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/icmp.h>
  42 #include <net/protocol.h>
  43 #include <net/addrconf.h>
  44 #include <net/arp.h>
  45 #include <net/checksum.h>
  46 #include <net/dsfield.h>
  47 #include <net/inet_ecn.h>
  48 #include <net/xfrm.h>
  49 #include <net/net_namespace.h>
  50 #include <net/netns/generic.h>
  51 #include <net/rtnetlink.h>
  52
  53 #include <net/ipv6.h>
  54 #include <net/ip6_fib.h>
  55 #include <net/ip6_route.h>
  56 #include <net/ip6_tunnel.h>
  57
  58
  59 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
  60 #define IPV6_TCLASS_SHIFT 20
  61
  62 #define HASH_SIZE_SHIFT  5
  63 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  64
  65 static int ip6gre_net_id __read_mostly;
  66 struct ip6gre_net {
  67         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
  68
  69         struct net_device *fb_tunnel_dev;
  70 };
  71
  72 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  73 static int ip6gre_tunnel_init(struct net_device *dev);
  74 static void ip6gre_tunnel_setup(struct net_device *dev);
  75 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  76 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  77
  78 /* Tunnel hash table */
  79
  80 /*
  81    4 hash tables:
  82
  83    3: (remote,local)
  84    2: (remote,*)
  85    1: (*,local)
  86    0: (*,*)
  87
  88    We require exact key match i.e. if a key is present in packet
  89    it will match only tunnel with the same key; if it is not present,
  90    it will match only keyless tunnel.
  91
  92    All keysless packets, if not matched configured keyless tunnels
  93    will match fallback tunnel.
  94  */
  95
  96 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
  97 static u32 HASH_ADDR(const struct in6_addr *addr)
  98 {
  99         u32 hash = ipv6_addr_hash(addr);
 100
 101         return hash_32(hash, HASH_SIZE_SHIFT);
 102 }
 103
 104 #define tunnels_r_l     tunnels[3]
 105 #define tunnels_r       tunnels[2]
 106 #define tunnels_l       tunnels[1]
 107 #define tunnels_wc      tunnels[0]
 108 /*
 109  * Locking : hash tables are protected by RCU and RTNL
 110  */
 111
 112 #define for_each_ip_tunnel_rcu(start) \
 113         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 114
 115 /* often modified stats are per cpu, other are shared (netdev->stats) */
 116 struct pcpu_tstats {
 117         u64     rx_packets;
 118         u64     rx_bytes;
 119         u64     tx_packets;
 120         u64     tx_bytes;
 121         struct u64_stats_sync   syncp;
 122 };
 123
 124 static struct rtnl_link_stats64 *ip6gre_get_stats64(struct net_device *dev,
 125                 struct rtnl_link_stats64 *tot)
 126 {
 127         int i;
 128
 129         for_each_possible_cpu(i) {
 130                 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
 131                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
 132                 unsigned int start;
 133
 134                 do {
 135                         start = u64_stats_fetch_begin_bh(&tstats->syncp);
 136                         rx_packets = tstats->rx_packets;
 137                         tx_packets = tstats->tx_packets;
 138                         rx_bytes = tstats->rx_bytes;
 139                         tx_bytes = tstats->tx_bytes;
 140                 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
 141
 142                 tot->rx_packets += rx_packets;
 143                 tot->tx_packets += tx_packets;
 144                 tot->rx_bytes   += rx_bytes;
 145                 tot->tx_bytes   += tx_bytes;
 146         }
 147
 148         tot->multicast = dev->stats.multicast;
 149         tot->rx_crc_errors = dev->stats.rx_crc_errors;
 150         tot->rx_fifo_errors = dev->stats.rx_fifo_errors;
 151         tot->rx_length_errors = dev->stats.rx_length_errors;
 152         tot->rx_errors = dev->stats.rx_errors;
 153         tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
 154         tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
 155         tot->tx_dropped = dev->stats.tx_dropped;
 156         tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
 157         tot->tx_errors = dev->stats.tx_errors;
 158
 159         return tot;
 160 }
 161
 162 /* Given src, dst and key, find appropriate for input tunnel. */
 163
 164 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 165                 const struct in6_addr *remote, const struct in6_addr *local,
 166                 __be32 key, __be16 gre_proto)
 167 {
 168         struct net *net = dev_net(dev);
 169         int link = dev->ifindex;
 170         unsigned int h0 = HASH_ADDR(remote);
 171         unsigned int h1 = HASH_KEY(key);
 172         struct ip6_tnl *t, *cand = NULL;
 173         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 174         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
 175                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 176         int score, cand_score = 4;
 177
 178         for_each_ip_tunnel_rcu(ign->tunnels_r_l[h0 ^ h1]) {
 179                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 180                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 181                     key != t->parms.i_key ||
 182                     !(t->dev->flags & IFF_UP))
 183                         continue;
 184
 185                 if (t->dev->type != ARPHRD_IP6GRE &&
 186                     t->dev->type != dev_type)
 187                         continue;
 188
 189                 score = 0;
 190                 if (t->parms.link != link)
 191                         score |= 1;
 192                 if (t->dev->type != dev_type)
 193                         score |= 2;
 194                 if (score == 0)
 195                         return t;
 196
 197                 if (score < cand_score) {
 198                         cand = t;
 199                         cand_score = score;
 200                 }
 201         }
 202
 203         for_each_ip_tunnel_rcu(ign->tunnels_r[h0 ^ h1]) {
 204                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 205                     key != t->parms.i_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         for_each_ip_tunnel_rcu(ign->tunnels_l[h1]) {
 228                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 229                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 230                                  !ipv6_addr_is_multicast(local))) ||
 231                     key != t->parms.i_key ||
 232                     !(t->dev->flags & IFF_UP))
 233                         continue;
 234
 235                 if (t->dev->type != ARPHRD_IP6GRE &&
 236                     t->dev->type != dev_type)
 237                         continue;
 238
 239                 score = 0;
 240                 if (t->parms.link != link)
 241                         score |= 1;
 242                 if (t->dev->type != dev_type)
 243                         score |= 2;
 244                 if (score == 0)
 245                         return t;
 246
 247                 if (score < cand_score) {
 248                         cand = t;
 249                         cand_score = score;
 250                 }
 251         }
 252
 253         for_each_ip_tunnel_rcu(ign->tunnels_wc[h1]) {
 254                 if (t->parms.i_key != key ||
 255                     !(t->dev->flags & IFF_UP))
 256                         continue;
 257
 258                 if (t->dev->type != ARPHRD_IP6GRE &&
 259                     t->dev->type != dev_type)
 260                         continue;
 261
 262                 score = 0;
 263                 if (t->parms.link != link)
 264                         score |= 1;
 265                 if (t->dev->type != dev_type)
 266                         score |= 2;
 267                 if (score == 0)
 268                         return t;
 269
 270                 if (score < cand_score) {
 271                         cand = t;
 272                         cand_score = score;
 273                 }
 274         }
 275
 276         if (cand != NULL)
 277                 return cand;
 278
 279         dev = ign->fb_tunnel_dev;
 280         if (dev->flags & IFF_UP)
 281                 return netdev_priv(dev);
 282
 283         return NULL;
 284 }
 285
 286 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 287                 const struct __ip6_tnl_parm *p)
 288 {
 289         const struct in6_addr *remote = &p->raddr;
 290         const struct in6_addr *local = &p->laddr;
 291         unsigned int h = HASH_KEY(p->i_key);
 292         int prio = 0;
 293
 294         if (!ipv6_addr_any(local))
 295                 prio |= 1;
 296         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 297                 prio |= 2;
 298                 h ^= HASH_ADDR(remote);
 299         }
 300
 301         return &ign->tunnels[prio][h];
 302 }
 303
 304 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 305                 const struct ip6_tnl *t)
 306 {
 307         return __ip6gre_bucket(ign, &t->parms);
 308 }
 309
 310 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 311 {
 312         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 313
 314         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 315         rcu_assign_pointer(*tp, t);
 316 }
 317
 318 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 319 {
 320         struct ip6_tnl __rcu **tp;
 321         struct ip6_tnl *iter;
 322
 323         for (tp = ip6gre_bucket(ign, t);
 324              (iter = rtnl_dereference(*tp)) != NULL;
 325              tp = &iter->next) {
 326                 if (t == iter) {
 327                         rcu_assign_pointer(*tp, t->next);
 328                         break;
 329                 }
 330         }
 331 }
 332
 333 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 334                                            const struct __ip6_tnl_parm *parms,
 335                                            int type)
 336 {
 337         const struct in6_addr *remote = &parms->raddr;
 338         const struct in6_addr *local = &parms->laddr;
 339         __be32 key = parms->i_key;
 340         int link = parms->link;
 341         struct ip6_tnl *t;
 342         struct ip6_tnl __rcu **tp;
 343         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 344
 345         for (tp = __ip6gre_bucket(ign, parms);
 346              (t = rtnl_dereference(*tp)) != NULL;
 347              tp = &t->next)
 348                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 349                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 350                     key == t->parms.i_key &&
 351                     link == t->parms.link &&
 352                     type == t->dev->type)
 353                         break;
 354
 355         return t;
 356 }
 357
 358 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 359                 const struct __ip6_tnl_parm *parms, int create)
 360 {
 361         struct ip6_tnl *t, *nt;
 362         struct net_device *dev;
 363         char name[IFNAMSIZ];
 364         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 365
 366         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 367         if (t || !create)
 368                 return t;
 369
 370         if (parms->name[0])
 371                 strlcpy(name, parms->name, IFNAMSIZ);
 372         else
 373                 strcpy(name, "ip6gre%d");
 374
 375         dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
 376         if (!dev)
 377                 return NULL;
 378
 379         dev_net_set(dev, net);
 380
 381         nt = netdev_priv(dev);
 382         nt->parms = *parms;
 383         dev->rtnl_link_ops = &ip6gre_link_ops;
 384
 385         nt->dev = dev;
 386         ip6gre_tnl_link_config(nt, 1);
 387
 388         if (register_netdevice(dev) < 0)
 389                 goto failed_free;
 390
 391         /* Can use a lockless transmit, unless we generate output sequences */
 392         if (!(nt->parms.o_flags & GRE_SEQ))
 393                 dev->features |= NETIF_F_LLTX;
 394
 395         dev_hold(dev);
 396         ip6gre_tunnel_link(ign, nt);
 397         return nt;
 398
 399 failed_free:
 400         free_netdev(dev);
 401         return NULL;
 402 }
 403
 404 static void ip6gre_tunnel_uninit(struct net_device *dev)
 405 {
 406         struct net *net = dev_net(dev);
 407         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 408
 409         ip6gre_tunnel_unlink(ign, netdev_priv(dev));
 410         dev_put(dev);
 411 }
 412
 413
 414 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 415                 u8 type, u8 code, int offset, __be32 info)
 416 {
 417         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 418         __be16 *p = (__be16 *)(skb->data + offset);
 419         int grehlen = offset + 4;
 420         struct ip6_tnl *t;
 421         __be16 flags;
 422
 423         flags = p[0];
 424         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
 425                 if (flags&(GRE_VERSION|GRE_ROUTING))
 426                         return;
 427                 if (flags&GRE_KEY) {
 428                         grehlen += 4;
 429                         if (flags&GRE_CSUM)
 430                                 grehlen += 4;
 431                 }
 432         }
 433
 434         /* If only 8 bytes returned, keyed message will be dropped here */
 435         if (!pskb_may_pull(skb, grehlen))
 436                 return;
 437         ipv6h = (const struct ipv6hdr *)skb->data;
 438         p = (__be16 *)(skb->data + offset);
 439
 440         rcu_read_lock();
 441
 442         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 443                                 flags & GRE_KEY ?
 444                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
 445                                 p[1]);
 446         if (t == NULL)
 447                 goto out;
 448
 449         switch (type) {
 450                 __u32 teli;
 451                 struct ipv6_tlv_tnl_enc_lim *tel;
 452                 __u32 mtu;
 453         case ICMPV6_DEST_UNREACH:
 454                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
 455                                      t->parms.name);
 456                 break;
 457         case ICMPV6_TIME_EXCEED:
 458                 if (code == ICMPV6_EXC_HOPLIMIT) {
 459                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 460                                              t->parms.name);
 461                 }
 462                 break;
 463         case ICMPV6_PARAMPROB:
 464                 teli = 0;
 465                 if (code == ICMPV6_HDR_FIELD)
 466                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 467
 468                 if (teli && teli == info - 2) {
 469                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 470                         if (tel->encap_limit == 0) {
 471                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 472                                                      t->parms.name);
 473                         }
 474                 } else {
 475                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 476                                              t->parms.name);
 477                 }
 478                 break;
 479         case ICMPV6_PKT_TOOBIG:
 480                 mtu = info - offset;
 481                 if (mtu < IPV6_MIN_MTU)
 482                         mtu = IPV6_MIN_MTU;
 483                 t->dev->mtu = mtu;
 484                 break;
 485         }
 486
 487         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 488                 t->err_count++;
 489         else
 490                 t->err_count = 1;
 491         t->err_time = jiffies;
 492 out:
 493         rcu_read_unlock();
 494 }
 495
 496 static inline void ip6gre_ecn_decapsulate_ipv4(const struct ip6_tnl *t,
 497                 const struct ipv6hdr *ipv6h, struct sk_buff *skb)
 498 {
 499         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 500
 501         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 502                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 503
 504         if (INET_ECN_is_ce(dsfield))
 505                 IP_ECN_set_ce(ip_hdr(skb));
 506 }
 507
 508 static inline void ip6gre_ecn_decapsulate_ipv6(const struct ip6_tnl *t,
 509                 const struct ipv6hdr *ipv6h, struct sk_buff *skb)
 510 {
 511         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 512                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 513
 514         if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
 515                 IP6_ECN_set_ce(ipv6_hdr(skb));
 516 }
 517
 518 static int ip6gre_rcv(struct sk_buff *skb)
 519 {
 520         const struct ipv6hdr *ipv6h;
 521         u8     *h;
 522         __be16    flags;
 523         __sum16   csum = 0;
 524         __be32 key = 0;
 525         u32    seqno = 0;
 526         struct ip6_tnl *tunnel;
 527         int    offset = 4;
 528         __be16 gre_proto;
 529
 530         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
 531                 goto drop_nolock;
 532
 533         ipv6h = ipv6_hdr(skb);
 534         h = skb->data;
 535         flags = *(__be16 *)h;
 536
 537         if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
 538                 /* - Version must be 0.
 539                    - We do not support routing headers.
 540                  */
 541                 if (flags&(GRE_VERSION|GRE_ROUTING))
 542                         goto drop_nolock;
 543
 544                 if (flags&GRE_CSUM) {
 545                         switch (skb->ip_summed) {
 546                         case CHECKSUM_COMPLETE:
 547                                 csum = csum_fold(skb->csum);
 548                                 if (!csum)
 549                                         break;
 550                                 /* fall through */
 551                         case CHECKSUM_NONE:
 552                                 skb->csum = 0;
 553                                 csum = __skb_checksum_complete(skb);
 554                                 skb->ip_summed = CHECKSUM_COMPLETE;
 555                         }
 556                         offset += 4;
 557                 }
 558                 if (flags&GRE_KEY) {
 559                         key = *(__be32 *)(h + offset);
 560                         offset += 4;
 561                 }
 562                 if (flags&GRE_SEQ) {
 563                         seqno = ntohl(*(__be32 *)(h + offset));
 564                         offset += 4;
 565                 }
 566         }
 567
 568         gre_proto = *(__be16 *)(h + 2);
 569
 570         rcu_read_lock();
 571         tunnel = ip6gre_tunnel_lookup(skb->dev,
 572                                           &ipv6h->saddr, &ipv6h->daddr, key,
 573                                           gre_proto);
 574         if (tunnel) {
 575                 struct pcpu_tstats *tstats;
 576
 577                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 578                         goto drop;
 579
 580                 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
 581                         tunnel->dev->stats.rx_dropped++;
 582                         goto drop;
 583                 }
 584
 585                 secpath_reset(skb);
 586
 587                 skb->protocol = gre_proto;
 588                 /* WCCP version 1 and 2 protocol decoding.
 589                  * - Change protocol to IP
 590                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
 591                  */
 592                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
 593                         skb->protocol = htons(ETH_P_IP);
 594                         if ((*(h + offset) & 0xF0) != 0x40)
 595                                 offset += 4;
 596                 }
 597
 598                 skb->mac_header = skb->network_header;
 599                 __pskb_pull(skb, offset);
 600                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
 601                 skb->pkt_type = PACKET_HOST;
 602
 603                 if (((flags&GRE_CSUM) && csum) ||
 604                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
 605                         tunnel->dev->stats.rx_crc_errors++;
 606                         tunnel->dev->stats.rx_errors++;
 607                         goto drop;
 608                 }
 609                 if (tunnel->parms.i_flags&GRE_SEQ) {
 610                         if (!(flags&GRE_SEQ) ||
 611                             (tunnel->i_seqno &&
 612                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
 613                                 tunnel->dev->stats.rx_fifo_errors++;
 614                                 tunnel->dev->stats.rx_errors++;
 615                                 goto drop;
 616                         }
 617                         tunnel->i_seqno = seqno + 1;
 618                 }
 619
 620                 /* Warning: All skb pointers will be invalidated! */
 621                 if (tunnel->dev->type == ARPHRD_ETHER) {
 622                         if (!pskb_may_pull(skb, ETH_HLEN)) {
 623                                 tunnel->dev->stats.rx_length_errors++;
 624                                 tunnel->dev->stats.rx_errors++;
 625                                 goto drop;
 626                         }
 627
 628                         ipv6h = ipv6_hdr(skb);
 629                         skb->protocol = eth_type_trans(skb, tunnel->dev);
 630                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 631                 }
 632
 633                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 634                 u64_stats_update_begin(&tstats->syncp);
 635                 tstats->rx_packets++;
 636                 tstats->rx_bytes += skb->len;
 637                 u64_stats_update_end(&tstats->syncp);
 638
 639                 __skb_tunnel_rx(skb, tunnel->dev);
 640
 641                 skb_reset_network_header(skb);
 642                 if (skb->protocol == htons(ETH_P_IP))
 643                         ip6gre_ecn_decapsulate_ipv4(tunnel, ipv6h, skb);
 644                 else if (skb->protocol == htons(ETH_P_IPV6))
 645                         ip6gre_ecn_decapsulate_ipv6(tunnel, ipv6h, skb);
 646
 647                 netif_rx(skb);
 648
 649                 rcu_read_unlock();
 650                 return 0;
 651         }
 652         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 653
 654 drop:
 655         rcu_read_unlock();
 656 drop_nolock:
 657         kfree_skb(skb);
 658         return 0;
 659 }
 660
 661 struct ipv6_tel_txoption {
 662         struct ipv6_txoptions ops;
 663         __u8 dst_opt[8];
 664 };
 665
 666 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 667 {
 668         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 669
 670         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 671         opt->dst_opt[3] = 1;
 672         opt->dst_opt[4] = encap_limit;
 673         opt->dst_opt[5] = IPV6_TLV_PADN;
 674         opt->dst_opt[6] = 1;
 675
 676         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 677         opt->ops.opt_nflen = 8;
 678 }
 679
 680 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 681                          struct net_device *dev,
 682                          __u8 dsfield,
 683                          struct flowi6 *fl6,
 684                          int encap_limit,
 685                          __u32 *pmtu)
 686 {
 687         struct net *net = dev_net(dev);
 688         struct ip6_tnl *tunnel = netdev_priv(dev);
 689         struct net_device *tdev;    /* Device to other host */
 690         struct ipv6hdr  *ipv6h;     /* Our new IP header */
 691         unsigned int max_headroom;  /* The extra header space needed */
 692         int    gre_hlen;
 693         struct ipv6_tel_txoption opt;
 694         int    mtu;
 695         struct dst_entry *dst = NULL, *ndst = NULL;
 696         struct net_device_stats *stats = &tunnel->dev->stats;
 697         int err = -1;
 698         u8 proto;
 699         int pkt_len;
 700         struct sk_buff *new_skb;
 701
 702         if (dev->type == ARPHRD_ETHER)
 703                 IPCB(skb)->flags = 0;
 704
 705         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 706                 gre_hlen = 0;
 707                 ipv6h = (struct ipv6hdr *)skb->data;
 708                 fl6->daddr = ipv6h->daddr;
 709         } else {
 710                 gre_hlen = tunnel->hlen;
 711                 fl6->daddr = tunnel->parms.raddr;
 712         }
 713
 714         if (!fl6->flowi6_mark)
 715                 dst = ip6_tnl_dst_check(tunnel);
 716
 717         if (!dst) {
 718                 ndst = ip6_route_output(net, NULL, fl6);
 719
 720                 if (ndst->error)
 721                         goto tx_err_link_failure;
 722                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
 723                 if (IS_ERR(ndst)) {
 724                         err = PTR_ERR(ndst);
 725                         ndst = NULL;
 726                         goto tx_err_link_failure;
 727                 }
 728                 dst = ndst;
 729         }
 730
 731         tdev = dst->dev;
 732
 733         if (tdev == dev) {
 734                 stats->collisions++;
 735                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 736                                      tunnel->parms.name);
 737                 goto tx_err_dst_release;
 738         }
 739
 740         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 741         if (encap_limit >= 0) {
 742                 max_headroom += 8;
 743                 mtu -= 8;
 744         }
 745         if (mtu < IPV6_MIN_MTU)
 746                 mtu = IPV6_MIN_MTU;
 747         if (skb_dst(skb))
 748                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 749         if (skb->len > mtu) {
 750                 *pmtu = mtu;
 751                 err = -EMSGSIZE;
 752                 goto tx_err_dst_release;
 753         }
 754
 755         if (tunnel->err_count > 0) {
 756                 if (time_before(jiffies,
 757                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 758                         tunnel->err_count--;
 759
 760                         dst_link_failure(skb);
 761                 } else
 762                         tunnel->err_count = 0;
 763         }
 764
 765         max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 766
 767         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 768             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 769                 new_skb = skb_realloc_headroom(skb, max_headroom);
 770                 if (max_headroom > dev->needed_headroom)
 771                         dev->needed_headroom = max_headroom;
 772                 if (!new_skb)
 773                         goto tx_err_dst_release;
 774
 775                 if (skb->sk)
 776                         skb_set_owner_w(new_skb, skb->sk);
 777                 consume_skb(skb);
 778                 skb = new_skb;
 779         }
 780
 781         skb_dst_drop(skb);
 782
 783         if (fl6->flowi6_mark) {
 784                 skb_dst_set(skb, dst);
 785                 ndst = NULL;
 786         } else {
 787                 skb_dst_set_noref(skb, dst);
 788         }
 789
 790         skb->transport_header = skb->network_header;
 791
 792         proto = NEXTHDR_GRE;
 793         if (encap_limit >= 0) {
 794                 init_tel_txopt(&opt, encap_limit);
 795                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 796         }
 797
 798         skb_push(skb, gre_hlen);
 799         skb_reset_network_header(skb);
 800
 801         /*
 802          *      Push down and install the IP header.
 803          */
 804         ipv6h = ipv6_hdr(skb);
 805         *(__be32 *)ipv6h = fl6->flowlabel | htonl(0x60000000);
 806         dsfield = INET_ECN_encapsulate(0, dsfield);
 807         ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
 808         ipv6h->hop_limit = tunnel->parms.hop_limit;
 809         ipv6h->nexthdr = proto;
 810         ipv6h->saddr = fl6->saddr;
 811         ipv6h->daddr = fl6->daddr;
 812
 813         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 814         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
 815                                    htons(ETH_P_TEB) : skb->protocol;
 816
 817         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 818                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 819
 820                 if (tunnel->parms.o_flags&GRE_SEQ) {
 821                         ++tunnel->o_seqno;
 822                         *ptr = htonl(tunnel->o_seqno);
 823                         ptr--;
 824                 }
 825                 if (tunnel->parms.o_flags&GRE_KEY) {
 826                         *ptr = tunnel->parms.o_key;
 827                         ptr--;
 828                 }
 829                 if (tunnel->parms.o_flags&GRE_CSUM) {
 830                         *ptr = 0;
 831                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
 832                                 skb->len - sizeof(struct ipv6hdr));
 833                 }
 834         }
 835
 836         nf_reset(skb);
 837         pkt_len = skb->len;
 838         err = ip6_local_out(skb);
 839
 840         if (net_xmit_eval(err) == 0) {
 841                 struct pcpu_tstats *tstats = this_cpu_ptr(tunnel->dev->tstats);
 842
 843                 tstats->tx_bytes += pkt_len;
 844                 tstats->tx_packets++;
 845         } else {
 846                 stats->tx_errors++;
 847                 stats->tx_aborted_errors++;
 848         }
 849
 850         if (ndst)
 851                 ip6_tnl_dst_store(tunnel, ndst);
 852
 853         return 0;
 854 tx_err_link_failure:
 855         stats->tx_carrier_errors++;
 856         dst_link_failure(skb);
 857 tx_err_dst_release:
 858         dst_release(ndst);
 859         return err;
 860 }
 861
 862 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 863 {
 864         struct ip6_tnl *t = netdev_priv(dev);
 865         const struct iphdr  *iph = ip_hdr(skb);
 866         int encap_limit = -1;
 867         struct flowi6 fl6;
 868         __u8 dsfield;
 869         __u32 mtu;
 870         int err;
 871
 872         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 873                 encap_limit = t->parms.encap_limit;
 874
 875         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 876         fl6.flowi6_proto = IPPROTO_IPIP;
 877
 878         dsfield = ipv4_get_dsfield(iph);
 879
 880         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 881                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 882                                           & IPV6_TCLASS_MASK;
 883         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 884                 fl6.flowi6_mark = skb->mark;
 885
 886         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 887         if (err != 0) {
 888                 /* XXX: send ICMP error even if DF is not set. */
 889                 if (err == -EMSGSIZE)
 890                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 891                                   htonl(mtu));
 892                 return -1;
 893         }
 894
 895         return 0;
 896 }
 897
 898 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 899 {
 900         struct ip6_tnl *t = netdev_priv(dev);
 901         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 902         int encap_limit = -1;
 903         __u16 offset;
 904         struct flowi6 fl6;
 905         __u8 dsfield;
 906         __u32 mtu;
 907         int err;
 908
 909         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 910                 return -1;
 911
 912         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 913         if (offset > 0) {
 914                 struct ipv6_tlv_tnl_enc_lim *tel;
 915                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 916                 if (tel->encap_limit == 0) {
 917                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 918                                     ICMPV6_HDR_FIELD, offset + 2);
 919                         return -1;
 920                 }
 921                 encap_limit = tel->encap_limit - 1;
 922         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 923                 encap_limit = t->parms.encap_limit;
 924
 925         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 926         fl6.flowi6_proto = IPPROTO_IPV6;
 927
 928         dsfield = ipv6_get_dsfield(ipv6h);
 929         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 930                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 931         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 932                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 933         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 934                 fl6.flowi6_mark = skb->mark;
 935
 936         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 937         if (err != 0) {
 938                 if (err == -EMSGSIZE)
 939                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 940                 return -1;
 941         }
 942
 943         return 0;
 944 }
 945
 946 /**
 947  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 948  *   @t: the outgoing tunnel device
 949  *   @hdr: IPv6 header from the incoming packet
 950  *
 951  * Description:
 952  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 953  *   doesn't match source of incoming packet.
 954  *
 955  * Return:
 956  *   1 if conflict,
 957  *   0 else
 958  **/
 959
 960 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 961         const struct ipv6hdr *hdr)
 962 {
 963         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 964 }
 965
 966 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 967 {
 968         struct ip6_tnl *t = netdev_priv(dev);
 969         int encap_limit = -1;
 970         struct flowi6 fl6;
 971         __u32 mtu;
 972         int err;
 973
 974         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 975                 encap_limit = t->parms.encap_limit;
 976
 977         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 978         fl6.flowi6_proto = skb->protocol;
 979
 980         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 981
 982         return err;
 983 }
 984
 985 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 986         struct net_device *dev)
 987 {
 988         struct ip6_tnl *t = netdev_priv(dev);
 989         struct net_device_stats *stats = &t->dev->stats;
 990         int ret;
 991
 992         if (!ip6_tnl_xmit_ctl(t))
 993                 return -1;
 994
 995         switch (skb->protocol) {
 996         case htons(ETH_P_IP):
 997                 ret = ip6gre_xmit_ipv4(skb, dev);
 998                 break;
 999         case htons(ETH_P_IPV6):
1000                 ret = ip6gre_xmit_ipv6(skb, dev);
1001                 break;
1002         default:
1003                 ret = ip6gre_xmit_other(skb, dev);
1004                 break;
1005         }
1006
1007         if (ret < 0)
1008                 goto tx_err;
1009
1010         return NETDEV_TX_OK;
1011
1012 tx_err:
1013         stats->tx_errors++;
1014         stats->tx_dropped++;
1015         kfree_skb(skb);
1016         return NETDEV_TX_OK;
1017 }
1018
1019 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1020 {
1021         struct net_device *dev = t->dev;
1022         struct __ip6_tnl_parm *p = &t->parms;
1023         struct flowi6 *fl6 = &t->fl.u.ip6;
1024         int addend = sizeof(struct ipv6hdr) + 4;
1025
1026         if (dev->type != ARPHRD_ETHER) {
1027                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1028                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1029         }
1030
1031         /* Set up flowi template */
1032         fl6->saddr = p->laddr;
1033         fl6->daddr = p->raddr;
1034         fl6->flowi6_oif = p->link;
1035         fl6->flowlabel = 0;
1036
1037         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1038                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1039         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1040                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1041
1042         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1043         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1044
1045         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1046                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1047                 dev->flags |= IFF_POINTOPOINT;
1048         else
1049                 dev->flags &= ~IFF_POINTOPOINT;
1050
1051         dev->iflink = p->link;
1052
1053         /* Precalculate GRE options length */
1054         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
1055                 if (t->parms.o_flags&GRE_CSUM)
1056                         addend += 4;
1057                 if (t->parms.o_flags&GRE_KEY)
1058                         addend += 4;
1059                 if (t->parms.o_flags&GRE_SEQ)
1060                         addend += 4;
1061         }
1062
1063         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1064                 int strict = (ipv6_addr_type(&p->raddr) &
1065                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1066
1067                 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1068                                                  &p->raddr, &p->laddr,
1069                                                  p->link, strict);
1070
1071                 if (rt == NULL)
1072                         return;
1073
1074                 if (rt->dst.dev) {
1075                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
1076
1077                         if (set_mtu) {
1078                                 dev->mtu = rt->dst.dev->mtu - addend;
1079                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1080                                         dev->mtu -= 8;
1081
1082                                 if (dev->mtu < IPV6_MIN_MTU)
1083                                         dev->mtu = IPV6_MIN_MTU;
1084                         }
1085                 }
1086                 dst_release(&rt->dst);
1087         }
1088
1089         t->hlen = addend;
1090 }
1091
1092 static int ip6gre_tnl_change(struct ip6_tnl *t,
1093         const struct __ip6_tnl_parm *p, int set_mtu)
1094 {
1095         t->parms.laddr = p->laddr;
1096         t->parms.raddr = p->raddr;
1097         t->parms.flags = p->flags;
1098         t->parms.hop_limit = p->hop_limit;
1099         t->parms.encap_limit = p->encap_limit;
1100         t->parms.flowinfo = p->flowinfo;
1101         t->parms.link = p->link;
1102         t->parms.proto = p->proto;
1103         t->parms.i_key = p->i_key;
1104         t->parms.o_key = p->o_key;
1105         t->parms.i_flags = p->i_flags;
1106         t->parms.o_flags = p->o_flags;
1107         ip6_tnl_dst_reset(t);
1108         ip6gre_tnl_link_config(t, set_mtu);
1109         return 0;
1110 }
1111
1112 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1113         const struct ip6_tnl_parm2 *u)
1114 {
1115         p->laddr = u->laddr;
1116         p->raddr = u->raddr;
1117         p->flags = u->flags;
1118         p->hop_limit = u->hop_limit;
1119         p->encap_limit = u->encap_limit;
1120         p->flowinfo = u->flowinfo;
1121         p->link = u->link;
1122         p->i_key = u->i_key;
1123         p->o_key = u->o_key;
1124         p->i_flags = u->i_flags;
1125         p->o_flags = u->o_flags;
1126         memcpy(p->name, u->name, sizeof(u->name));
1127 }
1128
1129 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1130         const struct __ip6_tnl_parm *p)
1131 {
1132         u->proto = IPPROTO_GRE;
1133         u->laddr = p->laddr;
1134         u->raddr = p->raddr;
1135         u->flags = p->flags;
1136         u->hop_limit = p->hop_limit;
1137         u->encap_limit = p->encap_limit;
1138         u->flowinfo = p->flowinfo;
1139         u->link = p->link;
1140         u->i_key = p->i_key;
1141         u->o_key = p->o_key;
1142         u->i_flags = p->i_flags;
1143         u->o_flags = p->o_flags;
1144         memcpy(u->name, p->name, sizeof(u->name));
1145 }
1146
1147 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1148         struct ifreq *ifr, int cmd)
1149 {
1150         int err = 0;
1151         struct ip6_tnl_parm2 p;
1152         struct __ip6_tnl_parm p1;
1153         struct ip6_tnl *t;
1154         struct net *net = dev_net(dev);
1155         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1156
1157         switch (cmd) {
1158         case SIOCGETTUNNEL:
1159                 t = NULL;
1160                 if (dev == ign->fb_tunnel_dev) {
1161                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1162                                 err = -EFAULT;
1163                                 break;
1164                         }
1165                         ip6gre_tnl_parm_from_user(&p1, &p);
1166                         t = ip6gre_tunnel_locate(net, &p1, 0);
1167                 }
1168                 if (t == NULL)
1169                         t = netdev_priv(dev);
1170                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1171                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1172                         err = -EFAULT;
1173                 break;
1174
1175         case SIOCADDTUNNEL:
1176         case SIOCCHGTUNNEL:
1177                 err = -EPERM;
1178                 if (!capable(CAP_NET_ADMIN))
1179                         goto done;
1180
1181                 err = -EFAULT;
1182                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1183                         goto done;
1184
1185                 err = -EINVAL;
1186                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1187                         goto done;
1188
1189                 if (!(p.i_flags&GRE_KEY))
1190                         p.i_key = 0;
1191                 if (!(p.o_flags&GRE_KEY))
1192                         p.o_key = 0;
1193
1194                 ip6gre_tnl_parm_from_user(&p1, &p);
1195                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1196
1197                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1198                         if (t != NULL) {
1199                                 if (t->dev != dev) {
1200                                         err = -EEXIST;
1201                                         break;
1202                                 }
1203                         } else {
1204                                 t = netdev_priv(dev);
1205
1206                                 ip6gre_tunnel_unlink(ign, t);
1207                                 synchronize_net();
1208                                 ip6gre_tnl_change(t, &p1, 1);
1209                                 ip6gre_tunnel_link(ign, t);
1210                                 netdev_state_change(dev);
1211                         }
1212                 }
1213
1214                 if (t) {
1215                         err = 0;
1216
1217                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1218                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1219                                 err = -EFAULT;
1220                 } else
1221                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1222                 break;
1223
1224         case SIOCDELTUNNEL:
1225                 err = -EPERM;
1226                 if (!capable(CAP_NET_ADMIN))
1227                         goto done;
1228
1229                 if (dev == ign->fb_tunnel_dev) {
1230                         err = -EFAULT;
1231                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1232                                 goto done;
1233                         err = -ENOENT;
1234                         ip6gre_tnl_parm_from_user(&p1, &p);
1235                         t = ip6gre_tunnel_locate(net, &p1, 0);
1236                         if (t == NULL)
1237                                 goto done;
1238                         err = -EPERM;
1239                         if (t == netdev_priv(ign->fb_tunnel_dev))
1240                                 goto done;
1241                         dev = t->dev;
1242                 }
1243                 unregister_netdevice(dev);
1244                 err = 0;
1245                 break;
1246
1247         default:
1248                 err = -EINVAL;
1249         }
1250
1251 done:
1252         return err;
1253 }
1254
1255 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1256 {
1257         struct ip6_tnl *tunnel = netdev_priv(dev);
1258         if (new_mtu < 68 ||
1259             new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1260                 return -EINVAL;
1261         dev->mtu = new_mtu;
1262         return 0;
1263 }
1264
1265 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1266                         unsigned short type,
1267                         const void *daddr, const void *saddr, unsigned int len)
1268 {
1269         struct ip6_tnl *t = netdev_priv(dev);
1270         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1271         __be16 *p = (__be16 *)(ipv6h+1);
1272
1273         *(__be32 *)ipv6h = t->fl.u.ip6.flowlabel | htonl(0x60000000);
1274         ipv6h->hop_limit = t->parms.hop_limit;
1275         ipv6h->nexthdr = NEXTHDR_GRE;
1276         ipv6h->saddr = t->parms.laddr;
1277         ipv6h->daddr = t->parms.raddr;
1278
1279         p[0]            = t->parms.o_flags;
1280         p[1]            = htons(type);
1281
1282         /*
1283          *      Set the source hardware address.
1284          */
1285
1286         if (saddr)
1287                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1288         if (daddr)
1289                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1290         if (!ipv6_addr_any(&ipv6h->daddr))
1291                 return t->hlen;
1292
1293         return -t->hlen;
1294 }
1295
1296 static int ip6gre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
1297 {
1298         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb_mac_header(skb);
1299         memcpy(haddr, &ipv6h->saddr, sizeof(struct in6_addr));
1300         return sizeof(struct in6_addr);
1301 }
1302
1303 static const struct header_ops ip6gre_header_ops = {
1304         .create = ip6gre_header,
1305         .parse  = ip6gre_header_parse,
1306 };
1307
1308 static const struct net_device_ops ip6gre_netdev_ops = {
1309         .ndo_init               = ip6gre_tunnel_init,
1310         .ndo_uninit             = ip6gre_tunnel_uninit,
1311         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1312         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1313         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1314         .ndo_get_stats64        = ip6gre_get_stats64,
1315 };
1316
1317 static void ip6gre_dev_free(struct net_device *dev)
1318 {
1319         free_percpu(dev->tstats);
1320         free_netdev(dev);
1321 }
1322
1323 static void ip6gre_tunnel_setup(struct net_device *dev)
1324 {
1325         struct ip6_tnl *t;
1326
1327         dev->netdev_ops = &ip6gre_netdev_ops;
1328         dev->destructor = ip6gre_dev_free;
1329
1330         dev->type = ARPHRD_IP6GRE;
1331         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1332         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1333         t = netdev_priv(dev);
1334         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1335                 dev->mtu -= 8;
1336         dev->flags |= IFF_NOARP;
1337         dev->iflink = 0;
1338         dev->addr_len = sizeof(struct in6_addr);
1339         dev->features |= NETIF_F_NETNS_LOCAL;
1340         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1341 }
1342
1343 static int ip6gre_tunnel_init(struct net_device *dev)
1344 {
1345         struct ip6_tnl *tunnel;
1346
1347         tunnel = netdev_priv(dev);
1348
1349         tunnel->dev = dev;
1350         strcpy(tunnel->parms.name, dev->name);
1351
1352         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1353         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1354
1355         if (ipv6_addr_any(&tunnel->parms.raddr))
1356                 dev->header_ops = &ip6gre_header_ops;
1357
1358         dev->tstats = alloc_percpu(struct pcpu_tstats);
1359         if (!dev->tstats)
1360                 return -ENOMEM;
1361
1362         return 0;
1363 }
1364
1365 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1366 {
1367         struct ip6_tnl *tunnel = netdev_priv(dev);
1368
1369         tunnel->dev = dev;
1370         strcpy(tunnel->parms.name, dev->name);
1371
1372         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1373
1374         dev_hold(dev);
1375 }
1376
1377
1378 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1379         .handler     = ip6gre_rcv,
1380         .err_handler = ip6gre_err,
1381         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1382 };
1383
1384 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1385         struct list_head *head)
1386 {
1387         int prio;
1388
1389         for (prio = 0; prio < 4; prio++) {
1390                 int h;
1391                 for (h = 0; h < HASH_SIZE; h++) {
1392                         struct ip6_tnl *t;
1393
1394                         t = rtnl_dereference(ign->tunnels[prio][h]);
1395
1396                         while (t != NULL) {
1397                                 unregister_netdevice_queue(t->dev, head);
1398                                 t = rtnl_dereference(t->next);
1399                         }
1400                 }
1401         }
1402 }
1403
1404 static int __net_init ip6gre_init_net(struct net *net)
1405 {
1406         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1407         int err;
1408
1409         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1410                                            ip6gre_tunnel_setup);
1411         if (!ign->fb_tunnel_dev) {
1412                 err = -ENOMEM;
1413                 goto err_alloc_dev;
1414         }
1415         dev_net_set(ign->fb_tunnel_dev, net);
1416
1417         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1418         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1419
1420         err = register_netdev(ign->fb_tunnel_dev);
1421         if (err)
1422                 goto err_reg_dev;
1423
1424         rcu_assign_pointer(ign->tunnels_wc[0],
1425                            netdev_priv(ign->fb_tunnel_dev));
1426         return 0;
1427
1428 err_reg_dev:
1429         ip6gre_dev_free(ign->fb_tunnel_dev);
1430 err_alloc_dev:
1431         return err;
1432 }
1433
1434 static void __net_exit ip6gre_exit_net(struct net *net)
1435 {
1436         struct ip6gre_net *ign;
1437         LIST_HEAD(list);
1438
1439         ign = net_generic(net, ip6gre_net_id);
1440         rtnl_lock();
1441         ip6gre_destroy_tunnels(ign, &list);
1442         unregister_netdevice_many(&list);
1443         rtnl_unlock();
1444 }
1445
1446 static struct pernet_operations ip6gre_net_ops = {
1447         .init = ip6gre_init_net,
1448         .exit = ip6gre_exit_net,
1449         .id   = &ip6gre_net_id,
1450         .size = sizeof(struct ip6gre_net),
1451 };
1452
1453 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1454 {
1455         __be16 flags;
1456
1457         if (!data)
1458                 return 0;
1459
1460         flags = 0;
1461         if (data[IFLA_GRE_IFLAGS])
1462                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1463         if (data[IFLA_GRE_OFLAGS])
1464                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1465         if (flags & (GRE_VERSION|GRE_ROUTING))
1466                 return -EINVAL;
1467
1468         return 0;
1469 }
1470
1471 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1472 {
1473         struct in6_addr daddr;
1474
1475         if (tb[IFLA_ADDRESS]) {
1476                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1477                         return -EINVAL;
1478                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1479                         return -EADDRNOTAVAIL;
1480         }
1481
1482         if (!data)
1483                 goto out;
1484
1485         if (data[IFLA_GRE_REMOTE]) {
1486                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1487                 if (ipv6_addr_any(&daddr))
1488                         return -EINVAL;
1489         }
1490
1491 out:
1492         return ip6gre_tunnel_validate(tb, data);
1493 }
1494
1495
1496 static void ip6gre_netlink_parms(struct nlattr *data[],
1497                                 struct __ip6_tnl_parm *parms)
1498 {
1499         memset(parms, 0, sizeof(*parms));
1500
1501         if (!data)
1502                 return;
1503
1504         if (data[IFLA_GRE_LINK])
1505                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1506
1507         if (data[IFLA_GRE_IFLAGS])
1508                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1509
1510         if (data[IFLA_GRE_OFLAGS])
1511                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1512
1513         if (data[IFLA_GRE_IKEY])
1514                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1515
1516         if (data[IFLA_GRE_OKEY])
1517                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1518
1519         if (data[IFLA_GRE_LOCAL])
1520                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1521
1522         if (data[IFLA_GRE_REMOTE])
1523                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1524
1525         if (data[IFLA_GRE_TTL])
1526                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1527
1528         if (data[IFLA_GRE_ENCAP_LIMIT])
1529                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1530
1531         if (data[IFLA_GRE_FLOWINFO])
1532                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1533
1534         if (data[IFLA_GRE_FLAGS])
1535                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1536 }
1537
1538 static int ip6gre_tap_init(struct net_device *dev)
1539 {
1540         struct ip6_tnl *tunnel;
1541
1542         tunnel = netdev_priv(dev);
1543
1544         tunnel->dev = dev;
1545         strcpy(tunnel->parms.name, dev->name);
1546
1547         ip6gre_tnl_link_config(tunnel, 1);
1548
1549         dev->tstats = alloc_percpu(struct pcpu_tstats);
1550         if (!dev->tstats)
1551                 return -ENOMEM;
1552
1553         return 0;
1554 }
1555
1556 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1557         .ndo_init = ip6gre_tap_init,
1558         .ndo_uninit = ip6gre_tunnel_uninit,
1559         .ndo_start_xmit = ip6gre_tunnel_xmit,
1560         .ndo_set_mac_address = eth_mac_addr,
1561         .ndo_validate_addr = eth_validate_addr,
1562         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1563         .ndo_get_stats64 = ip6gre_get_stats64,
1564 };
1565
1566 static void ip6gre_tap_setup(struct net_device *dev)
1567 {
1568
1569         ether_setup(dev);
1570
1571         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1572         dev->destructor = ip6gre_dev_free;
1573
1574         dev->iflink = 0;
1575         dev->features |= NETIF_F_NETNS_LOCAL;
1576 }
1577
1578 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1579         struct nlattr *tb[], struct nlattr *data[])
1580 {
1581         struct ip6_tnl *nt;
1582         struct net *net = dev_net(dev);
1583         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1584         int err;
1585
1586         nt = netdev_priv(dev);
1587         ip6gre_netlink_parms(data, &nt->parms);
1588
1589         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1590                 return -EEXIST;
1591
1592         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1593                 eth_hw_addr_random(dev);
1594
1595         nt->dev = dev;
1596         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1597
1598         /* Can use a lockless transmit, unless we generate output sequences */
1599         if (!(nt->parms.o_flags & GRE_SEQ))
1600                 dev->features |= NETIF_F_LLTX;
1601
1602         err = register_netdevice(dev);
1603         if (err)
1604                 goto out;
1605
1606         dev_hold(dev);
1607         ip6gre_tunnel_link(ign, nt);
1608
1609 out:
1610         return err;
1611 }
1612
1613 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1614                             struct nlattr *data[])
1615 {
1616         struct ip6_tnl *t, *nt;
1617         struct net *net = dev_net(dev);
1618         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1619         struct __ip6_tnl_parm p;
1620
1621         if (dev == ign->fb_tunnel_dev)
1622                 return -EINVAL;
1623
1624         nt = netdev_priv(dev);
1625         ip6gre_netlink_parms(data, &p);
1626
1627         t = ip6gre_tunnel_locate(net, &p, 0);
1628
1629         if (t) {
1630                 if (t->dev != dev)
1631                         return -EEXIST;
1632         } else {
1633                 t = nt;
1634
1635                 ip6gre_tunnel_unlink(ign, t);
1636                 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1637                 ip6gre_tunnel_link(ign, t);
1638                 netdev_state_change(dev);
1639         }
1640
1641         return 0;
1642 }
1643
1644 static size_t ip6gre_get_size(const struct net_device *dev)
1645 {
1646         return
1647                 /* IFLA_GRE_LINK */
1648                 nla_total_size(4) +
1649                 /* IFLA_GRE_IFLAGS */
1650                 nla_total_size(2) +
1651                 /* IFLA_GRE_OFLAGS */
1652                 nla_total_size(2) +
1653                 /* IFLA_GRE_IKEY */
1654                 nla_total_size(4) +
1655                 /* IFLA_GRE_OKEY */
1656                 nla_total_size(4) +
1657                 /* IFLA_GRE_LOCAL */
1658                 nla_total_size(4) +
1659                 /* IFLA_GRE_REMOTE */
1660                 nla_total_size(4) +
1661                 /* IFLA_GRE_TTL */
1662                 nla_total_size(1) +
1663                 /* IFLA_GRE_TOS */
1664                 nla_total_size(1) +
1665                 /* IFLA_GRE_ENCAP_LIMIT */
1666                 nla_total_size(1) +
1667                 /* IFLA_GRE_FLOWINFO */
1668                 nla_total_size(4) +
1669                 /* IFLA_GRE_FLAGS */
1670                 nla_total_size(4) +
1671                 0;
1672 }
1673
1674 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1675 {
1676         struct ip6_tnl *t = netdev_priv(dev);
1677         struct __ip6_tnl_parm *p = &t->parms;
1678
1679         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1680             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1681             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1682             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1683             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1684             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->raddr) ||
1685             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->laddr) ||
1686             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1687             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1688             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1689             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1690             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1691                 goto nla_put_failure;
1692         return 0;
1693
1694 nla_put_failure:
1695         return -EMSGSIZE;
1696 }
1697
1698 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1699         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1700         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1701         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1702         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1703         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1704         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1705         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1706         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1707         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1708         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1709         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1710 };
1711
1712 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1713         .kind           = "ip6gre",
1714         .maxtype        = IFLA_GRE_MAX,
1715         .policy         = ip6gre_policy,
1716         .priv_size      = sizeof(struct ip6_tnl),
1717         .setup          = ip6gre_tunnel_setup,
1718         .validate       = ip6gre_tunnel_validate,
1719         .newlink        = ip6gre_newlink,
1720         .changelink     = ip6gre_changelink,
1721         .get_size       = ip6gre_get_size,
1722         .fill_info      = ip6gre_fill_info,
1723 };
1724
1725 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1726         .kind           = "ip6gretap",
1727         .maxtype        = IFLA_GRE_MAX,
1728         .policy         = ip6gre_policy,
1729         .priv_size      = sizeof(struct ip6_tnl),
1730         .setup          = ip6gre_tap_setup,
1731         .validate       = ip6gre_tap_validate,
1732         .newlink        = ip6gre_newlink,
1733         .changelink     = ip6gre_changelink,
1734         .get_size       = ip6gre_get_size,
1735         .fill_info      = ip6gre_fill_info,
1736 };
1737
1738 /*
1739  *      And now the modules code and kernel interface.
1740  */
1741
1742 static int __init ip6gre_init(void)
1743 {
1744         int err;
1745
1746         pr_info("GRE over IPv6 tunneling driver\n");
1747
1748         err = register_pernet_device(&ip6gre_net_ops);
1749         if (err < 0)
1750                 return err;
1751
1752         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1753         if (err < 0) {
1754                 pr_info("%s: can't add protocol\n", __func__);
1755                 goto add_proto_failed;
1756         }
1757
1758         err = rtnl_link_register(&ip6gre_link_ops);
1759         if (err < 0)
1760                 goto rtnl_link_failed;
1761
1762         err = rtnl_link_register(&ip6gre_tap_ops);
1763         if (err < 0)
1764                 goto tap_ops_failed;
1765
1766 out:
1767         return err;
1768
1769 tap_ops_failed:
1770         rtnl_link_unregister(&ip6gre_link_ops);
1771 rtnl_link_failed:
1772         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1773 add_proto_failed:
1774         unregister_pernet_device(&ip6gre_net_ops);
1775         goto out;
1776 }
1777
1778 static void __exit ip6gre_fini(void)
1779 {
1780         rtnl_link_unregister(&ip6gre_tap_ops);
1781         rtnl_link_unregister(&ip6gre_link_ops);
1782         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1783         unregister_pernet_device(&ip6gre_net_ops);
1784 }
1785
1786 module_init(ip6gre_init);
1787 module_exit(ip6gre_fini);
1788 MODULE_LICENSE("GPL");
1789 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1790 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1791 MODULE_ALIAS_RTNL_LINK("ip6gre");
1792 MODULE_ALIAS_NETDEV("ip6gre0");