Pileus Git - ~andy/linux/blob - net/ipv6/ip6_gre.c

   1 /*
   2  *      GRE over IPv6 protocol decoder.
   3  *
   4  *      Authors: Dmitry Kozlov (xeb@mail.ru)
   5  *
   6  *      This program is free software; you can redistribute it and/or
   7  *      modify it under the terms of the GNU General Public License
   8  *      as published by the Free Software Foundation; either version
   9  *      2 of the License, or (at your option) any later version.
  10  *
  11  */
  12
  13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  14
  15 #include <linux/capability.h>
  16 #include <linux/module.h>
  17 #include <linux/types.h>
  18 #include <linux/kernel.h>
  19 #include <linux/slab.h>
  20 #include <linux/uaccess.h>
  21 #include <linux/skbuff.h>
  22 #include <linux/netdevice.h>
  23 #include <linux/in.h>
  24 #include <linux/tcp.h>
  25 #include <linux/udp.h>
  26 #include <linux/if_arp.h>
  27 #include <linux/mroute.h>
  28 #include <linux/init.h>
  29 #include <linux/in6.h>
  30 #include <linux/inetdevice.h>
  31 #include <linux/igmp.h>
  32 #include <linux/netfilter_ipv4.h>
  33 #include <linux/etherdevice.h>
  34 #include <linux/if_ether.h>
  35 #include <linux/hash.h>
  36 #include <linux/if_tunnel.h>
  37 #include <linux/ip6_tunnel.h>
  38
  39 #include <net/sock.h>
  40 #include <net/ip.h>
  41 #include <net/ip_tunnels.h>
  42 #include <net/icmp.h>
  43 #include <net/protocol.h>
  44 #include <net/addrconf.h>
  45 #include <net/arp.h>
  46 #include <net/checksum.h>
  47 #include <net/dsfield.h>
  48 #include <net/inet_ecn.h>
  49 #include <net/xfrm.h>
  50 #include <net/net_namespace.h>
  51 #include <net/netns/generic.h>
  52 #include <net/rtnetlink.h>
  53
  54 #include <net/ipv6.h>
  55 #include <net/ip6_fib.h>
  56 #include <net/ip6_route.h>
  57 #include <net/ip6_tunnel.h>
  58
  59
  60 static bool log_ecn_error = true;
  61 module_param(log_ecn_error, bool, 0644);
  62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  63
  64 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
  65 #define IPV6_TCLASS_SHIFT 20
  66
  67 #define HASH_SIZE_SHIFT  5
  68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  69
  70 static int ip6gre_net_id __read_mostly;
  71 struct ip6gre_net {
  72         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
  73
  74         struct net_device *fb_tunnel_dev;
  75 };
  76
  77 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  78 static int ip6gre_tunnel_init(struct net_device *dev);
  79 static void ip6gre_tunnel_setup(struct net_device *dev);
  80 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  81 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  82
  83 /* Tunnel hash table */
  84
  85 /*
  86    4 hash tables:
  87
  88    3: (remote,local)
  89    2: (remote,*)
  90    1: (*,local)
  91    0: (*,*)
  92
  93    We require exact key match i.e. if a key is present in packet
  94    it will match only tunnel with the same key; if it is not present,
  95    it will match only keyless tunnel.
  96
  97    All keysless packets, if not matched configured keyless tunnels
  98    will match fallback tunnel.
  99  */
 100
 101 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
 102 static u32 HASH_ADDR(const struct in6_addr *addr)
 103 {
 104         u32 hash = ipv6_addr_hash(addr);
 105
 106         return hash_32(hash, HASH_SIZE_SHIFT);
 107 }
 108
 109 #define tunnels_r_l     tunnels[3]
 110 #define tunnels_r       tunnels[2]
 111 #define tunnels_l       tunnels[1]
 112 #define tunnels_wc      tunnels[0]
 113
 114 /* Given src, dst and key, find appropriate for input tunnel. */
 115
 116 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 117                 const struct in6_addr *remote, const struct in6_addr *local,
 118                 __be32 key, __be16 gre_proto)
 119 {
 120         struct net *net = dev_net(dev);
 121         int link = dev->ifindex;
 122         unsigned int h0 = HASH_ADDR(remote);
 123         unsigned int h1 = HASH_KEY(key);
 124         struct ip6_tnl *t, *cand = NULL;
 125         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 126         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
 127                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 128         int score, cand_score = 4;
 129
 130         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 131                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 132                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 133                     key != t->parms.i_key ||
 134                     !(t->dev->flags & IFF_UP))
 135                         continue;
 136
 137                 if (t->dev->type != ARPHRD_IP6GRE &&
 138                     t->dev->type != dev_type)
 139                         continue;
 140
 141                 score = 0;
 142                 if (t->parms.link != link)
 143                         score |= 1;
 144                 if (t->dev->type != dev_type)
 145                         score |= 2;
 146                 if (score == 0)
 147                         return t;
 148
 149                 if (score < cand_score) {
 150                         cand = t;
 151                         cand_score = score;
 152                 }
 153         }
 154
 155         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 156                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 157                     key != t->parms.i_key ||
 158                     !(t->dev->flags & IFF_UP))
 159                         continue;
 160
 161                 if (t->dev->type != ARPHRD_IP6GRE &&
 162                     t->dev->type != dev_type)
 163                         continue;
 164
 165                 score = 0;
 166                 if (t->parms.link != link)
 167                         score |= 1;
 168                 if (t->dev->type != dev_type)
 169                         score |= 2;
 170                 if (score == 0)
 171                         return t;
 172
 173                 if (score < cand_score) {
 174                         cand = t;
 175                         cand_score = score;
 176                 }
 177         }
 178
 179         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 180                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 181                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 182                                  !ipv6_addr_is_multicast(local))) ||
 183                     key != t->parms.i_key ||
 184                     !(t->dev->flags & IFF_UP))
 185                         continue;
 186
 187                 if (t->dev->type != ARPHRD_IP6GRE &&
 188                     t->dev->type != dev_type)
 189                         continue;
 190
 191                 score = 0;
 192                 if (t->parms.link != link)
 193                         score |= 1;
 194                 if (t->dev->type != dev_type)
 195                         score |= 2;
 196                 if (score == 0)
 197                         return t;
 198
 199                 if (score < cand_score) {
 200                         cand = t;
 201                         cand_score = score;
 202                 }
 203         }
 204
 205         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 206                 if (t->parms.i_key != key ||
 207                     !(t->dev->flags & IFF_UP))
 208                         continue;
 209
 210                 if (t->dev->type != ARPHRD_IP6GRE &&
 211                     t->dev->type != dev_type)
 212                         continue;
 213
 214                 score = 0;
 215                 if (t->parms.link != link)
 216                         score |= 1;
 217                 if (t->dev->type != dev_type)
 218                         score |= 2;
 219                 if (score == 0)
 220                         return t;
 221
 222                 if (score < cand_score) {
 223                         cand = t;
 224                         cand_score = score;
 225                 }
 226         }
 227
 228         if (cand != NULL)
 229                 return cand;
 230
 231         dev = ign->fb_tunnel_dev;
 232         if (dev->flags & IFF_UP)
 233                 return netdev_priv(dev);
 234
 235         return NULL;
 236 }
 237
 238 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 239                 const struct __ip6_tnl_parm *p)
 240 {
 241         const struct in6_addr *remote = &p->raddr;
 242         const struct in6_addr *local = &p->laddr;
 243         unsigned int h = HASH_KEY(p->i_key);
 244         int prio = 0;
 245
 246         if (!ipv6_addr_any(local))
 247                 prio |= 1;
 248         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 249                 prio |= 2;
 250                 h ^= HASH_ADDR(remote);
 251         }
 252
 253         return &ign->tunnels[prio][h];
 254 }
 255
 256 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 257                 const struct ip6_tnl *t)
 258 {
 259         return __ip6gre_bucket(ign, &t->parms);
 260 }
 261
 262 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 263 {
 264         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 265
 266         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 267         rcu_assign_pointer(*tp, t);
 268 }
 269
 270 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 271 {
 272         struct ip6_tnl __rcu **tp;
 273         struct ip6_tnl *iter;
 274
 275         for (tp = ip6gre_bucket(ign, t);
 276              (iter = rtnl_dereference(*tp)) != NULL;
 277              tp = &iter->next) {
 278                 if (t == iter) {
 279                         rcu_assign_pointer(*tp, t->next);
 280                         break;
 281                 }
 282         }
 283 }
 284
 285 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 286                                            const struct __ip6_tnl_parm *parms,
 287                                            int type)
 288 {
 289         const struct in6_addr *remote = &parms->raddr;
 290         const struct in6_addr *local = &parms->laddr;
 291         __be32 key = parms->i_key;
 292         int link = parms->link;
 293         struct ip6_tnl *t;
 294         struct ip6_tnl __rcu **tp;
 295         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 296
 297         for (tp = __ip6gre_bucket(ign, parms);
 298              (t = rtnl_dereference(*tp)) != NULL;
 299              tp = &t->next)
 300                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 301                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 302                     key == t->parms.i_key &&
 303                     link == t->parms.link &&
 304                     type == t->dev->type)
 305                         break;
 306
 307         return t;
 308 }
 309
 310 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 311                 const struct __ip6_tnl_parm *parms, int create)
 312 {
 313         struct ip6_tnl *t, *nt;
 314         struct net_device *dev;
 315         char name[IFNAMSIZ];
 316         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 317
 318         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 319         if (t || !create)
 320                 return t;
 321
 322         if (parms->name[0])
 323                 strlcpy(name, parms->name, IFNAMSIZ);
 324         else
 325                 strcpy(name, "ip6gre%d");
 326
 327         dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
 328         if (!dev)
 329                 return NULL;
 330
 331         dev_net_set(dev, net);
 332
 333         nt = netdev_priv(dev);
 334         nt->parms = *parms;
 335         dev->rtnl_link_ops = &ip6gre_link_ops;
 336
 337         nt->dev = dev;
 338         ip6gre_tnl_link_config(nt, 1);
 339
 340         if (register_netdevice(dev) < 0)
 341                 goto failed_free;
 342
 343         /* Can use a lockless transmit, unless we generate output sequences */
 344         if (!(nt->parms.o_flags & GRE_SEQ))
 345                 dev->features |= NETIF_F_LLTX;
 346
 347         dev_hold(dev);
 348         ip6gre_tunnel_link(ign, nt);
 349         return nt;
 350
 351 failed_free:
 352         free_netdev(dev);
 353         return NULL;
 354 }
 355
 356 static void ip6gre_tunnel_uninit(struct net_device *dev)
 357 {
 358         struct net *net = dev_net(dev);
 359         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 360
 361         ip6gre_tunnel_unlink(ign, netdev_priv(dev));
 362         dev_put(dev);
 363 }
 364
 365
 366 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 367                 u8 type, u8 code, int offset, __be32 info)
 368 {
 369         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 370         __be16 *p = (__be16 *)(skb->data + offset);
 371         int grehlen = offset + 4;
 372         struct ip6_tnl *t;
 373         __be16 flags;
 374
 375         flags = p[0];
 376         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
 377                 if (flags&(GRE_VERSION|GRE_ROUTING))
 378                         return;
 379                 if (flags&GRE_KEY) {
 380                         grehlen += 4;
 381                         if (flags&GRE_CSUM)
 382                                 grehlen += 4;
 383                 }
 384         }
 385
 386         /* If only 8 bytes returned, keyed message will be dropped here */
 387         if (!pskb_may_pull(skb, grehlen))
 388                 return;
 389         ipv6h = (const struct ipv6hdr *)skb->data;
 390         p = (__be16 *)(skb->data + offset);
 391
 392         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 393                                 flags & GRE_KEY ?
 394                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
 395                                 p[1]);
 396         if (t == NULL)
 397                 return;
 398
 399         switch (type) {
 400                 __u32 teli;
 401                 struct ipv6_tlv_tnl_enc_lim *tel;
 402                 __u32 mtu;
 403         case ICMPV6_DEST_UNREACH:
 404                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
 405                                      t->parms.name);
 406                 break;
 407         case ICMPV6_TIME_EXCEED:
 408                 if (code == ICMPV6_EXC_HOPLIMIT) {
 409                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 410                                              t->parms.name);
 411                 }
 412                 break;
 413         case ICMPV6_PARAMPROB:
 414                 teli = 0;
 415                 if (code == ICMPV6_HDR_FIELD)
 416                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 417
 418                 if (teli && teli == info - 2) {
 419                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 420                         if (tel->encap_limit == 0) {
 421                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 422                                                      t->parms.name);
 423                         }
 424                 } else {
 425                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 426                                              t->parms.name);
 427                 }
 428                 break;
 429         case ICMPV6_PKT_TOOBIG:
 430                 mtu = info - offset;
 431                 if (mtu < IPV6_MIN_MTU)
 432                         mtu = IPV6_MIN_MTU;
 433                 t->dev->mtu = mtu;
 434                 break;
 435         }
 436
 437         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 438                 t->err_count++;
 439         else
 440                 t->err_count = 1;
 441         t->err_time = jiffies;
 442 }
 443
 444 static int ip6gre_rcv(struct sk_buff *skb)
 445 {
 446         const struct ipv6hdr *ipv6h;
 447         u8     *h;
 448         __be16    flags;
 449         __sum16   csum = 0;
 450         __be32 key = 0;
 451         u32    seqno = 0;
 452         struct ip6_tnl *tunnel;
 453         int    offset = 4;
 454         __be16 gre_proto;
 455         int err;
 456
 457         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
 458                 goto drop;
 459
 460         ipv6h = ipv6_hdr(skb);
 461         h = skb->data;
 462         flags = *(__be16 *)h;
 463
 464         if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
 465                 /* - Version must be 0.
 466                    - We do not support routing headers.
 467                  */
 468                 if (flags&(GRE_VERSION|GRE_ROUTING))
 469                         goto drop;
 470
 471                 if (flags&GRE_CSUM) {
 472                         switch (skb->ip_summed) {
 473                         case CHECKSUM_COMPLETE:
 474                                 csum = csum_fold(skb->csum);
 475                                 if (!csum)
 476                                         break;
 477                                 /* fall through */
 478                         case CHECKSUM_NONE:
 479                                 skb->csum = 0;
 480                                 csum = __skb_checksum_complete(skb);
 481                                 skb->ip_summed = CHECKSUM_COMPLETE;
 482                         }
 483                         offset += 4;
 484                 }
 485                 if (flags&GRE_KEY) {
 486                         key = *(__be32 *)(h + offset);
 487                         offset += 4;
 488                 }
 489                 if (flags&GRE_SEQ) {
 490                         seqno = ntohl(*(__be32 *)(h + offset));
 491                         offset += 4;
 492                 }
 493         }
 494
 495         gre_proto = *(__be16 *)(h + 2);
 496
 497         tunnel = ip6gre_tunnel_lookup(skb->dev,
 498                                           &ipv6h->saddr, &ipv6h->daddr, key,
 499                                           gre_proto);
 500         if (tunnel) {
 501                 struct pcpu_tstats *tstats;
 502
 503                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 504                         goto drop;
 505
 506                 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
 507                         tunnel->dev->stats.rx_dropped++;
 508                         goto drop;
 509                 }
 510
 511                 secpath_reset(skb);
 512
 513                 skb->protocol = gre_proto;
 514                 /* WCCP version 1 and 2 protocol decoding.
 515                  * - Change protocol to IP
 516                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
 517                  */
 518                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
 519                         skb->protocol = htons(ETH_P_IP);
 520                         if ((*(h + offset) & 0xF0) != 0x40)
 521                                 offset += 4;
 522                 }
 523
 524                 skb->mac_header = skb->network_header;
 525                 __pskb_pull(skb, offset);
 526                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
 527                 skb->pkt_type = PACKET_HOST;
 528
 529                 if (((flags&GRE_CSUM) && csum) ||
 530                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
 531                         tunnel->dev->stats.rx_crc_errors++;
 532                         tunnel->dev->stats.rx_errors++;
 533                         goto drop;
 534                 }
 535                 if (tunnel->parms.i_flags&GRE_SEQ) {
 536                         if (!(flags&GRE_SEQ) ||
 537                             (tunnel->i_seqno &&
 538                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
 539                                 tunnel->dev->stats.rx_fifo_errors++;
 540                                 tunnel->dev->stats.rx_errors++;
 541                                 goto drop;
 542                         }
 543                         tunnel->i_seqno = seqno + 1;
 544                 }
 545
 546                 /* Warning: All skb pointers will be invalidated! */
 547                 if (tunnel->dev->type == ARPHRD_ETHER) {
 548                         if (!pskb_may_pull(skb, ETH_HLEN)) {
 549                                 tunnel->dev->stats.rx_length_errors++;
 550                                 tunnel->dev->stats.rx_errors++;
 551                                 goto drop;
 552                         }
 553
 554                         ipv6h = ipv6_hdr(skb);
 555                         skb->protocol = eth_type_trans(skb, tunnel->dev);
 556                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 557                 }
 558
 559                 __skb_tunnel_rx(skb, tunnel->dev);
 560
 561                 skb_reset_network_header(skb);
 562
 563                 err = IP6_ECN_decapsulate(ipv6h, skb);
 564                 if (unlikely(err)) {
 565                         if (log_ecn_error)
 566                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 567                                                      &ipv6h->saddr,
 568                                                      ipv6_get_dsfield(ipv6h));
 569                         if (err > 1) {
 570                                 ++tunnel->dev->stats.rx_frame_errors;
 571                                 ++tunnel->dev->stats.rx_errors;
 572                                 goto drop;
 573                         }
 574                 }
 575
 576                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 577                 u64_stats_update_begin(&tstats->syncp);
 578                 tstats->rx_packets++;
 579                 tstats->rx_bytes += skb->len;
 580                 u64_stats_update_end(&tstats->syncp);
 581
 582                 netif_rx(skb);
 583
 584                 return 0;
 585         }
 586         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 587
 588 drop:
 589         kfree_skb(skb);
 590         return 0;
 591 }
 592
 593 struct ipv6_tel_txoption {
 594         struct ipv6_txoptions ops;
 595         __u8 dst_opt[8];
 596 };
 597
 598 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 599 {
 600         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 601
 602         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 603         opt->dst_opt[3] = 1;
 604         opt->dst_opt[4] = encap_limit;
 605         opt->dst_opt[5] = IPV6_TLV_PADN;
 606         opt->dst_opt[6] = 1;
 607
 608         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 609         opt->ops.opt_nflen = 8;
 610 }
 611
 612 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 613                          struct net_device *dev,
 614                          __u8 dsfield,
 615                          struct flowi6 *fl6,
 616                          int encap_limit,
 617                          __u32 *pmtu)
 618 {
 619         struct net *net = dev_net(dev);
 620         struct ip6_tnl *tunnel = netdev_priv(dev);
 621         struct net_device *tdev;    /* Device to other host */
 622         struct ipv6hdr  *ipv6h;     /* Our new IP header */
 623         unsigned int max_headroom;  /* The extra header space needed */
 624         int    gre_hlen;
 625         struct ipv6_tel_txoption opt;
 626         int    mtu;
 627         struct dst_entry *dst = NULL, *ndst = NULL;
 628         struct net_device_stats *stats = &tunnel->dev->stats;
 629         int err = -1;
 630         u8 proto;
 631         struct sk_buff *new_skb;
 632
 633         if (dev->type == ARPHRD_ETHER)
 634                 IPCB(skb)->flags = 0;
 635
 636         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 637                 gre_hlen = 0;
 638                 ipv6h = (struct ipv6hdr *)skb->data;
 639                 fl6->daddr = ipv6h->daddr;
 640         } else {
 641                 gre_hlen = tunnel->hlen;
 642                 fl6->daddr = tunnel->parms.raddr;
 643         }
 644
 645         if (!fl6->flowi6_mark)
 646                 dst = ip6_tnl_dst_check(tunnel);
 647
 648         if (!dst) {
 649                 ndst = ip6_route_output(net, NULL, fl6);
 650
 651                 if (ndst->error)
 652                         goto tx_err_link_failure;
 653                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
 654                 if (IS_ERR(ndst)) {
 655                         err = PTR_ERR(ndst);
 656                         ndst = NULL;
 657                         goto tx_err_link_failure;
 658                 }
 659                 dst = ndst;
 660         }
 661
 662         tdev = dst->dev;
 663
 664         if (tdev == dev) {
 665                 stats->collisions++;
 666                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 667                                      tunnel->parms.name);
 668                 goto tx_err_dst_release;
 669         }
 670
 671         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 672         if (encap_limit >= 0) {
 673                 max_headroom += 8;
 674                 mtu -= 8;
 675         }
 676         if (mtu < IPV6_MIN_MTU)
 677                 mtu = IPV6_MIN_MTU;
 678         if (skb_dst(skb))
 679                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 680         if (skb->len > mtu) {
 681                 *pmtu = mtu;
 682                 err = -EMSGSIZE;
 683                 goto tx_err_dst_release;
 684         }
 685
 686         if (tunnel->err_count > 0) {
 687                 if (time_before(jiffies,
 688                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 689                         tunnel->err_count--;
 690
 691                         dst_link_failure(skb);
 692                 } else
 693                         tunnel->err_count = 0;
 694         }
 695
 696         max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 697
 698         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 699             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 700                 new_skb = skb_realloc_headroom(skb, max_headroom);
 701                 if (max_headroom > dev->needed_headroom)
 702                         dev->needed_headroom = max_headroom;
 703                 if (!new_skb)
 704                         goto tx_err_dst_release;
 705
 706                 if (skb->sk)
 707                         skb_set_owner_w(new_skb, skb->sk);
 708                 consume_skb(skb);
 709                 skb = new_skb;
 710         }
 711
 712         skb_dst_drop(skb);
 713
 714         if (fl6->flowi6_mark) {
 715                 skb_dst_set(skb, dst);
 716                 ndst = NULL;
 717         } else {
 718                 skb_dst_set_noref(skb, dst);
 719         }
 720
 721         proto = NEXTHDR_GRE;
 722         if (encap_limit >= 0) {
 723                 init_tel_txopt(&opt, encap_limit);
 724                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 725         }
 726
 727         skb_push(skb, gre_hlen);
 728         skb_reset_network_header(skb);
 729         skb_set_transport_header(skb, sizeof(*ipv6h));
 730
 731         /*
 732          *      Push down and install the IP header.
 733          */
 734         ipv6h = ipv6_hdr(skb);
 735         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
 736         ipv6h->hop_limit = tunnel->parms.hop_limit;
 737         ipv6h->nexthdr = proto;
 738         ipv6h->saddr = fl6->saddr;
 739         ipv6h->daddr = fl6->daddr;
 740
 741         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 742         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
 743                                    htons(ETH_P_TEB) : skb->protocol;
 744
 745         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 746                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 747
 748                 if (tunnel->parms.o_flags&GRE_SEQ) {
 749                         ++tunnel->o_seqno;
 750                         *ptr = htonl(tunnel->o_seqno);
 751                         ptr--;
 752                 }
 753                 if (tunnel->parms.o_flags&GRE_KEY) {
 754                         *ptr = tunnel->parms.o_key;
 755                         ptr--;
 756                 }
 757                 if (tunnel->parms.o_flags&GRE_CSUM) {
 758                         *ptr = 0;
 759                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
 760                                 skb->len - sizeof(struct ipv6hdr));
 761                 }
 762         }
 763
 764         ip6tunnel_xmit(skb, dev);
 765         if (ndst)
 766                 ip6_tnl_dst_store(tunnel, ndst);
 767         return 0;
 768 tx_err_link_failure:
 769         stats->tx_carrier_errors++;
 770         dst_link_failure(skb);
 771 tx_err_dst_release:
 772         dst_release(ndst);
 773         return err;
 774 }
 775
 776 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 777 {
 778         struct ip6_tnl *t = netdev_priv(dev);
 779         const struct iphdr  *iph = ip_hdr(skb);
 780         int encap_limit = -1;
 781         struct flowi6 fl6;
 782         __u8 dsfield;
 783         __u32 mtu;
 784         int err;
 785
 786         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 787                 encap_limit = t->parms.encap_limit;
 788
 789         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 790         fl6.flowi6_proto = IPPROTO_IPIP;
 791
 792         dsfield = ipv4_get_dsfield(iph);
 793
 794         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 795                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 796                                           & IPV6_TCLASS_MASK;
 797         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 798                 fl6.flowi6_mark = skb->mark;
 799
 800         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 801         if (err != 0) {
 802                 /* XXX: send ICMP error even if DF is not set. */
 803                 if (err == -EMSGSIZE)
 804                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 805                                   htonl(mtu));
 806                 return -1;
 807         }
 808
 809         return 0;
 810 }
 811
 812 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 813 {
 814         struct ip6_tnl *t = netdev_priv(dev);
 815         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 816         int encap_limit = -1;
 817         __u16 offset;
 818         struct flowi6 fl6;
 819         __u8 dsfield;
 820         __u32 mtu;
 821         int err;
 822
 823         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 824                 return -1;
 825
 826         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 827         if (offset > 0) {
 828                 struct ipv6_tlv_tnl_enc_lim *tel;
 829                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 830                 if (tel->encap_limit == 0) {
 831                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 832                                     ICMPV6_HDR_FIELD, offset + 2);
 833                         return -1;
 834                 }
 835                 encap_limit = tel->encap_limit - 1;
 836         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 837                 encap_limit = t->parms.encap_limit;
 838
 839         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 840         fl6.flowi6_proto = IPPROTO_IPV6;
 841
 842         dsfield = ipv6_get_dsfield(ipv6h);
 843         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 844                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 845         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 846                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 847         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 848                 fl6.flowi6_mark = skb->mark;
 849
 850         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 851         if (err != 0) {
 852                 if (err == -EMSGSIZE)
 853                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 854                 return -1;
 855         }
 856
 857         return 0;
 858 }
 859
 860 /**
 861  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 862  *   @t: the outgoing tunnel device
 863  *   @hdr: IPv6 header from the incoming packet
 864  *
 865  * Description:
 866  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 867  *   doesn't match source of incoming packet.
 868  *
 869  * Return:
 870  *   1 if conflict,
 871  *   0 else
 872  **/
 873
 874 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 875         const struct ipv6hdr *hdr)
 876 {
 877         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 878 }
 879
 880 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 881 {
 882         struct ip6_tnl *t = netdev_priv(dev);
 883         int encap_limit = -1;
 884         struct flowi6 fl6;
 885         __u32 mtu;
 886         int err;
 887
 888         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 889                 encap_limit = t->parms.encap_limit;
 890
 891         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 892         fl6.flowi6_proto = skb->protocol;
 893
 894         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 895
 896         return err;
 897 }
 898
 899 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 900         struct net_device *dev)
 901 {
 902         struct ip6_tnl *t = netdev_priv(dev);
 903         struct net_device_stats *stats = &t->dev->stats;
 904         int ret;
 905
 906         if (!ip6_tnl_xmit_ctl(t))
 907                 goto tx_err;
 908
 909         switch (skb->protocol) {
 910         case htons(ETH_P_IP):
 911                 ret = ip6gre_xmit_ipv4(skb, dev);
 912                 break;
 913         case htons(ETH_P_IPV6):
 914                 ret = ip6gre_xmit_ipv6(skb, dev);
 915                 break;
 916         default:
 917                 ret = ip6gre_xmit_other(skb, dev);
 918                 break;
 919         }
 920
 921         if (ret < 0)
 922                 goto tx_err;
 923
 924         return NETDEV_TX_OK;
 925
 926 tx_err:
 927         stats->tx_errors++;
 928         stats->tx_dropped++;
 929         kfree_skb(skb);
 930         return NETDEV_TX_OK;
 931 }
 932
 933 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 934 {
 935         struct net_device *dev = t->dev;
 936         struct __ip6_tnl_parm *p = &t->parms;
 937         struct flowi6 *fl6 = &t->fl.u.ip6;
 938         int addend = sizeof(struct ipv6hdr) + 4;
 939
 940         if (dev->type != ARPHRD_ETHER) {
 941                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 942                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 943         }
 944
 945         /* Set up flowi template */
 946         fl6->saddr = p->laddr;
 947         fl6->daddr = p->raddr;
 948         fl6->flowi6_oif = p->link;
 949         fl6->flowlabel = 0;
 950
 951         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 952                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 953         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 954                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 955
 956         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 957         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 958
 959         if (p->flags&IP6_TNL_F_CAP_XMIT &&
 960                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
 961                 dev->flags |= IFF_POINTOPOINT;
 962         else
 963                 dev->flags &= ~IFF_POINTOPOINT;
 964
 965         dev->iflink = p->link;
 966
 967         /* Precalculate GRE options length */
 968         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
 969                 if (t->parms.o_flags&GRE_CSUM)
 970                         addend += 4;
 971                 if (t->parms.o_flags&GRE_KEY)
 972                         addend += 4;
 973                 if (t->parms.o_flags&GRE_SEQ)
 974                         addend += 4;
 975         }
 976
 977         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 978                 int strict = (ipv6_addr_type(&p->raddr) &
 979                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 980
 981                 struct rt6_info *rt = rt6_lookup(dev_net(dev),
 982                                                  &p->raddr, &p->laddr,
 983                                                  p->link, strict);
 984
 985                 if (rt == NULL)
 986                         return;
 987
 988                 if (rt->dst.dev) {
 989                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
 990
 991                         if (set_mtu) {
 992                                 dev->mtu = rt->dst.dev->mtu - addend;
 993                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 994                                         dev->mtu -= 8;
 995
 996                                 if (dev->mtu < IPV6_MIN_MTU)
 997                                         dev->mtu = IPV6_MIN_MTU;
 998                         }
 999                 }
1000                 ip6_rt_put(rt);
1001         }
1002
1003         t->hlen = addend;
1004 }
1005
1006 static int ip6gre_tnl_change(struct ip6_tnl *t,
1007         const struct __ip6_tnl_parm *p, int set_mtu)
1008 {
1009         t->parms.laddr = p->laddr;
1010         t->parms.raddr = p->raddr;
1011         t->parms.flags = p->flags;
1012         t->parms.hop_limit = p->hop_limit;
1013         t->parms.encap_limit = p->encap_limit;
1014         t->parms.flowinfo = p->flowinfo;
1015         t->parms.link = p->link;
1016         t->parms.proto = p->proto;
1017         t->parms.i_key = p->i_key;
1018         t->parms.o_key = p->o_key;
1019         t->parms.i_flags = p->i_flags;
1020         t->parms.o_flags = p->o_flags;
1021         ip6_tnl_dst_reset(t);
1022         ip6gre_tnl_link_config(t, set_mtu);
1023         return 0;
1024 }
1025
1026 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1027         const struct ip6_tnl_parm2 *u)
1028 {
1029         p->laddr = u->laddr;
1030         p->raddr = u->raddr;
1031         p->flags = u->flags;
1032         p->hop_limit = u->hop_limit;
1033         p->encap_limit = u->encap_limit;
1034         p->flowinfo = u->flowinfo;
1035         p->link = u->link;
1036         p->i_key = u->i_key;
1037         p->o_key = u->o_key;
1038         p->i_flags = u->i_flags;
1039         p->o_flags = u->o_flags;
1040         memcpy(p->name, u->name, sizeof(u->name));
1041 }
1042
1043 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1044         const struct __ip6_tnl_parm *p)
1045 {
1046         u->proto = IPPROTO_GRE;
1047         u->laddr = p->laddr;
1048         u->raddr = p->raddr;
1049         u->flags = p->flags;
1050         u->hop_limit = p->hop_limit;
1051         u->encap_limit = p->encap_limit;
1052         u->flowinfo = p->flowinfo;
1053         u->link = p->link;
1054         u->i_key = p->i_key;
1055         u->o_key = p->o_key;
1056         u->i_flags = p->i_flags;
1057         u->o_flags = p->o_flags;
1058         memcpy(u->name, p->name, sizeof(u->name));
1059 }
1060
1061 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1062         struct ifreq *ifr, int cmd)
1063 {
1064         int err = 0;
1065         struct ip6_tnl_parm2 p;
1066         struct __ip6_tnl_parm p1;
1067         struct ip6_tnl *t;
1068         struct net *net = dev_net(dev);
1069         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1070
1071         switch (cmd) {
1072         case SIOCGETTUNNEL:
1073                 t = NULL;
1074                 if (dev == ign->fb_tunnel_dev) {
1075                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1076                                 err = -EFAULT;
1077                                 break;
1078                         }
1079                         ip6gre_tnl_parm_from_user(&p1, &p);
1080                         t = ip6gre_tunnel_locate(net, &p1, 0);
1081                 }
1082                 if (t == NULL)
1083                         t = netdev_priv(dev);
1084                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1085                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1086                         err = -EFAULT;
1087                 break;
1088
1089         case SIOCADDTUNNEL:
1090         case SIOCCHGTUNNEL:
1091                 err = -EPERM;
1092                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1093                         goto done;
1094
1095                 err = -EFAULT;
1096                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1097                         goto done;
1098
1099                 err = -EINVAL;
1100                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1101                         goto done;
1102
1103                 if (!(p.i_flags&GRE_KEY))
1104                         p.i_key = 0;
1105                 if (!(p.o_flags&GRE_KEY))
1106                         p.o_key = 0;
1107
1108                 ip6gre_tnl_parm_from_user(&p1, &p);
1109                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1110
1111                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1112                         if (t != NULL) {
1113                                 if (t->dev != dev) {
1114                                         err = -EEXIST;
1115                                         break;
1116                                 }
1117                         } else {
1118                                 t = netdev_priv(dev);
1119
1120                                 ip6gre_tunnel_unlink(ign, t);
1121                                 synchronize_net();
1122                                 ip6gre_tnl_change(t, &p1, 1);
1123                                 ip6gre_tunnel_link(ign, t);
1124                                 netdev_state_change(dev);
1125                         }
1126                 }
1127
1128                 if (t) {
1129                         err = 0;
1130
1131                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1132                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1133                                 err = -EFAULT;
1134                 } else
1135                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1136                 break;
1137
1138         case SIOCDELTUNNEL:
1139                 err = -EPERM;
1140                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1141                         goto done;
1142
1143                 if (dev == ign->fb_tunnel_dev) {
1144                         err = -EFAULT;
1145                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1146                                 goto done;
1147                         err = -ENOENT;
1148                         ip6gre_tnl_parm_from_user(&p1, &p);
1149                         t = ip6gre_tunnel_locate(net, &p1, 0);
1150                         if (t == NULL)
1151                                 goto done;
1152                         err = -EPERM;
1153                         if (t == netdev_priv(ign->fb_tunnel_dev))
1154                                 goto done;
1155                         dev = t->dev;
1156                 }
1157                 unregister_netdevice(dev);
1158                 err = 0;
1159                 break;
1160
1161         default:
1162                 err = -EINVAL;
1163         }
1164
1165 done:
1166         return err;
1167 }
1168
1169 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1170 {
1171         struct ip6_tnl *tunnel = netdev_priv(dev);
1172         if (new_mtu < 68 ||
1173             new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1174                 return -EINVAL;
1175         dev->mtu = new_mtu;
1176         return 0;
1177 }
1178
1179 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1180                         unsigned short type,
1181                         const void *daddr, const void *saddr, unsigned int len)
1182 {
1183         struct ip6_tnl *t = netdev_priv(dev);
1184         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1185         __be16 *p = (__be16 *)(ipv6h+1);
1186
1187         ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
1188         ipv6h->hop_limit = t->parms.hop_limit;
1189         ipv6h->nexthdr = NEXTHDR_GRE;
1190         ipv6h->saddr = t->parms.laddr;
1191         ipv6h->daddr = t->parms.raddr;
1192
1193         p[0]            = t->parms.o_flags;
1194         p[1]            = htons(type);
1195
1196         /*
1197          *      Set the source hardware address.
1198          */
1199
1200         if (saddr)
1201                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1202         if (daddr)
1203                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1204         if (!ipv6_addr_any(&ipv6h->daddr))
1205                 return t->hlen;
1206
1207         return -t->hlen;
1208 }
1209
1210 static const struct header_ops ip6gre_header_ops = {
1211         .create = ip6gre_header,
1212 };
1213
1214 static const struct net_device_ops ip6gre_netdev_ops = {
1215         .ndo_init               = ip6gre_tunnel_init,
1216         .ndo_uninit             = ip6gre_tunnel_uninit,
1217         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1218         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1219         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1220         .ndo_get_stats64        = ip_tunnel_get_stats64,
1221 };
1222
1223 static void ip6gre_dev_free(struct net_device *dev)
1224 {
1225         free_percpu(dev->tstats);
1226         free_netdev(dev);
1227 }
1228
1229 static void ip6gre_tunnel_setup(struct net_device *dev)
1230 {
1231         struct ip6_tnl *t;
1232
1233         dev->netdev_ops = &ip6gre_netdev_ops;
1234         dev->destructor = ip6gre_dev_free;
1235
1236         dev->type = ARPHRD_IP6GRE;
1237         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1238         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1239         t = netdev_priv(dev);
1240         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1241                 dev->mtu -= 8;
1242         dev->flags |= IFF_NOARP;
1243         dev->iflink = 0;
1244         dev->addr_len = sizeof(struct in6_addr);
1245         dev->features |= NETIF_F_NETNS_LOCAL;
1246         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1247 }
1248
1249 static int ip6gre_tunnel_init(struct net_device *dev)
1250 {
1251         struct ip6_tnl *tunnel;
1252
1253         tunnel = netdev_priv(dev);
1254
1255         tunnel->dev = dev;
1256         strcpy(tunnel->parms.name, dev->name);
1257
1258         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1259         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1260
1261         if (ipv6_addr_any(&tunnel->parms.raddr))
1262                 dev->header_ops = &ip6gre_header_ops;
1263
1264         dev->tstats = alloc_percpu(struct pcpu_tstats);
1265         if (!dev->tstats)
1266                 return -ENOMEM;
1267
1268         return 0;
1269 }
1270
1271 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1272 {
1273         struct ip6_tnl *tunnel = netdev_priv(dev);
1274
1275         tunnel->dev = dev;
1276         strcpy(tunnel->parms.name, dev->name);
1277
1278         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1279
1280         dev_hold(dev);
1281 }
1282
1283
1284 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1285         .handler     = ip6gre_rcv,
1286         .err_handler = ip6gre_err,
1287         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1288 };
1289
1290 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1291         struct list_head *head)
1292 {
1293         int prio;
1294
1295         for (prio = 0; prio < 4; prio++) {
1296                 int h;
1297                 for (h = 0; h < HASH_SIZE; h++) {
1298                         struct ip6_tnl *t;
1299
1300                         t = rtnl_dereference(ign->tunnels[prio][h]);
1301
1302                         while (t != NULL) {
1303                                 unregister_netdevice_queue(t->dev, head);
1304                                 t = rtnl_dereference(t->next);
1305                         }
1306                 }
1307         }
1308 }
1309
1310 static int __net_init ip6gre_init_net(struct net *net)
1311 {
1312         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1313         int err;
1314
1315         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1316                                            ip6gre_tunnel_setup);
1317         if (!ign->fb_tunnel_dev) {
1318                 err = -ENOMEM;
1319                 goto err_alloc_dev;
1320         }
1321         dev_net_set(ign->fb_tunnel_dev, net);
1322
1323         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1324         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1325
1326         err = register_netdev(ign->fb_tunnel_dev);
1327         if (err)
1328                 goto err_reg_dev;
1329
1330         rcu_assign_pointer(ign->tunnels_wc[0],
1331                            netdev_priv(ign->fb_tunnel_dev));
1332         return 0;
1333
1334 err_reg_dev:
1335         ip6gre_dev_free(ign->fb_tunnel_dev);
1336 err_alloc_dev:
1337         return err;
1338 }
1339
1340 static void __net_exit ip6gre_exit_net(struct net *net)
1341 {
1342         struct ip6gre_net *ign;
1343         LIST_HEAD(list);
1344
1345         ign = net_generic(net, ip6gre_net_id);
1346         rtnl_lock();
1347         ip6gre_destroy_tunnels(ign, &list);
1348         unregister_netdevice_many(&list);
1349         rtnl_unlock();
1350 }
1351
1352 static struct pernet_operations ip6gre_net_ops = {
1353         .init = ip6gre_init_net,
1354         .exit = ip6gre_exit_net,
1355         .id   = &ip6gre_net_id,
1356         .size = sizeof(struct ip6gre_net),
1357 };
1358
1359 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1360 {
1361         __be16 flags;
1362
1363         if (!data)
1364                 return 0;
1365
1366         flags = 0;
1367         if (data[IFLA_GRE_IFLAGS])
1368                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1369         if (data[IFLA_GRE_OFLAGS])
1370                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1371         if (flags & (GRE_VERSION|GRE_ROUTING))
1372                 return -EINVAL;
1373
1374         return 0;
1375 }
1376
1377 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1378 {
1379         struct in6_addr daddr;
1380
1381         if (tb[IFLA_ADDRESS]) {
1382                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1383                         return -EINVAL;
1384                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1385                         return -EADDRNOTAVAIL;
1386         }
1387
1388         if (!data)
1389                 goto out;
1390
1391         if (data[IFLA_GRE_REMOTE]) {
1392                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1393                 if (ipv6_addr_any(&daddr))
1394                         return -EINVAL;
1395         }
1396
1397 out:
1398         return ip6gre_tunnel_validate(tb, data);
1399 }
1400
1401
1402 static void ip6gre_netlink_parms(struct nlattr *data[],
1403                                 struct __ip6_tnl_parm *parms)
1404 {
1405         memset(parms, 0, sizeof(*parms));
1406
1407         if (!data)
1408                 return;
1409
1410         if (data[IFLA_GRE_LINK])
1411                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1412
1413         if (data[IFLA_GRE_IFLAGS])
1414                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1415
1416         if (data[IFLA_GRE_OFLAGS])
1417                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1418
1419         if (data[IFLA_GRE_IKEY])
1420                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1421
1422         if (data[IFLA_GRE_OKEY])
1423                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1424
1425         if (data[IFLA_GRE_LOCAL])
1426                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1427
1428         if (data[IFLA_GRE_REMOTE])
1429                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1430
1431         if (data[IFLA_GRE_TTL])
1432                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1433
1434         if (data[IFLA_GRE_ENCAP_LIMIT])
1435                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1436
1437         if (data[IFLA_GRE_FLOWINFO])
1438                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1439
1440         if (data[IFLA_GRE_FLAGS])
1441                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1442 }
1443
1444 static int ip6gre_tap_init(struct net_device *dev)
1445 {
1446         struct ip6_tnl *tunnel;
1447
1448         tunnel = netdev_priv(dev);
1449
1450         tunnel->dev = dev;
1451         strcpy(tunnel->parms.name, dev->name);
1452
1453         ip6gre_tnl_link_config(tunnel, 1);
1454
1455         dev->tstats = alloc_percpu(struct pcpu_tstats);
1456         if (!dev->tstats)
1457                 return -ENOMEM;
1458
1459         return 0;
1460 }
1461
1462 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1463         .ndo_init = ip6gre_tap_init,
1464         .ndo_uninit = ip6gre_tunnel_uninit,
1465         .ndo_start_xmit = ip6gre_tunnel_xmit,
1466         .ndo_set_mac_address = eth_mac_addr,
1467         .ndo_validate_addr = eth_validate_addr,
1468         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1469         .ndo_get_stats64 = ip_tunnel_get_stats64,
1470 };
1471
1472 static void ip6gre_tap_setup(struct net_device *dev)
1473 {
1474
1475         ether_setup(dev);
1476
1477         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1478         dev->destructor = ip6gre_dev_free;
1479
1480         dev->iflink = 0;
1481         dev->features |= NETIF_F_NETNS_LOCAL;
1482 }
1483
1484 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1485         struct nlattr *tb[], struct nlattr *data[])
1486 {
1487         struct ip6_tnl *nt;
1488         struct net *net = dev_net(dev);
1489         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1490         int err;
1491
1492         nt = netdev_priv(dev);
1493         ip6gre_netlink_parms(data, &nt->parms);
1494
1495         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1496                 return -EEXIST;
1497
1498         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1499                 eth_hw_addr_random(dev);
1500
1501         nt->dev = dev;
1502         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1503
1504         /* Can use a lockless transmit, unless we generate output sequences */
1505         if (!(nt->parms.o_flags & GRE_SEQ))
1506                 dev->features |= NETIF_F_LLTX;
1507
1508         err = register_netdevice(dev);
1509         if (err)
1510                 goto out;
1511
1512         dev_hold(dev);
1513         ip6gre_tunnel_link(ign, nt);
1514
1515 out:
1516         return err;
1517 }
1518
1519 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1520                             struct nlattr *data[])
1521 {
1522         struct ip6_tnl *t, *nt;
1523         struct net *net = dev_net(dev);
1524         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1525         struct __ip6_tnl_parm p;
1526
1527         if (dev == ign->fb_tunnel_dev)
1528                 return -EINVAL;
1529
1530         nt = netdev_priv(dev);
1531         ip6gre_netlink_parms(data, &p);
1532
1533         t = ip6gre_tunnel_locate(net, &p, 0);
1534
1535         if (t) {
1536                 if (t->dev != dev)
1537                         return -EEXIST;
1538         } else {
1539                 t = nt;
1540
1541                 ip6gre_tunnel_unlink(ign, t);
1542                 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1543                 ip6gre_tunnel_link(ign, t);
1544                 netdev_state_change(dev);
1545         }
1546
1547         return 0;
1548 }
1549
1550 static size_t ip6gre_get_size(const struct net_device *dev)
1551 {
1552         return
1553                 /* IFLA_GRE_LINK */
1554                 nla_total_size(4) +
1555                 /* IFLA_GRE_IFLAGS */
1556                 nla_total_size(2) +
1557                 /* IFLA_GRE_OFLAGS */
1558                 nla_total_size(2) +
1559                 /* IFLA_GRE_IKEY */
1560                 nla_total_size(4) +
1561                 /* IFLA_GRE_OKEY */
1562                 nla_total_size(4) +
1563                 /* IFLA_GRE_LOCAL */
1564                 nla_total_size(sizeof(struct in6_addr)) +
1565                 /* IFLA_GRE_REMOTE */
1566                 nla_total_size(sizeof(struct in6_addr)) +
1567                 /* IFLA_GRE_TTL */
1568                 nla_total_size(1) +
1569                 /* IFLA_GRE_TOS */
1570                 nla_total_size(1) +
1571                 /* IFLA_GRE_ENCAP_LIMIT */
1572                 nla_total_size(1) +
1573                 /* IFLA_GRE_FLOWINFO */
1574                 nla_total_size(4) +
1575                 /* IFLA_GRE_FLAGS */
1576                 nla_total_size(4) +
1577                 0;
1578 }
1579
1580 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1581 {
1582         struct ip6_tnl *t = netdev_priv(dev);
1583         struct __ip6_tnl_parm *p = &t->parms;
1584
1585         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1586             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1587             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1588             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1589             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1590             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1591             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1592             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1593             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1594             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1595             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1596             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1597                 goto nla_put_failure;
1598         return 0;
1599
1600 nla_put_failure:
1601         return -EMSGSIZE;
1602 }
1603
1604 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1605         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1606         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1607         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1608         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1609         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1610         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1611         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1612         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1613         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1614         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1615         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1616 };
1617
1618 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1619         .kind           = "ip6gre",
1620         .maxtype        = IFLA_GRE_MAX,
1621         .policy         = ip6gre_policy,
1622         .priv_size      = sizeof(struct ip6_tnl),
1623         .setup          = ip6gre_tunnel_setup,
1624         .validate       = ip6gre_tunnel_validate,
1625         .newlink        = ip6gre_newlink,
1626         .changelink     = ip6gre_changelink,
1627         .get_size       = ip6gre_get_size,
1628         .fill_info      = ip6gre_fill_info,
1629 };
1630
1631 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1632         .kind           = "ip6gretap",
1633         .maxtype        = IFLA_GRE_MAX,
1634         .policy         = ip6gre_policy,
1635         .priv_size      = sizeof(struct ip6_tnl),
1636         .setup          = ip6gre_tap_setup,
1637         .validate       = ip6gre_tap_validate,
1638         .newlink        = ip6gre_newlink,
1639         .changelink     = ip6gre_changelink,
1640         .get_size       = ip6gre_get_size,
1641         .fill_info      = ip6gre_fill_info,
1642 };
1643
1644 /*
1645  *      And now the modules code and kernel interface.
1646  */
1647
1648 static int __init ip6gre_init(void)
1649 {
1650         int err;
1651
1652         pr_info("GRE over IPv6 tunneling driver\n");
1653
1654         err = register_pernet_device(&ip6gre_net_ops);
1655         if (err < 0)
1656                 return err;
1657
1658         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1659         if (err < 0) {
1660                 pr_info("%s: can't add protocol\n", __func__);
1661                 goto add_proto_failed;
1662         }
1663
1664         err = rtnl_link_register(&ip6gre_link_ops);
1665         if (err < 0)
1666                 goto rtnl_link_failed;
1667
1668         err = rtnl_link_register(&ip6gre_tap_ops);
1669         if (err < 0)
1670                 goto tap_ops_failed;
1671
1672 out:
1673         return err;
1674
1675 tap_ops_failed:
1676         rtnl_link_unregister(&ip6gre_link_ops);
1677 rtnl_link_failed:
1678         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1679 add_proto_failed:
1680         unregister_pernet_device(&ip6gre_net_ops);
1681         goto out;
1682 }
1683
1684 static void __exit ip6gre_fini(void)
1685 {
1686         rtnl_link_unregister(&ip6gre_tap_ops);
1687         rtnl_link_unregister(&ip6gre_link_ops);
1688         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1689         unregister_pernet_device(&ip6gre_net_ops);
1690 }
1691
1692 module_init(ip6gre_init);
1693 module_exit(ip6gre_fini);
1694 MODULE_LICENSE("GPL");
1695 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1696 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1697 MODULE_ALIAS_RTNL_LINK("ip6gre");
1698 MODULE_ALIAS_NETDEV("ip6gre0");