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/init.h>
  28 #include <linux/in6.h>
  29 #include <linux/inetdevice.h>
  30 #include <linux/igmp.h>
  31 #include <linux/netfilter_ipv4.h>
  32 #include <linux/etherdevice.h>
  33 #include <linux/if_ether.h>
  34 #include <linux/hash.h>
  35 #include <linux/if_tunnel.h>
  36 #include <linux/ip6_tunnel.h>
  37
  38 #include <net/sock.h>
  39 #include <net/ip.h>
  40 #include <net/ip_tunnels.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 static bool log_ecn_error = true;
  60 module_param(log_ecn_error, bool, 0644);
  61 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  62
  63 #define HASH_SIZE_SHIFT  5
  64 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  65
  66 static int ip6gre_net_id __read_mostly;
  67 struct ip6gre_net {
  68         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
  69
  70         struct net_device *fb_tunnel_dev;
  71 };
  72
  73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
  75 static int ip6gre_tunnel_init(struct net_device *dev);
  76 static void ip6gre_tunnel_setup(struct net_device *dev);
  77 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  78 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  79
  80 /* Tunnel hash table */
  81
  82 /*
  83    4 hash tables:
  84
  85    3: (remote,local)
  86    2: (remote,*)
  87    1: (*,local)
  88    0: (*,*)
  89
  90    We require exact key match i.e. if a key is present in packet
  91    it will match only tunnel with the same key; if it is not present,
  92    it will match only keyless tunnel.
  93
  94    All keysless packets, if not matched configured keyless tunnels
  95    will match fallback tunnel.
  96  */
  97
  98 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
  99 static u32 HASH_ADDR(const struct in6_addr *addr)
 100 {
 101         u32 hash = ipv6_addr_hash(addr);
 102
 103         return hash_32(hash, HASH_SIZE_SHIFT);
 104 }
 105
 106 #define tunnels_r_l     tunnels[3]
 107 #define tunnels_r       tunnels[2]
 108 #define tunnels_l       tunnels[1]
 109 #define tunnels_wc      tunnels[0]
 110
 111 /* Given src, dst and key, find appropriate for input tunnel. */
 112
 113 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 114                 const struct in6_addr *remote, const struct in6_addr *local,
 115                 __be32 key, __be16 gre_proto)
 116 {
 117         struct net *net = dev_net(dev);
 118         int link = dev->ifindex;
 119         unsigned int h0 = HASH_ADDR(remote);
 120         unsigned int h1 = HASH_KEY(key);
 121         struct ip6_tnl *t, *cand = NULL;
 122         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 123         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
 124                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 125         int score, cand_score = 4;
 126
 127         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 128                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 129                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 130                     key != t->parms.i_key ||
 131                     !(t->dev->flags & IFF_UP))
 132                         continue;
 133
 134                 if (t->dev->type != ARPHRD_IP6GRE &&
 135                     t->dev->type != dev_type)
 136                         continue;
 137
 138                 score = 0;
 139                 if (t->parms.link != link)
 140                         score |= 1;
 141                 if (t->dev->type != dev_type)
 142                         score |= 2;
 143                 if (score == 0)
 144                         return t;
 145
 146                 if (score < cand_score) {
 147                         cand = t;
 148                         cand_score = score;
 149                 }
 150         }
 151
 152         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 153                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 154                     key != t->parms.i_key ||
 155                     !(t->dev->flags & IFF_UP))
 156                         continue;
 157
 158                 if (t->dev->type != ARPHRD_IP6GRE &&
 159                     t->dev->type != dev_type)
 160                         continue;
 161
 162                 score = 0;
 163                 if (t->parms.link != link)
 164                         score |= 1;
 165                 if (t->dev->type != dev_type)
 166                         score |= 2;
 167                 if (score == 0)
 168                         return t;
 169
 170                 if (score < cand_score) {
 171                         cand = t;
 172                         cand_score = score;
 173                 }
 174         }
 175
 176         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 177                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 178                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 179                                  !ipv6_addr_is_multicast(local))) ||
 180                     key != t->parms.i_key ||
 181                     !(t->dev->flags & IFF_UP))
 182                         continue;
 183
 184                 if (t->dev->type != ARPHRD_IP6GRE &&
 185                     t->dev->type != dev_type)
 186                         continue;
 187
 188                 score = 0;
 189                 if (t->parms.link != link)
 190                         score |= 1;
 191                 if (t->dev->type != dev_type)
 192                         score |= 2;
 193                 if (score == 0)
 194                         return t;
 195
 196                 if (score < cand_score) {
 197                         cand = t;
 198                         cand_score = score;
 199                 }
 200         }
 201
 202         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 203                 if (t->parms.i_key != key ||
 204                     !(t->dev->flags & IFF_UP))
 205                         continue;
 206
 207                 if (t->dev->type != ARPHRD_IP6GRE &&
 208                     t->dev->type != dev_type)
 209                         continue;
 210
 211                 score = 0;
 212                 if (t->parms.link != link)
 213                         score |= 1;
 214                 if (t->dev->type != dev_type)
 215                         score |= 2;
 216                 if (score == 0)
 217                         return t;
 218
 219                 if (score < cand_score) {
 220                         cand = t;
 221                         cand_score = score;
 222                 }
 223         }
 224
 225         if (cand)
 226                 return cand;
 227
 228         dev = ign->fb_tunnel_dev;
 229         if (dev->flags & IFF_UP)
 230                 return netdev_priv(dev);
 231
 232         return NULL;
 233 }
 234
 235 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 236                 const struct __ip6_tnl_parm *p)
 237 {
 238         const struct in6_addr *remote = &p->raddr;
 239         const struct in6_addr *local = &p->laddr;
 240         unsigned int h = HASH_KEY(p->i_key);
 241         int prio = 0;
 242
 243         if (!ipv6_addr_any(local))
 244                 prio |= 1;
 245         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 246                 prio |= 2;
 247                 h ^= HASH_ADDR(remote);
 248         }
 249
 250         return &ign->tunnels[prio][h];
 251 }
 252
 253 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 254                 const struct ip6_tnl *t)
 255 {
 256         return __ip6gre_bucket(ign, &t->parms);
 257 }
 258
 259 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 260 {
 261         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 262
 263         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 264         rcu_assign_pointer(*tp, t);
 265 }
 266
 267 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 268 {
 269         struct ip6_tnl __rcu **tp;
 270         struct ip6_tnl *iter;
 271
 272         for (tp = ip6gre_bucket(ign, t);
 273              (iter = rtnl_dereference(*tp)) != NULL;
 274              tp = &iter->next) {
 275                 if (t == iter) {
 276                         rcu_assign_pointer(*tp, t->next);
 277                         break;
 278                 }
 279         }
 280 }
 281
 282 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 283                                            const struct __ip6_tnl_parm *parms,
 284                                            int type)
 285 {
 286         const struct in6_addr *remote = &parms->raddr;
 287         const struct in6_addr *local = &parms->laddr;
 288         __be32 key = parms->i_key;
 289         int link = parms->link;
 290         struct ip6_tnl *t;
 291         struct ip6_tnl __rcu **tp;
 292         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 293
 294         for (tp = __ip6gre_bucket(ign, parms);
 295              (t = rtnl_dereference(*tp)) != NULL;
 296              tp = &t->next)
 297                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 298                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 299                     key == t->parms.i_key &&
 300                     link == t->parms.link &&
 301                     type == t->dev->type)
 302                         break;
 303
 304         return t;
 305 }
 306
 307 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 308                 const struct __ip6_tnl_parm *parms, int create)
 309 {
 310         struct ip6_tnl *t, *nt;
 311         struct net_device *dev;
 312         char name[IFNAMSIZ];
 313         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 314
 315         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 316         if (t && create)
 317                 return NULL;
 318         if (t || !create)
 319                 return t;
 320
 321         if (parms->name[0])
 322                 strlcpy(name, parms->name, IFNAMSIZ);
 323         else
 324                 strcpy(name, "ip6gre%d");
 325
 326         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 327                            ip6gre_tunnel_setup);
 328         if (!dev)
 329                 return NULL;
 330
 331         dev_net_set(dev, net);
 332
 333         nt = netdev_priv(dev);
 334         nt->parms = *parms;
 335         dev->rtnl_link_ops = &ip6gre_link_ops;
 336
 337         nt->dev = dev;
 338         nt->net = dev_net(dev);
 339         ip6gre_tnl_link_config(nt, 1);
 340
 341         if (register_netdevice(dev) < 0)
 342                 goto failed_free;
 343
 344         /* Can use a lockless transmit, unless we generate output sequences */
 345         if (!(nt->parms.o_flags & GRE_SEQ))
 346                 dev->features |= NETIF_F_LLTX;
 347
 348         dev_hold(dev);
 349         ip6gre_tunnel_link(ign, nt);
 350         return nt;
 351
 352 failed_free:
 353         free_netdev(dev);
 354         return NULL;
 355 }
 356
 357 static void ip6gre_tunnel_uninit(struct net_device *dev)
 358 {
 359         struct ip6_tnl *t = netdev_priv(dev);
 360         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 361
 362         ip6gre_tunnel_unlink(ign, t);
 363         dst_cache_reset(&t->dst_cache);
 364         dev_put(dev);
 365 }
 366
 367
 368 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 369                 u8 type, u8 code, int offset, __be32 info)
 370 {
 371         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 372         __be16 *p = (__be16 *)(skb->data + offset);
 373         int grehlen = offset + 4;
 374         struct ip6_tnl *t;
 375         __be16 flags;
 376
 377         flags = p[0];
 378         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
 379                 if (flags&(GRE_VERSION|GRE_ROUTING))
 380                         return;
 381                 if (flags&GRE_KEY) {
 382                         grehlen += 4;
 383                         if (flags&GRE_CSUM)
 384                                 grehlen += 4;
 385                 }
 386         }
 387
 388         /* If only 8 bytes returned, keyed message will be dropped here */
 389         if (!pskb_may_pull(skb, grehlen))
 390                 return;
 391         ipv6h = (const struct ipv6hdr *)skb->data;
 392         p = (__be16 *)(skb->data + offset);
 393
 394         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 395                                 flags & GRE_KEY ?
 396                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
 397                                 p[1]);
 398         if (!t)
 399                 return;
 400
 401         switch (type) {
 402                 __u32 teli;
 403                 struct ipv6_tlv_tnl_enc_lim *tel;
 404                 __u32 mtu;
 405         case ICMPV6_DEST_UNREACH:
 406                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 407                                     t->parms.name);
 408                 break;
 409         case ICMPV6_TIME_EXCEED:
 410                 if (code == ICMPV6_EXC_HOPLIMIT) {
 411                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 412                                             t->parms.name);
 413                 }
 414                 break;
 415         case ICMPV6_PARAMPROB:
 416                 teli = 0;
 417                 if (code == ICMPV6_HDR_FIELD)
 418                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 419
 420                 if (teli && teli == be32_to_cpu(info) - 2) {
 421                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 422                         if (tel->encap_limit == 0) {
 423                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 424                                                     t->parms.name);
 425                         }
 426                 } else {
 427                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 428                                             t->parms.name);
 429                 }
 430                 break;
 431         case ICMPV6_PKT_TOOBIG:
 432                 mtu = be32_to_cpu(info) - offset;
 433                 if (mtu < IPV6_MIN_MTU)
 434                         mtu = IPV6_MIN_MTU;
 435                 t->dev->mtu = mtu;
 436                 break;
 437         }
 438
 439         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 440                 t->err_count++;
 441         else
 442                 t->err_count = 1;
 443         t->err_time = jiffies;
 444 }
 445
 446 static int ip6gre_rcv(struct sk_buff *skb)
 447 {
 448         const struct ipv6hdr *ipv6h;
 449         u8     *h;
 450         __be16    flags;
 451         __sum16   csum = 0;
 452         __be32 key = 0;
 453         u32    seqno = 0;
 454         struct ip6_tnl *tunnel;
 455         int    offset = 4;
 456         __be16 gre_proto;
 457         int err;
 458
 459         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
 460                 goto drop;
 461
 462         ipv6h = ipv6_hdr(skb);
 463         h = skb->data;
 464         flags = *(__be16 *)h;
 465
 466         if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
 467                 /* - Version must be 0.
 468                    - We do not support routing headers.
 469                  */
 470                 if (flags&(GRE_VERSION|GRE_ROUTING))
 471                         goto drop;
 472
 473                 if (flags&GRE_CSUM) {
 474                         csum = skb_checksum_simple_validate(skb);
 475                         offset += 4;
 476                 }
 477                 if (flags&GRE_KEY) {
 478                         key = *(__be32 *)(h + offset);
 479                         offset += 4;
 480                 }
 481                 if (flags&GRE_SEQ) {
 482                         seqno = ntohl(*(__be32 *)(h + offset));
 483                         offset += 4;
 484                 }
 485         }
 486
 487         gre_proto = *(__be16 *)(h + 2);
 488
 489         tunnel = ip6gre_tunnel_lookup(skb->dev,
 490                                           &ipv6h->saddr, &ipv6h->daddr, key,
 491                                           gre_proto);
 492         if (tunnel) {
 493                 struct pcpu_sw_netstats *tstats;
 494
 495                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 496                         goto drop;
 497
 498                 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
 499                         tunnel->dev->stats.rx_dropped++;
 500                         goto drop;
 501                 }
 502
 503                 skb->protocol = gre_proto;
 504                 /* WCCP version 1 and 2 protocol decoding.
 505                  * - Change protocol to IPv6
 506                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
 507                  */
 508                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
 509                         skb->protocol = htons(ETH_P_IPV6);
 510                         if ((*(h + offset) & 0xF0) != 0x40)
 511                                 offset += 4;
 512                 }
 513
 514                 skb->mac_header = skb->network_header;
 515                 __pskb_pull(skb, offset);
 516                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
 517
 518                 if (((flags&GRE_CSUM) && csum) ||
 519                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
 520                         tunnel->dev->stats.rx_crc_errors++;
 521                         tunnel->dev->stats.rx_errors++;
 522                         goto drop;
 523                 }
 524                 if (tunnel->parms.i_flags&GRE_SEQ) {
 525                         if (!(flags&GRE_SEQ) ||
 526                             (tunnel->i_seqno &&
 527                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
 528                                 tunnel->dev->stats.rx_fifo_errors++;
 529                                 tunnel->dev->stats.rx_errors++;
 530                                 goto drop;
 531                         }
 532                         tunnel->i_seqno = seqno + 1;
 533                 }
 534
 535                 /* Warning: All skb pointers will be invalidated! */
 536                 if (tunnel->dev->type == ARPHRD_ETHER) {
 537                         if (!pskb_may_pull(skb, ETH_HLEN)) {
 538                                 tunnel->dev->stats.rx_length_errors++;
 539                                 tunnel->dev->stats.rx_errors++;
 540                                 goto drop;
 541                         }
 542
 543                         ipv6h = ipv6_hdr(skb);
 544                         skb->protocol = eth_type_trans(skb, tunnel->dev);
 545                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 546                 }
 547
 548                 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 549
 550                 skb_reset_network_header(skb);
 551
 552                 err = IP6_ECN_decapsulate(ipv6h, skb);
 553                 if (unlikely(err)) {
 554                         if (log_ecn_error)
 555                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 556                                                      &ipv6h->saddr,
 557                                                      ipv6_get_dsfield(ipv6h));
 558                         if (err > 1) {
 559                                 ++tunnel->dev->stats.rx_frame_errors;
 560                                 ++tunnel->dev->stats.rx_errors;
 561                                 goto drop;
 562                         }
 563                 }
 564
 565                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 566                 u64_stats_update_begin(&tstats->syncp);
 567                 tstats->rx_packets++;
 568                 tstats->rx_bytes += skb->len;
 569                 u64_stats_update_end(&tstats->syncp);
 570
 571                 netif_rx(skb);
 572
 573                 return 0;
 574         }
 575         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 576
 577 drop:
 578         kfree_skb(skb);
 579         return 0;
 580 }
 581
 582 struct ipv6_tel_txoption {
 583         struct ipv6_txoptions ops;
 584         __u8 dst_opt[8];
 585 };
 586
 587 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 588 {
 589         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 590
 591         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 592         opt->dst_opt[3] = 1;
 593         opt->dst_opt[4] = encap_limit;
 594         opt->dst_opt[5] = IPV6_TLV_PADN;
 595         opt->dst_opt[6] = 1;
 596
 597         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 598         opt->ops.opt_nflen = 8;
 599 }
 600
 601 static __sum16 gre6_checksum(struct sk_buff *skb)
 602 {
 603         __wsum csum;
 604
 605         if (skb->ip_summed == CHECKSUM_PARTIAL)
 606                 csum = lco_csum(skb);
 607         else
 608                 csum = skb_checksum(skb, sizeof(struct ipv6hdr),
 609                                     skb->len - sizeof(struct ipv6hdr), 0);
 610         return csum_fold(csum);
 611 }
 612
 613 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 614                          struct net_device *dev,
 615                          __u8 dsfield,
 616                          struct flowi6 *fl6,
 617                          int encap_limit,
 618                          __u32 *pmtu)
 619 {
 620         struct ip6_tnl *tunnel = netdev_priv(dev);
 621         struct net *net = tunnel->net;
 622         struct net_device *tdev;    /* Device to other host */
 623         struct ipv6hdr  *ipv6h;     /* Our new IP header */
 624         unsigned int min_headroom = 0; /* The extra header space needed */
 625         int    gre_hlen;
 626         struct ipv6_tel_txoption opt;
 627         int    mtu;
 628         struct dst_entry *dst = NULL, *ndst = NULL;
 629         struct net_device_stats *stats = &tunnel->dev->stats;
 630         int err = -1;
 631         u8 proto;
 632         __be16 protocol;
 633
 634         if (dev->type == ARPHRD_ETHER)
 635                 IPCB(skb)->flags = 0;
 636
 637         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 638                 gre_hlen = 0;
 639                 ipv6h = (struct ipv6hdr *)skb->data;
 640                 fl6->daddr = ipv6h->daddr;
 641         } else {
 642                 gre_hlen = tunnel->hlen;
 643                 fl6->daddr = tunnel->parms.raddr;
 644         }
 645
 646         if (!fl6->flowi6_mark)
 647                 dst = dst_cache_get(&tunnel->dst_cache);
 648
 649         if (!dst) {
 650                 dst = ip6_route_output(net, NULL, fl6);
 651
 652                 if (dst->error)
 653                         goto tx_err_link_failure;
 654                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
 655                 if (IS_ERR(dst)) {
 656                         err = PTR_ERR(dst);
 657                         dst = NULL;
 658                         goto tx_err_link_failure;
 659                 }
 660                 ndst = dst;
 661         }
 662
 663         tdev = dst->dev;
 664
 665         if (tdev == dev) {
 666                 stats->collisions++;
 667                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 668                                      tunnel->parms.name);
 669                 goto tx_err_dst_release;
 670         }
 671
 672         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 673         if (encap_limit >= 0) {
 674                 min_headroom += 8;
 675                 mtu -= 8;
 676         }
 677         if (mtu < IPV6_MIN_MTU)
 678                 mtu = IPV6_MIN_MTU;
 679         if (skb_dst(skb))
 680                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 681         if (skb->len > mtu && !skb_is_gso(skb)) {
 682                 *pmtu = mtu;
 683                 err = -EMSGSIZE;
 684                 goto tx_err_dst_release;
 685         }
 686
 687         if (tunnel->err_count > 0) {
 688                 if (time_before(jiffies,
 689                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 690                         tunnel->err_count--;
 691
 692                         dst_link_failure(skb);
 693                 } else
 694                         tunnel->err_count = 0;
 695         }
 696
 697         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
 698
 699         min_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 700
 701         if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
 702                 int head_delta = SKB_DATA_ALIGN(min_headroom -
 703                                                 skb_headroom(skb) +
 704                                                 16);
 705
 706                 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
 707                                        0, GFP_ATOMIC);
 708                 if (min_headroom > dev->needed_headroom)
 709                         dev->needed_headroom = min_headroom;
 710                 if (unlikely(err))
 711                         goto tx_err_dst_release;
 712         }
 713
 714         if (!fl6->flowi6_mark && ndst)
 715                 dst_cache_set_ip6(&tunnel->dst_cache, ndst, &fl6->saddr);
 716         skb_dst_set(skb, dst);
 717
 718         proto = NEXTHDR_GRE;
 719         if (encap_limit >= 0) {
 720                 init_tel_txopt(&opt, encap_limit);
 721                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 722         }
 723
 724         err = iptunnel_handle_offloads(skb,
 725                                        (tunnel->parms.o_flags & GRE_CSUM) ?
 726                                        SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 727         if (err)
 728                 goto tx_err_dst_release;
 729
 730         skb_push(skb, gre_hlen);
 731         skb_reset_network_header(skb);
 732         skb_set_transport_header(skb, sizeof(*ipv6h));
 733
 734         /*
 735          *      Push down and install the IP header.
 736          */
 737         ipv6h = ipv6_hdr(skb);
 738         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
 739                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
 740         ipv6h->hop_limit = tunnel->parms.hop_limit;
 741         ipv6h->nexthdr = proto;
 742         ipv6h->saddr = fl6->saddr;
 743         ipv6h->daddr = fl6->daddr;
 744
 745         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 746         protocol = (dev->type == ARPHRD_ETHER) ?
 747                     htons(ETH_P_TEB) : skb->protocol;
 748         ((__be16 *)(ipv6h + 1))[1] = protocol;
 749
 750         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 751                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 752
 753                 if (tunnel->parms.o_flags&GRE_SEQ) {
 754                         ++tunnel->o_seqno;
 755                         *ptr = htonl(tunnel->o_seqno);
 756                         ptr--;
 757                 }
 758                 if (tunnel->parms.o_flags&GRE_KEY) {
 759                         *ptr = tunnel->parms.o_key;
 760                         ptr--;
 761                 }
 762                 if ((tunnel->parms.o_flags & GRE_CSUM) &&
 763                     !(skb_shinfo(skb)->gso_type &
 764                       (SKB_GSO_GRE | SKB_GSO_GRE_CSUM))) {
 765                         *ptr = 0;
 766                         *(__sum16 *)ptr = gre6_checksum(skb);
 767                 }
 768         }
 769
 770         skb_set_inner_protocol(skb, protocol);
 771
 772         ip6tunnel_xmit(NULL, skb, dev);
 773         return 0;
 774 tx_err_link_failure:
 775         stats->tx_carrier_errors++;
 776         dst_link_failure(skb);
 777 tx_err_dst_release:
 778         dst_release(dst);
 779         return err;
 780 }
 781
 782 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 783 {
 784         struct ip6_tnl *t = netdev_priv(dev);
 785         const struct iphdr  *iph = ip_hdr(skb);
 786         int encap_limit = -1;
 787         struct flowi6 fl6;
 788         __u8 dsfield;
 789         __u32 mtu;
 790         int err;
 791
 792         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 793
 794         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 795                 encap_limit = t->parms.encap_limit;
 796
 797         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 798         fl6.flowi6_proto = IPPROTO_GRE;
 799
 800         dsfield = ipv4_get_dsfield(iph);
 801
 802         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 803                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 804                                           & IPV6_TCLASS_MASK;
 805         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 806                 fl6.flowi6_mark = skb->mark;
 807
 808         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 809         if (err != 0) {
 810                 /* XXX: send ICMP error even if DF is not set. */
 811                 if (err == -EMSGSIZE)
 812                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 813                                   htonl(mtu));
 814                 return -1;
 815         }
 816
 817         return 0;
 818 }
 819
 820 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 821 {
 822         struct ip6_tnl *t = netdev_priv(dev);
 823         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 824         int encap_limit = -1;
 825         __u16 offset;
 826         struct flowi6 fl6;
 827         __u8 dsfield;
 828         __u32 mtu;
 829         int err;
 830
 831         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 832                 return -1;
 833
 834         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 835         if (offset > 0) {
 836                 struct ipv6_tlv_tnl_enc_lim *tel;
 837                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 838                 if (tel->encap_limit == 0) {
 839                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 840                                     ICMPV6_HDR_FIELD, offset + 2);
 841                         return -1;
 842                 }
 843                 encap_limit = tel->encap_limit - 1;
 844         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 845                 encap_limit = t->parms.encap_limit;
 846
 847         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 848         fl6.flowi6_proto = IPPROTO_GRE;
 849
 850         dsfield = ipv6_get_dsfield(ipv6h);
 851         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 852                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 853         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 854                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
 855         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 856                 fl6.flowi6_mark = skb->mark;
 857
 858         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 859         if (err != 0) {
 860                 if (err == -EMSGSIZE)
 861                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 862                 return -1;
 863         }
 864
 865         return 0;
 866 }
 867
 868 /**
 869  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 870  *   @t: the outgoing tunnel device
 871  *   @hdr: IPv6 header from the incoming packet
 872  *
 873  * Description:
 874  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 875  *   doesn't match source of incoming packet.
 876  *
 877  * Return:
 878  *   1 if conflict,
 879  *   0 else
 880  **/
 881
 882 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 883         const struct ipv6hdr *hdr)
 884 {
 885         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 886 }
 887
 888 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 889 {
 890         struct ip6_tnl *t = netdev_priv(dev);
 891         int encap_limit = -1;
 892         struct flowi6 fl6;
 893         __u32 mtu;
 894         int err;
 895
 896         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 897                 encap_limit = t->parms.encap_limit;
 898
 899         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 900         fl6.flowi6_proto = skb->protocol;
 901
 902         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 903
 904         return err;
 905 }
 906
 907 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 908         struct net_device *dev)
 909 {
 910         struct ip6_tnl *t = netdev_priv(dev);
 911         struct net_device_stats *stats = &t->dev->stats;
 912         int ret;
 913
 914         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 915                 goto tx_err;
 916
 917         switch (skb->protocol) {
 918         case htons(ETH_P_IP):
 919                 ret = ip6gre_xmit_ipv4(skb, dev);
 920                 break;
 921         case htons(ETH_P_IPV6):
 922                 ret = ip6gre_xmit_ipv6(skb, dev);
 923                 break;
 924         default:
 925                 ret = ip6gre_xmit_other(skb, dev);
 926                 break;
 927         }
 928
 929         if (ret < 0)
 930                 goto tx_err;
 931
 932         return NETDEV_TX_OK;
 933
 934 tx_err:
 935         stats->tx_errors++;
 936         stats->tx_dropped++;
 937         kfree_skb(skb);
 938         return NETDEV_TX_OK;
 939 }
 940
 941 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 942 {
 943         struct net_device *dev = t->dev;
 944         struct __ip6_tnl_parm *p = &t->parms;
 945         struct flowi6 *fl6 = &t->fl.u.ip6;
 946         int addend = sizeof(struct ipv6hdr) + 4;
 947
 948         if (dev->type != ARPHRD_ETHER) {
 949                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 950                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 951         }
 952
 953         /* Set up flowi template */
 954         fl6->saddr = p->laddr;
 955         fl6->daddr = p->raddr;
 956         fl6->flowi6_oif = p->link;
 957         fl6->flowlabel = 0;
 958
 959         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 960                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 961         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 962                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 963
 964         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 965         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 966
 967         if (p->flags&IP6_TNL_F_CAP_XMIT &&
 968                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
 969                 dev->flags |= IFF_POINTOPOINT;
 970         else
 971                 dev->flags &= ~IFF_POINTOPOINT;
 972
 973         /* Precalculate GRE options length */
 974         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
 975                 if (t->parms.o_flags&GRE_CSUM)
 976                         addend += 4;
 977                 if (t->parms.o_flags&GRE_KEY)
 978                         addend += 4;
 979                 if (t->parms.o_flags&GRE_SEQ)
 980                         addend += 4;
 981         }
 982         t->hlen = addend;
 983
 984         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 985                 int strict = (ipv6_addr_type(&p->raddr) &
 986                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 987
 988                 struct rt6_info *rt = rt6_lookup(t->net,
 989                                                  &p->raddr, &p->laddr,
 990                                                  p->link, strict);
 991
 992                 if (!rt)
 993                         return;
 994
 995                 if (rt->dst.dev) {
 996                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
 997
 998                         if (set_mtu) {
 999                                 dev->mtu = rt->dst.dev->mtu - addend;
1000                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1001                                         dev->mtu -= 8;
1002                                 if (dev->type == ARPHRD_ETHER)
1003                                         dev->mtu -= ETH_HLEN;
1004
1005                                 if (dev->mtu < IPV6_MIN_MTU)
1006                                         dev->mtu = IPV6_MIN_MTU;
1007                         }
1008                 }
1009                 ip6_rt_put(rt);
1010         }
1011 }
1012
1013 static int ip6gre_tnl_change(struct ip6_tnl *t,
1014         const struct __ip6_tnl_parm *p, int set_mtu)
1015 {
1016         t->parms.laddr = p->laddr;
1017         t->parms.raddr = p->raddr;
1018         t->parms.flags = p->flags;
1019         t->parms.hop_limit = p->hop_limit;
1020         t->parms.encap_limit = p->encap_limit;
1021         t->parms.flowinfo = p->flowinfo;
1022         t->parms.link = p->link;
1023         t->parms.proto = p->proto;
1024         t->parms.i_key = p->i_key;
1025         t->parms.o_key = p->o_key;
1026         t->parms.i_flags = p->i_flags;
1027         t->parms.o_flags = p->o_flags;
1028         dst_cache_reset(&t->dst_cache);
1029         ip6gre_tnl_link_config(t, set_mtu);
1030         return 0;
1031 }
1032
1033 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1034         const struct ip6_tnl_parm2 *u)
1035 {
1036         p->laddr = u->laddr;
1037         p->raddr = u->raddr;
1038         p->flags = u->flags;
1039         p->hop_limit = u->hop_limit;
1040         p->encap_limit = u->encap_limit;
1041         p->flowinfo = u->flowinfo;
1042         p->link = u->link;
1043         p->i_key = u->i_key;
1044         p->o_key = u->o_key;
1045         p->i_flags = u->i_flags;
1046         p->o_flags = u->o_flags;
1047         memcpy(p->name, u->name, sizeof(u->name));
1048 }
1049
1050 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1051         const struct __ip6_tnl_parm *p)
1052 {
1053         u->proto = IPPROTO_GRE;
1054         u->laddr = p->laddr;
1055         u->raddr = p->raddr;
1056         u->flags = p->flags;
1057         u->hop_limit = p->hop_limit;
1058         u->encap_limit = p->encap_limit;
1059         u->flowinfo = p->flowinfo;
1060         u->link = p->link;
1061         u->i_key = p->i_key;
1062         u->o_key = p->o_key;
1063         u->i_flags = p->i_flags;
1064         u->o_flags = p->o_flags;
1065         memcpy(u->name, p->name, sizeof(u->name));
1066 }
1067
1068 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1069         struct ifreq *ifr, int cmd)
1070 {
1071         int err = 0;
1072         struct ip6_tnl_parm2 p;
1073         struct __ip6_tnl_parm p1;
1074         struct ip6_tnl *t = netdev_priv(dev);
1075         struct net *net = t->net;
1076         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1077
1078         switch (cmd) {
1079         case SIOCGETTUNNEL:
1080                 if (dev == ign->fb_tunnel_dev) {
1081                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1082                                 err = -EFAULT;
1083                                 break;
1084                         }
1085                         ip6gre_tnl_parm_from_user(&p1, &p);
1086                         t = ip6gre_tunnel_locate(net, &p1, 0);
1087                         if (!t)
1088                                 t = netdev_priv(dev);
1089                 }
1090                 memset(&p, 0, sizeof(p));
1091                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1092                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1093                         err = -EFAULT;
1094                 break;
1095
1096         case SIOCADDTUNNEL:
1097         case SIOCCHGTUNNEL:
1098                 err = -EPERM;
1099                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1100                         goto done;
1101
1102                 err = -EFAULT;
1103                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1104                         goto done;
1105
1106                 err = -EINVAL;
1107                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1108                         goto done;
1109
1110                 if (!(p.i_flags&GRE_KEY))
1111                         p.i_key = 0;
1112                 if (!(p.o_flags&GRE_KEY))
1113                         p.o_key = 0;
1114
1115                 ip6gre_tnl_parm_from_user(&p1, &p);
1116                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1117
1118                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1119                         if (t) {
1120                                 if (t->dev != dev) {
1121                                         err = -EEXIST;
1122                                         break;
1123                                 }
1124                         } else {
1125                                 t = netdev_priv(dev);
1126
1127                                 ip6gre_tunnel_unlink(ign, t);
1128                                 synchronize_net();
1129                                 ip6gre_tnl_change(t, &p1, 1);
1130                                 ip6gre_tunnel_link(ign, t);
1131                                 netdev_state_change(dev);
1132                         }
1133                 }
1134
1135                 if (t) {
1136                         err = 0;
1137
1138                         memset(&p, 0, sizeof(p));
1139                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1140                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1141                                 err = -EFAULT;
1142                 } else
1143                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1144                 break;
1145
1146         case SIOCDELTUNNEL:
1147                 err = -EPERM;
1148                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1149                         goto done;
1150
1151                 if (dev == ign->fb_tunnel_dev) {
1152                         err = -EFAULT;
1153                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1154                                 goto done;
1155                         err = -ENOENT;
1156                         ip6gre_tnl_parm_from_user(&p1, &p);
1157                         t = ip6gre_tunnel_locate(net, &p1, 0);
1158                         if (!t)
1159                                 goto done;
1160                         err = -EPERM;
1161                         if (t == netdev_priv(ign->fb_tunnel_dev))
1162                                 goto done;
1163                         dev = t->dev;
1164                 }
1165                 unregister_netdevice(dev);
1166                 err = 0;
1167                 break;
1168
1169         default:
1170                 err = -EINVAL;
1171         }
1172
1173 done:
1174         return err;
1175 }
1176
1177 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1178 {
1179         if (new_mtu < 68 ||
1180             new_mtu > 0xFFF8 - dev->hard_header_len)
1181                 return -EINVAL;
1182         dev->mtu = new_mtu;
1183         return 0;
1184 }
1185
1186 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1187                         unsigned short type,
1188                         const void *daddr, const void *saddr, unsigned int len)
1189 {
1190         struct ip6_tnl *t = netdev_priv(dev);
1191         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1192         __be16 *p = (__be16 *)(ipv6h+1);
1193
1194         ip6_flow_hdr(ipv6h, 0,
1195                      ip6_make_flowlabel(dev_net(dev), skb,
1196                                         t->fl.u.ip6.flowlabel, true,
1197                                         &t->fl.u.ip6));
1198         ipv6h->hop_limit = t->parms.hop_limit;
1199         ipv6h->nexthdr = NEXTHDR_GRE;
1200         ipv6h->saddr = t->parms.laddr;
1201         ipv6h->daddr = t->parms.raddr;
1202
1203         p[0]            = t->parms.o_flags;
1204         p[1]            = htons(type);
1205
1206         /*
1207          *      Set the source hardware address.
1208          */
1209
1210         if (saddr)
1211                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1212         if (daddr)
1213                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1214         if (!ipv6_addr_any(&ipv6h->daddr))
1215                 return t->hlen;
1216
1217         return -t->hlen;
1218 }
1219
1220 static const struct header_ops ip6gre_header_ops = {
1221         .create = ip6gre_header,
1222 };
1223
1224 static const struct net_device_ops ip6gre_netdev_ops = {
1225         .ndo_init               = ip6gre_tunnel_init,
1226         .ndo_uninit             = ip6gre_tunnel_uninit,
1227         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1228         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1229         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1230         .ndo_get_stats64        = ip_tunnel_get_stats64,
1231         .ndo_get_iflink         = ip6_tnl_get_iflink,
1232 };
1233
1234 static void ip6gre_dev_free(struct net_device *dev)
1235 {
1236         struct ip6_tnl *t = netdev_priv(dev);
1237
1238         dst_cache_destroy(&t->dst_cache);
1239         free_percpu(dev->tstats);
1240         free_netdev(dev);
1241 }
1242
1243 static void ip6gre_tunnel_setup(struct net_device *dev)
1244 {
1245         struct ip6_tnl *t;
1246
1247         dev->netdev_ops = &ip6gre_netdev_ops;
1248         dev->destructor = ip6gre_dev_free;
1249
1250         dev->type = ARPHRD_IP6GRE;
1251         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1252         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1253         t = netdev_priv(dev);
1254         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1255                 dev->mtu -= 8;
1256         dev->flags |= IFF_NOARP;
1257         dev->addr_len = sizeof(struct in6_addr);
1258         netif_keep_dst(dev);
1259 }
1260
1261 static int ip6gre_tunnel_init_common(struct net_device *dev)
1262 {
1263         struct ip6_tnl *tunnel;
1264         int ret;
1265
1266         tunnel = netdev_priv(dev);
1267
1268         tunnel->dev = dev;
1269         tunnel->net = dev_net(dev);
1270         strcpy(tunnel->parms.name, dev->name);
1271
1272         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1273         if (!dev->tstats)
1274                 return -ENOMEM;
1275
1276         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1277         if (ret) {
1278                 free_percpu(dev->tstats);
1279                 dev->tstats = NULL;
1280                 return ret;
1281         }
1282
1283         return 0;
1284 }
1285
1286 static int ip6gre_tunnel_init(struct net_device *dev)
1287 {
1288         struct ip6_tnl *tunnel;
1289         int ret;
1290
1291         ret = ip6gre_tunnel_init_common(dev);
1292         if (ret)
1293                 return ret;
1294
1295         tunnel = netdev_priv(dev);
1296
1297         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1298         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1299
1300         if (ipv6_addr_any(&tunnel->parms.raddr))
1301                 dev->header_ops = &ip6gre_header_ops;
1302
1303         return 0;
1304 }
1305
1306 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1307 {
1308         struct ip6_tnl *tunnel = netdev_priv(dev);
1309
1310         tunnel->dev = dev;
1311         tunnel->net = dev_net(dev);
1312         strcpy(tunnel->parms.name, dev->name);
1313
1314         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1315
1316         dev_hold(dev);
1317 }
1318
1319
1320 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1321         .handler     = ip6gre_rcv,
1322         .err_handler = ip6gre_err,
1323         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1324 };
1325
1326 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1327 {
1328         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1329         struct net_device *dev, *aux;
1330         int prio;
1331
1332         for_each_netdev_safe(net, dev, aux)
1333                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1334                     dev->rtnl_link_ops == &ip6gre_tap_ops)
1335                         unregister_netdevice_queue(dev, head);
1336
1337         for (prio = 0; prio < 4; prio++) {
1338                 int h;
1339                 for (h = 0; h < HASH_SIZE; h++) {
1340                         struct ip6_tnl *t;
1341
1342                         t = rtnl_dereference(ign->tunnels[prio][h]);
1343
1344                         while (t) {
1345                                 /* If dev is in the same netns, it has already
1346                                  * been added to the list by the previous loop.
1347                                  */
1348                                 if (!net_eq(dev_net(t->dev), net))
1349                                         unregister_netdevice_queue(t->dev,
1350                                                                    head);
1351                                 t = rtnl_dereference(t->next);
1352                         }
1353                 }
1354         }
1355 }
1356
1357 static int __net_init ip6gre_init_net(struct net *net)
1358 {
1359         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1360         int err;
1361
1362         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1363                                           NET_NAME_UNKNOWN,
1364                                           ip6gre_tunnel_setup);
1365         if (!ign->fb_tunnel_dev) {
1366                 err = -ENOMEM;
1367                 goto err_alloc_dev;
1368         }
1369         dev_net_set(ign->fb_tunnel_dev, net);
1370         /* FB netdevice is special: we have one, and only one per netns.
1371          * Allowing to move it to another netns is clearly unsafe.
1372          */
1373         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1374
1375
1376         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1377         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1378
1379         err = register_netdev(ign->fb_tunnel_dev);
1380         if (err)
1381                 goto err_reg_dev;
1382
1383         rcu_assign_pointer(ign->tunnels_wc[0],
1384                            netdev_priv(ign->fb_tunnel_dev));
1385         return 0;
1386
1387 err_reg_dev:
1388         ip6gre_dev_free(ign->fb_tunnel_dev);
1389 err_alloc_dev:
1390         return err;
1391 }
1392
1393 static void __net_exit ip6gre_exit_net(struct net *net)
1394 {
1395         LIST_HEAD(list);
1396
1397         rtnl_lock();
1398         ip6gre_destroy_tunnels(net, &list);
1399         unregister_netdevice_many(&list);
1400         rtnl_unlock();
1401 }
1402
1403 static struct pernet_operations ip6gre_net_ops = {
1404         .init = ip6gre_init_net,
1405         .exit = ip6gre_exit_net,
1406         .id   = &ip6gre_net_id,
1407         .size = sizeof(struct ip6gre_net),
1408 };
1409
1410 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1411 {
1412         __be16 flags;
1413
1414         if (!data)
1415                 return 0;
1416
1417         flags = 0;
1418         if (data[IFLA_GRE_IFLAGS])
1419                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1420         if (data[IFLA_GRE_OFLAGS])
1421                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1422         if (flags & (GRE_VERSION|GRE_ROUTING))
1423                 return -EINVAL;
1424
1425         return 0;
1426 }
1427
1428 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1429 {
1430         struct in6_addr daddr;
1431
1432         if (tb[IFLA_ADDRESS]) {
1433                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1434                         return -EINVAL;
1435                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1436                         return -EADDRNOTAVAIL;
1437         }
1438
1439         if (!data)
1440                 goto out;
1441
1442         if (data[IFLA_GRE_REMOTE]) {
1443                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1444                 if (ipv6_addr_any(&daddr))
1445                         return -EINVAL;
1446         }
1447
1448 out:
1449         return ip6gre_tunnel_validate(tb, data);
1450 }
1451
1452
1453 static void ip6gre_netlink_parms(struct nlattr *data[],
1454                                 struct __ip6_tnl_parm *parms)
1455 {
1456         memset(parms, 0, sizeof(*parms));
1457
1458         if (!data)
1459                 return;
1460
1461         if (data[IFLA_GRE_LINK])
1462                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1463
1464         if (data[IFLA_GRE_IFLAGS])
1465                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1466
1467         if (data[IFLA_GRE_OFLAGS])
1468                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1469
1470         if (data[IFLA_GRE_IKEY])
1471                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1472
1473         if (data[IFLA_GRE_OKEY])
1474                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1475
1476         if (data[IFLA_GRE_LOCAL])
1477                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1478
1479         if (data[IFLA_GRE_REMOTE])
1480                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1481
1482         if (data[IFLA_GRE_TTL])
1483                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1484
1485         if (data[IFLA_GRE_ENCAP_LIMIT])
1486                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1487
1488         if (data[IFLA_GRE_FLOWINFO])
1489                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1490
1491         if (data[IFLA_GRE_FLAGS])
1492                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1493 }
1494
1495 static int ip6gre_tap_init(struct net_device *dev)
1496 {
1497         struct ip6_tnl *tunnel;
1498         int ret;
1499
1500         ret = ip6gre_tunnel_init_common(dev);
1501         if (ret)
1502                 return ret;
1503
1504         tunnel = netdev_priv(dev);
1505
1506         ip6gre_tnl_link_config(tunnel, 1);
1507
1508         return 0;
1509 }
1510
1511 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1512         .ndo_init = ip6gre_tap_init,
1513         .ndo_uninit = ip6gre_tunnel_uninit,
1514         .ndo_start_xmit = ip6gre_tunnel_xmit,
1515         .ndo_set_mac_address = eth_mac_addr,
1516         .ndo_validate_addr = eth_validate_addr,
1517         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1518         .ndo_get_stats64 = ip_tunnel_get_stats64,
1519         .ndo_get_iflink = ip6_tnl_get_iflink,
1520 };
1521
1522 #define GRE6_FEATURES (NETIF_F_SG |             \
1523                        NETIF_F_FRAGLIST |       \
1524                        NETIF_F_HIGHDMA |                \
1525                        NETIF_F_HW_CSUM)
1526
1527 static void ip6gre_tap_setup(struct net_device *dev)
1528 {
1529
1530         ether_setup(dev);
1531
1532         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1533         dev->destructor = ip6gre_dev_free;
1534
1535         dev->features |= NETIF_F_NETNS_LOCAL;
1536         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1537 }
1538
1539 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1540         struct nlattr *tb[], struct nlattr *data[])
1541 {
1542         struct ip6_tnl *nt;
1543         struct net *net = dev_net(dev);
1544         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1545         int err;
1546
1547         nt = netdev_priv(dev);
1548         ip6gre_netlink_parms(data, &nt->parms);
1549
1550         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1551                 return -EEXIST;
1552
1553         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1554                 eth_hw_addr_random(dev);
1555
1556         nt->dev = dev;
1557         nt->net = dev_net(dev);
1558         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1559
1560         dev->features           |= GRE6_FEATURES;
1561         dev->hw_features        |= GRE6_FEATURES;
1562
1563         if (!(nt->parms.o_flags & GRE_SEQ)) {
1564                 /* TCP segmentation offload is not supported when we
1565                  * generate output sequences.
1566                  */
1567                 dev->features    |= NETIF_F_GSO_SOFTWARE;
1568                 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1569
1570                 /* Can use a lockless transmit, unless we generate
1571                  * output sequences
1572                  */
1573                 dev->features |= NETIF_F_LLTX;
1574         }
1575
1576         err = register_netdevice(dev);
1577         if (err)
1578                 goto out;
1579
1580         dev_hold(dev);
1581         ip6gre_tunnel_link(ign, nt);
1582
1583 out:
1584         return err;
1585 }
1586
1587 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1588                             struct nlattr *data[])
1589 {
1590         struct ip6_tnl *t, *nt = netdev_priv(dev);
1591         struct net *net = nt->net;
1592         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1593         struct __ip6_tnl_parm p;
1594
1595         if (dev == ign->fb_tunnel_dev)
1596                 return -EINVAL;
1597
1598         ip6gre_netlink_parms(data, &p);
1599
1600         t = ip6gre_tunnel_locate(net, &p, 0);
1601
1602         if (t) {
1603                 if (t->dev != dev)
1604                         return -EEXIST;
1605         } else {
1606                 t = nt;
1607         }
1608
1609         ip6gre_tunnel_unlink(ign, t);
1610         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1611         ip6gre_tunnel_link(ign, t);
1612         return 0;
1613 }
1614
1615 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1616 {
1617         struct net *net = dev_net(dev);
1618         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1619
1620         if (dev != ign->fb_tunnel_dev)
1621                 unregister_netdevice_queue(dev, head);
1622 }
1623
1624 static size_t ip6gre_get_size(const struct net_device *dev)
1625 {
1626         return
1627                 /* IFLA_GRE_LINK */
1628                 nla_total_size(4) +
1629                 /* IFLA_GRE_IFLAGS */
1630                 nla_total_size(2) +
1631                 /* IFLA_GRE_OFLAGS */
1632                 nla_total_size(2) +
1633                 /* IFLA_GRE_IKEY */
1634                 nla_total_size(4) +
1635                 /* IFLA_GRE_OKEY */
1636                 nla_total_size(4) +
1637                 /* IFLA_GRE_LOCAL */
1638                 nla_total_size(sizeof(struct in6_addr)) +
1639                 /* IFLA_GRE_REMOTE */
1640                 nla_total_size(sizeof(struct in6_addr)) +
1641                 /* IFLA_GRE_TTL */
1642                 nla_total_size(1) +
1643                 /* IFLA_GRE_TOS */
1644                 nla_total_size(1) +
1645                 /* IFLA_GRE_ENCAP_LIMIT */
1646                 nla_total_size(1) +
1647                 /* IFLA_GRE_FLOWINFO */
1648                 nla_total_size(4) +
1649                 /* IFLA_GRE_FLAGS */
1650                 nla_total_size(4) +
1651                 0;
1652 }
1653
1654 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1655 {
1656         struct ip6_tnl *t = netdev_priv(dev);
1657         struct __ip6_tnl_parm *p = &t->parms;
1658
1659         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1660             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1661             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1662             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1663             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1664             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1665             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1666             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1667             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1668             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1669             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1670             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1671                 goto nla_put_failure;
1672         return 0;
1673
1674 nla_put_failure:
1675         return -EMSGSIZE;
1676 }
1677
1678 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1679         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1680         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1681         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1682         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1683         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1684         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1685         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1686         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1687         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1688         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1689         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1690 };
1691
1692 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1693         .kind           = "ip6gre",
1694         .maxtype        = IFLA_GRE_MAX,
1695         .policy         = ip6gre_policy,
1696         .priv_size      = sizeof(struct ip6_tnl),
1697         .setup          = ip6gre_tunnel_setup,
1698         .validate       = ip6gre_tunnel_validate,
1699         .newlink        = ip6gre_newlink,
1700         .changelink     = ip6gre_changelink,
1701         .dellink        = ip6gre_dellink,
1702         .get_size       = ip6gre_get_size,
1703         .fill_info      = ip6gre_fill_info,
1704         .get_link_net   = ip6_tnl_get_link_net,
1705 };
1706
1707 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1708         .kind           = "ip6gretap",
1709         .maxtype        = IFLA_GRE_MAX,
1710         .policy         = ip6gre_policy,
1711         .priv_size      = sizeof(struct ip6_tnl),
1712         .setup          = ip6gre_tap_setup,
1713         .validate       = ip6gre_tap_validate,
1714         .newlink        = ip6gre_newlink,
1715         .changelink     = ip6gre_changelink,
1716         .get_size       = ip6gre_get_size,
1717         .fill_info      = ip6gre_fill_info,
1718         .get_link_net   = ip6_tnl_get_link_net,
1719 };
1720
1721 /*
1722  *      And now the modules code and kernel interface.
1723  */
1724
1725 static int __init ip6gre_init(void)
1726 {
1727         int err;
1728
1729         pr_info("GRE over IPv6 tunneling driver\n");
1730
1731         err = register_pernet_device(&ip6gre_net_ops);
1732         if (err < 0)
1733                 return err;
1734
1735         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1736         if (err < 0) {
1737                 pr_info("%s: can't add protocol\n", __func__);
1738                 goto add_proto_failed;
1739         }
1740
1741         err = rtnl_link_register(&ip6gre_link_ops);
1742         if (err < 0)
1743                 goto rtnl_link_failed;
1744
1745         err = rtnl_link_register(&ip6gre_tap_ops);
1746         if (err < 0)
1747                 goto tap_ops_failed;
1748
1749 out:
1750         return err;
1751
1752 tap_ops_failed:
1753         rtnl_link_unregister(&ip6gre_link_ops);
1754 rtnl_link_failed:
1755         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1756 add_proto_failed:
1757         unregister_pernet_device(&ip6gre_net_ops);
1758         goto out;
1759 }
1760
1761 static void __exit ip6gre_fini(void)
1762 {
1763         rtnl_link_unregister(&ip6gre_tap_ops);
1764         rtnl_link_unregister(&ip6gre_link_ops);
1765         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1766         unregister_pernet_device(&ip6gre_net_ops);
1767 }
1768
1769 module_init(ip6gre_init);
1770 module_exit(ip6gre_fini);
1771 MODULE_LICENSE("GPL");
1772 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1773 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1774 MODULE_ALIAS_RTNL_LINK("ip6gre");
1775 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1776 MODULE_ALIAS_NETDEV("ip6gre0");