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 static struct rtnl_link_stats64 *ip6gre_get_stats64(struct net_device *dev,
 115                 struct rtnl_link_stats64 *tot)
 116 {
 117         int i;
 118
 119         for_each_possible_cpu(i) {
 120                 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
 121                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
 122                 unsigned int start;
 123
 124                 do {
 125                         start = u64_stats_fetch_begin_bh(&tstats->syncp);
 126                         rx_packets = tstats->rx_packets;
 127                         tx_packets = tstats->tx_packets;
 128                         rx_bytes = tstats->rx_bytes;
 129                         tx_bytes = tstats->tx_bytes;
 130                 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
 131
 132                 tot->rx_packets += rx_packets;
 133                 tot->tx_packets += tx_packets;
 134                 tot->rx_bytes   += rx_bytes;
 135                 tot->tx_bytes   += tx_bytes;
 136         }
 137
 138         tot->multicast = dev->stats.multicast;
 139         tot->rx_crc_errors = dev->stats.rx_crc_errors;
 140         tot->rx_fifo_errors = dev->stats.rx_fifo_errors;
 141         tot->rx_length_errors = dev->stats.rx_length_errors;
 142         tot->rx_frame_errors = dev->stats.rx_frame_errors;
 143         tot->rx_errors = dev->stats.rx_errors;
 144
 145         tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
 146         tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
 147         tot->tx_dropped = dev->stats.tx_dropped;
 148         tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
 149         tot->tx_errors = dev->stats.tx_errors;
 150
 151         return tot;
 152 }
 153
 154 /* Given src, dst and key, find appropriate for input tunnel. */
 155
 156 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 157                 const struct in6_addr *remote, const struct in6_addr *local,
 158                 __be32 key, __be16 gre_proto)
 159 {
 160         struct net *net = dev_net(dev);
 161         int link = dev->ifindex;
 162         unsigned int h0 = HASH_ADDR(remote);
 163         unsigned int h1 = HASH_KEY(key);
 164         struct ip6_tnl *t, *cand = NULL;
 165         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 166         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
 167                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 168         int score, cand_score = 4;
 169
 170         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 171                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 172                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 173                     key != t->parms.i_key ||
 174                     !(t->dev->flags & IFF_UP))
 175                         continue;
 176
 177                 if (t->dev->type != ARPHRD_IP6GRE &&
 178                     t->dev->type != dev_type)
 179                         continue;
 180
 181                 score = 0;
 182                 if (t->parms.link != link)
 183                         score |= 1;
 184                 if (t->dev->type != dev_type)
 185                         score |= 2;
 186                 if (score == 0)
 187                         return t;
 188
 189                 if (score < cand_score) {
 190                         cand = t;
 191                         cand_score = score;
 192                 }
 193         }
 194
 195         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 196                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 197                     key != t->parms.i_key ||
 198                     !(t->dev->flags & IFF_UP))
 199                         continue;
 200
 201                 if (t->dev->type != ARPHRD_IP6GRE &&
 202                     t->dev->type != dev_type)
 203                         continue;
 204
 205                 score = 0;
 206                 if (t->parms.link != link)
 207                         score |= 1;
 208                 if (t->dev->type != dev_type)
 209                         score |= 2;
 210                 if (score == 0)
 211                         return t;
 212
 213                 if (score < cand_score) {
 214                         cand = t;
 215                         cand_score = score;
 216                 }
 217         }
 218
 219         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 220                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 221                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 222                                  !ipv6_addr_is_multicast(local))) ||
 223                     key != t->parms.i_key ||
 224                     !(t->dev->flags & IFF_UP))
 225                         continue;
 226
 227                 if (t->dev->type != ARPHRD_IP6GRE &&
 228                     t->dev->type != dev_type)
 229                         continue;
 230
 231                 score = 0;
 232                 if (t->parms.link != link)
 233                         score |= 1;
 234                 if (t->dev->type != dev_type)
 235                         score |= 2;
 236                 if (score == 0)
 237                         return t;
 238
 239                 if (score < cand_score) {
 240                         cand = t;
 241                         cand_score = score;
 242                 }
 243         }
 244
 245         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 246                 if (t->parms.i_key != key ||
 247                     !(t->dev->flags & IFF_UP))
 248                         continue;
 249
 250                 if (t->dev->type != ARPHRD_IP6GRE &&
 251                     t->dev->type != dev_type)
 252                         continue;
 253
 254                 score = 0;
 255                 if (t->parms.link != link)
 256                         score |= 1;
 257                 if (t->dev->type != dev_type)
 258                         score |= 2;
 259                 if (score == 0)
 260                         return t;
 261
 262                 if (score < cand_score) {
 263                         cand = t;
 264                         cand_score = score;
 265                 }
 266         }
 267
 268         if (cand != NULL)
 269                 return cand;
 270
 271         dev = ign->fb_tunnel_dev;
 272         if (dev->flags & IFF_UP)
 273                 return netdev_priv(dev);
 274
 275         return NULL;
 276 }
 277
 278 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 279                 const struct __ip6_tnl_parm *p)
 280 {
 281         const struct in6_addr *remote = &p->raddr;
 282         const struct in6_addr *local = &p->laddr;
 283         unsigned int h = HASH_KEY(p->i_key);
 284         int prio = 0;
 285
 286         if (!ipv6_addr_any(local))
 287                 prio |= 1;
 288         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 289                 prio |= 2;
 290                 h ^= HASH_ADDR(remote);
 291         }
 292
 293         return &ign->tunnels[prio][h];
 294 }
 295
 296 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 297                 const struct ip6_tnl *t)
 298 {
 299         return __ip6gre_bucket(ign, &t->parms);
 300 }
 301
 302 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 303 {
 304         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 305
 306         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 307         rcu_assign_pointer(*tp, t);
 308 }
 309
 310 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 311 {
 312         struct ip6_tnl __rcu **tp;
 313         struct ip6_tnl *iter;
 314
 315         for (tp = ip6gre_bucket(ign, t);
 316              (iter = rtnl_dereference(*tp)) != NULL;
 317              tp = &iter->next) {
 318                 if (t == iter) {
 319                         rcu_assign_pointer(*tp, t->next);
 320                         break;
 321                 }
 322         }
 323 }
 324
 325 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 326                                            const struct __ip6_tnl_parm *parms,
 327                                            int type)
 328 {
 329         const struct in6_addr *remote = &parms->raddr;
 330         const struct in6_addr *local = &parms->laddr;
 331         __be32 key = parms->i_key;
 332         int link = parms->link;
 333         struct ip6_tnl *t;
 334         struct ip6_tnl __rcu **tp;
 335         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 336
 337         for (tp = __ip6gre_bucket(ign, parms);
 338              (t = rtnl_dereference(*tp)) != NULL;
 339              tp = &t->next)
 340                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 341                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 342                     key == t->parms.i_key &&
 343                     link == t->parms.link &&
 344                     type == t->dev->type)
 345                         break;
 346
 347         return t;
 348 }
 349
 350 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 351                 const struct __ip6_tnl_parm *parms, int create)
 352 {
 353         struct ip6_tnl *t, *nt;
 354         struct net_device *dev;
 355         char name[IFNAMSIZ];
 356         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 357
 358         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 359         if (t || !create)
 360                 return t;
 361
 362         if (parms->name[0])
 363                 strlcpy(name, parms->name, IFNAMSIZ);
 364         else
 365                 strcpy(name, "ip6gre%d");
 366
 367         dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
 368         if (!dev)
 369                 return NULL;
 370
 371         dev_net_set(dev, net);
 372
 373         nt = netdev_priv(dev);
 374         nt->parms = *parms;
 375         dev->rtnl_link_ops = &ip6gre_link_ops;
 376
 377         nt->dev = dev;
 378         ip6gre_tnl_link_config(nt, 1);
 379
 380         if (register_netdevice(dev) < 0)
 381                 goto failed_free;
 382
 383         /* Can use a lockless transmit, unless we generate output sequences */
 384         if (!(nt->parms.o_flags & GRE_SEQ))
 385                 dev->features |= NETIF_F_LLTX;
 386
 387         dev_hold(dev);
 388         ip6gre_tunnel_link(ign, nt);
 389         return nt;
 390
 391 failed_free:
 392         free_netdev(dev);
 393         return NULL;
 394 }
 395
 396 static void ip6gre_tunnel_uninit(struct net_device *dev)
 397 {
 398         struct net *net = dev_net(dev);
 399         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 400
 401         ip6gre_tunnel_unlink(ign, netdev_priv(dev));
 402         dev_put(dev);
 403 }
 404
 405
 406 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 407                 u8 type, u8 code, int offset, __be32 info)
 408 {
 409         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 410         __be16 *p = (__be16 *)(skb->data + offset);
 411         int grehlen = offset + 4;
 412         struct ip6_tnl *t;
 413         __be16 flags;
 414
 415         flags = p[0];
 416         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
 417                 if (flags&(GRE_VERSION|GRE_ROUTING))
 418                         return;
 419                 if (flags&GRE_KEY) {
 420                         grehlen += 4;
 421                         if (flags&GRE_CSUM)
 422                                 grehlen += 4;
 423                 }
 424         }
 425
 426         /* If only 8 bytes returned, keyed message will be dropped here */
 427         if (!pskb_may_pull(skb, grehlen))
 428                 return;
 429         ipv6h = (const struct ipv6hdr *)skb->data;
 430         p = (__be16 *)(skb->data + offset);
 431
 432         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 433                                 flags & GRE_KEY ?
 434                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
 435                                 p[1]);
 436         if (t == NULL)
 437                 return;
 438
 439         switch (type) {
 440                 __u32 teli;
 441                 struct ipv6_tlv_tnl_enc_lim *tel;
 442                 __u32 mtu;
 443         case ICMPV6_DEST_UNREACH:
 444                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
 445                                      t->parms.name);
 446                 break;
 447         case ICMPV6_TIME_EXCEED:
 448                 if (code == ICMPV6_EXC_HOPLIMIT) {
 449                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 450                                              t->parms.name);
 451                 }
 452                 break;
 453         case ICMPV6_PARAMPROB:
 454                 teli = 0;
 455                 if (code == ICMPV6_HDR_FIELD)
 456                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 457
 458                 if (teli && teli == info - 2) {
 459                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 460                         if (tel->encap_limit == 0) {
 461                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 462                                                      t->parms.name);
 463                         }
 464                 } else {
 465                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 466                                              t->parms.name);
 467                 }
 468                 break;
 469         case ICMPV6_PKT_TOOBIG:
 470                 mtu = info - offset;
 471                 if (mtu < IPV6_MIN_MTU)
 472                         mtu = IPV6_MIN_MTU;
 473                 t->dev->mtu = mtu;
 474                 break;
 475         }
 476
 477         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 478                 t->err_count++;
 479         else
 480                 t->err_count = 1;
 481         t->err_time = jiffies;
 482 }
 483
 484 static int ip6gre_rcv(struct sk_buff *skb)
 485 {
 486         const struct ipv6hdr *ipv6h;
 487         u8     *h;
 488         __be16    flags;
 489         __sum16   csum = 0;
 490         __be32 key = 0;
 491         u32    seqno = 0;
 492         struct ip6_tnl *tunnel;
 493         int    offset = 4;
 494         __be16 gre_proto;
 495         int err;
 496
 497         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
 498                 goto drop;
 499
 500         ipv6h = ipv6_hdr(skb);
 501         h = skb->data;
 502         flags = *(__be16 *)h;
 503
 504         if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
 505                 /* - Version must be 0.
 506                    - We do not support routing headers.
 507                  */
 508                 if (flags&(GRE_VERSION|GRE_ROUTING))
 509                         goto drop;
 510
 511                 if (flags&GRE_CSUM) {
 512                         switch (skb->ip_summed) {
 513                         case CHECKSUM_COMPLETE:
 514                                 csum = csum_fold(skb->csum);
 515                                 if (!csum)
 516                                         break;
 517                                 /* fall through */
 518                         case CHECKSUM_NONE:
 519                                 skb->csum = 0;
 520                                 csum = __skb_checksum_complete(skb);
 521                                 skb->ip_summed = CHECKSUM_COMPLETE;
 522                         }
 523                         offset += 4;
 524                 }
 525                 if (flags&GRE_KEY) {
 526                         key = *(__be32 *)(h + offset);
 527                         offset += 4;
 528                 }
 529                 if (flags&GRE_SEQ) {
 530                         seqno = ntohl(*(__be32 *)(h + offset));
 531                         offset += 4;
 532                 }
 533         }
 534
 535         gre_proto = *(__be16 *)(h + 2);
 536
 537         tunnel = ip6gre_tunnel_lookup(skb->dev,
 538                                           &ipv6h->saddr, &ipv6h->daddr, key,
 539                                           gre_proto);
 540         if (tunnel) {
 541                 struct pcpu_tstats *tstats;
 542
 543                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 544                         goto drop;
 545
 546                 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
 547                         tunnel->dev->stats.rx_dropped++;
 548                         goto drop;
 549                 }
 550
 551                 secpath_reset(skb);
 552
 553                 skb->protocol = gre_proto;
 554                 /* WCCP version 1 and 2 protocol decoding.
 555                  * - Change protocol to IP
 556                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
 557                  */
 558                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
 559                         skb->protocol = htons(ETH_P_IP);
 560                         if ((*(h + offset) & 0xF0) != 0x40)
 561                                 offset += 4;
 562                 }
 563
 564                 skb->mac_header = skb->network_header;
 565                 __pskb_pull(skb, offset);
 566                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
 567                 skb->pkt_type = PACKET_HOST;
 568
 569                 if (((flags&GRE_CSUM) && csum) ||
 570                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
 571                         tunnel->dev->stats.rx_crc_errors++;
 572                         tunnel->dev->stats.rx_errors++;
 573                         goto drop;
 574                 }
 575                 if (tunnel->parms.i_flags&GRE_SEQ) {
 576                         if (!(flags&GRE_SEQ) ||
 577                             (tunnel->i_seqno &&
 578                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
 579                                 tunnel->dev->stats.rx_fifo_errors++;
 580                                 tunnel->dev->stats.rx_errors++;
 581                                 goto drop;
 582                         }
 583                         tunnel->i_seqno = seqno + 1;
 584                 }
 585
 586                 /* Warning: All skb pointers will be invalidated! */
 587                 if (tunnel->dev->type == ARPHRD_ETHER) {
 588                         if (!pskb_may_pull(skb, ETH_HLEN)) {
 589                                 tunnel->dev->stats.rx_length_errors++;
 590                                 tunnel->dev->stats.rx_errors++;
 591                                 goto drop;
 592                         }
 593
 594                         ipv6h = ipv6_hdr(skb);
 595                         skb->protocol = eth_type_trans(skb, tunnel->dev);
 596                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 597                 }
 598
 599                 __skb_tunnel_rx(skb, tunnel->dev);
 600
 601                 skb_reset_network_header(skb);
 602
 603                 err = IP6_ECN_decapsulate(ipv6h, skb);
 604                 if (unlikely(err)) {
 605                         if (log_ecn_error)
 606                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 607                                                      &ipv6h->saddr,
 608                                                      ipv6_get_dsfield(ipv6h));
 609                         if (err > 1) {
 610                                 ++tunnel->dev->stats.rx_frame_errors;
 611                                 ++tunnel->dev->stats.rx_errors;
 612                                 goto drop;
 613                         }
 614                 }
 615
 616                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 617                 u64_stats_update_begin(&tstats->syncp);
 618                 tstats->rx_packets++;
 619                 tstats->rx_bytes += skb->len;
 620                 u64_stats_update_end(&tstats->syncp);
 621
 622                 netif_rx(skb);
 623
 624                 return 0;
 625         }
 626         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 627
 628 drop:
 629         kfree_skb(skb);
 630         return 0;
 631 }
 632
 633 struct ipv6_tel_txoption {
 634         struct ipv6_txoptions ops;
 635         __u8 dst_opt[8];
 636 };
 637
 638 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 639 {
 640         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 641
 642         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 643         opt->dst_opt[3] = 1;
 644         opt->dst_opt[4] = encap_limit;
 645         opt->dst_opt[5] = IPV6_TLV_PADN;
 646         opt->dst_opt[6] = 1;
 647
 648         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 649         opt->ops.opt_nflen = 8;
 650 }
 651
 652 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 653                          struct net_device *dev,
 654                          __u8 dsfield,
 655                          struct flowi6 *fl6,
 656                          int encap_limit,
 657                          __u32 *pmtu)
 658 {
 659         struct net *net = dev_net(dev);
 660         struct ip6_tnl *tunnel = netdev_priv(dev);
 661         struct net_device *tdev;    /* Device to other host */
 662         struct ipv6hdr  *ipv6h;     /* Our new IP header */
 663         unsigned int max_headroom;  /* The extra header space needed */
 664         int    gre_hlen;
 665         struct ipv6_tel_txoption opt;
 666         int    mtu;
 667         struct dst_entry *dst = NULL, *ndst = NULL;
 668         struct net_device_stats *stats = &tunnel->dev->stats;
 669         int err = -1;
 670         u8 proto;
 671         struct sk_buff *new_skb;
 672
 673         if (dev->type == ARPHRD_ETHER)
 674                 IPCB(skb)->flags = 0;
 675
 676         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 677                 gre_hlen = 0;
 678                 ipv6h = (struct ipv6hdr *)skb->data;
 679                 fl6->daddr = ipv6h->daddr;
 680         } else {
 681                 gre_hlen = tunnel->hlen;
 682                 fl6->daddr = tunnel->parms.raddr;
 683         }
 684
 685         if (!fl6->flowi6_mark)
 686                 dst = ip6_tnl_dst_check(tunnel);
 687
 688         if (!dst) {
 689                 ndst = ip6_route_output(net, NULL, fl6);
 690
 691                 if (ndst->error)
 692                         goto tx_err_link_failure;
 693                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
 694                 if (IS_ERR(ndst)) {
 695                         err = PTR_ERR(ndst);
 696                         ndst = NULL;
 697                         goto tx_err_link_failure;
 698                 }
 699                 dst = ndst;
 700         }
 701
 702         tdev = dst->dev;
 703
 704         if (tdev == dev) {
 705                 stats->collisions++;
 706                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 707                                      tunnel->parms.name);
 708                 goto tx_err_dst_release;
 709         }
 710
 711         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 712         if (encap_limit >= 0) {
 713                 max_headroom += 8;
 714                 mtu -= 8;
 715         }
 716         if (mtu < IPV6_MIN_MTU)
 717                 mtu = IPV6_MIN_MTU;
 718         if (skb_dst(skb))
 719                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 720         if (skb->len > mtu) {
 721                 *pmtu = mtu;
 722                 err = -EMSGSIZE;
 723                 goto tx_err_dst_release;
 724         }
 725
 726         if (tunnel->err_count > 0) {
 727                 if (time_before(jiffies,
 728                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 729                         tunnel->err_count--;
 730
 731                         dst_link_failure(skb);
 732                 } else
 733                         tunnel->err_count = 0;
 734         }
 735
 736         max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 737
 738         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 739             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 740                 new_skb = skb_realloc_headroom(skb, max_headroom);
 741                 if (max_headroom > dev->needed_headroom)
 742                         dev->needed_headroom = max_headroom;
 743                 if (!new_skb)
 744                         goto tx_err_dst_release;
 745
 746                 if (skb->sk)
 747                         skb_set_owner_w(new_skb, skb->sk);
 748                 consume_skb(skb);
 749                 skb = new_skb;
 750         }
 751
 752         skb_dst_drop(skb);
 753
 754         if (fl6->flowi6_mark) {
 755                 skb_dst_set(skb, dst);
 756                 ndst = NULL;
 757         } else {
 758                 skb_dst_set_noref(skb, dst);
 759         }
 760
 761         proto = NEXTHDR_GRE;
 762         if (encap_limit >= 0) {
 763                 init_tel_txopt(&opt, encap_limit);
 764                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 765         }
 766
 767         skb_push(skb, gre_hlen);
 768         skb_reset_network_header(skb);
 769         skb_set_transport_header(skb, sizeof(*ipv6h));
 770
 771         /*
 772          *      Push down and install the IP header.
 773          */
 774         ipv6h = ipv6_hdr(skb);
 775         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
 776         ipv6h->hop_limit = tunnel->parms.hop_limit;
 777         ipv6h->nexthdr = proto;
 778         ipv6h->saddr = fl6->saddr;
 779         ipv6h->daddr = fl6->daddr;
 780
 781         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 782         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
 783                                    htons(ETH_P_TEB) : skb->protocol;
 784
 785         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 786                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 787
 788                 if (tunnel->parms.o_flags&GRE_SEQ) {
 789                         ++tunnel->o_seqno;
 790                         *ptr = htonl(tunnel->o_seqno);
 791                         ptr--;
 792                 }
 793                 if (tunnel->parms.o_flags&GRE_KEY) {
 794                         *ptr = tunnel->parms.o_key;
 795                         ptr--;
 796                 }
 797                 if (tunnel->parms.o_flags&GRE_CSUM) {
 798                         *ptr = 0;
 799                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
 800                                 skb->len - sizeof(struct ipv6hdr));
 801                 }
 802         }
 803
 804         ip6tunnel_xmit(skb, dev);
 805         if (ndst)
 806                 ip6_tnl_dst_store(tunnel, ndst);
 807         return 0;
 808 tx_err_link_failure:
 809         stats->tx_carrier_errors++;
 810         dst_link_failure(skb);
 811 tx_err_dst_release:
 812         dst_release(ndst);
 813         return err;
 814 }
 815
 816 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 817 {
 818         struct ip6_tnl *t = netdev_priv(dev);
 819         const struct iphdr  *iph = ip_hdr(skb);
 820         int encap_limit = -1;
 821         struct flowi6 fl6;
 822         __u8 dsfield;
 823         __u32 mtu;
 824         int err;
 825
 826         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 827                 encap_limit = t->parms.encap_limit;
 828
 829         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 830         fl6.flowi6_proto = IPPROTO_IPIP;
 831
 832         dsfield = ipv4_get_dsfield(iph);
 833
 834         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 835                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 836                                           & IPV6_TCLASS_MASK;
 837         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 838                 fl6.flowi6_mark = skb->mark;
 839
 840         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 841         if (err != 0) {
 842                 /* XXX: send ICMP error even if DF is not set. */
 843                 if (err == -EMSGSIZE)
 844                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 845                                   htonl(mtu));
 846                 return -1;
 847         }
 848
 849         return 0;
 850 }
 851
 852 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 853 {
 854         struct ip6_tnl *t = netdev_priv(dev);
 855         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 856         int encap_limit = -1;
 857         __u16 offset;
 858         struct flowi6 fl6;
 859         __u8 dsfield;
 860         __u32 mtu;
 861         int err;
 862
 863         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 864                 return -1;
 865
 866         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 867         if (offset > 0) {
 868                 struct ipv6_tlv_tnl_enc_lim *tel;
 869                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 870                 if (tel->encap_limit == 0) {
 871                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 872                                     ICMPV6_HDR_FIELD, offset + 2);
 873                         return -1;
 874                 }
 875                 encap_limit = tel->encap_limit - 1;
 876         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 877                 encap_limit = t->parms.encap_limit;
 878
 879         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 880         fl6.flowi6_proto = IPPROTO_IPV6;
 881
 882         dsfield = ipv6_get_dsfield(ipv6h);
 883         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 884                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 885         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 886                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 887         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 888                 fl6.flowi6_mark = skb->mark;
 889
 890         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 891         if (err != 0) {
 892                 if (err == -EMSGSIZE)
 893                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 894                 return -1;
 895         }
 896
 897         return 0;
 898 }
 899
 900 /**
 901  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 902  *   @t: the outgoing tunnel device
 903  *   @hdr: IPv6 header from the incoming packet
 904  *
 905  * Description:
 906  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 907  *   doesn't match source of incoming packet.
 908  *
 909  * Return:
 910  *   1 if conflict,
 911  *   0 else
 912  **/
 913
 914 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 915         const struct ipv6hdr *hdr)
 916 {
 917         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 918 }
 919
 920 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 921 {
 922         struct ip6_tnl *t = netdev_priv(dev);
 923         int encap_limit = -1;
 924         struct flowi6 fl6;
 925         __u32 mtu;
 926         int err;
 927
 928         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 929                 encap_limit = t->parms.encap_limit;
 930
 931         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 932         fl6.flowi6_proto = skb->protocol;
 933
 934         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 935
 936         return err;
 937 }
 938
 939 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 940         struct net_device *dev)
 941 {
 942         struct ip6_tnl *t = netdev_priv(dev);
 943         struct net_device_stats *stats = &t->dev->stats;
 944         int ret;
 945
 946         if (!ip6_tnl_xmit_ctl(t))
 947                 goto tx_err;
 948
 949         switch (skb->protocol) {
 950         case htons(ETH_P_IP):
 951                 ret = ip6gre_xmit_ipv4(skb, dev);
 952                 break;
 953         case htons(ETH_P_IPV6):
 954                 ret = ip6gre_xmit_ipv6(skb, dev);
 955                 break;
 956         default:
 957                 ret = ip6gre_xmit_other(skb, dev);
 958                 break;
 959         }
 960
 961         if (ret < 0)
 962                 goto tx_err;
 963
 964         return NETDEV_TX_OK;
 965
 966 tx_err:
 967         stats->tx_errors++;
 968         stats->tx_dropped++;
 969         kfree_skb(skb);
 970         return NETDEV_TX_OK;
 971 }
 972
 973 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 974 {
 975         struct net_device *dev = t->dev;
 976         struct __ip6_tnl_parm *p = &t->parms;
 977         struct flowi6 *fl6 = &t->fl.u.ip6;
 978         int addend = sizeof(struct ipv6hdr) + 4;
 979
 980         if (dev->type != ARPHRD_ETHER) {
 981                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 982                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 983         }
 984
 985         /* Set up flowi template */
 986         fl6->saddr = p->laddr;
 987         fl6->daddr = p->raddr;
 988         fl6->flowi6_oif = p->link;
 989         fl6->flowlabel = 0;
 990
 991         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 992                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 993         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 994                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 995
 996         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 997         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 998
 999         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1000                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1001                 dev->flags |= IFF_POINTOPOINT;
1002         else
1003                 dev->flags &= ~IFF_POINTOPOINT;
1004
1005         dev->iflink = p->link;
1006
1007         /* Precalculate GRE options length */
1008         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
1009                 if (t->parms.o_flags&GRE_CSUM)
1010                         addend += 4;
1011                 if (t->parms.o_flags&GRE_KEY)
1012                         addend += 4;
1013                 if (t->parms.o_flags&GRE_SEQ)
1014                         addend += 4;
1015         }
1016
1017         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1018                 int strict = (ipv6_addr_type(&p->raddr) &
1019                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1020
1021                 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1022                                                  &p->raddr, &p->laddr,
1023                                                  p->link, strict);
1024
1025                 if (rt == NULL)
1026                         return;
1027
1028                 if (rt->dst.dev) {
1029                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
1030
1031                         if (set_mtu) {
1032                                 dev->mtu = rt->dst.dev->mtu - addend;
1033                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1034                                         dev->mtu -= 8;
1035
1036                                 if (dev->mtu < IPV6_MIN_MTU)
1037                                         dev->mtu = IPV6_MIN_MTU;
1038                         }
1039                 }
1040                 ip6_rt_put(rt);
1041         }
1042
1043         t->hlen = addend;
1044 }
1045
1046 static int ip6gre_tnl_change(struct ip6_tnl *t,
1047         const struct __ip6_tnl_parm *p, int set_mtu)
1048 {
1049         t->parms.laddr = p->laddr;
1050         t->parms.raddr = p->raddr;
1051         t->parms.flags = p->flags;
1052         t->parms.hop_limit = p->hop_limit;
1053         t->parms.encap_limit = p->encap_limit;
1054         t->parms.flowinfo = p->flowinfo;
1055         t->parms.link = p->link;
1056         t->parms.proto = p->proto;
1057         t->parms.i_key = p->i_key;
1058         t->parms.o_key = p->o_key;
1059         t->parms.i_flags = p->i_flags;
1060         t->parms.o_flags = p->o_flags;
1061         ip6_tnl_dst_reset(t);
1062         ip6gre_tnl_link_config(t, set_mtu);
1063         return 0;
1064 }
1065
1066 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1067         const struct ip6_tnl_parm2 *u)
1068 {
1069         p->laddr = u->laddr;
1070         p->raddr = u->raddr;
1071         p->flags = u->flags;
1072         p->hop_limit = u->hop_limit;
1073         p->encap_limit = u->encap_limit;
1074         p->flowinfo = u->flowinfo;
1075         p->link = u->link;
1076         p->i_key = u->i_key;
1077         p->o_key = u->o_key;
1078         p->i_flags = u->i_flags;
1079         p->o_flags = u->o_flags;
1080         memcpy(p->name, u->name, sizeof(u->name));
1081 }
1082
1083 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1084         const struct __ip6_tnl_parm *p)
1085 {
1086         u->proto = IPPROTO_GRE;
1087         u->laddr = p->laddr;
1088         u->raddr = p->raddr;
1089         u->flags = p->flags;
1090         u->hop_limit = p->hop_limit;
1091         u->encap_limit = p->encap_limit;
1092         u->flowinfo = p->flowinfo;
1093         u->link = p->link;
1094         u->i_key = p->i_key;
1095         u->o_key = p->o_key;
1096         u->i_flags = p->i_flags;
1097         u->o_flags = p->o_flags;
1098         memcpy(u->name, p->name, sizeof(u->name));
1099 }
1100
1101 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1102         struct ifreq *ifr, int cmd)
1103 {
1104         int err = 0;
1105         struct ip6_tnl_parm2 p;
1106         struct __ip6_tnl_parm p1;
1107         struct ip6_tnl *t;
1108         struct net *net = dev_net(dev);
1109         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1110
1111         switch (cmd) {
1112         case SIOCGETTUNNEL:
1113                 t = NULL;
1114                 if (dev == ign->fb_tunnel_dev) {
1115                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1116                                 err = -EFAULT;
1117                                 break;
1118                         }
1119                         ip6gre_tnl_parm_from_user(&p1, &p);
1120                         t = ip6gre_tunnel_locate(net, &p1, 0);
1121                 }
1122                 if (t == NULL)
1123                         t = netdev_priv(dev);
1124                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1125                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1126                         err = -EFAULT;
1127                 break;
1128
1129         case SIOCADDTUNNEL:
1130         case SIOCCHGTUNNEL:
1131                 err = -EPERM;
1132                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1133                         goto done;
1134
1135                 err = -EFAULT;
1136                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1137                         goto done;
1138
1139                 err = -EINVAL;
1140                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1141                         goto done;
1142
1143                 if (!(p.i_flags&GRE_KEY))
1144                         p.i_key = 0;
1145                 if (!(p.o_flags&GRE_KEY))
1146                         p.o_key = 0;
1147
1148                 ip6gre_tnl_parm_from_user(&p1, &p);
1149                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1150
1151                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1152                         if (t != NULL) {
1153                                 if (t->dev != dev) {
1154                                         err = -EEXIST;
1155                                         break;
1156                                 }
1157                         } else {
1158                                 t = netdev_priv(dev);
1159
1160                                 ip6gre_tunnel_unlink(ign, t);
1161                                 synchronize_net();
1162                                 ip6gre_tnl_change(t, &p1, 1);
1163                                 ip6gre_tunnel_link(ign, t);
1164                                 netdev_state_change(dev);
1165                         }
1166                 }
1167
1168                 if (t) {
1169                         err = 0;
1170
1171                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1172                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1173                                 err = -EFAULT;
1174                 } else
1175                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1176                 break;
1177
1178         case SIOCDELTUNNEL:
1179                 err = -EPERM;
1180                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1181                         goto done;
1182
1183                 if (dev == ign->fb_tunnel_dev) {
1184                         err = -EFAULT;
1185                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1186                                 goto done;
1187                         err = -ENOENT;
1188                         ip6gre_tnl_parm_from_user(&p1, &p);
1189                         t = ip6gre_tunnel_locate(net, &p1, 0);
1190                         if (t == NULL)
1191                                 goto done;
1192                         err = -EPERM;
1193                         if (t == netdev_priv(ign->fb_tunnel_dev))
1194                                 goto done;
1195                         dev = t->dev;
1196                 }
1197                 unregister_netdevice(dev);
1198                 err = 0;
1199                 break;
1200
1201         default:
1202                 err = -EINVAL;
1203         }
1204
1205 done:
1206         return err;
1207 }
1208
1209 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1210 {
1211         struct ip6_tnl *tunnel = netdev_priv(dev);
1212         if (new_mtu < 68 ||
1213             new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1214                 return -EINVAL;
1215         dev->mtu = new_mtu;
1216         return 0;
1217 }
1218
1219 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1220                         unsigned short type,
1221                         const void *daddr, const void *saddr, unsigned int len)
1222 {
1223         struct ip6_tnl *t = netdev_priv(dev);
1224         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1225         __be16 *p = (__be16 *)(ipv6h+1);
1226
1227         ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
1228         ipv6h->hop_limit = t->parms.hop_limit;
1229         ipv6h->nexthdr = NEXTHDR_GRE;
1230         ipv6h->saddr = t->parms.laddr;
1231         ipv6h->daddr = t->parms.raddr;
1232
1233         p[0]            = t->parms.o_flags;
1234         p[1]            = htons(type);
1235
1236         /*
1237          *      Set the source hardware address.
1238          */
1239
1240         if (saddr)
1241                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1242         if (daddr)
1243                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1244         if (!ipv6_addr_any(&ipv6h->daddr))
1245                 return t->hlen;
1246
1247         return -t->hlen;
1248 }
1249
1250 static const struct header_ops ip6gre_header_ops = {
1251         .create = ip6gre_header,
1252 };
1253
1254 static const struct net_device_ops ip6gre_netdev_ops = {
1255         .ndo_init               = ip6gre_tunnel_init,
1256         .ndo_uninit             = ip6gre_tunnel_uninit,
1257         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1258         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1259         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1260         .ndo_get_stats64        = ip6gre_get_stats64,
1261 };
1262
1263 static void ip6gre_dev_free(struct net_device *dev)
1264 {
1265         free_percpu(dev->tstats);
1266         free_netdev(dev);
1267 }
1268
1269 static void ip6gre_tunnel_setup(struct net_device *dev)
1270 {
1271         struct ip6_tnl *t;
1272
1273         dev->netdev_ops = &ip6gre_netdev_ops;
1274         dev->destructor = ip6gre_dev_free;
1275
1276         dev->type = ARPHRD_IP6GRE;
1277         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1278         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1279         t = netdev_priv(dev);
1280         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1281                 dev->mtu -= 8;
1282         dev->flags |= IFF_NOARP;
1283         dev->iflink = 0;
1284         dev->addr_len = sizeof(struct in6_addr);
1285         dev->features |= NETIF_F_NETNS_LOCAL;
1286         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1287 }
1288
1289 static int ip6gre_tunnel_init(struct net_device *dev)
1290 {
1291         struct ip6_tnl *tunnel;
1292
1293         tunnel = netdev_priv(dev);
1294
1295         tunnel->dev = dev;
1296         strcpy(tunnel->parms.name, dev->name);
1297
1298         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1299         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1300
1301         if (ipv6_addr_any(&tunnel->parms.raddr))
1302                 dev->header_ops = &ip6gre_header_ops;
1303
1304         dev->tstats = alloc_percpu(struct pcpu_tstats);
1305         if (!dev->tstats)
1306                 return -ENOMEM;
1307
1308         return 0;
1309 }
1310
1311 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1312 {
1313         struct ip6_tnl *tunnel = netdev_priv(dev);
1314
1315         tunnel->dev = dev;
1316         strcpy(tunnel->parms.name, dev->name);
1317
1318         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1319
1320         dev_hold(dev);
1321 }
1322
1323
1324 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1325         .handler     = ip6gre_rcv,
1326         .err_handler = ip6gre_err,
1327         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1328 };
1329
1330 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1331         struct list_head *head)
1332 {
1333         int prio;
1334
1335         for (prio = 0; prio < 4; prio++) {
1336                 int h;
1337                 for (h = 0; h < HASH_SIZE; h++) {
1338                         struct ip6_tnl *t;
1339
1340                         t = rtnl_dereference(ign->tunnels[prio][h]);
1341
1342                         while (t != NULL) {
1343                                 unregister_netdevice_queue(t->dev, head);
1344                                 t = rtnl_dereference(t->next);
1345                         }
1346                 }
1347         }
1348 }
1349
1350 static int __net_init ip6gre_init_net(struct net *net)
1351 {
1352         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1353         int err;
1354
1355         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1356                                            ip6gre_tunnel_setup);
1357         if (!ign->fb_tunnel_dev) {
1358                 err = -ENOMEM;
1359                 goto err_alloc_dev;
1360         }
1361         dev_net_set(ign->fb_tunnel_dev, net);
1362
1363         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1364         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1365
1366         err = register_netdev(ign->fb_tunnel_dev);
1367         if (err)
1368                 goto err_reg_dev;
1369
1370         rcu_assign_pointer(ign->tunnels_wc[0],
1371                            netdev_priv(ign->fb_tunnel_dev));
1372         return 0;
1373
1374 err_reg_dev:
1375         ip6gre_dev_free(ign->fb_tunnel_dev);
1376 err_alloc_dev:
1377         return err;
1378 }
1379
1380 static void __net_exit ip6gre_exit_net(struct net *net)
1381 {
1382         struct ip6gre_net *ign;
1383         LIST_HEAD(list);
1384
1385         ign = net_generic(net, ip6gre_net_id);
1386         rtnl_lock();
1387         ip6gre_destroy_tunnels(ign, &list);
1388         unregister_netdevice_many(&list);
1389         rtnl_unlock();
1390 }
1391
1392 static struct pernet_operations ip6gre_net_ops = {
1393         .init = ip6gre_init_net,
1394         .exit = ip6gre_exit_net,
1395         .id   = &ip6gre_net_id,
1396         .size = sizeof(struct ip6gre_net),
1397 };
1398
1399 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1400 {
1401         __be16 flags;
1402
1403         if (!data)
1404                 return 0;
1405
1406         flags = 0;
1407         if (data[IFLA_GRE_IFLAGS])
1408                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1409         if (data[IFLA_GRE_OFLAGS])
1410                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1411         if (flags & (GRE_VERSION|GRE_ROUTING))
1412                 return -EINVAL;
1413
1414         return 0;
1415 }
1416
1417 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1418 {
1419         struct in6_addr daddr;
1420
1421         if (tb[IFLA_ADDRESS]) {
1422                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1423                         return -EINVAL;
1424                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1425                         return -EADDRNOTAVAIL;
1426         }
1427
1428         if (!data)
1429                 goto out;
1430
1431         if (data[IFLA_GRE_REMOTE]) {
1432                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1433                 if (ipv6_addr_any(&daddr))
1434                         return -EINVAL;
1435         }
1436
1437 out:
1438         return ip6gre_tunnel_validate(tb, data);
1439 }
1440
1441
1442 static void ip6gre_netlink_parms(struct nlattr *data[],
1443                                 struct __ip6_tnl_parm *parms)
1444 {
1445         memset(parms, 0, sizeof(*parms));
1446
1447         if (!data)
1448                 return;
1449
1450         if (data[IFLA_GRE_LINK])
1451                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1452
1453         if (data[IFLA_GRE_IFLAGS])
1454                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1455
1456         if (data[IFLA_GRE_OFLAGS])
1457                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1458
1459         if (data[IFLA_GRE_IKEY])
1460                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1461
1462         if (data[IFLA_GRE_OKEY])
1463                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1464
1465         if (data[IFLA_GRE_LOCAL])
1466                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1467
1468         if (data[IFLA_GRE_REMOTE])
1469                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1470
1471         if (data[IFLA_GRE_TTL])
1472                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1473
1474         if (data[IFLA_GRE_ENCAP_LIMIT])
1475                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1476
1477         if (data[IFLA_GRE_FLOWINFO])
1478                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1479
1480         if (data[IFLA_GRE_FLAGS])
1481                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1482 }
1483
1484 static int ip6gre_tap_init(struct net_device *dev)
1485 {
1486         struct ip6_tnl *tunnel;
1487
1488         tunnel = netdev_priv(dev);
1489
1490         tunnel->dev = dev;
1491         strcpy(tunnel->parms.name, dev->name);
1492
1493         ip6gre_tnl_link_config(tunnel, 1);
1494
1495         dev->tstats = alloc_percpu(struct pcpu_tstats);
1496         if (!dev->tstats)
1497                 return -ENOMEM;
1498
1499         return 0;
1500 }
1501
1502 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1503         .ndo_init = ip6gre_tap_init,
1504         .ndo_uninit = ip6gre_tunnel_uninit,
1505         .ndo_start_xmit = ip6gre_tunnel_xmit,
1506         .ndo_set_mac_address = eth_mac_addr,
1507         .ndo_validate_addr = eth_validate_addr,
1508         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1509         .ndo_get_stats64 = ip6gre_get_stats64,
1510 };
1511
1512 static void ip6gre_tap_setup(struct net_device *dev)
1513 {
1514
1515         ether_setup(dev);
1516
1517         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1518         dev->destructor = ip6gre_dev_free;
1519
1520         dev->iflink = 0;
1521         dev->features |= NETIF_F_NETNS_LOCAL;
1522 }
1523
1524 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1525         struct nlattr *tb[], struct nlattr *data[])
1526 {
1527         struct ip6_tnl *nt;
1528         struct net *net = dev_net(dev);
1529         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1530         int err;
1531
1532         nt = netdev_priv(dev);
1533         ip6gre_netlink_parms(data, &nt->parms);
1534
1535         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1536                 return -EEXIST;
1537
1538         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1539                 eth_hw_addr_random(dev);
1540
1541         nt->dev = dev;
1542         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1543
1544         /* Can use a lockless transmit, unless we generate output sequences */
1545         if (!(nt->parms.o_flags & GRE_SEQ))
1546                 dev->features |= NETIF_F_LLTX;
1547
1548         err = register_netdevice(dev);
1549         if (err)
1550                 goto out;
1551
1552         dev_hold(dev);
1553         ip6gre_tunnel_link(ign, nt);
1554
1555 out:
1556         return err;
1557 }
1558
1559 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1560                             struct nlattr *data[])
1561 {
1562         struct ip6_tnl *t, *nt;
1563         struct net *net = dev_net(dev);
1564         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1565         struct __ip6_tnl_parm p;
1566
1567         if (dev == ign->fb_tunnel_dev)
1568                 return -EINVAL;
1569
1570         nt = netdev_priv(dev);
1571         ip6gre_netlink_parms(data, &p);
1572
1573         t = ip6gre_tunnel_locate(net, &p, 0);
1574
1575         if (t) {
1576                 if (t->dev != dev)
1577                         return -EEXIST;
1578         } else {
1579                 t = nt;
1580
1581                 ip6gre_tunnel_unlink(ign, t);
1582                 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1583                 ip6gre_tunnel_link(ign, t);
1584                 netdev_state_change(dev);
1585         }
1586
1587         return 0;
1588 }
1589
1590 static size_t ip6gre_get_size(const struct net_device *dev)
1591 {
1592         return
1593                 /* IFLA_GRE_LINK */
1594                 nla_total_size(4) +
1595                 /* IFLA_GRE_IFLAGS */
1596                 nla_total_size(2) +
1597                 /* IFLA_GRE_OFLAGS */
1598                 nla_total_size(2) +
1599                 /* IFLA_GRE_IKEY */
1600                 nla_total_size(4) +
1601                 /* IFLA_GRE_OKEY */
1602                 nla_total_size(4) +
1603                 /* IFLA_GRE_LOCAL */
1604                 nla_total_size(sizeof(struct in6_addr)) +
1605                 /* IFLA_GRE_REMOTE */
1606                 nla_total_size(sizeof(struct in6_addr)) +
1607                 /* IFLA_GRE_TTL */
1608                 nla_total_size(1) +
1609                 /* IFLA_GRE_TOS */
1610                 nla_total_size(1) +
1611                 /* IFLA_GRE_ENCAP_LIMIT */
1612                 nla_total_size(1) +
1613                 /* IFLA_GRE_FLOWINFO */
1614                 nla_total_size(4) +
1615                 /* IFLA_GRE_FLAGS */
1616                 nla_total_size(4) +
1617                 0;
1618 }
1619
1620 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1621 {
1622         struct ip6_tnl *t = netdev_priv(dev);
1623         struct __ip6_tnl_parm *p = &t->parms;
1624
1625         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1626             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1627             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1628             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1629             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1630             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1631             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1632             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1633             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1634             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1635             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1636             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1637                 goto nla_put_failure;
1638         return 0;
1639
1640 nla_put_failure:
1641         return -EMSGSIZE;
1642 }
1643
1644 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1645         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1646         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1647         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1648         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1649         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1650         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1651         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1652         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1653         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1654         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1655         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1656 };
1657
1658 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1659         .kind           = "ip6gre",
1660         .maxtype        = IFLA_GRE_MAX,
1661         .policy         = ip6gre_policy,
1662         .priv_size      = sizeof(struct ip6_tnl),
1663         .setup          = ip6gre_tunnel_setup,
1664         .validate       = ip6gre_tunnel_validate,
1665         .newlink        = ip6gre_newlink,
1666         .changelink     = ip6gre_changelink,
1667         .get_size       = ip6gre_get_size,
1668         .fill_info      = ip6gre_fill_info,
1669 };
1670
1671 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1672         .kind           = "ip6gretap",
1673         .maxtype        = IFLA_GRE_MAX,
1674         .policy         = ip6gre_policy,
1675         .priv_size      = sizeof(struct ip6_tnl),
1676         .setup          = ip6gre_tap_setup,
1677         .validate       = ip6gre_tap_validate,
1678         .newlink        = ip6gre_newlink,
1679         .changelink     = ip6gre_changelink,
1680         .get_size       = ip6gre_get_size,
1681         .fill_info      = ip6gre_fill_info,
1682 };
1683
1684 /*
1685  *      And now the modules code and kernel interface.
1686  */
1687
1688 static int __init ip6gre_init(void)
1689 {
1690         int err;
1691
1692         pr_info("GRE over IPv6 tunneling driver\n");
1693
1694         err = register_pernet_device(&ip6gre_net_ops);
1695         if (err < 0)
1696                 return err;
1697
1698         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1699         if (err < 0) {
1700                 pr_info("%s: can't add protocol\n", __func__);
1701                 goto add_proto_failed;
1702         }
1703
1704         err = rtnl_link_register(&ip6gre_link_ops);
1705         if (err < 0)
1706                 goto rtnl_link_failed;
1707
1708         err = rtnl_link_register(&ip6gre_tap_ops);
1709         if (err < 0)
1710                 goto tap_ops_failed;
1711
1712 out:
1713         return err;
1714
1715 tap_ops_failed:
1716         rtnl_link_unregister(&ip6gre_link_ops);
1717 rtnl_link_failed:
1718         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1719 add_proto_failed:
1720         unregister_pernet_device(&ip6gre_net_ops);
1721         goto out;
1722 }
1723
1724 static void __exit ip6gre_fini(void)
1725 {
1726         rtnl_link_unregister(&ip6gre_tap_ops);
1727         rtnl_link_unregister(&ip6gre_link_ops);
1728         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1729         unregister_pernet_device(&ip6gre_net_ops);
1730 }
1731
1732 module_init(ip6gre_init);
1733 module_exit(ip6gre_fini);
1734 MODULE_LICENSE("GPL");
1735 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1736 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1737 MODULE_ALIAS_RTNL_LINK("ip6gre");
1738 MODULE_ALIAS_NETDEV("ip6gre0");