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
  59
  60 static bool log_ecn_error = true;
  61 module_param(log_ecn_error, bool, 0644);
  62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  63
  64 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
  65 #define IPV6_TCLASS_SHIFT 20
  66
  67 #define HASH_SIZE_SHIFT  5
  68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  69
  70 static int ip6gre_net_id __read_mostly;
  71 struct ip6gre_net {
  72         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
  73
  74         struct net_device *fb_tunnel_dev;
  75 };
  76
  77 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  78 static int ip6gre_tunnel_init(struct net_device *dev);
  79 static void ip6gre_tunnel_setup(struct net_device *dev);
  80 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  81 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  82
  83 /* Tunnel hash table */
  84
  85 /*
  86    4 hash tables:
  87
  88    3: (remote,local)
  89    2: (remote,*)
  90    1: (*,local)
  91    0: (*,*)
  92
  93    We require exact key match i.e. if a key is present in packet
  94    it will match only tunnel with the same key; if it is not present,
  95    it will match only keyless tunnel.
  96
  97    All keysless packets, if not matched configured keyless tunnels
  98    will match fallback tunnel.
  99  */
 100
 101 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
 102 static u32 HASH_ADDR(const struct in6_addr *addr)
 103 {
 104         u32 hash = ipv6_addr_hash(addr);
 105
 106         return hash_32(hash, HASH_SIZE_SHIFT);
 107 }
 108
 109 #define tunnels_r_l     tunnels[3]
 110 #define tunnels_r       tunnels[2]
 111 #define tunnels_l       tunnels[1]
 112 #define tunnels_wc      tunnels[0]
 113
 114 /* Given src, dst and key, find appropriate for input tunnel. */
 115
 116 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 117                 const struct in6_addr *remote, const struct in6_addr *local,
 118                 __be32 key, __be16 gre_proto)
 119 {
 120         struct net *net = dev_net(dev);
 121         int link = dev->ifindex;
 122         unsigned int h0 = HASH_ADDR(remote);
 123         unsigned int h1 = HASH_KEY(key);
 124         struct ip6_tnl *t, *cand = NULL;
 125         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 126         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
 127                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 128         int score, cand_score = 4;
 129
 130         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 131                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 132                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 133                     key != t->parms.i_key ||
 134                     !(t->dev->flags & IFF_UP))
 135                         continue;
 136
 137                 if (t->dev->type != ARPHRD_IP6GRE &&
 138                     t->dev->type != dev_type)
 139                         continue;
 140
 141                 score = 0;
 142                 if (t->parms.link != link)
 143                         score |= 1;
 144                 if (t->dev->type != dev_type)
 145                         score |= 2;
 146                 if (score == 0)
 147                         return t;
 148
 149                 if (score < cand_score) {
 150                         cand = t;
 151                         cand_score = score;
 152                 }
 153         }
 154
 155         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 156                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 157                     key != t->parms.i_key ||
 158                     !(t->dev->flags & IFF_UP))
 159                         continue;
 160
 161                 if (t->dev->type != ARPHRD_IP6GRE &&
 162                     t->dev->type != dev_type)
 163                         continue;
 164
 165                 score = 0;
 166                 if (t->parms.link != link)
 167                         score |= 1;
 168                 if (t->dev->type != dev_type)
 169                         score |= 2;
 170                 if (score == 0)
 171                         return t;
 172
 173                 if (score < cand_score) {
 174                         cand = t;
 175                         cand_score = score;
 176                 }
 177         }
 178
 179         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 180                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 181                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 182                                  !ipv6_addr_is_multicast(local))) ||
 183                     key != t->parms.i_key ||
 184                     !(t->dev->flags & IFF_UP))
 185                         continue;
 186
 187                 if (t->dev->type != ARPHRD_IP6GRE &&
 188                     t->dev->type != dev_type)
 189                         continue;
 190
 191                 score = 0;
 192                 if (t->parms.link != link)
 193                         score |= 1;
 194                 if (t->dev->type != dev_type)
 195                         score |= 2;
 196                 if (score == 0)
 197                         return t;
 198
 199                 if (score < cand_score) {
 200                         cand = t;
 201                         cand_score = score;
 202                 }
 203         }
 204
 205         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 206                 if (t->parms.i_key != key ||
 207                     !(t->dev->flags & IFF_UP))
 208                         continue;
 209
 210                 if (t->dev->type != ARPHRD_IP6GRE &&
 211                     t->dev->type != dev_type)
 212                         continue;
 213
 214                 score = 0;
 215                 if (t->parms.link != link)
 216                         score |= 1;
 217                 if (t->dev->type != dev_type)
 218                         score |= 2;
 219                 if (score == 0)
 220                         return t;
 221
 222                 if (score < cand_score) {
 223                         cand = t;
 224                         cand_score = score;
 225                 }
 226         }
 227
 228         if (cand != NULL)
 229                 return cand;
 230
 231         dev = ign->fb_tunnel_dev;
 232         if (dev->flags & IFF_UP)
 233                 return netdev_priv(dev);
 234
 235         return NULL;
 236 }
 237
 238 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 239                 const struct __ip6_tnl_parm *p)
 240 {
 241         const struct in6_addr *remote = &p->raddr;
 242         const struct in6_addr *local = &p->laddr;
 243         unsigned int h = HASH_KEY(p->i_key);
 244         int prio = 0;
 245
 246         if (!ipv6_addr_any(local))
 247                 prio |= 1;
 248         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 249                 prio |= 2;
 250                 h ^= HASH_ADDR(remote);
 251         }
 252
 253         return &ign->tunnels[prio][h];
 254 }
 255
 256 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 257                 const struct ip6_tnl *t)
 258 {
 259         return __ip6gre_bucket(ign, &t->parms);
 260 }
 261
 262 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 263 {
 264         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 265
 266         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 267         rcu_assign_pointer(*tp, t);
 268 }
 269
 270 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 271 {
 272         struct ip6_tnl __rcu **tp;
 273         struct ip6_tnl *iter;
 274
 275         for (tp = ip6gre_bucket(ign, t);
 276              (iter = rtnl_dereference(*tp)) != NULL;
 277              tp = &iter->next) {
 278                 if (t == iter) {
 279                         rcu_assign_pointer(*tp, t->next);
 280                         break;
 281                 }
 282         }
 283 }
 284
 285 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 286                                            const struct __ip6_tnl_parm *parms,
 287                                            int type)
 288 {
 289         const struct in6_addr *remote = &parms->raddr;
 290         const struct in6_addr *local = &parms->laddr;
 291         __be32 key = parms->i_key;
 292         int link = parms->link;
 293         struct ip6_tnl *t;
 294         struct ip6_tnl __rcu **tp;
 295         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 296
 297         for (tp = __ip6gre_bucket(ign, parms);
 298              (t = rtnl_dereference(*tp)) != NULL;
 299              tp = &t->next)
 300                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 301                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 302                     key == t->parms.i_key &&
 303                     link == t->parms.link &&
 304                     type == t->dev->type)
 305                         break;
 306
 307         return t;
 308 }
 309
 310 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 311                 const struct __ip6_tnl_parm *parms, int create)
 312 {
 313         struct ip6_tnl *t, *nt;
 314         struct net_device *dev;
 315         char name[IFNAMSIZ];
 316         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 317
 318         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 319         if (t || !create)
 320                 return t;
 321
 322         if (parms->name[0])
 323                 strlcpy(name, parms->name, IFNAMSIZ);
 324         else
 325                 strcpy(name, "ip6gre%d");
 326
 327         dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
 328         if (!dev)
 329                 return NULL;
 330
 331         dev_net_set(dev, net);
 332
 333         nt = netdev_priv(dev);
 334         nt->parms = *parms;
 335         dev->rtnl_link_ops = &ip6gre_link_ops;
 336
 337         nt->dev = dev;
 338         ip6gre_tnl_link_config(nt, 1);
 339
 340         if (register_netdevice(dev) < 0)
 341                 goto failed_free;
 342
 343         /* Can use a lockless transmit, unless we generate output sequences */
 344         if (!(nt->parms.o_flags & GRE_SEQ))
 345                 dev->features |= NETIF_F_LLTX;
 346
 347         dev_hold(dev);
 348         ip6gre_tunnel_link(ign, nt);
 349         return nt;
 350
 351 failed_free:
 352         free_netdev(dev);
 353         return NULL;
 354 }
 355
 356 static void ip6gre_tunnel_uninit(struct net_device *dev)
 357 {
 358         struct net *net = dev_net(dev);
 359         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 360
 361         ip6gre_tunnel_unlink(ign, netdev_priv(dev));
 362         ip6_tnl_dst_reset(netdev_priv(dev));
 363         dev_put(dev);
 364 }
 365
 366
 367 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 368                 u8 type, u8 code, int offset, __be32 info)
 369 {
 370         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 371         __be16 *p = (__be16 *)(skb->data + offset);
 372         int grehlen = offset + 4;
 373         struct ip6_tnl *t;
 374         __be16 flags;
 375
 376         flags = p[0];
 377         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
 378                 if (flags&(GRE_VERSION|GRE_ROUTING))
 379                         return;
 380                 if (flags&GRE_KEY) {
 381                         grehlen += 4;
 382                         if (flags&GRE_CSUM)
 383                                 grehlen += 4;
 384                 }
 385         }
 386
 387         /* If only 8 bytes returned, keyed message will be dropped here */
 388         if (!pskb_may_pull(skb, grehlen))
 389                 return;
 390         ipv6h = (const struct ipv6hdr *)skb->data;
 391         p = (__be16 *)(skb->data + offset);
 392
 393         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 394                                 flags & GRE_KEY ?
 395                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
 396                                 p[1]);
 397         if (t == NULL)
 398                 return;
 399
 400         switch (type) {
 401                 __u32 teli;
 402                 struct ipv6_tlv_tnl_enc_lim *tel;
 403                 __u32 mtu;
 404         case ICMPV6_DEST_UNREACH:
 405                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
 406                                      t->parms.name);
 407                 break;
 408         case ICMPV6_TIME_EXCEED:
 409                 if (code == ICMPV6_EXC_HOPLIMIT) {
 410                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 411                                              t->parms.name);
 412                 }
 413                 break;
 414         case ICMPV6_PARAMPROB:
 415                 teli = 0;
 416                 if (code == ICMPV6_HDR_FIELD)
 417                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 418
 419                 if (teli && teli == info - 2) {
 420                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 421                         if (tel->encap_limit == 0) {
 422                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 423                                                      t->parms.name);
 424                         }
 425                 } else {
 426                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 427                                              t->parms.name);
 428                 }
 429                 break;
 430         case ICMPV6_PKT_TOOBIG:
 431                 mtu = info - offset;
 432                 if (mtu < IPV6_MIN_MTU)
 433                         mtu = IPV6_MIN_MTU;
 434                 t->dev->mtu = mtu;
 435                 break;
 436         }
 437
 438         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 439                 t->err_count++;
 440         else
 441                 t->err_count = 1;
 442         t->err_time = jiffies;
 443 }
 444
 445 static int ip6gre_rcv(struct sk_buff *skb)
 446 {
 447         const struct ipv6hdr *ipv6h;
 448         u8     *h;
 449         __be16    flags;
 450         __sum16   csum = 0;
 451         __be32 key = 0;
 452         u32    seqno = 0;
 453         struct ip6_tnl *tunnel;
 454         int    offset = 4;
 455         __be16 gre_proto;
 456         int err;
 457
 458         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
 459                 goto drop;
 460
 461         ipv6h = ipv6_hdr(skb);
 462         h = skb->data;
 463         flags = *(__be16 *)h;
 464
 465         if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
 466                 /* - Version must be 0.
 467                    - We do not support routing headers.
 468                  */
 469                 if (flags&(GRE_VERSION|GRE_ROUTING))
 470                         goto drop;
 471
 472                 if (flags&GRE_CSUM) {
 473                         switch (skb->ip_summed) {
 474                         case CHECKSUM_COMPLETE:
 475                                 csum = csum_fold(skb->csum);
 476                                 if (!csum)
 477                                         break;
 478                                 /* fall through */
 479                         case CHECKSUM_NONE:
 480                                 skb->csum = 0;
 481                                 csum = __skb_checksum_complete(skb);
 482                                 skb->ip_summed = CHECKSUM_COMPLETE;
 483                         }
 484                         offset += 4;
 485                 }
 486                 if (flags&GRE_KEY) {
 487                         key = *(__be32 *)(h + offset);
 488                         offset += 4;
 489                 }
 490                 if (flags&GRE_SEQ) {
 491                         seqno = ntohl(*(__be32 *)(h + offset));
 492                         offset += 4;
 493                 }
 494         }
 495
 496         gre_proto = *(__be16 *)(h + 2);
 497
 498         tunnel = ip6gre_tunnel_lookup(skb->dev,
 499                                           &ipv6h->saddr, &ipv6h->daddr, key,
 500                                           gre_proto);
 501         if (tunnel) {
 502                 struct pcpu_tstats *tstats;
 503
 504                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 505                         goto drop;
 506
 507                 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
 508                         tunnel->dev->stats.rx_dropped++;
 509                         goto drop;
 510                 }
 511
 512                 secpath_reset(skb);
 513
 514                 skb->protocol = gre_proto;
 515                 /* WCCP version 1 and 2 protocol decoding.
 516                  * - Change protocol to IPv6
 517                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
 518                  */
 519                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
 520                         skb->protocol = htons(ETH_P_IPV6);
 521                         if ((*(h + offset) & 0xF0) != 0x40)
 522                                 offset += 4;
 523                 }
 524
 525                 skb->mac_header = skb->network_header;
 526                 __pskb_pull(skb, offset);
 527                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
 528                 skb->pkt_type = PACKET_HOST;
 529
 530                 if (((flags&GRE_CSUM) && csum) ||
 531                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
 532                         tunnel->dev->stats.rx_crc_errors++;
 533                         tunnel->dev->stats.rx_errors++;
 534                         goto drop;
 535                 }
 536                 if (tunnel->parms.i_flags&GRE_SEQ) {
 537                         if (!(flags&GRE_SEQ) ||
 538                             (tunnel->i_seqno &&
 539                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
 540                                 tunnel->dev->stats.rx_fifo_errors++;
 541                                 tunnel->dev->stats.rx_errors++;
 542                                 goto drop;
 543                         }
 544                         tunnel->i_seqno = seqno + 1;
 545                 }
 546
 547                 /* Warning: All skb pointers will be invalidated! */
 548                 if (tunnel->dev->type == ARPHRD_ETHER) {
 549                         if (!pskb_may_pull(skb, ETH_HLEN)) {
 550                                 tunnel->dev->stats.rx_length_errors++;
 551                                 tunnel->dev->stats.rx_errors++;
 552                                 goto drop;
 553                         }
 554
 555                         ipv6h = ipv6_hdr(skb);
 556                         skb->protocol = eth_type_trans(skb, tunnel->dev);
 557                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 558                 }
 559
 560                 __skb_tunnel_rx(skb, tunnel->dev);
 561
 562                 skb_reset_network_header(skb);
 563
 564                 err = IP6_ECN_decapsulate(ipv6h, skb);
 565                 if (unlikely(err)) {
 566                         if (log_ecn_error)
 567                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 568                                                      &ipv6h->saddr,
 569                                                      ipv6_get_dsfield(ipv6h));
 570                         if (err > 1) {
 571                                 ++tunnel->dev->stats.rx_frame_errors;
 572                                 ++tunnel->dev->stats.rx_errors;
 573                                 goto drop;
 574                         }
 575                 }
 576
 577                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 578                 u64_stats_update_begin(&tstats->syncp);
 579                 tstats->rx_packets++;
 580                 tstats->rx_bytes += skb->len;
 581                 u64_stats_update_end(&tstats->syncp);
 582
 583                 netif_rx(skb);
 584
 585                 return 0;
 586         }
 587         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 588
 589 drop:
 590         kfree_skb(skb);
 591         return 0;
 592 }
 593
 594 struct ipv6_tel_txoption {
 595         struct ipv6_txoptions ops;
 596         __u8 dst_opt[8];
 597 };
 598
 599 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 600 {
 601         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 602
 603         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 604         opt->dst_opt[3] = 1;
 605         opt->dst_opt[4] = encap_limit;
 606         opt->dst_opt[5] = IPV6_TLV_PADN;
 607         opt->dst_opt[6] = 1;
 608
 609         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 610         opt->ops.opt_nflen = 8;
 611 }
 612
 613 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 614                          struct net_device *dev,
 615                          __u8 dsfield,
 616                          struct flowi6 *fl6,
 617                          int encap_limit,
 618                          __u32 *pmtu)
 619 {
 620         struct net *net = dev_net(dev);
 621         struct ip6_tnl *tunnel = netdev_priv(dev);
 622         struct net_device *tdev;    /* Device to other host */
 623         struct ipv6hdr  *ipv6h;     /* Our new IP header */
 624         unsigned int max_headroom = 0; /* The extra header space needed */
 625         int    gre_hlen;
 626         struct ipv6_tel_txoption opt;
 627         int    mtu;
 628         struct dst_entry *dst = NULL, *ndst = NULL;
 629         struct net_device_stats *stats = &tunnel->dev->stats;
 630         int err = -1;
 631         u8 proto;
 632         struct sk_buff *new_skb;
 633
 634         if (dev->type == ARPHRD_ETHER)
 635                 IPCB(skb)->flags = 0;
 636
 637         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 638                 gre_hlen = 0;
 639                 ipv6h = (struct ipv6hdr *)skb->data;
 640                 fl6->daddr = ipv6h->daddr;
 641         } else {
 642                 gre_hlen = tunnel->hlen;
 643                 fl6->daddr = tunnel->parms.raddr;
 644         }
 645
 646         if (!fl6->flowi6_mark)
 647                 dst = ip6_tnl_dst_check(tunnel);
 648
 649         if (!dst) {
 650                 ndst = ip6_route_output(net, NULL, fl6);
 651
 652                 if (ndst->error)
 653                         goto tx_err_link_failure;
 654                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
 655                 if (IS_ERR(ndst)) {
 656                         err = PTR_ERR(ndst);
 657                         ndst = NULL;
 658                         goto tx_err_link_failure;
 659                 }
 660                 dst = ndst;
 661         }
 662
 663         tdev = dst->dev;
 664
 665         if (tdev == dev) {
 666                 stats->collisions++;
 667                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 668                                      tunnel->parms.name);
 669                 goto tx_err_dst_release;
 670         }
 671
 672         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 673         if (encap_limit >= 0) {
 674                 max_headroom += 8;
 675                 mtu -= 8;
 676         }
 677         if (mtu < IPV6_MIN_MTU)
 678                 mtu = IPV6_MIN_MTU;
 679         if (skb_dst(skb))
 680                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 681         if (skb->len > mtu) {
 682                 *pmtu = mtu;
 683                 err = -EMSGSIZE;
 684                 goto tx_err_dst_release;
 685         }
 686
 687         if (tunnel->err_count > 0) {
 688                 if (time_before(jiffies,
 689                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 690                         tunnel->err_count--;
 691
 692                         dst_link_failure(skb);
 693                 } else
 694                         tunnel->err_count = 0;
 695         }
 696
 697         max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 698
 699         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 700             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 701                 new_skb = skb_realloc_headroom(skb, max_headroom);
 702                 if (max_headroom > dev->needed_headroom)
 703                         dev->needed_headroom = max_headroom;
 704                 if (!new_skb)
 705                         goto tx_err_dst_release;
 706
 707                 if (skb->sk)
 708                         skb_set_owner_w(new_skb, skb->sk);
 709                 consume_skb(skb);
 710                 skb = new_skb;
 711         }
 712
 713         skb_dst_drop(skb);
 714
 715         if (fl6->flowi6_mark) {
 716                 skb_dst_set(skb, dst);
 717                 ndst = NULL;
 718         } else {
 719                 skb_dst_set_noref(skb, dst);
 720         }
 721
 722         proto = NEXTHDR_GRE;
 723         if (encap_limit >= 0) {
 724                 init_tel_txopt(&opt, encap_limit);
 725                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 726         }
 727
 728         skb_push(skb, gre_hlen);
 729         skb_reset_network_header(skb);
 730         skb_set_transport_header(skb, sizeof(*ipv6h));
 731
 732         /*
 733          *      Push down and install the IP header.
 734          */
 735         ipv6h = ipv6_hdr(skb);
 736         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
 737         ipv6h->hop_limit = tunnel->parms.hop_limit;
 738         ipv6h->nexthdr = proto;
 739         ipv6h->saddr = fl6->saddr;
 740         ipv6h->daddr = fl6->daddr;
 741
 742         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 743         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
 744                                    htons(ETH_P_TEB) : skb->protocol;
 745
 746         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 747                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 748
 749                 if (tunnel->parms.o_flags&GRE_SEQ) {
 750                         ++tunnel->o_seqno;
 751                         *ptr = htonl(tunnel->o_seqno);
 752                         ptr--;
 753                 }
 754                 if (tunnel->parms.o_flags&GRE_KEY) {
 755                         *ptr = tunnel->parms.o_key;
 756                         ptr--;
 757                 }
 758                 if (tunnel->parms.o_flags&GRE_CSUM) {
 759                         *ptr = 0;
 760                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
 761                                 skb->len - sizeof(struct ipv6hdr));
 762                 }
 763         }
 764
 765         ip6tunnel_xmit(skb, dev);
 766         if (ndst)
 767                 ip6_tnl_dst_store(tunnel, ndst);
 768         return 0;
 769 tx_err_link_failure:
 770         stats->tx_carrier_errors++;
 771         dst_link_failure(skb);
 772 tx_err_dst_release:
 773         dst_release(ndst);
 774         return err;
 775 }
 776
 777 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 778 {
 779         struct ip6_tnl *t = netdev_priv(dev);
 780         const struct iphdr  *iph = ip_hdr(skb);
 781         int encap_limit = -1;
 782         struct flowi6 fl6;
 783         __u8 dsfield;
 784         __u32 mtu;
 785         int err;
 786
 787         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 788                 encap_limit = t->parms.encap_limit;
 789
 790         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 791         fl6.flowi6_proto = IPPROTO_GRE;
 792
 793         dsfield = ipv4_get_dsfield(iph);
 794
 795         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 796                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 797                                           & IPV6_TCLASS_MASK;
 798         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 799                 fl6.flowi6_mark = skb->mark;
 800
 801         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 802         if (err != 0) {
 803                 /* XXX: send ICMP error even if DF is not set. */
 804                 if (err == -EMSGSIZE)
 805                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 806                                   htonl(mtu));
 807                 return -1;
 808         }
 809
 810         return 0;
 811 }
 812
 813 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 814 {
 815         struct ip6_tnl *t = netdev_priv(dev);
 816         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 817         int encap_limit = -1;
 818         __u16 offset;
 819         struct flowi6 fl6;
 820         __u8 dsfield;
 821         __u32 mtu;
 822         int err;
 823
 824         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 825                 return -1;
 826
 827         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 828         if (offset > 0) {
 829                 struct ipv6_tlv_tnl_enc_lim *tel;
 830                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 831                 if (tel->encap_limit == 0) {
 832                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 833                                     ICMPV6_HDR_FIELD, offset + 2);
 834                         return -1;
 835                 }
 836                 encap_limit = tel->encap_limit - 1;
 837         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 838                 encap_limit = t->parms.encap_limit;
 839
 840         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 841         fl6.flowi6_proto = IPPROTO_GRE;
 842
 843         dsfield = ipv6_get_dsfield(ipv6h);
 844         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 845                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 846         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 847                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 848         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 849                 fl6.flowi6_mark = skb->mark;
 850
 851         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 852         if (err != 0) {
 853                 if (err == -EMSGSIZE)
 854                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 855                 return -1;
 856         }
 857
 858         return 0;
 859 }
 860
 861 /**
 862  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 863  *   @t: the outgoing tunnel device
 864  *   @hdr: IPv6 header from the incoming packet
 865  *
 866  * Description:
 867  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 868  *   doesn't match source of incoming packet.
 869  *
 870  * Return:
 871  *   1 if conflict,
 872  *   0 else
 873  **/
 874
 875 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 876         const struct ipv6hdr *hdr)
 877 {
 878         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 879 }
 880
 881 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 882 {
 883         struct ip6_tnl *t = netdev_priv(dev);
 884         int encap_limit = -1;
 885         struct flowi6 fl6;
 886         __u32 mtu;
 887         int err;
 888
 889         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 890                 encap_limit = t->parms.encap_limit;
 891
 892         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 893         fl6.flowi6_proto = skb->protocol;
 894
 895         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 896
 897         return err;
 898 }
 899
 900 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 901         struct net_device *dev)
 902 {
 903         struct ip6_tnl *t = netdev_priv(dev);
 904         struct net_device_stats *stats = &t->dev->stats;
 905         int ret;
 906
 907         if (!ip6_tnl_xmit_ctl(t))
 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         stats->tx_errors++;
 929         stats->tx_dropped++;
 930         kfree_skb(skb);
 931         return NETDEV_TX_OK;
 932 }
 933
 934 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 935 {
 936         struct net_device *dev = t->dev;
 937         struct __ip6_tnl_parm *p = &t->parms;
 938         struct flowi6 *fl6 = &t->fl.u.ip6;
 939         int addend = sizeof(struct ipv6hdr) + 4;
 940
 941         if (dev->type != ARPHRD_ETHER) {
 942                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 943                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 944         }
 945
 946         /* Set up flowi template */
 947         fl6->saddr = p->laddr;
 948         fl6->daddr = p->raddr;
 949         fl6->flowi6_oif = p->link;
 950         fl6->flowlabel = 0;
 951
 952         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 953                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 954         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 955                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 956
 957         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 958         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 959
 960         if (p->flags&IP6_TNL_F_CAP_XMIT &&
 961                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
 962                 dev->flags |= IFF_POINTOPOINT;
 963         else
 964                 dev->flags &= ~IFF_POINTOPOINT;
 965
 966         /* Precalculate GRE options length */
 967         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
 968                 if (t->parms.o_flags&GRE_CSUM)
 969                         addend += 4;
 970                 if (t->parms.o_flags&GRE_KEY)
 971                         addend += 4;
 972                 if (t->parms.o_flags&GRE_SEQ)
 973                         addend += 4;
 974         }
 975
 976         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 977                 int strict = (ipv6_addr_type(&p->raddr) &
 978                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 979
 980                 struct rt6_info *rt = rt6_lookup(dev_net(dev),
 981                                                  &p->raddr, &p->laddr,
 982                                                  p->link, strict);
 983
 984                 if (rt == NULL)
 985                         return;
 986
 987                 if (rt->dst.dev) {
 988                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
 989
 990                         if (set_mtu) {
 991                                 dev->mtu = rt->dst.dev->mtu - addend;
 992                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 993                                         dev->mtu -= 8;
 994
 995                                 if (dev->mtu < IPV6_MIN_MTU)
 996                                         dev->mtu = IPV6_MIN_MTU;
 997                         }
 998                 }
 999                 ip6_rt_put(rt);
1000         }
1001
1002         t->hlen = addend;
1003 }
1004
1005 static int ip6gre_tnl_change(struct ip6_tnl *t,
1006         const struct __ip6_tnl_parm *p, int set_mtu)
1007 {
1008         t->parms.laddr = p->laddr;
1009         t->parms.raddr = p->raddr;
1010         t->parms.flags = p->flags;
1011         t->parms.hop_limit = p->hop_limit;
1012         t->parms.encap_limit = p->encap_limit;
1013         t->parms.flowinfo = p->flowinfo;
1014         t->parms.link = p->link;
1015         t->parms.proto = p->proto;
1016         t->parms.i_key = p->i_key;
1017         t->parms.o_key = p->o_key;
1018         t->parms.i_flags = p->i_flags;
1019         t->parms.o_flags = p->o_flags;
1020         ip6_tnl_dst_reset(t);
1021         ip6gre_tnl_link_config(t, set_mtu);
1022         return 0;
1023 }
1024
1025 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1026         const struct ip6_tnl_parm2 *u)
1027 {
1028         p->laddr = u->laddr;
1029         p->raddr = u->raddr;
1030         p->flags = u->flags;
1031         p->hop_limit = u->hop_limit;
1032         p->encap_limit = u->encap_limit;
1033         p->flowinfo = u->flowinfo;
1034         p->link = u->link;
1035         p->i_key = u->i_key;
1036         p->o_key = u->o_key;
1037         p->i_flags = u->i_flags;
1038         p->o_flags = u->o_flags;
1039         memcpy(p->name, u->name, sizeof(u->name));
1040 }
1041
1042 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1043         const struct __ip6_tnl_parm *p)
1044 {
1045         u->proto = IPPROTO_GRE;
1046         u->laddr = p->laddr;
1047         u->raddr = p->raddr;
1048         u->flags = p->flags;
1049         u->hop_limit = p->hop_limit;
1050         u->encap_limit = p->encap_limit;
1051         u->flowinfo = p->flowinfo;
1052         u->link = p->link;
1053         u->i_key = p->i_key;
1054         u->o_key = p->o_key;
1055         u->i_flags = p->i_flags;
1056         u->o_flags = p->o_flags;
1057         memcpy(u->name, p->name, sizeof(u->name));
1058 }
1059
1060 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1061         struct ifreq *ifr, int cmd)
1062 {
1063         int err = 0;
1064         struct ip6_tnl_parm2 p;
1065         struct __ip6_tnl_parm p1;
1066         struct ip6_tnl *t;
1067         struct net *net = dev_net(dev);
1068         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1069
1070         switch (cmd) {
1071         case SIOCGETTUNNEL:
1072                 t = NULL;
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                 }
1081                 if (t == NULL)
1082                         t = netdev_priv(dev);
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         struct ip6_tnl *tunnel = netdev_priv(dev);
1173         if (new_mtu < 68 ||
1174             new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1175                 return -EINVAL;
1176         dev->mtu = new_mtu;
1177         return 0;
1178 }
1179
1180 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1181                         unsigned short type,
1182                         const void *daddr, const void *saddr, unsigned int len)
1183 {
1184         struct ip6_tnl *t = netdev_priv(dev);
1185         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1186         __be16 *p = (__be16 *)(ipv6h+1);
1187
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[0]            = t->parms.o_flags;
1195         p[1]            = htons(type);
1196
1197         /*
1198          *      Set the source hardware address.
1199          */
1200
1201         if (saddr)
1202                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1203         if (daddr)
1204                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1205         if (!ipv6_addr_any(&ipv6h->daddr))
1206                 return t->hlen;
1207
1208         return -t->hlen;
1209 }
1210
1211 static const struct header_ops ip6gre_header_ops = {
1212         .create = ip6gre_header,
1213 };
1214
1215 static const struct net_device_ops ip6gre_netdev_ops = {
1216         .ndo_init               = ip6gre_tunnel_init,
1217         .ndo_uninit             = ip6gre_tunnel_uninit,
1218         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1219         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1220         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1221         .ndo_get_stats64        = ip_tunnel_get_stats64,
1222 };
1223
1224 static void ip6gre_dev_free(struct net_device *dev)
1225 {
1226         free_percpu(dev->tstats);
1227         free_netdev(dev);
1228 }
1229
1230 static void ip6gre_tunnel_setup(struct net_device *dev)
1231 {
1232         struct ip6_tnl *t;
1233
1234         dev->netdev_ops = &ip6gre_netdev_ops;
1235         dev->destructor = ip6gre_dev_free;
1236
1237         dev->type = ARPHRD_IP6GRE;
1238         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1239         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1240         t = netdev_priv(dev);
1241         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1242                 dev->mtu -= 8;
1243         dev->flags |= IFF_NOARP;
1244         dev->iflink = 0;
1245         dev->addr_len = sizeof(struct in6_addr);
1246         dev->features |= NETIF_F_NETNS_LOCAL;
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
1254         tunnel = netdev_priv(dev);
1255
1256         tunnel->dev = dev;
1257         strcpy(tunnel->parms.name, dev->name);
1258
1259         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1260         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1261
1262         if (ipv6_addr_any(&tunnel->parms.raddr))
1263                 dev->header_ops = &ip6gre_header_ops;
1264
1265         dev->tstats = alloc_percpu(struct pcpu_tstats);
1266         if (!dev->tstats)
1267                 return -ENOMEM;
1268
1269         dev->iflink = tunnel->parms.link;
1270
1271         return 0;
1272 }
1273
1274 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1275 {
1276         struct ip6_tnl *tunnel = netdev_priv(dev);
1277
1278         tunnel->dev = dev;
1279         strcpy(tunnel->parms.name, dev->name);
1280
1281         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1282
1283         dev_hold(dev);
1284 }
1285
1286 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1287         .handler     = ip6gre_rcv,
1288         .err_handler = ip6gre_err,
1289         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1290 };
1291
1292 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1293         struct list_head *head)
1294 {
1295         int prio;
1296
1297         for (prio = 0; prio < 4; prio++) {
1298                 int h;
1299                 for (h = 0; h < HASH_SIZE; h++) {
1300                         struct ip6_tnl *t;
1301
1302                         t = rtnl_dereference(ign->tunnels[prio][h]);
1303
1304                         while (t != NULL) {
1305                                 unregister_netdevice_queue(t->dev, head);
1306                                 t = rtnl_dereference(t->next);
1307                         }
1308                 }
1309         }
1310 }
1311
1312 static int __net_init ip6gre_init_net(struct net *net)
1313 {
1314         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1315         int err;
1316
1317         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1318                                            ip6gre_tunnel_setup);
1319         if (!ign->fb_tunnel_dev) {
1320                 err = -ENOMEM;
1321                 goto err_alloc_dev;
1322         }
1323         dev_net_set(ign->fb_tunnel_dev, net);
1324
1325         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1326         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1327
1328         err = register_netdev(ign->fb_tunnel_dev);
1329         if (err)
1330                 goto err_reg_dev;
1331
1332         rcu_assign_pointer(ign->tunnels_wc[0],
1333                            netdev_priv(ign->fb_tunnel_dev));
1334         return 0;
1335
1336 err_reg_dev:
1337         ip6gre_dev_free(ign->fb_tunnel_dev);
1338 err_alloc_dev:
1339         return err;
1340 }
1341
1342 static void __net_exit ip6gre_exit_net(struct net *net)
1343 {
1344         struct ip6gre_net *ign;
1345         LIST_HEAD(list);
1346
1347         ign = net_generic(net, ip6gre_net_id);
1348         rtnl_lock();
1349         ip6gre_destroy_tunnels(ign, &list);
1350         unregister_netdevice_many(&list);
1351         rtnl_unlock();
1352 }
1353
1354 static struct pernet_operations ip6gre_net_ops = {
1355         .init = ip6gre_init_net,
1356         .exit = ip6gre_exit_net,
1357         .id   = &ip6gre_net_id,
1358         .size = sizeof(struct ip6gre_net),
1359 };
1360
1361 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1362 {
1363         __be16 flags;
1364
1365         if (!data)
1366                 return 0;
1367
1368         flags = 0;
1369         if (data[IFLA_GRE_IFLAGS])
1370                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1371         if (data[IFLA_GRE_OFLAGS])
1372                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1373         if (flags & (GRE_VERSION|GRE_ROUTING))
1374                 return -EINVAL;
1375
1376         return 0;
1377 }
1378
1379 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1380 {
1381         struct in6_addr daddr;
1382
1383         if (tb[IFLA_ADDRESS]) {
1384                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1385                         return -EINVAL;
1386                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1387                         return -EADDRNOTAVAIL;
1388         }
1389
1390         if (!data)
1391                 goto out;
1392
1393         if (data[IFLA_GRE_REMOTE]) {
1394                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1395                 if (ipv6_addr_any(&daddr))
1396                         return -EINVAL;
1397         }
1398
1399 out:
1400         return ip6gre_tunnel_validate(tb, data);
1401 }
1402
1403
1404 static void ip6gre_netlink_parms(struct nlattr *data[],
1405                                 struct __ip6_tnl_parm *parms)
1406 {
1407         memset(parms, 0, sizeof(*parms));
1408
1409         if (!data)
1410                 return;
1411
1412         if (data[IFLA_GRE_LINK])
1413                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1414
1415         if (data[IFLA_GRE_IFLAGS])
1416                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1417
1418         if (data[IFLA_GRE_OFLAGS])
1419                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1420
1421         if (data[IFLA_GRE_IKEY])
1422                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1423
1424         if (data[IFLA_GRE_OKEY])
1425                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1426
1427         if (data[IFLA_GRE_LOCAL])
1428                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1429
1430         if (data[IFLA_GRE_REMOTE])
1431                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1432
1433         if (data[IFLA_GRE_TTL])
1434                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1435
1436         if (data[IFLA_GRE_ENCAP_LIMIT])
1437                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1438
1439         if (data[IFLA_GRE_FLOWINFO])
1440                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1441
1442         if (data[IFLA_GRE_FLAGS])
1443                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1444 }
1445
1446 static int ip6gre_tap_init(struct net_device *dev)
1447 {
1448         struct ip6_tnl *tunnel;
1449
1450         tunnel = netdev_priv(dev);
1451
1452         tunnel->dev = dev;
1453         strcpy(tunnel->parms.name, dev->name);
1454
1455         ip6gre_tnl_link_config(tunnel, 1);
1456
1457         dev->tstats = alloc_percpu(struct pcpu_tstats);
1458         if (!dev->tstats)
1459                 return -ENOMEM;
1460
1461         dev->iflink = tunnel->parms.link;
1462
1463         return 0;
1464 }
1465
1466 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1467         .ndo_init = ip6gre_tap_init,
1468         .ndo_uninit = ip6gre_tunnel_uninit,
1469         .ndo_start_xmit = ip6gre_tunnel_xmit,
1470         .ndo_set_mac_address = eth_mac_addr,
1471         .ndo_validate_addr = eth_validate_addr,
1472         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1473         .ndo_get_stats64 = ip_tunnel_get_stats64,
1474 };
1475
1476 static void ip6gre_tap_setup(struct net_device *dev)
1477 {
1478
1479         ether_setup(dev);
1480
1481         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1482         dev->destructor = ip6gre_dev_free;
1483
1484         dev->iflink = 0;
1485         dev->features |= NETIF_F_NETNS_LOCAL;
1486 }
1487
1488 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1489         struct nlattr *tb[], struct nlattr *data[])
1490 {
1491         struct ip6_tnl *nt;
1492         struct net *net = dev_net(dev);
1493         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1494         int err;
1495
1496         nt = netdev_priv(dev);
1497         ip6gre_netlink_parms(data, &nt->parms);
1498
1499         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1500                 return -EEXIST;
1501
1502         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1503                 eth_hw_addr_random(dev);
1504
1505         nt->dev = dev;
1506         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1507
1508         /* Can use a lockless transmit, unless we generate output sequences */
1509         if (!(nt->parms.o_flags & GRE_SEQ))
1510                 dev->features |= NETIF_F_LLTX;
1511
1512         err = register_netdevice(dev);
1513         if (err)
1514                 goto out;
1515
1516         dev_hold(dev);
1517         ip6gre_tunnel_link(ign, nt);
1518
1519 out:
1520         return err;
1521 }
1522
1523 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1524                             struct nlattr *data[])
1525 {
1526         struct ip6_tnl *t, *nt;
1527         struct net *net = dev_net(dev);
1528         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1529         struct __ip6_tnl_parm p;
1530
1531         if (dev == ign->fb_tunnel_dev)
1532                 return -EINVAL;
1533
1534         nt = netdev_priv(dev);
1535         ip6gre_netlink_parms(data, &p);
1536
1537         t = ip6gre_tunnel_locate(net, &p, 0);
1538
1539         if (t) {
1540                 if (t->dev != dev)
1541                         return -EEXIST;
1542         } else {
1543                 t = nt;
1544
1545                 ip6gre_tunnel_unlink(ign, t);
1546                 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1547                 ip6gre_tunnel_link(ign, t);
1548                 netdev_state_change(dev);
1549         }
1550
1551         return 0;
1552 }
1553
1554 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1555 {
1556         struct net *net = dev_net(dev);
1557         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1558
1559         if (dev != ign->fb_tunnel_dev)
1560                 unregister_netdevice_queue(dev, head);
1561 }
1562
1563 static size_t ip6gre_get_size(const struct net_device *dev)
1564 {
1565         return
1566                 /* IFLA_GRE_LINK */
1567                 nla_total_size(4) +
1568                 /* IFLA_GRE_IFLAGS */
1569                 nla_total_size(2) +
1570                 /* IFLA_GRE_OFLAGS */
1571                 nla_total_size(2) +
1572                 /* IFLA_GRE_IKEY */
1573                 nla_total_size(4) +
1574                 /* IFLA_GRE_OKEY */
1575                 nla_total_size(4) +
1576                 /* IFLA_GRE_LOCAL */
1577                 nla_total_size(sizeof(struct in6_addr)) +
1578                 /* IFLA_GRE_REMOTE */
1579                 nla_total_size(sizeof(struct in6_addr)) +
1580                 /* IFLA_GRE_TTL */
1581                 nla_total_size(1) +
1582                 /* IFLA_GRE_TOS */
1583                 nla_total_size(1) +
1584                 /* IFLA_GRE_ENCAP_LIMIT */
1585                 nla_total_size(1) +
1586                 /* IFLA_GRE_FLOWINFO */
1587                 nla_total_size(4) +
1588                 /* IFLA_GRE_FLAGS */
1589                 nla_total_size(4) +
1590                 0;
1591 }
1592
1593 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1594 {
1595         struct ip6_tnl *t = netdev_priv(dev);
1596         struct __ip6_tnl_parm *p = &t->parms;
1597
1598         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1599             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1600             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1601             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1602             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1603             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1604             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1605             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1606             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1607             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1608             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1609             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1610                 goto nla_put_failure;
1611         return 0;
1612
1613 nla_put_failure:
1614         return -EMSGSIZE;
1615 }
1616
1617 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1618         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1619         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1620         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1621         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1622         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1623         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1624         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1625         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1626         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1627         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1628         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1629 };
1630
1631 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1632         .kind           = "ip6gre",
1633         .maxtype        = IFLA_GRE_MAX,
1634         .policy         = ip6gre_policy,
1635         .priv_size      = sizeof(struct ip6_tnl),
1636         .setup          = ip6gre_tunnel_setup,
1637         .validate       = ip6gre_tunnel_validate,
1638         .newlink        = ip6gre_newlink,
1639         .changelink     = ip6gre_changelink,
1640         .dellink        = ip6gre_dellink,
1641         .get_size       = ip6gre_get_size,
1642         .fill_info      = ip6gre_fill_info,
1643 };
1644
1645 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1646         .kind           = "ip6gretap",
1647         .maxtype        = IFLA_GRE_MAX,
1648         .policy         = ip6gre_policy,
1649         .priv_size      = sizeof(struct ip6_tnl),
1650         .setup          = ip6gre_tap_setup,
1651         .validate       = ip6gre_tap_validate,
1652         .newlink        = ip6gre_newlink,
1653         .changelink     = ip6gre_changelink,
1654         .get_size       = ip6gre_get_size,
1655         .fill_info      = ip6gre_fill_info,
1656 };
1657
1658 /*
1659  *      And now the modules code and kernel interface.
1660  */
1661
1662 static int __init ip6gre_init(void)
1663 {
1664         int err;
1665
1666         pr_info("GRE over IPv6 tunneling driver\n");
1667
1668         err = register_pernet_device(&ip6gre_net_ops);
1669         if (err < 0)
1670                 return err;
1671
1672         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1673         if (err < 0) {
1674                 pr_info("%s: can't add protocol\n", __func__);
1675                 goto add_proto_failed;
1676         }
1677
1678         err = rtnl_link_register(&ip6gre_link_ops);
1679         if (err < 0)
1680                 goto rtnl_link_failed;
1681
1682         err = rtnl_link_register(&ip6gre_tap_ops);
1683         if (err < 0)
1684                 goto tap_ops_failed;
1685
1686 out:
1687         return err;
1688
1689 tap_ops_failed:
1690         rtnl_link_unregister(&ip6gre_link_ops);
1691 rtnl_link_failed:
1692         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1693 add_proto_failed:
1694         unregister_pernet_device(&ip6gre_net_ops);
1695         goto out;
1696 }
1697
1698 static void __exit ip6gre_fini(void)
1699 {
1700         rtnl_link_unregister(&ip6gre_tap_ops);
1701         rtnl_link_unregister(&ip6gre_link_ops);
1702         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1703         unregister_pernet_device(&ip6gre_net_ops);
1704 }
1705
1706 module_init(ip6gre_init);
1707 module_exit(ip6gre_fini);
1708 MODULE_LICENSE("GPL");
1709 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1710 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1711 MODULE_ALIAS_RTNL_LINK("ip6gre");
1712 MODULE_ALIAS_NETDEV("ip6gre0");