p = (__be16 *)(ipv6h + 1);
1397         p[0] = t->parms.o_flags;
1398         p[1] = htons(type);
1399
1400         /*
1401          *      Set the source hardware address.
1402          */
1403
1404         if (saddr)
1405                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1406         if (daddr)
1407                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1408         if (!ipv6_addr_any(&ipv6h->daddr))
1409                 return t->hlen;
1410
1411         return -t->hlen;
1412 }
1413
1414 static const struct header_ops ip6gre_header_ops = {
1415         .create = ip6gre_header,
1416 };
1417
1418 static const struct net_device_ops ip6gre_netdev_ops = {
1419         .ndo_init               = ip6gre_tunnel_init,
1420         .ndo_uninit             = ip6gre_tunnel_uninit,
1421         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1422         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1423         .ndo_change_mtu         = ip6_tnl_change_mtu,
1424         .ndo_get_stats64        = ip_tunnel_get_stats64,
1425         .ndo_get_iflink         = ip6_tnl_get_iflink,
1426 };
1427
1428 static void ip6gre_dev_free(struct net_device *dev)
1429 {
1430         struct ip6_tnl *t = netdev_priv(dev);
1431
1432         gro_cells_destroy(&t->gro_cells);
1433         dst_cache_destroy(&t->dst_cache);
1434         free_percpu(dev->tstats);
1435 }
1436
1437 static void ip6gre_tunnel_setup(struct net_device *dev)
1438 {
1439         dev->netdev_ops = &ip6gre_netdev_ops;
1440         dev->needs_free_netdev = true;
1441         dev->priv_destructor = ip6gre_dev_free;
1442
1443         dev->type = ARPHRD_IP6GRE;
1444
1445         dev->flags |= IFF_NOARP;
1446         dev->addr_len = sizeof(struct in6_addr);
1447         netif_keep_dst(dev);
1448         /* This perm addr will be used as interface identifier by IPv6 */
1449         dev->addr_assign_type = NET_ADDR_RANDOM;
1450         eth_random_addr(dev->perm_addr);
1451 }
1452
1453 #define GRE6_FEATURES (NETIF_F_SG |             \
1454                        NETIF_F_FRAGLIST |       \
1455                        NETIF_F_HIGHDMA |        \
1456                        NETIF_F_HW_CSUM)
1457
1458 static void ip6gre_tnl_init_features(struct net_device *dev)
1459 {
1460         struct ip6_tnl *nt = netdev_priv(dev);
1461
1462         dev->features           |= GRE6_FEATURES;
1463         dev->hw_features        |= GRE6_FEATURES;
1464
1465         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1466                 /* TCP offload with GRE SEQ is not supported, nor
1467                  * can we support 2 levels of outer headers requiring
1468                  * an update.
1469                  */
1470                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1471                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1472                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1473                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1474                 }
1475
1476                 /* Can use a lockless transmit, unless we generate
1477                  * output sequences
1478                  */
1479                 dev->features |= NETIF_F_LLTX;
1480         }
1481 }
1482
1483 static int ip6gre_tunnel_init_common(struct net_device *dev)
1484 {
1485         struct ip6_tnl *tunnel;
1486         int ret;
1487         int t_hlen;
1488
1489         tunnel = netdev_priv(dev);
1490
1491         tunnel->dev = dev;
1492         tunnel->net = dev_net(dev);
1493         strcpy(tunnel->parms.name, dev->name);
1494
1495         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1496         if (!dev->tstats)
1497                 return -ENOMEM;
1498
1499         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1500         if (ret)
1501                 goto cleanup_alloc_pcpu_stats;
1502
1503         ret = gro_cells_init(&tunnel->gro_cells, dev);
1504         if (ret)
1505                 goto cleanup_dst_cache_init;
1506
1507         t_hlen = ip6gre_calc_hlen(tunnel);
1508         dev->mtu = ETH_DATA_LEN - t_hlen;
1509         if (dev->type == ARPHRD_ETHER)
1510                 dev->mtu -= ETH_HLEN;
1511         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1512                 dev->mtu -= 8;
1513
1514         if (tunnel->parms.collect_md) {
1515                 netif_keep_dst(dev);
1516         }
1517         ip6gre_tnl_init_features(dev);
1518
1519         dev_hold(dev);
1520         return 0;
1521
1522 cleanup_dst_cache_init:
1523         dst_cache_destroy(&tunnel->dst_cache);
1524 cleanup_alloc_pcpu_stats:
1525         free_percpu(dev->tstats);
1526         dev->tstats = NULL;
1527         return ret;
1528 }
1529
1530 static int ip6gre_tunnel_init(struct net_device *dev)
1531 {
1532         struct ip6_tnl *tunnel;
1533         int ret;
1534
1535         ret = ip6gre_tunnel_init_common(dev);
1536         if (ret)
1537                 return ret;
1538
1539         tunnel = netdev_priv(dev);
1540
1541         if (tunnel->parms.collect_md)
1542                 return 0;
1543
1544         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1545         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1546
1547         if (ipv6_addr_any(&tunnel->parms.raddr))
1548                 dev->header_ops = &ip6gre_header_ops;
1549
1550         return 0;
1551 }
1552
1553 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1554 {
1555         struct ip6_tnl *tunnel = netdev_priv(dev);
1556
1557         tunnel->dev = dev;
1558         tunnel->net = dev_net(dev);
1559         strcpy(tunnel->parms.name, dev->name);
1560
1561         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1562 }
1563
1564 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1565         .handler     = gre_rcv,
1566         .err_handler = ip6gre_err,
1567         .flags       = INET6_PROTO_FINAL,
1568 };
1569
1570 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1571 {
1572         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1573         struct net_device *dev, *aux;
1574         int prio;
1575
1576         for_each_netdev_safe(net, dev, aux)
1577                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1578                     dev->rtnl_link_ops == &ip6gre_tap_ops ||
1579                     dev->rtnl_link_ops == &ip6erspan_tap_ops)
1580                         unregister_netdevice_queue(dev, head);
1581
1582         for (prio = 0; prio < 4; prio++) {
1583                 int h;
1584                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1585                         struct ip6_tnl *t;
1586
1587                         t = rtnl_dereference(ign->tunnels[prio][h]);
1588
1589                         while (t) {
1590                                 /* If dev is in the same netns, it has already
1591                                  * been added to the list by the previous loop.
1592                                  */
1593                                 if (!net_eq(dev_net(t->dev), net))
1594                                         unregister_netdevice_queue(t->dev,
1595                                                                    head);
1596                                 t = rtnl_dereference(t->next);
1597                         }
1598                 }
1599         }
1600 }
1601
1602 static int __net_init ip6gre_init_net(struct net *net)
1603 {
1604         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1605         struct net_device *ndev;
1606         int err;
1607
1608         if (!net_has_fallback_tunnels(net))
1609                 return 0;
1610         ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1611                             NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1612         if (!ndev) {
1613                 err = -ENOMEM;
1614                 goto err_alloc_dev;
1615         }
1616         ign->fb_tunnel_dev = ndev;
1617         dev_net_set(ign->fb_tunnel_dev, net);
1618         /* FB netdevice is special: we have one, and only one per netns.
1619          * Allowing to move it to another netns is clearly unsafe.
1620          */
1621         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1622
1623
1624         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1625         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1626
1627         err = register_netdev(ign->fb_tunnel_dev);
1628         if (err)
1629                 goto err_reg_dev;
1630
1631         rcu_assign_pointer(ign->tunnels_wc[0],
1632                            netdev_priv(ign->fb_tunnel_dev));
1633         return 0;
1634
1635 err_reg_dev:
1636         free_netdev(ndev);
1637 err_alloc_dev:
1638         return err;
1639 }
1640
1641 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1642 {
1643         struct net *net;
1644         LIST_HEAD(list);
1645
1646         rtnl_lock();
1647         list_for_each_entry(net, net_list, exit_list)
1648                 ip6gre_destroy_tunnels(net, &list);
1649         unregister_netdevice_many(&list);
1650         rtnl_unlock();
1651 }
1652
1653 static struct pernet_operations ip6gre_net_ops = {
1654         .init = ip6gre_init_net,
1655         .exit_batch = ip6gre_exit_batch_net,
1656         .id   = &ip6gre_net_id,
1657         .size = sizeof(struct ip6gre_net),
1658 };
1659
1660 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1661                                   struct netlink_ext_ack *extack)
1662 {
1663         __be16 flags;
1664
1665         if (!data)
1666                 return 0;
1667
1668         flags = 0;
1669         if (data[IFLA_GRE_IFLAGS])
1670                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1671         if (data[IFLA_GRE_OFLAGS])
1672                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1673         if (flags & (GRE_VERSION|GRE_ROUTING))
1674                 return -EINVAL;
1675
1676         return 0;
1677 }
1678
1679 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1680                                struct netlink_ext_ack *extack)
1681 {
1682         struct in6_addr daddr;
1683
1684         if (tb[IFLA_ADDRESS]) {
1685                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1686                         return -EINVAL;
1687                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1688                         return -EADDRNOTAVAIL;
1689         }
1690
1691         if (!data)
1692                 goto out;
1693
1694         if (data[IFLA_GRE_REMOTE]) {
1695                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1696                 if (ipv6_addr_any(&daddr))
1697                         return -EINVAL;
1698         }
1699
1700 out:
1701         return ip6gre_tunnel_validate(tb, data, extack);
1702 }
1703
1704 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1705                                   struct netlink_ext_ack *extack)
1706 {
1707         __be16 flags = 0;
1708         int ret, ver = 0;
1709
1710         if (!data)
1711                 return 0;
1712
1713         ret = ip6gre_tap_validate(tb, data, extack);
1714         if (ret)
1715                 return ret;
1716
1717         /* ERSPAN should only have GRE sequence and key flag */
1718         if (data[IFLA_GRE_OFLAGS])
1719                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1720         if (data[IFLA_GRE_IFLAGS])
1721                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1722         if (!data[IFLA_GRE_COLLECT_METADATA] &&
1723             flags != (GRE_SEQ | GRE_KEY))
1724                 return -EINVAL;
1725
1726         /* ERSPAN Session ID only has 10-bit. Since we reuse
1727          * 32-bit key field as ID, check it's range.
1728          */
1729         if (data[IFLA_GRE_IKEY] &&
1730             (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1731                 return -EINVAL;
1732
1733         if (data[IFLA_GRE_OKEY] &&
1734             (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1735                 return -EINVAL;
1736
1737         if (data[IFLA_GRE_ERSPAN_VER]) {
1738                 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1739                 if (ver != 1 && ver != 2)
1740                         return -EINVAL;
1741         }
1742
1743         if (ver == 1) {
1744                 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1745                         u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1746
1747                         if (index & ~INDEX_MASK)
1748                                 return -EINVAL;
1749                 }
1750         } else if (ver == 2) {
1751                 if (data[IFLA_GRE_ERSPAN_DIR]) {
1752                         u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1753
1754                         if (dir & ~(DIR_MASK >> DIR_OFFSET))
1755                                 return -EINVAL;
1756                 }
1757
1758                 if (data[IFLA_GRE_ERSPAN_HWID]) {
1759                         u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1760
1761                         if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1762                                 return -EINVAL;
1763                 }
1764         }
1765
1766         return 0;
1767 }
1768
1769 static void ip6erspan_set_version(struct nlattr *data[],
1770                                   struct __ip6_tnl_parm *parms)
1771 {
1772         if (!data)
1773                 return;
1774
1775         parms->erspan_ver = 1;
1776         if (data[IFLA_GRE_ERSPAN_VER])
1777                 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1778
1779         if (parms->erspan_ver == 1) {
1780                 if (data[IFLA_GRE_ERSPAN_INDEX])
1781                         parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1782         } else if (parms->erspan_ver == 2) {
1783                 if (data[IFLA_GRE_ERSPAN_DIR])
1784                         parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1785                 if (data[IFLA_GRE_ERSPAN_HWID])
1786                         parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1787         }
1788 }
1789
1790 static void ip6gre_netlink_parms(struct nlattr *data[],
1791                                 struct __ip6_tnl_parm *parms)
1792 {
1793         memset(parms, 0, sizeof(*parms));
1794
1795         if (!data)
1796                 return;
1797
1798         if (data[IFLA_GRE_LINK])
1799                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1800
1801         if (data[IFLA_GRE_IFLAGS])
1802                 parms->i_flags = gre_flags_to_tnl_flags(
1803                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1804
1805         if (data[IFLA_GRE_OFLAGS])
1806                 parms->o_flags = gre_flags_to_tnl_flags(
1807                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1808
1809         if (data[IFLA_GRE_IKEY])
1810                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1811
1812         if (data[IFLA_GRE_OKEY])
1813                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1814
1815         if (data[IFLA_GRE_LOCAL])
1816                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1817
1818         if (data[IFLA_GRE_REMOTE])
1819                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1820
1821         if (data[IFLA_GRE_TTL])
1822                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1823
1824         if (data[IFLA_GRE_ENCAP_LIMIT])
1825                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1826
1827         if (data[IFLA_GRE_FLOWINFO])
1828                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1829
1830         if (data[IFLA_GRE_FLAGS])
1831                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1832
1833         if (data[IFLA_GRE_FWMARK])
1834                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1835
1836         if (data[IFLA_GRE_COLLECT_METADATA])
1837                 parms->collect_md = true;
1838 }
1839
1840 static int ip6gre_tap_init(struct net_device *dev)
1841 {
1842         int ret;
1843
1844         ret = ip6gre_tunnel_init_common(dev);
1845         if (ret)
1846                 return ret;
1847
1848         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1849
1850         return 0;
1851 }
1852
1853 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1854         .ndo_init = ip6gre_tap_init,
1855         .ndo_uninit = ip6gre_tunnel_uninit,
1856         .ndo_start_xmit = ip6gre_tunnel_xmit,
1857         .ndo_set_mac_address = eth_mac_addr,
1858         .ndo_validate_addr = eth_validate_addr,
1859         .ndo_change_mtu = ip6_tnl_change_mtu,
1860         .ndo_get_stats64 = ip_tunnel_get_stats64,
1861         .ndo_get_iflink = ip6_tnl_get_iflink,
1862 };
1863
1864 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1865 {
1866         int t_hlen;
1867
1868         tunnel->tun_hlen = 8;
1869         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1870                        erspan_hdr_len(tunnel->parms.erspan_ver);
1871
1872         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1873         tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1874         return t_hlen;
1875 }
1876
1877 static int ip6erspan_tap_init(struct net_device *dev)
1878 {
1879         struct ip6_tnl *tunnel;
1880         int t_hlen;
1881         int ret;
1882
1883         tunnel = netdev_priv(dev);
1884
1885         tunnel->dev = dev;
1886         tunnel->net = dev_net(dev);
1887         strcpy(tunnel->parms.name, dev->name);
1888
1889         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1890         if (!dev->tstats)
1891                 return -ENOMEM;
1892
1893         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1894         if (ret)
1895                 goto cleanup_alloc_pcpu_stats;
1896
1897         ret = gro_cells_init(&tunnel->gro_cells, dev);
1898         if (ret)
1899                 goto cleanup_dst_cache_init;
1900
1901         t_hlen = ip6erspan_calc_hlen(tunnel);
1902         dev->mtu = ETH_DATA_LEN - t_hlen;
1903         if (dev->type == ARPHRD_ETHER)
1904                 dev->mtu -= ETH_HLEN;
1905         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1906                 dev->mtu -= 8;
1907
1908         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1909         ip6erspan_tnl_link_config(tunnel, 1);
1910
1911         dev_hold(dev);
1912         return 0;
1913
1914 cleanup_dst_cache_init:
1915         dst_cache_destroy(&tunnel->dst_cache);
1916 cleanup_alloc_pcpu_stats:
1917         free_percpu(dev->tstats);
1918         dev->tstats = NULL;
1919         return ret;
1920 }
1921
1922 static const struct net_device_ops ip6erspan_netdev_ops = {
1923         .ndo_init =             ip6erspan_tap_init,
1924         .ndo_uninit =           ip6erspan_tunnel_uninit,
1925         .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1926         .ndo_set_mac_address =  eth_mac_addr,
1927         .ndo_validate_addr =    eth_validate_addr,
1928         .ndo_change_mtu =       ip6_tnl_change_mtu,
1929         .ndo_get_stats64 =      ip_tunnel_get_stats64,
1930         .ndo_get_iflink =       ip6_tnl_get_iflink,
1931 };
1932
1933 static void ip6gre_tap_setup(struct net_device *dev)
1934 {
1935
1936         ether_setup(dev);
1937
1938         dev->max_mtu = 0;
1939         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1940         dev->needs_free_netdev = true;
1941         dev->priv_destructor = ip6gre_dev_free;
1942
1943         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1944         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1945         netif_keep_dst(dev);
1946 }
1947
1948 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1949                                        struct ip_tunnel_encap *ipencap)
1950 {
1951         bool ret = false;
1952
1953         memset(ipencap, 0, sizeof(*ipencap));
1954
1955         if (!data)
1956                 return ret;
1957
1958         if (data[IFLA_GRE_ENCAP_TYPE]) {
1959                 ret = true;
1960                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1961         }
1962
1963         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1964                 ret = true;
1965                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1966         }
1967
1968         if (data[IFLA_GRE_ENCAP_SPORT]) {
1969                 ret = true;
1970                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1971         }
1972
1973         if (data[IFLA_GRE_ENCAP_DPORT]) {
1974                 ret = true;
1975                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1976         }
1977
1978         return ret;
1979 }
1980
1981 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1982                                  struct nlattr *tb[], struct nlattr *data[],
1983                                  struct netlink_ext_ack *extack)
1984 {
1985         struct ip6_tnl *nt;
1986         struct ip_tunnel_encap ipencap;
1987         int err;
1988
1989         nt = netdev_priv(dev);
1990
1991         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1992                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1993
1994                 if (err < 0)
1995                         return err;
1996         }
1997
1998         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1999                 eth_hw_addr_random(dev);
2000
2001         nt->dev = dev;
2002         nt->net = dev_net(dev);
2003
2004         err = register_netdevice(dev);
2005         if (err)
2006                 goto out;
2007
2008         if (tb[IFLA_MTU])
2009                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2010
2011 out:
2012         return err;
2013 }
2014
2015 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2016                           struct nlattr *tb[], struct nlattr *data[],
2017                           struct netlink_ext_ack *extack)
2018 {
2019         struct ip6_tnl *nt = netdev_priv(dev);
2020         struct net *net = dev_net(dev);
2021         struct ip6gre_net *ign;
2022         int err;
2023
2024         ip6gre_netlink_parms(data, &nt->parms);
2025         ign = net_generic(net, ip6gre_net_id);
2026
2027         if (nt->parms.collect_md) {
2028                 if (rtnl_dereference(ign->collect_md_tun))
2029                         return -EEXIST;
2030         } else {
2031                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2032                         return -EEXIST;
2033         }
2034
2035         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2036         if (!err) {
2037                 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2038                 ip6gre_tunnel_link_md(ign, nt);
2039                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2040         }
2041         return err;
2042 }
2043
2044 static struct ip6_tnl *
2045 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2046                          struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2047                          struct netlink_ext_ack *extack)
2048 {
2049         struct ip6_tnl *t, *nt = netdev_priv(dev);
2050         struct net *net = nt->net;
2051         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2052         struct ip_tunnel_encap ipencap;
2053
2054         if (dev == ign->fb_tunnel_dev)
2055                 return ERR_PTR(-EINVAL);
2056
2057         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2058                 int err = ip6_tnl_encap_setup(nt, &ipencap);
2059
2060                 if (err < 0)
2061                         return ERR_PTR(err);
2062         }
2063
2064         ip6gre_netlink_parms(data, p_p);
2065
2066         t = ip6gre_tunnel_locate(net, p_p, 0);
2067
2068         if (t) {
2069                 if (t->dev != dev)
2070                         return ERR_PTR(-EEXIST);
2071         } else {
2072                 t = nt;
2073         }
2074
2075         return t;
2076 }
2077
2078 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2079                              struct nlattr *data[],
2080                              struct netlink_ext_ack *extack)
2081 {
2082         struct ip6_tnl *t = netdev_priv(dev);
2083         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2084         struct __ip6_tnl_parm p;
2085
2086         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2087         if (IS_ERR(t))
2088                 return PTR_ERR(t);
2089
2090         ip6gre_tunnel_unlink_md(ign, t);
2091         ip6gre_tunnel_unlink(ign, t);
2092         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2093         ip6gre_tunnel_link_md(ign, t);
2094         ip6gre_tunnel_link(ign, t);
2095         return 0;
2096 }
2097
2098 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2099 {
2100         struct net *net = dev_net(dev);
2101         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2102
2103         if (dev != ign->fb_tunnel_dev)
2104                 unregister_netdevice_queue(dev, head);
2105 }
2106
2107 static size_t ip6gre_get_size(const struct net_device *dev)
2108 {
2109         return
2110                 /* IFLA_GRE_LINK */
2111                 nla_total_size(4) +
2112                 /* IFLA_GRE_IFLAGS */
2113                 nla_total_size(2) +
2114                 /* IFLA_GRE_OFLAGS */
2115                 nla_total_size(2) +
2116                 /* IFLA_GRE_IKEY */
2117                 nla_total_size(4) +
2118                 /* IFLA_GRE_OKEY */
2119                 nla_total_size(4) +
2120                 /* IFLA_GRE_LOCAL */
2121                 nla_total_size(sizeof(struct in6_addr)) +
2122                 /* IFLA_GRE_REMOTE */
2123                 nla_total_size(sizeof(struct in6_addr)) +
2124                 /* IFLA_GRE_TTL */
2125                 nla_total_size(1) +
2126                 /* IFLA_GRE_ENCAP_LIMIT */
2127                 nla_total_size(1) +
2128                 /* IFLA_GRE_FLOWINFO */
2129                 nla_total_size(4) +
2130                 /* IFLA_GRE_FLAGS */
2131                 nla_total_size(4) +
2132                 /* IFLA_GRE_ENCAP_TYPE */
2133                 nla_total_size(2) +
2134                 /* IFLA_GRE_ENCAP_FLAGS */
2135                 nla_total_size(2) +
2136                 /* IFLA_GRE_ENCAP_SPORT */
2137                 nla_total_size(2) +
2138                 /* IFLA_GRE_ENCAP_DPORT */
2139                 nla_total_size(2) +
2140                 /* IFLA_GRE_COLLECT_METADATA */
2141                 nla_total_size(0) +
2142                 /* IFLA_GRE_FWMARK */
2143                 nla_total_size(4) +
2144                 /* IFLA_GRE_ERSPAN_INDEX */
2145                 nla_total_size(4) +
2146                 0;
2147 }
2148
2149 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2150 {
2151         struct ip6_tnl *t = netdev_priv(dev);
2152         struct __ip6_tnl_parm *p = &t->parms;
2153         __be16 o_flags = p->o_flags;
2154
2155         if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2156                 if (!p->collect_md)
2157                         o_flags |= TUNNEL_KEY;
2158
2159                 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2160                         goto nla_put_failure;
2161
2162                 if (p->erspan_ver == 1) {
2163                         if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2164                                 goto nla_put_failure;
2165                 } else {
2166                         if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2167                                 goto nla_put_failure;
2168                         if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2169                                 goto nla_put_failure;
2170                 }
2171         }
2172
2173         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2174             nla_put_be16(skb, IFLA_GRE_IFLAGS,
2175                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2176             nla_put_be16(skb, IFLA_GRE_OFLAGS,
2177                          gre_tnl_flags_to_gre_flags(o_flags)) ||
2178             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2179             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2180             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2181             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2182             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2183             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2184             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2185             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2186             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2187                 goto nla_put_failure;
2188
2189         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2190                         t->encap.type) ||
2191             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2192                          t->encap.sport) ||
2193             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2194                          t->encap.dport) ||
2195             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2196                         t->encap.flags))
2197                 goto nla_put_failure;
2198
2199         if (p->collect_md) {
2200                 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2201                         goto nla_put_failure;
2202         }
2203
2204         return 0;
2205
2206 nla_put_failure:
2207         return -EMSGSIZE;
2208 }
2209
2210 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2211         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2212         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2213         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2214         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2215         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2216         [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2217         [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2218         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2219         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2220         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2221         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2222         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2223         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2224         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2225         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2226         [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2227         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2228         [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2229         [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2230         [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2231         [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2232 };
2233
2234 static void ip6erspan_tap_setup(struct net_device *dev)
2235 {
2236         ether_setup(dev);
2237
2238         dev->max_mtu = 0;
2239         dev->netdev_ops = &ip6erspan_netdev_ops;
2240         dev->needs_free_netdev = true;
2241         dev->priv_destructor = ip6gre_dev_free;
2242
2243         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2244         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2245         netif_keep_dst(dev);
2246 }
2247
2248 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2249                              struct nlattr *tb[], struct nlattr *data[],
2250                              struct netlink_ext_ack *extack)
2251 {
2252         struct ip6_tnl *nt = netdev_priv(dev);
2253         struct net *net = dev_net(dev);
2254         struct ip6gre_net *ign;
2255         int err;
2256
2257         ip6gre_netlink_parms(data, &nt->parms);
2258         ip6erspan_set_version(data, &nt->parms);
2259         ign = net_generic(net, ip6gre_net_id);
2260
2261         if (nt->parms.collect_md) {
2262                 if (rtnl_dereference(ign->collect_md_tun_erspan))
2263                         return -EEXIST;
2264         } else {
2265                 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2266                         return -EEXIST;
2267         }
2268
2269         err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2270         if (!err) {
2271                 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2272                 ip6erspan_tunnel_link_md(ign, nt);
2273                 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2274         }
2275         return err;
2276 }
2277
2278 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2279 {
2280         ip6gre_tnl_link_config_common(t);
2281         ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2282 }
2283
2284 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2285                                 const struct __ip6_tnl_parm *p, int set_mtu)
2286 {
2287         ip6gre_tnl_copy_tnl_parm(t, p);
2288         ip6erspan_tnl_link_config(t, set_mtu);
2289         return 0;
2290 }
2291
2292 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2293                                 struct nlattr *data[],
2294                                 struct netlink_ext_ack *extack)
2295 {
2296         struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2297         struct __ip6_tnl_parm p;
2298         struct ip6_tnl *t;
2299
2300         t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2301         if (IS_ERR(t))
2302                 return PTR_ERR(t);
2303
2304         ip6erspan_set_version(data, &p);
2305         ip6gre_tunnel_unlink_md(ign, t);
2306         ip6gre_tunnel_unlink(ign, t);
2307         ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2308         ip6erspan_tunnel_link_md(ign, t);
2309         ip6gre_tunnel_link(ign, t);
2310         return 0;
2311 }
2312
2313 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2314         .kind           = "ip6gre",
2315         .maxtype        = IFLA_GRE_MAX,
2316         .policy         = ip6gre_policy,
2317         .priv_size      = sizeof(struct ip6_tnl),
2318         .setup          = ip6gre_tunnel_setup,
2319         .validate       = ip6gre_tunnel_validate,
2320         .newlink        = ip6gre_newlink,
2321         .changelink     = ip6gre_changelink,
2322         .dellink        = ip6gre_dellink,
2323         .get_size       = ip6gre_get_size,
2324         .fill_info      = ip6gre_fill_info,
2325         .get_link_net   = ip6_tnl_get_link_net,
2326 };
2327
2328 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2329         .kind           = "ip6gretap",
2330         .maxtype        = IFLA_GRE_MAX,
2331         .policy         = ip6gre_policy,
2332         .priv_size      = sizeof(struct ip6_tnl),
2333         .setup          = ip6gre_tap_setup,
2334         .validate       = ip6gre_tap_validate,
2335         .newlink        = ip6gre_newlink,
2336         .changelink     = ip6gre_changelink,
2337         .get_size       = ip6gre_get_size,
2338         .fill_info      = ip6gre_fill_info,
2339         .get_link_net   = ip6_tnl_get_link_net,
2340 };
2341
2342 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2343         .kind           = "ip6erspan",
2344         .maxtype        = IFLA_GRE_MAX,
2345         .policy         = ip6gre_policy,
2346         .priv_size      = sizeof(struct ip6_tnl),
2347         .setup          = ip6erspan_tap_setup,
2348         .validate       = ip6erspan_tap_validate,
2349         .newlink        = ip6erspan_newlink,
2350         .changelink     = ip6erspan_changelink,
2351         .get_size       = ip6gre_get_size,
2352         .fill_info      = ip6gre_fill_info,
2353         .get_link_net   = ip6_tnl_get_link_net,
2354 };
2355
2356 /*
2357  *      And now the modules code and kernel interface.
2358  */
2359
2360 static int __init ip6gre_init(void)
2361 {
2362         int err;
2363
2364         pr_info("GRE over IPv6 tunneling driver\n");
2365
2366         err = register_pernet_device(&ip6gre_net_ops);
2367         if (err < 0)
2368                 return err;
2369
2370         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2371         if (err < 0) {
2372                 pr_info("%s: can't add protocol\n", __func__);
2373                 goto add_proto_failed;
2374         }
2375
2376         err = rtnl_link_register(&ip6gre_link_ops);
2377         if (err < 0)
2378                 goto rtnl_link_failed;
2379
2380         err = rtnl_link_register(&ip6gre_tap_ops);
2381         if (err < 0)
2382                 goto tap_ops_failed;
2383
2384         err = rtnl_link_register(&ip6erspan_tap_ops);
2385         if (err < 0)
2386                 goto erspan_link_failed;
2387
2388 out:
2389         return err;
2390
2391 erspan_link_failed:
2392         rtnl_link_unregister(&ip6gre_tap_ops);
2393 tap_ops_failed:
2394         rtnl_link_unregister(&ip6gre_link_ops);
2395 rtnl_link_failed:
2396         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2397 add_proto_failed:
2398         unregister_pernet_device(&ip6gre_net_ops);
2399         goto out;
2400 }
2401
2402 static void __exit ip6gre_fini(void)
2403 {
2404         rtnl_link_unregister(&ip6gre_tap_ops);
2405         rtnl_link_unregister(&ip6gre_link_ops);
2406         rtnl_link_unregister(&ip6erspan_tap_ops);
2407         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2408         unregister_pernet_device(&ip6gre_net_ops);
2409 }
2410
2411 module_init(ip6gre_init);
2412 module_exit(ip6gre_fini);
2413 MODULE_LICENSE("GPL");
2414 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2415 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2416 MODULE_ALIAS_RTNL_LINK("ip6gre");
2417 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2418 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2419 MODULE_ALIAS_NETDEV("ip6gre0");