net/ipv6/ip6_gre.c

   1 /*
   2  *      GRE over IPv6 protocol decoder.
   3  *
   4  *      Authors: Dmitry Kozlov (xeb@mail.ru)
   5  *
   6  *      This program is free software; you can redistribute it and/or
   7  *      modify it under the terms of the GNU General Public License
   8  *      as published by the Free Software Foundation; either version
   9  *      2 of the License, or (at your option) any later version.
  10  *
  11  */
  12
  13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  14
  15 #include <linux/capability.h>
  16 #include <linux/module.h>
  17 #include <linux/types.h>
  18 #include <linux/kernel.h>
  19 #include <linux/slab.h>
  20 #include <linux/uaccess.h>
  21 #include <linux/skbuff.h>
  22 #include <linux/netdevice.h>
  23 #include <linux/in.h>
  24 #include <linux/tcp.h>
  25 #include <linux/udp.h>
  26 #include <linux/if_arp.h>
  27 #include <linux/mroute.h>
  28 #include <linux/init.h>
  29 #include <linux/in6.h>
  30 #include <linux/inetdevice.h>
  31 #include <linux/igmp.h>
  32 #include <linux/netfilter_ipv4.h>
  33 #include <linux/etherdevice.h>
  34 #include <linux/if_ether.h>
  35 #include <linux/hash.h>
  36 #include <linux/if_tunnel.h>
  37 #include <linux/ip6_tunnel.h>
  38
  39 #include <net/sock.h>
  40 #include <net/ip.h>
  41 #include <net/ip_tunnels.h>
  42 #include <net/icmp.h>
  43 #include <net/protocol.h>
  44 #include <net/addrconf.h>
  45 #include <net/arp.h>
  46 #include <net/checksum.h>
  47 #include <net/dsfield.h>
  48 #include <net/inet_ecn.h>
  49 #include <net/xfrm.h>
  50 #include <net/net_namespace.h>
  51 #include <net/netns/generic.h>
  52 #include <net/rtnetlink.h>
  53
  54 #include <net/ipv6.h>
  55 #include <net/ip6_fib.h>
  56 #include <net/ip6_route.h>
  57 #include <net/ip6_tunnel.h>
  58 #include <net/gre.h>
  59
  60
  61 static bool log_ecn_error = true;
  62 module_param(log_ecn_error, bool, 0644);
  63 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  64
  65 #define HASH_SIZE_SHIFT  5
  66 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  67
  68 static int ip6gre_net_id __read_mostly;
  69 struct ip6gre_net {
  70         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
  71
  72         struct net_device *fb_tunnel_dev;
  73 };
  74
  75 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  76 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
  77 static int ip6gre_tunnel_init(struct net_device *dev);
  78 static void ip6gre_tunnel_setup(struct net_device *dev);
  79 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  80 static void ip6gre_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))&(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, 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                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 127         int score, cand_score = 4;
 128
 129         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 130                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 131                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 132                     key != t->parms.i_key ||
 133                     !(t->dev->flags & IFF_UP))
 134                         continue;
 135
 136                 if (t->dev->type != ARPHRD_IP6GRE &&
 137                     t->dev->type != dev_type)
 138                         continue;
 139
 140                 score = 0;
 141                 if (t->parms.link != link)
 142                         score |= 1;
 143                 if (t->dev->type != dev_type)
 144                         score |= 2;
 145                 if (score == 0)
 146                         return t;
 147
 148                 if (score < cand_score) {
 149                         cand = t;
 150                         cand_score = score;
 151                 }
 152         }
 153
 154         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 155                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 156                     key != t->parms.i_key ||
 157                     !(t->dev->flags & IFF_UP))
 158                         continue;
 159
 160                 if (t->dev->type != ARPHRD_IP6GRE &&
 161                     t->dev->type != dev_type)
 162                         continue;
 163
 164                 score = 0;
 165                 if (t->parms.link != link)
 166                         score |= 1;
 167                 if (t->dev->type != dev_type)
 168                         score |= 2;
 169                 if (score == 0)
 170                         return t;
 171
 172                 if (score < cand_score) {
 173                         cand = t;
 174                         cand_score = score;
 175                 }
 176         }
 177
 178         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 179                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 180                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 181                                  !ipv6_addr_is_multicast(local))) ||
 182                     key != t->parms.i_key ||
 183                     !(t->dev->flags & IFF_UP))
 184                         continue;
 185
 186                 if (t->dev->type != ARPHRD_IP6GRE &&
 187                     t->dev->type != dev_type)
 188                         continue;
 189
 190                 score = 0;
 191                 if (t->parms.link != link)
 192                         score |= 1;
 193                 if (t->dev->type != dev_type)
 194                         score |= 2;
 195                 if (score == 0)
 196                         return t;
 197
 198                 if (score < cand_score) {
 199                         cand = t;
 200                         cand_score = score;
 201                 }
 202         }
 203
 204         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 205                 if (t->parms.i_key != key ||
 206                     !(t->dev->flags & IFF_UP))
 207                         continue;
 208
 209                 if (t->dev->type != ARPHRD_IP6GRE &&
 210                     t->dev->type != dev_type)
 211                         continue;
 212
 213                 score = 0;
 214                 if (t->parms.link != link)
 215                         score |= 1;
 216                 if (t->dev->type != dev_type)
 217                         score |= 2;
 218                 if (score == 0)
 219                         return t;
 220
 221                 if (score < cand_score) {
 222                         cand = t;
 223                         cand_score = score;
 224                 }
 225         }
 226
 227         if (cand != NULL)
 228                 return cand;
 229
 230         dev = ign->fb_tunnel_dev;
 231         if (dev->flags & IFF_UP)
 232                 return netdev_priv(dev);
 233
 234         return NULL;
 235 }
 236
 237 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 238                 const struct __ip6_tnl_parm *p)
 239 {
 240         const struct in6_addr *remote = &p->raddr;
 241         const struct in6_addr *local = &p->laddr;
 242         unsigned int h = HASH_KEY(p->i_key);
 243         int prio = 0;
 244
 245         if (!ipv6_addr_any(local))
 246                 prio |= 1;
 247         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 248                 prio |= 2;
 249                 h ^= HASH_ADDR(remote);
 250         }
 251
 252         return &ign->tunnels[prio][h];
 253 }
 254
 255 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 256                 const struct ip6_tnl *t)
 257 {
 258         return __ip6gre_bucket(ign, &t->parms);
 259 }
 260
 261 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 262 {
 263         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 264
 265         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 266         rcu_assign_pointer(*tp, t);
 267 }
 268
 269 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 270 {
 271         struct ip6_tnl __rcu **tp;
 272         struct ip6_tnl *iter;
 273
 274         for (tp = ip6gre_bucket(ign, t);
 275              (iter = rtnl_dereference(*tp)) != NULL;
 276              tp = &iter->next) {
 277                 if (t == iter) {
 278                         rcu_assign_pointer(*tp, t->next);
 279                         break;
 280                 }
 281         }
 282 }
 283
 284 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 285                                            const struct __ip6_tnl_parm *parms,
 286                                            int type)
 287 {
 288         const struct in6_addr *remote = &parms->raddr;
 289         const struct in6_addr *local = &parms->laddr;
 290         __be32 key = parms->i_key;
 291         int link = parms->link;
 292         struct ip6_tnl *t;
 293         struct ip6_tnl __rcu **tp;
 294         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 295
 296         for (tp = __ip6gre_bucket(ign, parms);
 297              (t = rtnl_dereference(*tp)) != NULL;
 298              tp = &t->next)
 299                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 300                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 301                     key == t->parms.i_key &&
 302                     link == t->parms.link &&
 303                     type == t->dev->type)
 304                         break;
 305
 306         return t;
 307 }
 308
 309 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 310                 const struct __ip6_tnl_parm *parms, int create)
 311 {
 312         struct ip6_tnl *t, *nt;
 313         struct net_device *dev;
 314         char name[IFNAMSIZ];
 315         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 316
 317         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 318         if (t || !create)
 319                 return t;
 320
 321         if (parms->name[0]) {
 322                 if (!dev_valid_name(parms->name))
 323                         return NULL;
 324                 strlcpy(name, parms->name, IFNAMSIZ);
 325         } else {
 326                 strcpy(name, "ip6gre%d");
 327         }
 328         dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
 329         if (!dev)
 330                 return NULL;
 331
 332         dev_net_set(dev, net);
 333
 334         nt = netdev_priv(dev);
 335         nt->parms = *parms;
 336         dev->rtnl_link_ops = &ip6gre_link_ops;
 337
 338         nt->dev = dev;
 339         nt->net = dev_net(dev);
 340         ip6gre_tnl_link_config(nt, 1);
 341
 342         if (register_netdevice(dev) < 0)
 343                 goto failed_free;
 344
 345         /* Can use a lockless transmit, unless we generate output sequences */
 346         if (!(nt->parms.o_flags & GRE_SEQ))
 347                 dev->features |= NETIF_F_LLTX;
 348
 349         dev_hold(dev);
 350         ip6gre_tunnel_link(ign, nt);
 351         return nt;
 352
 353 failed_free:
 354         free_netdev(dev);
 355         return NULL;
 356 }
 357
 358 static void ip6gre_tunnel_uninit(struct net_device *dev)
 359 {
 360         struct ip6_tnl *t = netdev_priv(dev);
 361         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 362
 363         ip6gre_tunnel_unlink(ign, t);
 364         ip6_tnl_dst_reset(t);
 365         dev_put(dev);
 366 }
 367
 368
 369 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 370                        u8 type, u8 code, int offset, __be32 info)
 371 {
 372         const struct gre_base_hdr *greh;
 373         const struct ipv6hdr *ipv6h;
 374         int grehlen = sizeof(*greh);
 375         struct ip6_tnl *t;
 376         int key_off = 0;
 377         __be16 flags;
 378         __be32 key;
 379
 380         if (!pskb_may_pull(skb, offset + grehlen))
 381                 return;
 382         greh = (const struct gre_base_hdr *)(skb->data + offset);
 383         flags = greh->flags;
 384         if (flags & (GRE_VERSION | GRE_ROUTING))
 385                 return;
 386         if (flags & GRE_CSUM)
 387                 grehlen += 4;
 388         if (flags & GRE_KEY) {
 389                 key_off = grehlen + offset;
 390                 grehlen += 4;
 391         }
 392
 393         if (!pskb_may_pull(skb, offset + grehlen))
 394                 return;
 395         ipv6h = (const struct ipv6hdr *)skb->data;
 396         greh = (const struct gre_base_hdr *)(skb->data + offset);
 397         key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
 398
 399         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 400                                  key, greh->protocol);
 401         if (t == NULL)
 402                 return;
 403
 404         switch (type) {
 405                 __u32 teli;
 406                 struct ipv6_tlv_tnl_enc_lim *tel;
 407                 __u32 mtu;
 408         case ICMPV6_DEST_UNREACH:
 409                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 410                                     t->parms.name);
 411                 if (code != ICMPV6_PORT_UNREACH)
 412                         break;
 413                 return;
 414         case ICMPV6_TIME_EXCEED:
 415                 if (code == ICMPV6_EXC_HOPLIMIT) {
 416                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 417                                             t->parms.name);
 418                         break;
 419                 }
 420                 return;
 421         case ICMPV6_PARAMPROB:
 422                 teli = 0;
 423                 if (code == ICMPV6_HDR_FIELD)
 424                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 425
 426                 if (teli && teli == be32_to_cpu(info) - 2) {
 427                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 428                         if (tel->encap_limit == 0) {
 429                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 430                                                     t->parms.name);
 431                         }
 432                 } else {
 433                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 434                                             t->parms.name);
 435                 }
 436                 return;
 437         case ICMPV6_PKT_TOOBIG:
 438                 mtu = be32_to_cpu(info) - offset;
 439                 if (mtu < IPV6_MIN_MTU)
 440                         mtu = IPV6_MIN_MTU;
 441                 t->dev->mtu = mtu;
 442                 return;
 443         }
 444
 445         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 446                 t->err_count++;
 447         else
 448                 t->err_count = 1;
 449         t->err_time = jiffies;
 450 }
 451
 452 static int ip6gre_rcv(struct sk_buff *skb)
 453 {
 454         const struct ipv6hdr *ipv6h;
 455         u8     *h;
 456         __be16    flags;
 457         __sum16   csum = 0;
 458         __be32 key = 0;
 459         u32    seqno = 0;
 460         struct ip6_tnl *tunnel;
 461         int    offset = 4;
 462         __be16 gre_proto;
 463         int err;
 464
 465         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
 466                 goto drop;
 467
 468         ipv6h = ipv6_hdr(skb);
 469         h = skb->data;
 470         flags = *(__be16 *)h;
 471
 472         if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
 473                 /* - Version must be 0.
 474                    - We do not support routing headers.
 475                  */
 476                 if (flags&(GRE_VERSION|GRE_ROUTING))
 477                         goto drop;
 478
 479                 if (flags&GRE_CSUM) {
 480                         csum = skb_checksum_simple_validate(skb);
 481                         offset += 4;
 482                 }
 483                 if (flags&GRE_KEY) {
 484                         key = *(__be32 *)(h + offset);
 485                         offset += 4;
 486                 }
 487                 if (flags&GRE_SEQ) {
 488                         seqno = ntohl(*(__be32 *)(h + offset));
 489                         offset += 4;
 490                 }
 491         }
 492
 493         gre_proto = *(__be16 *)(h + 2);
 494
 495         tunnel = ip6gre_tunnel_lookup(skb->dev,
 496                                           &ipv6h->saddr, &ipv6h->daddr, key,
 497                                           gre_proto);
 498         if (tunnel) {
 499                 struct pcpu_sw_netstats *tstats;
 500
 501                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 502                         goto drop;
 503
 504                 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
 505                         tunnel->dev->stats.rx_dropped++;
 506                         goto drop;
 507                 }
 508
 509                 skb->protocol = gre_proto;
 510                 /* WCCP version 1 and 2 protocol decoding.
 511                  * - Change protocol to IPv6
 512                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
 513                  */
 514                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
 515                         skb->protocol = htons(ETH_P_IPV6);
 516                         if ((*(h + offset) & 0xF0) != 0x40)
 517                                 offset += 4;
 518                 }
 519
 520                 skb->mac_header = skb->network_header;
 521                 __pskb_pull(skb, offset);
 522                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
 523
 524                 if (((flags&GRE_CSUM) && csum) ||
 525                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
 526                         tunnel->dev->stats.rx_crc_errors++;
 527                         tunnel->dev->stats.rx_errors++;
 528                         goto drop;
 529                 }
 530                 if (tunnel->parms.i_flags&GRE_SEQ) {
 531                         if (!(flags&GRE_SEQ) ||
 532                             (tunnel->i_seqno &&
 533                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
 534                                 tunnel->dev->stats.rx_fifo_errors++;
 535                                 tunnel->dev->stats.rx_errors++;
 536                                 goto drop;
 537                         }
 538                         tunnel->i_seqno = seqno + 1;
 539                 }
 540
 541                 /* Warning: All skb pointers will be invalidated! */
 542                 if (tunnel->dev->type == ARPHRD_ETHER) {
 543                         if (!pskb_may_pull(skb, ETH_HLEN)) {
 544                                 tunnel->dev->stats.rx_length_errors++;
 545                                 tunnel->dev->stats.rx_errors++;
 546                                 goto drop;
 547                         }
 548
 549                         ipv6h = ipv6_hdr(skb);
 550                         skb->protocol = eth_type_trans(skb, tunnel->dev);
 551                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 552                 }
 553
 554                 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 555
 556                 skb_reset_network_header(skb);
 557
 558                 err = IP6_ECN_decapsulate(ipv6h, skb);
 559                 if (unlikely(err)) {
 560                         if (log_ecn_error)
 561                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 562                                                      &ipv6h->saddr,
 563                                                      ipv6_get_dsfield(ipv6h));
 564                         if (err > 1) {
 565                                 ++tunnel->dev->stats.rx_frame_errors;
 566                                 ++tunnel->dev->stats.rx_errors;
 567                                 goto drop;
 568                         }
 569                 }
 570
 571                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 572                 u64_stats_update_begin(&tstats->syncp);
 573                 tstats->rx_packets++;
 574                 tstats->rx_bytes += skb->len;
 575                 u64_stats_update_end(&tstats->syncp);
 576
 577                 netif_rx(skb);
 578
 579                 return 0;
 580         }
 581         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 582
 583 drop:
 584         kfree_skb(skb);
 585         return 0;
 586 }
 587
 588 struct ipv6_tel_txoption {
 589         struct ipv6_txoptions ops;
 590         __u8 dst_opt[8];
 591 };
 592
 593 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 594 {
 595         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 596
 597         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 598         opt->dst_opt[3] = 1;
 599         opt->dst_opt[4] = encap_limit;
 600         opt->dst_opt[5] = IPV6_TLV_PADN;
 601         opt->dst_opt[6] = 1;
 602
 603         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 604         opt->ops.opt_nflen = 8;
 605 }
 606
 607 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 608                          struct net_device *dev,
 609                          __u8 dsfield,
 610                          struct flowi6 *fl6,
 611                          int encap_limit,
 612                          __u32 *pmtu)
 613 {
 614         struct ip6_tnl *tunnel = netdev_priv(dev);
 615         struct net *net = tunnel->net;
 616         struct net_device *tdev;    /* Device to other host */
 617         struct ipv6hdr  *ipv6h;     /* Our new IP header */
 618         unsigned int max_headroom = 0; /* The extra header space needed */
 619         int    gre_hlen;
 620         struct ipv6_tel_txoption opt;
 621         int    mtu;
 622         struct dst_entry *dst = NULL, *ndst = NULL;
 623         struct net_device_stats *stats = &tunnel->dev->stats;
 624         int err = -1;
 625         u8 proto;
 626         struct sk_buff *new_skb;
 627
 628         if (dev->type == ARPHRD_ETHER)
 629                 IPCB(skb)->flags = 0;
 630
 631         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 632                 gre_hlen = 0;
 633                 ipv6h = (struct ipv6hdr *)skb->data;
 634                 fl6->daddr = ipv6h->daddr;
 635         } else {
 636                 gre_hlen = tunnel->hlen;
 637                 fl6->daddr = tunnel->parms.raddr;
 638         }
 639
 640         if (!fl6->flowi6_mark)
 641                 dst = ip6_tnl_dst_check(tunnel);
 642
 643         if (!dst) {
 644                 ndst = ip6_route_output(net, NULL, fl6);
 645
 646                 if (ndst->error)
 647                         goto tx_err_link_failure;
 648                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
 649                 if (IS_ERR(ndst)) {
 650                         err = PTR_ERR(ndst);
 651                         ndst = NULL;
 652                         goto tx_err_link_failure;
 653                 }
 654                 dst = ndst;
 655         }
 656
 657         tdev = dst->dev;
 658
 659         if (tdev == dev) {
 660                 stats->collisions++;
 661                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 662                                      tunnel->parms.name);
 663                 goto tx_err_dst_release;
 664         }
 665
 666         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 667         if (encap_limit >= 0) {
 668                 max_headroom += 8;
 669                 mtu -= 8;
 670         }
 671         if (mtu < IPV6_MIN_MTU)
 672                 mtu = IPV6_MIN_MTU;
 673         if (skb_dst(skb))
 674                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 675         if (skb->len > mtu) {
 676                 *pmtu = mtu;
 677                 err = -EMSGSIZE;
 678                 goto tx_err_dst_release;
 679         }
 680
 681         if (tunnel->err_count > 0) {
 682                 if (time_before(jiffies,
 683                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 684                         tunnel->err_count--;
 685
 686                         dst_link_failure(skb);
 687                 } else
 688                         tunnel->err_count = 0;
 689         }
 690
 691         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
 692
 693         max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 694
 695         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 696             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 697                 new_skb = skb_realloc_headroom(skb, max_headroom);
 698                 if (max_headroom > dev->needed_headroom)
 699                         dev->needed_headroom = max_headroom;
 700                 if (!new_skb)
 701                         goto tx_err_dst_release;
 702
 703                 if (skb->sk)
 704                         skb_set_owner_w(new_skb, skb->sk);
 705                 consume_skb(skb);
 706                 skb = new_skb;
 707         }
 708
 709         if (fl6->flowi6_mark) {
 710                 skb_dst_set(skb, dst);
 711                 ndst = NULL;
 712         } else {
 713                 skb_dst_set_noref(skb, dst);
 714         }
 715
 716         proto = NEXTHDR_GRE;
 717         if (encap_limit >= 0) {
 718                 init_tel_txopt(&opt, encap_limit);
 719                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 720         }
 721
 722         if (likely(!skb->encapsulation)) {
 723                 skb_reset_inner_headers(skb);
 724                 skb->encapsulation = 1;
 725         }
 726
 727         skb_push(skb, gre_hlen);
 728         skb_reset_network_header(skb);
 729         skb_set_transport_header(skb, sizeof(*ipv6h));
 730
 731         /*
 732          *      Push down and install the IP header.
 733          */
 734         ipv6h = ipv6_hdr(skb);
 735         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
 736         ipv6h->hop_limit = tunnel->parms.hop_limit;
 737         ipv6h->nexthdr = proto;
 738         ipv6h->saddr = fl6->saddr;
 739         ipv6h->daddr = fl6->daddr;
 740
 741         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 742         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
 743                                    htons(ETH_P_TEB) : skb->protocol;
 744
 745         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 746                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 747
 748                 if (tunnel->parms.o_flags&GRE_SEQ) {
 749                         ++tunnel->o_seqno;
 750                         *ptr = htonl(tunnel->o_seqno);
 751                         ptr--;
 752                 }
 753                 if (tunnel->parms.o_flags&GRE_KEY) {
 754                         *ptr = tunnel->parms.o_key;
 755                         ptr--;
 756                 }
 757                 if (tunnel->parms.o_flags&GRE_CSUM) {
 758                         *ptr = 0;
 759                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
 760                                 skb->len - sizeof(struct ipv6hdr));
 761                 }
 762         }
 763
 764         ip6tunnel_xmit(skb, dev);
 765         if (ndst)
 766                 ip6_tnl_dst_store(tunnel, ndst);
 767         return 0;
 768 tx_err_link_failure:
 769         stats->tx_carrier_errors++;
 770         dst_link_failure(skb);
 771 tx_err_dst_release:
 772         dst_release(ndst);
 773         return err;
 774 }
 775
 776 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 777 {
 778         struct ip6_tnl *t = netdev_priv(dev);
 779         const struct iphdr  *iph = ip_hdr(skb);
 780         int encap_limit = -1;
 781         struct flowi6 fl6;
 782         __u8 dsfield;
 783         __u32 mtu;
 784         int err;
 785
 786         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 787                 encap_limit = t->parms.encap_limit;
 788
 789         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 790         fl6.flowi6_proto = IPPROTO_GRE;
 791
 792         dsfield = ipv4_get_dsfield(iph);
 793
 794         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 795                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 796                                           & IPV6_TCLASS_MASK;
 797         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 798                 fl6.flowi6_mark = skb->mark;
 799
 800         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 801         if (err != 0) {
 802                 /* XXX: send ICMP error even if DF is not set. */
 803                 if (err == -EMSGSIZE)
 804                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 805                                   htonl(mtu));
 806                 return -1;
 807         }
 808
 809         return 0;
 810 }
 811
 812 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 813 {
 814         struct ip6_tnl *t = netdev_priv(dev);
 815         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 816         int encap_limit = -1;
 817         __u16 offset;
 818         struct flowi6 fl6;
 819         __u8 dsfield;
 820         __u32 mtu;
 821         int err;
 822
 823         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 824                 return -1;
 825
 826         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 827         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 828         ipv6h = ipv6_hdr(skb);
 829
 830         if (offset > 0) {
 831                 struct ipv6_tlv_tnl_enc_lim *tel;
 832                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 833                 if (tel->encap_limit == 0) {
 834                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 835                                     ICMPV6_HDR_FIELD, offset + 2);
 836                         return -1;
 837                 }
 838                 encap_limit = tel->encap_limit - 1;
 839         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 840                 encap_limit = t->parms.encap_limit;
 841
 842         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 843         fl6.flowi6_proto = IPPROTO_GRE;
 844
 845         dsfield = ipv6_get_dsfield(ipv6h);
 846         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 847                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 848         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 849                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
 850         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 851                 fl6.flowi6_mark = skb->mark;
 852
 853         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 854         if (err != 0) {
 855                 if (err == -EMSGSIZE)
 856                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 857                 return -1;
 858         }
 859
 860         return 0;
 861 }
 862
 863 /**
 864  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 865  *   @t: the outgoing tunnel device
 866  *   @hdr: IPv6 header from the incoming packet
 867  *
 868  * Description:
 869  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 870  *   doesn't match source of incoming packet.
 871  *
 872  * Return:
 873  *   1 if conflict,
 874  *   0 else
 875  **/
 876
 877 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 878         const struct ipv6hdr *hdr)
 879 {
 880         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 881 }
 882
 883 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 884 {
 885         struct ip6_tnl *t = netdev_priv(dev);
 886         int encap_limit = -1;
 887         struct flowi6 fl6;
 888         __u32 mtu;
 889         int err;
 890
 891         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 892                 encap_limit = t->parms.encap_limit;
 893
 894         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 895
 896         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 897
 898         return err;
 899 }
 900
 901 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 902         struct net_device *dev)
 903 {
 904         struct ip6_tnl *t = netdev_priv(dev);
 905         struct net_device_stats *stats = &t->dev->stats;
 906         int ret;
 907
 908         if (!ip6_tnl_xmit_ctl(t))
 909                 goto tx_err;
 910
 911         switch (skb->protocol) {
 912         case htons(ETH_P_IP):
 913                 ret = ip6gre_xmit_ipv4(skb, dev);
 914                 break;
 915         case htons(ETH_P_IPV6):
 916                 ret = ip6gre_xmit_ipv6(skb, dev);
 917                 break;
 918         default:
 919                 ret = ip6gre_xmit_other(skb, dev);
 920                 break;
 921         }
 922
 923         if (ret < 0)
 924                 goto tx_err;
 925
 926         return NETDEV_TX_OK;
 927
 928 tx_err:
 929         stats->tx_errors++;
 930         stats->tx_dropped++;
 931         kfree_skb(skb);
 932         return NETDEV_TX_OK;
 933 }
 934
 935 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 936 {
 937         struct net_device *dev = t->dev;
 938         struct __ip6_tnl_parm *p = &t->parms;
 939         struct flowi6 *fl6 = &t->fl.u.ip6;
 940         int addend = sizeof(struct ipv6hdr) + 4;
 941
 942         if (dev->type != ARPHRD_ETHER) {
 943                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 944                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 945         }
 946
 947         /* Set up flowi template */
 948         fl6->saddr = p->laddr;
 949         fl6->daddr = p->raddr;
 950         fl6->flowi6_oif = p->link;
 951         fl6->flowlabel = 0;
 952         fl6->flowi6_proto = IPPROTO_GRE;
 953
 954         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 955                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 956         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 957                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 958
 959         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 960         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 961
 962         if (p->flags&IP6_TNL_F_CAP_XMIT &&
 963                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
 964                 dev->flags |= IFF_POINTOPOINT;
 965         else
 966                 dev->flags &= ~IFF_POINTOPOINT;
 967
 968         /* Precalculate GRE options length */
 969         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
 970                 if (t->parms.o_flags&GRE_CSUM)
 971                         addend += 4;
 972                 if (t->parms.o_flags&GRE_KEY)
 973                         addend += 4;
 974                 if (t->parms.o_flags&GRE_SEQ)
 975                         addend += 4;
 976         }
 977         t->hlen = addend;
 978
 979         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 980                 int strict = (ipv6_addr_type(&p->raddr) &
 981                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 982
 983                 struct rt6_info *rt = rt6_lookup(t->net,
 984                                                  &p->raddr, &p->laddr,
 985                                                  p->link, strict);
 986
 987                 if (rt == NULL)
 988                         return;
 989
 990                 if (rt->dst.dev) {
 991                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
 992
 993                         if (set_mtu) {
 994                                 dev->mtu = rt->dst.dev->mtu - addend;
 995                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 996                                         dev->mtu -= 8;
 997
 998                                 if (dev->mtu < IPV6_MIN_MTU)
 999                                         dev->mtu = IPV6_MIN_MTU;
1000                         }
1001                 }
1002                 ip6_rt_put(rt);
1003         }
1004 }
1005
1006 static int ip6gre_tnl_change(struct ip6_tnl *t,
1007         const struct __ip6_tnl_parm *p, int set_mtu)
1008 {
1009         t->parms.laddr = p->laddr;
1010         t->parms.raddr = p->raddr;
1011         t->parms.flags = p->flags;
1012         t->parms.hop_limit = p->hop_limit;
1013         t->parms.encap_limit = p->encap_limit;
1014         t->parms.flowinfo = p->flowinfo;
1015         t->parms.link = p->link;
1016         t->parms.proto = p->proto;
1017         t->parms.i_key = p->i_key;
1018         t->parms.o_key = p->o_key;
1019         t->parms.i_flags = p->i_flags;
1020         t->parms.o_flags = p->o_flags;
1021         ip6_tnl_dst_reset(t);
1022         ip6gre_tnl_link_config(t, set_mtu);
1023         return 0;
1024 }
1025
1026 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1027         const struct ip6_tnl_parm2 *u)
1028 {
1029         p->laddr = u->laddr;
1030         p->raddr = u->raddr;
1031         p->flags = u->flags;
1032         p->hop_limit = u->hop_limit;
1033         p->encap_limit = u->encap_limit;
1034         p->flowinfo = u->flowinfo;
1035         p->link = u->link;
1036         p->i_key = u->i_key;
1037         p->o_key = u->o_key;
1038         p->i_flags = u->i_flags;
1039         p->o_flags = u->o_flags;
1040         memcpy(p->name, u->name, sizeof(u->name));
1041 }
1042
1043 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1044         const struct __ip6_tnl_parm *p)
1045 {
1046         u->proto = IPPROTO_GRE;
1047         u->laddr = p->laddr;
1048         u->raddr = p->raddr;
1049         u->flags = p->flags;
1050         u->hop_limit = p->hop_limit;
1051         u->encap_limit = p->encap_limit;
1052         u->flowinfo = p->flowinfo;
1053         u->link = p->link;
1054         u->i_key = p->i_key;
1055         u->o_key = p->o_key;
1056         u->i_flags = p->i_flags;
1057         u->o_flags = p->o_flags;
1058         memcpy(u->name, p->name, sizeof(u->name));
1059 }
1060
1061 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1062         struct ifreq *ifr, int cmd)
1063 {
1064         int err = 0;
1065         struct ip6_tnl_parm2 p;
1066         struct __ip6_tnl_parm p1;
1067         struct ip6_tnl *t = netdev_priv(dev);
1068         struct net *net = t->net;
1069         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1070
1071         switch (cmd) {
1072         case SIOCGETTUNNEL:
1073                 if (dev == ign->fb_tunnel_dev) {
1074                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1075                                 err = -EFAULT;
1076                                 break;
1077                         }
1078                         ip6gre_tnl_parm_from_user(&p1, &p);
1079                         t = ip6gre_tunnel_locate(net, &p1, 0);
1080                         if (t == NULL)
1081                                 t = netdev_priv(dev);
1082                 }
1083                 memset(&p, 0, sizeof(p));
1084                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1085                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1086                         err = -EFAULT;
1087                 break;
1088
1089         case SIOCADDTUNNEL:
1090         case SIOCCHGTUNNEL:
1091                 err = -EPERM;
1092                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1093                         goto done;
1094
1095                 err = -EFAULT;
1096                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1097                         goto done;
1098
1099                 err = -EINVAL;
1100                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1101                         goto done;
1102
1103                 if (!(p.i_flags&GRE_KEY))
1104                         p.i_key = 0;
1105                 if (!(p.o_flags&GRE_KEY))
1106                         p.o_key = 0;
1107
1108                 ip6gre_tnl_parm_from_user(&p1, &p);
1109                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1110
1111                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1112                         if (t != NULL) {
1113                                 if (t->dev != dev) {
1114                                         err = -EEXIST;
1115                                         break;
1116                                 }
1117                         } else {
1118                                 t = netdev_priv(dev);
1119
1120                                 ip6gre_tunnel_unlink(ign, t);
1121                                 synchronize_net();
1122                                 ip6gre_tnl_change(t, &p1, 1);
1123                                 ip6gre_tunnel_link(ign, t);
1124                                 netdev_state_change(dev);
1125                         }
1126                 }
1127
1128                 if (t) {
1129                         err = 0;
1130
1131                         memset(&p, 0, sizeof(p));
1132                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1133                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1134                                 err = -EFAULT;
1135                 } else
1136                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1137                 break;
1138
1139         case SIOCDELTUNNEL:
1140                 err = -EPERM;
1141                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1142                         goto done;
1143
1144                 if (dev == ign->fb_tunnel_dev) {
1145                         err = -EFAULT;
1146                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1147                                 goto done;
1148                         err = -ENOENT;
1149                         ip6gre_tnl_parm_from_user(&p1, &p);
1150                         t = ip6gre_tunnel_locate(net, &p1, 0);
1151                         if (t == NULL)
1152                                 goto done;
1153                         err = -EPERM;
1154                         if (t == netdev_priv(ign->fb_tunnel_dev))
1155                                 goto done;
1156                         dev = t->dev;
1157                 }
1158                 unregister_netdevice(dev);
1159                 err = 0;
1160                 break;
1161
1162         default:
1163                 err = -EINVAL;
1164         }
1165
1166 done:
1167         return err;
1168 }
1169
1170 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1171 {
1172         if (new_mtu < 68 ||
1173             new_mtu > 0xFFF8 - dev->hard_header_len)
1174                 return -EINVAL;
1175         dev->mtu = new_mtu;
1176         return 0;
1177 }
1178
1179 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1180                          unsigned short type, const void *daddr,
1181                          const void *saddr, unsigned int len)
1182 {
1183         struct ip6_tnl *t = netdev_priv(dev);
1184         struct ipv6hdr *ipv6h;
1185         __be16 *p;
1186
1187         ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen + sizeof(*ipv6h));
1188         ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
1189         ipv6h->hop_limit = t->parms.hop_limit;
1190         ipv6h->nexthdr = NEXTHDR_GRE;
1191         ipv6h->saddr = t->parms.laddr;
1192         ipv6h->daddr = t->parms.raddr;
1193
1194         p = (__be16 *)(ipv6h + 1);
1195         p[0] = t->parms.o_flags;
1196         p[1] = htons(type);
1197
1198         /*
1199          *      Set the source hardware address.
1200          */
1201
1202         if (saddr)
1203                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1204         if (daddr)
1205                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1206         if (!ipv6_addr_any(&ipv6h->daddr))
1207                 return t->hlen;
1208
1209         return -t->hlen;
1210 }
1211
1212 static const struct header_ops ip6gre_header_ops = {
1213         .create = ip6gre_header,
1214 };
1215
1216 static const struct net_device_ops ip6gre_netdev_ops = {
1217         .ndo_init               = ip6gre_tunnel_init,
1218         .ndo_uninit             = ip6gre_tunnel_uninit,
1219         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1220         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1221         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1222         .ndo_get_stats64        = ip_tunnel_get_stats64,
1223 };
1224
1225 static void ip6gre_dev_free(struct net_device *dev)
1226 {
1227         free_percpu(dev->tstats);
1228         free_netdev(dev);
1229 }
1230
1231 static void ip6gre_tunnel_setup(struct net_device *dev)
1232 {
1233         struct ip6_tnl *t;
1234
1235         dev->netdev_ops = &ip6gre_netdev_ops;
1236         dev->destructor = ip6gre_dev_free;
1237
1238         dev->type = ARPHRD_IP6GRE;
1239         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1240         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1241         t = netdev_priv(dev);
1242         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1243                 dev->mtu -= 8;
1244         dev->flags |= IFF_NOARP;
1245         dev->iflink = 0;
1246         dev->addr_len = sizeof(struct in6_addr);
1247         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1248 }
1249
1250 static int ip6gre_tunnel_init(struct net_device *dev)
1251 {
1252         struct ip6_tnl *tunnel;
1253         int i;
1254
1255         tunnel = netdev_priv(dev);
1256
1257         tunnel->dev = dev;
1258         tunnel->net = dev_net(dev);
1259         strcpy(tunnel->parms.name, dev->name);
1260
1261         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1262         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1263
1264         if (ipv6_addr_any(&tunnel->parms.raddr))
1265                 dev->header_ops = &ip6gre_header_ops;
1266
1267         dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1268         if (!dev->tstats)
1269                 return -ENOMEM;
1270
1271         for_each_possible_cpu(i) {
1272                 struct pcpu_sw_netstats *ip6gre_tunnel_stats;
1273                 ip6gre_tunnel_stats = per_cpu_ptr(dev->tstats, i);
1274                 u64_stats_init(&ip6gre_tunnel_stats->syncp);
1275         }
1276
1277         dev->iflink = tunnel->parms.link;
1278
1279         return 0;
1280 }
1281
1282 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1283 {
1284         struct ip6_tnl *tunnel = netdev_priv(dev);
1285
1286         tunnel->dev = dev;
1287         tunnel->net = dev_net(dev);
1288         strcpy(tunnel->parms.name, dev->name);
1289
1290         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1291
1292         dev_hold(dev);
1293 }
1294
1295
1296 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1297         .handler     = ip6gre_rcv,
1298         .err_handler = ip6gre_err,
1299         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1300 };
1301
1302 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1303 {
1304         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1305         struct net_device *dev, *aux;
1306         int prio;
1307
1308         for_each_netdev_safe(net, dev, aux)
1309                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1310                     dev->rtnl_link_ops == &ip6gre_tap_ops)
1311                         unregister_netdevice_queue(dev, head);
1312
1313         for (prio = 0; prio < 4; prio++) {
1314                 int h;
1315                 for (h = 0; h < HASH_SIZE; h++) {
1316                         struct ip6_tnl *t;
1317
1318                         t = rtnl_dereference(ign->tunnels[prio][h]);
1319
1320                         while (t != NULL) {
1321                                 /* If dev is in the same netns, it has already
1322                                  * been added to the list by the previous loop.
1323                                  */
1324                                 if (!net_eq(dev_net(t->dev), net))
1325                                         unregister_netdevice_queue(t->dev,
1326                                                                    head);
1327                                 t = rtnl_dereference(t->next);
1328                         }
1329                 }
1330         }
1331 }
1332
1333 static int __net_init ip6gre_init_net(struct net *net)
1334 {
1335         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1336         int err;
1337
1338         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1339                                            ip6gre_tunnel_setup);
1340         if (!ign->fb_tunnel_dev) {
1341                 err = -ENOMEM;
1342                 goto err_alloc_dev;
1343         }
1344         dev_net_set(ign->fb_tunnel_dev, net);
1345         /* FB netdevice is special: we have one, and only one per netns.
1346          * Allowing to move it to another netns is clearly unsafe.
1347          */
1348         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1349
1350
1351         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1352         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1353
1354         err = register_netdev(ign->fb_tunnel_dev);
1355         if (err)
1356                 goto err_reg_dev;
1357
1358         rcu_assign_pointer(ign->tunnels_wc[0],
1359                            netdev_priv(ign->fb_tunnel_dev));
1360         return 0;
1361
1362 err_reg_dev:
1363         ip6gre_dev_free(ign->fb_tunnel_dev);
1364 err_alloc_dev:
1365         return err;
1366 }
1367
1368 static void __net_exit ip6gre_exit_net(struct net *net)
1369 {
1370         LIST_HEAD(list);
1371
1372         rtnl_lock();
1373         ip6gre_destroy_tunnels(net, &list);
1374         unregister_netdevice_many(&list);
1375         rtnl_unlock();
1376 }
1377
1378 static struct pernet_operations ip6gre_net_ops = {
1379         .init = ip6gre_init_net,
1380         .exit = ip6gre_exit_net,
1381         .id   = &ip6gre_net_id,
1382         .size = sizeof(struct ip6gre_net),
1383 };
1384
1385 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1386 {
1387         __be16 flags;
1388
1389         if (!data)
1390                 return 0;
1391
1392         flags = 0;
1393         if (data[IFLA_GRE_IFLAGS])
1394                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1395         if (data[IFLA_GRE_OFLAGS])
1396                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1397         if (flags & (GRE_VERSION|GRE_ROUTING))
1398                 return -EINVAL;
1399
1400         return 0;
1401 }
1402
1403 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1404 {
1405         struct in6_addr daddr;
1406
1407         if (tb[IFLA_ADDRESS]) {
1408                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1409                         return -EINVAL;
1410                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1411                         return -EADDRNOTAVAIL;
1412         }
1413
1414         if (!data)
1415                 goto out;
1416
1417         if (data[IFLA_GRE_REMOTE]) {
1418                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1419                 if (ipv6_addr_any(&daddr))
1420                         return -EINVAL;
1421         }
1422
1423 out:
1424         return ip6gre_tunnel_validate(tb, data);
1425 }
1426
1427
1428 static void ip6gre_netlink_parms(struct nlattr *data[],
1429                                 struct __ip6_tnl_parm *parms)
1430 {
1431         memset(parms, 0, sizeof(*parms));
1432
1433         if (!data)
1434                 return;
1435
1436         if (data[IFLA_GRE_LINK])
1437                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1438
1439         if (data[IFLA_GRE_IFLAGS])
1440                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1441
1442         if (data[IFLA_GRE_OFLAGS])
1443                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1444
1445         if (data[IFLA_GRE_IKEY])
1446                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1447
1448         if (data[IFLA_GRE_OKEY])
1449                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1450
1451         if (data[IFLA_GRE_LOCAL])
1452                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1453
1454         if (data[IFLA_GRE_REMOTE])
1455                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1456
1457         if (data[IFLA_GRE_TTL])
1458                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1459
1460         if (data[IFLA_GRE_ENCAP_LIMIT])
1461                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1462
1463         if (data[IFLA_GRE_FLOWINFO])
1464                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1465
1466         if (data[IFLA_GRE_FLAGS])
1467                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1468 }
1469
1470 static int ip6gre_tap_init(struct net_device *dev)
1471 {
1472         struct ip6_tnl *tunnel;
1473
1474         tunnel = netdev_priv(dev);
1475
1476         tunnel->dev = dev;
1477         tunnel->net = dev_net(dev);
1478         strcpy(tunnel->parms.name, dev->name);
1479
1480         ip6gre_tnl_link_config(tunnel, 1);
1481
1482         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1483         if (!dev->tstats)
1484                 return -ENOMEM;
1485
1486         dev->iflink = tunnel->parms.link;
1487
1488         return 0;
1489 }
1490
1491 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1492         .ndo_init = ip6gre_tap_init,
1493         .ndo_uninit = ip6gre_tunnel_uninit,
1494         .ndo_start_xmit = ip6gre_tunnel_xmit,
1495         .ndo_set_mac_address = eth_mac_addr,
1496         .ndo_validate_addr = eth_validate_addr,
1497         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1498         .ndo_get_stats64 = ip_tunnel_get_stats64,
1499 };
1500
1501 static void ip6gre_tap_setup(struct net_device *dev)
1502 {
1503
1504         ether_setup(dev);
1505
1506         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1507         dev->destructor = ip6gre_dev_free;
1508
1509         dev->iflink = 0;
1510         dev->features |= NETIF_F_NETNS_LOCAL;
1511 }
1512
1513 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1514         struct nlattr *tb[], struct nlattr *data[])
1515 {
1516         struct ip6_tnl *nt;
1517         struct net *net = dev_net(dev);
1518         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1519         int err;
1520
1521         nt = netdev_priv(dev);
1522         ip6gre_netlink_parms(data, &nt->parms);
1523
1524         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1525                 return -EEXIST;
1526
1527         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1528                 eth_hw_addr_random(dev);
1529
1530         nt->dev = dev;
1531         nt->net = dev_net(dev);
1532         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1533
1534         /* Can use a lockless transmit, unless we generate output sequences */
1535         if (!(nt->parms.o_flags & GRE_SEQ))
1536                 dev->features |= NETIF_F_LLTX;
1537
1538         err = register_netdevice(dev);
1539         if (err)
1540                 goto out;
1541
1542         dev_hold(dev);
1543         ip6gre_tunnel_link(ign, nt);
1544
1545 out:
1546         return err;
1547 }
1548
1549 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1550                             struct nlattr *data[])
1551 {
1552         struct ip6_tnl *t, *nt = netdev_priv(dev);
1553         struct net *net = nt->net;
1554         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1555         struct __ip6_tnl_parm p;
1556
1557         if (dev == ign->fb_tunnel_dev)
1558                 return -EINVAL;
1559
1560         ip6gre_netlink_parms(data, &p);
1561
1562         t = ip6gre_tunnel_locate(net, &p, 0);
1563
1564         if (t) {
1565                 if (t->dev != dev)
1566                         return -EEXIST;
1567         } else {
1568                 t = nt;
1569         }
1570
1571         ip6gre_tunnel_unlink(ign, t);
1572         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1573         ip6gre_tunnel_link(ign, t);
1574         return 0;
1575 }
1576
1577 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1578 {
1579         struct net *net = dev_net(dev);
1580         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1581
1582         if (dev != ign->fb_tunnel_dev)
1583                 unregister_netdevice_queue(dev, head);
1584 }
1585
1586 static size_t ip6gre_get_size(const struct net_device *dev)
1587 {
1588         return
1589                 /* IFLA_GRE_LINK */
1590                 nla_total_size(4) +
1591                 /* IFLA_GRE_IFLAGS */
1592                 nla_total_size(2) +
1593                 /* IFLA_GRE_OFLAGS */
1594                 nla_total_size(2) +
1595                 /* IFLA_GRE_IKEY */
1596                 nla_total_size(4) +
1597                 /* IFLA_GRE_OKEY */
1598                 nla_total_size(4) +
1599                 /* IFLA_GRE_LOCAL */
1600                 nla_total_size(sizeof(struct in6_addr)) +
1601                 /* IFLA_GRE_REMOTE */
1602                 nla_total_size(sizeof(struct in6_addr)) +
1603                 /* IFLA_GRE_TTL */
1604                 nla_total_size(1) +
1605                 /* IFLA_GRE_TOS */
1606                 nla_total_size(1) +
1607                 /* IFLA_GRE_ENCAP_LIMIT */
1608                 nla_total_size(1) +
1609                 /* IFLA_GRE_FLOWINFO */
1610                 nla_total_size(4) +
1611                 /* IFLA_GRE_FLAGS */
1612                 nla_total_size(4) +
1613                 0;
1614 }
1615
1616 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1617 {
1618         struct ip6_tnl *t = netdev_priv(dev);
1619         struct __ip6_tnl_parm *p = &t->parms;
1620
1621         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1622             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1623             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1624             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1625             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1626             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1627             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1628             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1629             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1630             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1631             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1632             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1633                 goto nla_put_failure;
1634         return 0;
1635
1636 nla_put_failure:
1637         return -EMSGSIZE;
1638 }
1639
1640 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1641         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1642         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1643         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1644         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1645         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1646         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1647         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1648         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1649         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1650         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1651         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1652 };
1653
1654 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1655         .kind           = "ip6gre",
1656         .maxtype        = IFLA_GRE_MAX,
1657         .policy         = ip6gre_policy,
1658         .priv_size      = sizeof(struct ip6_tnl),
1659         .setup          = ip6gre_tunnel_setup,
1660         .validate       = ip6gre_tunnel_validate,
1661         .newlink        = ip6gre_newlink,
1662         .changelink     = ip6gre_changelink,
1663         .dellink        = ip6gre_dellink,
1664         .get_size       = ip6gre_get_size,
1665         .fill_info      = ip6gre_fill_info,
1666 };
1667
1668 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1669         .kind           = "ip6gretap",
1670         .maxtype        = IFLA_GRE_MAX,
1671         .policy         = ip6gre_policy,
1672         .priv_size      = sizeof(struct ip6_tnl),
1673         .setup          = ip6gre_tap_setup,
1674         .validate       = ip6gre_tap_validate,
1675         .newlink        = ip6gre_newlink,
1676         .changelink     = ip6gre_changelink,
1677         .get_size       = ip6gre_get_size,
1678         .fill_info      = ip6gre_fill_info,
1679 };
1680
1681 /*
1682  *      And now the modules code and kernel interface.
1683  */
1684
1685 static int __init ip6gre_init(void)
1686 {
1687         int err;
1688
1689         pr_info("GRE over IPv6 tunneling driver\n");
1690
1691         err = register_pernet_device(&ip6gre_net_ops);
1692         if (err < 0)
1693                 return err;
1694
1695         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1696         if (err < 0) {
1697                 pr_info("%s: can't add protocol\n", __func__);
1698                 goto add_proto_failed;
1699         }
1700
1701         err = rtnl_link_register(&ip6gre_link_ops);
1702         if (err < 0)
1703                 goto rtnl_link_failed;
1704
1705         err = rtnl_link_register(&ip6gre_tap_ops);
1706         if (err < 0)
1707                 goto tap_ops_failed;
1708
1709 out:
1710         return err;
1711
1712 tap_ops_failed:
1713         rtnl_link_unregister(&ip6gre_link_ops);
1714 rtnl_link_failed:
1715         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1716 add_proto_failed:
1717         unregister_pernet_device(&ip6gre_net_ops);
1718         goto out;
1719 }
1720
1721 static void __exit ip6gre_fini(void)
1722 {
1723         rtnl_link_unregister(&ip6gre_tap_ops);
1724         rtnl_link_unregister(&ip6gre_link_ops);
1725         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1726         unregister_pernet_device(&ip6gre_net_ops);
1727 }
1728
1729 module_init(ip6gre_init);
1730 module_exit(ip6gre_fini);
1731 MODULE_LICENSE("GPL");
1732 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1733 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1734 MODULE_ALIAS_RTNL_LINK("ip6gre");
1735 MODULE_ALIAS_NETDEV("ip6gre0");