net/ipv6/ip6_gre.c

   1 /*
   2  *      GRE over IPv6 protocol decoder.
   3  *
   4  *      Authors: Dmitry Kozlov (xeb@mail.ru)
   5  *
   6  *      This program is free software; you can redistribute it and/or
   7  *      modify it under the terms of the GNU General Public License
   8  *      as published by the Free Software Foundation; either version
   9  *      2 of the License, or (at your option) any later version.
  10  *
  11  */
  12
  13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  14
  15 #include <linux/capability.h>
  16 #include <linux/module.h>
  17 #include <linux/types.h>
  18 #include <linux/kernel.h>
  19 #include <linux/slab.h>
  20 #include <linux/uaccess.h>
  21 #include <linux/skbuff.h>
  22 #include <linux/netdevice.h>
  23 #include <linux/in.h>
  24 #include <linux/tcp.h>
  25 #include <linux/udp.h>
  26 #include <linux/if_arp.h>
  27 #include <linux/mroute.h>
  28 #include <linux/init.h>
  29 #include <linux/in6.h>
  30 #include <linux/inetdevice.h>
  31 #include <linux/igmp.h>
  32 #include <linux/netfilter_ipv4.h>
  33 #include <linux/etherdevice.h>
  34 #include <linux/if_ether.h>
  35 #include <linux/hash.h>
  36 #include <linux/if_tunnel.h>
  37 #include <linux/ip6_tunnel.h>
  38
  39 #include <net/sock.h>
  40 #include <net/ip.h>
  41 #include <net/icmp.h>
  42 #include <net/protocol.h>
  43 #include <net/addrconf.h>
  44 #include <net/arp.h>
  45 #include <net/checksum.h>
  46 #include <net/dsfield.h>
  47 #include <net/inet_ecn.h>
  48 #include <net/xfrm.h>
  49 #include <net/net_namespace.h>
  50 #include <net/netns/generic.h>
  51 #include <net/rtnetlink.h>
  52
  53 #include <net/ipv6.h>
  54 #include <net/ip6_fib.h>
  55 #include <net/ip6_route.h>
  56 #include <net/ip6_tunnel.h>
  57
  58
  59 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 IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
  64 #define IPV6_TCLASS_SHIFT 20
  65
  66 #define HASH_SIZE_SHIFT  5
  67 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  68
  69 static int ip6gre_net_id __read_mostly;
  70 struct ip6gre_net {
  71         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
  72
  73         struct net_device *fb_tunnel_dev;
  74 };
  75
  76 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  77 static int ip6gre_tunnel_init(struct net_device *dev);
  78 static void ip6gre_tunnel_setup(struct net_device *dev);
  79 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  80 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  81
  82 /* Tunnel hash table */
  83
  84 /*
  85    4 hash tables:
  86
  87    3: (remote,local)
  88    2: (remote,*)
  89    1: (*,local)
  90    0: (*,*)
  91
  92    We require exact key match i.e. if a key is present in packet
  93    it will match only tunnel with the same key; if it is not present,
  94    it will match only keyless tunnel.
  95
  96    All keysless packets, if not matched configured keyless tunnels
  97    will match fallback tunnel.
  98  */
  99
 100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
 101 static u32 HASH_ADDR(const struct in6_addr *addr)
 102 {
 103         u32 hash = ipv6_addr_hash(addr);
 104
 105         return hash_32(hash, HASH_SIZE_SHIFT);
 106 }
 107
 108 #define tunnels_r_l     tunnels[3]
 109 #define tunnels_r       tunnels[2]
 110 #define tunnels_l       tunnels[1]
 111 #define tunnels_wc      tunnels[0]
 112 /*
 113  * Locking : hash tables are protected by RCU and RTNL
 114  */
 115
 116 #define for_each_ip_tunnel_rcu(start) \
 117         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 118
 119 /* often modified stats are per cpu, other are shared (netdev->stats) */
 120 struct pcpu_tstats {
 121         u64     rx_packets;
 122         u64     rx_bytes;
 123         u64     tx_packets;
 124         u64     tx_bytes;
 125         struct u64_stats_sync   syncp;
 126 };
 127
 128 static struct rtnl_link_stats64 *ip6gre_get_stats64(struct net_device *dev,
 129                 struct rtnl_link_stats64 *tot)
 130 {
 131         int i;
 132
 133         for_each_possible_cpu(i) {
 134                 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
 135                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
 136                 unsigned int start;
 137
 138                 do {
 139                         start = u64_stats_fetch_begin_bh(&tstats->syncp);
 140                         rx_packets = tstats->rx_packets;
 141                         tx_packets = tstats->tx_packets;
 142                         rx_bytes = tstats->rx_bytes;
 143                         tx_bytes = tstats->tx_bytes;
 144                 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
 145
 146                 tot->rx_packets += rx_packets;
 147                 tot->tx_packets += tx_packets;
 148                 tot->rx_bytes   += rx_bytes;
 149                 tot->tx_bytes   += tx_bytes;
 150         }
 151
 152         tot->multicast = dev->stats.multicast;
 153         tot->rx_crc_errors = dev->stats.rx_crc_errors;
 154         tot->rx_fifo_errors = dev->stats.rx_fifo_errors;
 155         tot->rx_length_errors = dev->stats.rx_length_errors;
 156         tot->rx_frame_errors = dev->stats.rx_frame_errors;
 157         tot->rx_errors = dev->stats.rx_errors;
 158
 159         tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
 160         tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
 161         tot->tx_dropped = dev->stats.tx_dropped;
 162         tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
 163         tot->tx_errors = dev->stats.tx_errors;
 164
 165         return tot;
 166 }
 167
 168 /* Given src, dst and key, find appropriate for input tunnel. */
 169
 170 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 171                 const struct in6_addr *remote, const struct in6_addr *local,
 172                 __be32 key, __be16 gre_proto)
 173 {
 174         struct net *net = dev_net(dev);
 175         int link = dev->ifindex;
 176         unsigned int h0 = HASH_ADDR(remote);
 177         unsigned int h1 = HASH_KEY(key);
 178         struct ip6_tnl *t, *cand = NULL;
 179         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 180         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
 181                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 182         int score, cand_score = 4;
 183
 184         for_each_ip_tunnel_rcu(ign->tunnels_r_l[h0 ^ h1]) {
 185                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 186                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 187                     key != t->parms.i_key ||
 188                     !(t->dev->flags & IFF_UP))
 189                         continue;
 190
 191                 if (t->dev->type != ARPHRD_IP6GRE &&
 192                     t->dev->type != dev_type)
 193                         continue;
 194
 195                 score = 0;
 196                 if (t->parms.link != link)
 197                         score |= 1;
 198                 if (t->dev->type != dev_type)
 199                         score |= 2;
 200                 if (score == 0)
 201                         return t;
 202
 203                 if (score < cand_score) {
 204                         cand = t;
 205                         cand_score = score;
 206                 }
 207         }
 208
 209         for_each_ip_tunnel_rcu(ign->tunnels_r[h0 ^ h1]) {
 210                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 211                     key != t->parms.i_key ||
 212                     !(t->dev->flags & IFF_UP))
 213                         continue;
 214
 215                 if (t->dev->type != ARPHRD_IP6GRE &&
 216                     t->dev->type != dev_type)
 217                         continue;
 218
 219                 score = 0;
 220                 if (t->parms.link != link)
 221                         score |= 1;
 222                 if (t->dev->type != dev_type)
 223                         score |= 2;
 224                 if (score == 0)
 225                         return t;
 226
 227                 if (score < cand_score) {
 228                         cand = t;
 229                         cand_score = score;
 230                 }
 231         }
 232
 233         for_each_ip_tunnel_rcu(ign->tunnels_l[h1]) {
 234                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 235                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 236                                  !ipv6_addr_is_multicast(local))) ||
 237                     key != t->parms.i_key ||
 238                     !(t->dev->flags & IFF_UP))
 239                         continue;
 240
 241                 if (t->dev->type != ARPHRD_IP6GRE &&
 242                     t->dev->type != dev_type)
 243                         continue;
 244
 245                 score = 0;
 246                 if (t->parms.link != link)
 247                         score |= 1;
 248                 if (t->dev->type != dev_type)
 249                         score |= 2;
 250                 if (score == 0)
 251                         return t;
 252
 253                 if (score < cand_score) {
 254                         cand = t;
 255                         cand_score = score;
 256                 }
 257         }
 258
 259         for_each_ip_tunnel_rcu(ign->tunnels_wc[h1]) {
 260                 if (t->parms.i_key != key ||
 261                     !(t->dev->flags & IFF_UP))
 262                         continue;
 263
 264                 if (t->dev->type != ARPHRD_IP6GRE &&
 265                     t->dev->type != dev_type)
 266                         continue;
 267
 268                 score = 0;
 269                 if (t->parms.link != link)
 270                         score |= 1;
 271                 if (t->dev->type != dev_type)
 272                         score |= 2;
 273                 if (score == 0)
 274                         return t;
 275
 276                 if (score < cand_score) {
 277                         cand = t;
 278                         cand_score = score;
 279                 }
 280         }
 281
 282         if (cand != NULL)
 283                 return cand;
 284
 285         dev = ign->fb_tunnel_dev;
 286         if (dev->flags & IFF_UP)
 287                 return netdev_priv(dev);
 288
 289         return NULL;
 290 }
 291
 292 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 293                 const struct __ip6_tnl_parm *p)
 294 {
 295         const struct in6_addr *remote = &p->raddr;
 296         const struct in6_addr *local = &p->laddr;
 297         unsigned int h = HASH_KEY(p->i_key);
 298         int prio = 0;
 299
 300         if (!ipv6_addr_any(local))
 301                 prio |= 1;
 302         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 303                 prio |= 2;
 304                 h ^= HASH_ADDR(remote);
 305         }
 306
 307         return &ign->tunnels[prio][h];
 308 }
 309
 310 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 311                 const struct ip6_tnl *t)
 312 {
 313         return __ip6gre_bucket(ign, &t->parms);
 314 }
 315
 316 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 317 {
 318         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 319
 320         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 321         rcu_assign_pointer(*tp, t);
 322 }
 323
 324 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 325 {
 326         struct ip6_tnl __rcu **tp;
 327         struct ip6_tnl *iter;
 328
 329         for (tp = ip6gre_bucket(ign, t);
 330              (iter = rtnl_dereference(*tp)) != NULL;
 331              tp = &iter->next) {
 332                 if (t == iter) {
 333                         rcu_assign_pointer(*tp, t->next);
 334                         break;
 335                 }
 336         }
 337 }
 338
 339 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 340                                            const struct __ip6_tnl_parm *parms,
 341                                            int type)
 342 {
 343         const struct in6_addr *remote = &parms->raddr;
 344         const struct in6_addr *local = &parms->laddr;
 345         __be32 key = parms->i_key;
 346         int link = parms->link;
 347         struct ip6_tnl *t;
 348         struct ip6_tnl __rcu **tp;
 349         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 350
 351         for (tp = __ip6gre_bucket(ign, parms);
 352              (t = rtnl_dereference(*tp)) != NULL;
 353              tp = &t->next)
 354                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 355                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 356                     key == t->parms.i_key &&
 357                     link == t->parms.link &&
 358                     type == t->dev->type)
 359                         break;
 360
 361         return t;
 362 }
 363
 364 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 365                 const struct __ip6_tnl_parm *parms, int create)
 366 {
 367         struct ip6_tnl *t, *nt;
 368         struct net_device *dev;
 369         char name[IFNAMSIZ];
 370         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 371
 372         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 373         if (t || !create)
 374                 return t;
 375
 376         if (parms->name[0])
 377                 strlcpy(name, parms->name, IFNAMSIZ);
 378         else
 379                 strcpy(name, "ip6gre%d");
 380
 381         dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
 382         if (!dev)
 383                 return NULL;
 384
 385         dev_net_set(dev, net);
 386
 387         nt = netdev_priv(dev);
 388         nt->parms = *parms;
 389         dev->rtnl_link_ops = &ip6gre_link_ops;
 390
 391         nt->dev = dev;
 392         ip6gre_tnl_link_config(nt, 1);
 393
 394         if (register_netdevice(dev) < 0)
 395                 goto failed_free;
 396
 397         /* Can use a lockless transmit, unless we generate output sequences */
 398         if (!(nt->parms.o_flags & GRE_SEQ))
 399                 dev->features |= NETIF_F_LLTX;
 400
 401         dev_hold(dev);
 402         ip6gre_tunnel_link(ign, nt);
 403         return nt;
 404
 405 failed_free:
 406         free_netdev(dev);
 407         return NULL;
 408 }
 409
 410 static void ip6gre_tunnel_uninit(struct net_device *dev)
 411 {
 412         struct net *net = dev_net(dev);
 413         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 414
 415         ip6gre_tunnel_unlink(ign, netdev_priv(dev));
 416         dev_put(dev);
 417 }
 418
 419
 420 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 421                 u8 type, u8 code, int offset, __be32 info)
 422 {
 423         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 424         __be16 *p = (__be16 *)(skb->data + offset);
 425         int grehlen = offset + 4;
 426         struct ip6_tnl *t;
 427         __be16 flags;
 428
 429         flags = p[0];
 430         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
 431                 if (flags&(GRE_VERSION|GRE_ROUTING))
 432                         return;
 433                 if (flags&GRE_KEY) {
 434                         grehlen += 4;
 435                         if (flags&GRE_CSUM)
 436                                 grehlen += 4;
 437                 }
 438         }
 439
 440         /* If only 8 bytes returned, keyed message will be dropped here */
 441         if (!pskb_may_pull(skb, grehlen))
 442                 return;
 443         ipv6h = (const struct ipv6hdr *)skb->data;
 444         p = (__be16 *)(skb->data + offset);
 445
 446         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 447                                 flags & GRE_KEY ?
 448                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
 449                                 p[1]);
 450         if (t == NULL)
 451                 return;
 452
 453         switch (type) {
 454                 __u32 teli;
 455                 struct ipv6_tlv_tnl_enc_lim *tel;
 456                 __u32 mtu;
 457         case ICMPV6_DEST_UNREACH:
 458                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
 459                                      t->parms.name);
 460                 break;
 461         case ICMPV6_TIME_EXCEED:
 462                 if (code == ICMPV6_EXC_HOPLIMIT) {
 463                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 464                                              t->parms.name);
 465                 }
 466                 break;
 467         case ICMPV6_PARAMPROB:
 468                 teli = 0;
 469                 if (code == ICMPV6_HDR_FIELD)
 470                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 471
 472                 if (teli && teli == info - 2) {
 473                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 474                         if (tel->encap_limit == 0) {
 475                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 476                                                      t->parms.name);
 477                         }
 478                 } else {
 479                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 480                                              t->parms.name);
 481                 }
 482                 break;
 483         case ICMPV6_PKT_TOOBIG:
 484                 mtu = info - offset;
 485                 if (mtu < IPV6_MIN_MTU)
 486                         mtu = IPV6_MIN_MTU;
 487                 t->dev->mtu = mtu;
 488                 break;
 489         }
 490
 491         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 492                 t->err_count++;
 493         else
 494                 t->err_count = 1;
 495         t->err_time = jiffies;
 496 }
 497
 498 static int ip6gre_rcv(struct sk_buff *skb)
 499 {
 500         const struct ipv6hdr *ipv6h;
 501         u8     *h;
 502         __be16    flags;
 503         __sum16   csum = 0;
 504         __be32 key = 0;
 505         u32    seqno = 0;
 506         struct ip6_tnl *tunnel;
 507         int    offset = 4;
 508         __be16 gre_proto;
 509         int err;
 510
 511         if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
 512                 goto drop;
 513
 514         ipv6h = ipv6_hdr(skb);
 515         h = skb->data;
 516         flags = *(__be16 *)h;
 517
 518         if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
 519                 /* - Version must be 0.
 520                    - We do not support routing headers.
 521                  */
 522                 if (flags&(GRE_VERSION|GRE_ROUTING))
 523                         goto drop;
 524
 525                 if (flags&GRE_CSUM) {
 526                         switch (skb->ip_summed) {
 527                         case CHECKSUM_COMPLETE:
 528                                 csum = csum_fold(skb->csum);
 529                                 if (!csum)
 530                                         break;
 531                                 /* fall through */
 532                         case CHECKSUM_NONE:
 533                                 skb->csum = 0;
 534                                 csum = __skb_checksum_complete(skb);
 535                                 skb->ip_summed = CHECKSUM_COMPLETE;
 536                         }
 537                         offset += 4;
 538                 }
 539                 if (flags&GRE_KEY) {
 540                         key = *(__be32 *)(h + offset);
 541                         offset += 4;
 542                 }
 543                 if (flags&GRE_SEQ) {
 544                         seqno = ntohl(*(__be32 *)(h + offset));
 545                         offset += 4;
 546                 }
 547         }
 548
 549         gre_proto = *(__be16 *)(h + 2);
 550
 551         tunnel = ip6gre_tunnel_lookup(skb->dev,
 552                                           &ipv6h->saddr, &ipv6h->daddr, key,
 553                                           gre_proto);
 554         if (tunnel) {
 555                 struct pcpu_tstats *tstats;
 556
 557                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 558                         goto drop;
 559
 560                 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
 561                         tunnel->dev->stats.rx_dropped++;
 562                         goto drop;
 563                 }
 564
 565                 secpath_reset(skb);
 566
 567                 skb->protocol = gre_proto;
 568                 /* WCCP version 1 and 2 protocol decoding.
 569                  * - Change protocol to IP
 570                  * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
 571                  */
 572                 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
 573                         skb->protocol = htons(ETH_P_IP);
 574                         if ((*(h + offset) & 0xF0) != 0x40)
 575                                 offset += 4;
 576                 }
 577
 578                 skb->mac_header = skb->network_header;
 579                 __pskb_pull(skb, offset);
 580                 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
 581                 skb->pkt_type = PACKET_HOST;
 582
 583                 if (((flags&GRE_CSUM) && csum) ||
 584                     (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
 585                         tunnel->dev->stats.rx_crc_errors++;
 586                         tunnel->dev->stats.rx_errors++;
 587                         goto drop;
 588                 }
 589                 if (tunnel->parms.i_flags&GRE_SEQ) {
 590                         if (!(flags&GRE_SEQ) ||
 591                             (tunnel->i_seqno &&
 592                                         (s32)(seqno - tunnel->i_seqno) < 0)) {
 593                                 tunnel->dev->stats.rx_fifo_errors++;
 594                                 tunnel->dev->stats.rx_errors++;
 595                                 goto drop;
 596                         }
 597                         tunnel->i_seqno = seqno + 1;
 598                 }
 599
 600                 /* Warning: All skb pointers will be invalidated! */
 601                 if (tunnel->dev->type == ARPHRD_ETHER) {
 602                         if (!pskb_may_pull(skb, ETH_HLEN)) {
 603                                 tunnel->dev->stats.rx_length_errors++;
 604                                 tunnel->dev->stats.rx_errors++;
 605                                 goto drop;
 606                         }
 607
 608                         ipv6h = ipv6_hdr(skb);
 609                         skb->protocol = eth_type_trans(skb, tunnel->dev);
 610                         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 611                 }
 612
 613                 __skb_tunnel_rx(skb, tunnel->dev);
 614
 615                 skb_reset_network_header(skb);
 616
 617                 err = IP6_ECN_decapsulate(ipv6h, skb);
 618                 if (unlikely(err)) {
 619                         if (log_ecn_error)
 620                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 621                                                      &ipv6h->saddr,
 622                                                      ipv6_get_dsfield(ipv6h));
 623                         if (err > 1) {
 624                                 ++tunnel->dev->stats.rx_frame_errors;
 625                                 ++tunnel->dev->stats.rx_errors;
 626                                 goto drop;
 627                         }
 628                 }
 629
 630                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 631                 u64_stats_update_begin(&tstats->syncp);
 632                 tstats->rx_packets++;
 633                 tstats->rx_bytes += skb->len;
 634                 u64_stats_update_end(&tstats->syncp);
 635
 636                 netif_rx(skb);
 637
 638                 return 0;
 639         }
 640         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 641
 642 drop:
 643         kfree_skb(skb);
 644         return 0;
 645 }
 646
 647 struct ipv6_tel_txoption {
 648         struct ipv6_txoptions ops;
 649         __u8 dst_opt[8];
 650 };
 651
 652 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 653 {
 654         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 655
 656         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 657         opt->dst_opt[3] = 1;
 658         opt->dst_opt[4] = encap_limit;
 659         opt->dst_opt[5] = IPV6_TLV_PADN;
 660         opt->dst_opt[6] = 1;
 661
 662         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 663         opt->ops.opt_nflen = 8;
 664 }
 665
 666 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 667                          struct net_device *dev,
 668                          __u8 dsfield,
 669                          struct flowi6 *fl6,
 670                          int encap_limit,
 671                          __u32 *pmtu)
 672 {
 673         struct net *net = dev_net(dev);
 674         struct ip6_tnl *tunnel = netdev_priv(dev);
 675         struct net_device *tdev;    /* Device to other host */
 676         struct ipv6hdr  *ipv6h;     /* Our new IP header */
 677         unsigned int max_headroom;  /* The extra header space needed */
 678         int    gre_hlen;
 679         struct ipv6_tel_txoption opt;
 680         int    mtu;
 681         struct dst_entry *dst = NULL, *ndst = NULL;
 682         struct net_device_stats *stats = &tunnel->dev->stats;
 683         int err = -1;
 684         u8 proto;
 685         int pkt_len;
 686         struct sk_buff *new_skb;
 687
 688         if (dev->type == ARPHRD_ETHER)
 689                 IPCB(skb)->flags = 0;
 690
 691         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 692                 gre_hlen = 0;
 693                 ipv6h = (struct ipv6hdr *)skb->data;
 694                 fl6->daddr = ipv6h->daddr;
 695         } else {
 696                 gre_hlen = tunnel->hlen;
 697                 fl6->daddr = tunnel->parms.raddr;
 698         }
 699
 700         if (!fl6->flowi6_mark)
 701                 dst = ip6_tnl_dst_check(tunnel);
 702
 703         if (!dst) {
 704                 ndst = ip6_route_output(net, NULL, fl6);
 705
 706                 if (ndst->error)
 707                         goto tx_err_link_failure;
 708                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
 709                 if (IS_ERR(ndst)) {
 710                         err = PTR_ERR(ndst);
 711                         ndst = NULL;
 712                         goto tx_err_link_failure;
 713                 }
 714                 dst = ndst;
 715         }
 716
 717         tdev = dst->dev;
 718
 719         if (tdev == dev) {
 720                 stats->collisions++;
 721                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 722                                      tunnel->parms.name);
 723                 goto tx_err_dst_release;
 724         }
 725
 726         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 727         if (encap_limit >= 0) {
 728                 max_headroom += 8;
 729                 mtu -= 8;
 730         }
 731         if (mtu < IPV6_MIN_MTU)
 732                 mtu = IPV6_MIN_MTU;
 733         if (skb_dst(skb))
 734                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 735         if (skb->len > mtu) {
 736                 *pmtu = mtu;
 737                 err = -EMSGSIZE;
 738                 goto tx_err_dst_release;
 739         }
 740
 741         if (tunnel->err_count > 0) {
 742                 if (time_before(jiffies,
 743                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 744                         tunnel->err_count--;
 745
 746                         dst_link_failure(skb);
 747                 } else
 748                         tunnel->err_count = 0;
 749         }
 750
 751         max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 752
 753         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 754             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 755                 new_skb = skb_realloc_headroom(skb, max_headroom);
 756                 if (max_headroom > dev->needed_headroom)
 757                         dev->needed_headroom = max_headroom;
 758                 if (!new_skb)
 759                         goto tx_err_dst_release;
 760
 761                 if (skb->sk)
 762                         skb_set_owner_w(new_skb, skb->sk);
 763                 consume_skb(skb);
 764                 skb = new_skb;
 765         }
 766
 767         skb_dst_drop(skb);
 768
 769         if (fl6->flowi6_mark) {
 770                 skb_dst_set(skb, dst);
 771                 ndst = NULL;
 772         } else {
 773                 skb_dst_set_noref(skb, dst);
 774         }
 775
 776         skb->transport_header = skb->network_header;
 777
 778         proto = NEXTHDR_GRE;
 779         if (encap_limit >= 0) {
 780                 init_tel_txopt(&opt, encap_limit);
 781                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 782         }
 783
 784         skb_push(skb, gre_hlen);
 785         skb_reset_network_header(skb);
 786
 787         /*
 788          *      Push down and install the IP header.
 789          */
 790         ipv6h = ipv6_hdr(skb);
 791         *(__be32 *)ipv6h = fl6->flowlabel | htonl(0x60000000);
 792         dsfield = INET_ECN_encapsulate(0, dsfield);
 793         ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
 794         ipv6h->hop_limit = tunnel->parms.hop_limit;
 795         ipv6h->nexthdr = proto;
 796         ipv6h->saddr = fl6->saddr;
 797         ipv6h->daddr = fl6->daddr;
 798
 799         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 800         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
 801                                    htons(ETH_P_TEB) : skb->protocol;
 802
 803         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 804                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 805
 806                 if (tunnel->parms.o_flags&GRE_SEQ) {
 807                         ++tunnel->o_seqno;
 808                         *ptr = htonl(tunnel->o_seqno);
 809                         ptr--;
 810                 }
 811                 if (tunnel->parms.o_flags&GRE_KEY) {
 812                         *ptr = tunnel->parms.o_key;
 813                         ptr--;
 814                 }
 815                 if (tunnel->parms.o_flags&GRE_CSUM) {
 816                         *ptr = 0;
 817                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
 818                                 skb->len - sizeof(struct ipv6hdr));
 819                 }
 820         }
 821
 822         nf_reset(skb);
 823         pkt_len = skb->len;
 824         err = ip6_local_out(skb);
 825
 826         if (net_xmit_eval(err) == 0) {
 827                 struct pcpu_tstats *tstats = this_cpu_ptr(tunnel->dev->tstats);
 828
 829                 tstats->tx_bytes += pkt_len;
 830                 tstats->tx_packets++;
 831         } else {
 832                 stats->tx_errors++;
 833                 stats->tx_aborted_errors++;
 834         }
 835
 836         if (ndst)
 837                 ip6_tnl_dst_store(tunnel, ndst);
 838
 839         return 0;
 840 tx_err_link_failure:
 841         stats->tx_carrier_errors++;
 842         dst_link_failure(skb);
 843 tx_err_dst_release:
 844         dst_release(ndst);
 845         return err;
 846 }
 847
 848 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 849 {
 850         struct ip6_tnl *t = netdev_priv(dev);
 851         const struct iphdr  *iph = ip_hdr(skb);
 852         int encap_limit = -1;
 853         struct flowi6 fl6;
 854         __u8 dsfield;
 855         __u32 mtu;
 856         int err;
 857
 858         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 859                 encap_limit = t->parms.encap_limit;
 860
 861         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 862         fl6.flowi6_proto = IPPROTO_IPIP;
 863
 864         dsfield = ipv4_get_dsfield(iph);
 865
 866         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 867                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 868                                           & IPV6_TCLASS_MASK;
 869         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 870                 fl6.flowi6_mark = skb->mark;
 871
 872         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 873         if (err != 0) {
 874                 /* XXX: send ICMP error even if DF is not set. */
 875                 if (err == -EMSGSIZE)
 876                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 877                                   htonl(mtu));
 878                 return -1;
 879         }
 880
 881         return 0;
 882 }
 883
 884 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 885 {
 886         struct ip6_tnl *t = netdev_priv(dev);
 887         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 888         int encap_limit = -1;
 889         __u16 offset;
 890         struct flowi6 fl6;
 891         __u8 dsfield;
 892         __u32 mtu;
 893         int err;
 894
 895         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 896                 return -1;
 897
 898         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 899         if (offset > 0) {
 900                 struct ipv6_tlv_tnl_enc_lim *tel;
 901                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 902                 if (tel->encap_limit == 0) {
 903                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 904                                     ICMPV6_HDR_FIELD, offset + 2);
 905                         return -1;
 906                 }
 907                 encap_limit = tel->encap_limit - 1;
 908         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 909                 encap_limit = t->parms.encap_limit;
 910
 911         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 912         fl6.flowi6_proto = IPPROTO_IPV6;
 913
 914         dsfield = ipv6_get_dsfield(ipv6h);
 915         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 916                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 917         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 918                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 919         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 920                 fl6.flowi6_mark = skb->mark;
 921
 922         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 923         if (err != 0) {
 924                 if (err == -EMSGSIZE)
 925                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 926                 return -1;
 927         }
 928
 929         return 0;
 930 }
 931
 932 /**
 933  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 934  *   @t: the outgoing tunnel device
 935  *   @hdr: IPv6 header from the incoming packet
 936  *
 937  * Description:
 938  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 939  *   doesn't match source of incoming packet.
 940  *
 941  * Return:
 942  *   1 if conflict,
 943  *   0 else
 944  **/
 945
 946 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 947         const struct ipv6hdr *hdr)
 948 {
 949         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 950 }
 951
 952 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 953 {
 954         struct ip6_tnl *t = netdev_priv(dev);
 955         int encap_limit = -1;
 956         struct flowi6 fl6;
 957         __u32 mtu;
 958         int err;
 959
 960         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 961                 encap_limit = t->parms.encap_limit;
 962
 963         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 964         fl6.flowi6_proto = skb->protocol;
 965
 966         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 967
 968         return err;
 969 }
 970
 971 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 972         struct net_device *dev)
 973 {
 974         struct ip6_tnl *t = netdev_priv(dev);
 975         struct net_device_stats *stats = &t->dev->stats;
 976         int ret;
 977
 978         if (!ip6_tnl_xmit_ctl(t))
 979                 return -1;
 980
 981         switch (skb->protocol) {
 982         case htons(ETH_P_IP):
 983                 ret = ip6gre_xmit_ipv4(skb, dev);
 984                 break;
 985         case htons(ETH_P_IPV6):
 986                 ret = ip6gre_xmit_ipv6(skb, dev);
 987                 break;
 988         default:
 989                 ret = ip6gre_xmit_other(skb, dev);
 990                 break;
 991         }
 992
 993         if (ret < 0)
 994                 goto tx_err;
 995
 996         return NETDEV_TX_OK;
 997
 998 tx_err:
 999         stats->tx_errors++;
1000         stats->tx_dropped++;
1001         kfree_skb(skb);
1002         return NETDEV_TX_OK;
1003 }
1004
1005 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1006 {
1007         struct net_device *dev = t->dev;
1008         struct __ip6_tnl_parm *p = &t->parms;
1009         struct flowi6 *fl6 = &t->fl.u.ip6;
1010         int addend = sizeof(struct ipv6hdr) + 4;
1011
1012         if (dev->type != ARPHRD_ETHER) {
1013                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1014                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1015         }
1016
1017         /* Set up flowi template */
1018         fl6->saddr = p->laddr;
1019         fl6->daddr = p->raddr;
1020         fl6->flowi6_oif = p->link;
1021         fl6->flowlabel = 0;
1022
1023         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1024                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1025         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1026                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1027
1028         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1029         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1030
1031         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1032                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1033                 dev->flags |= IFF_POINTOPOINT;
1034         else
1035                 dev->flags &= ~IFF_POINTOPOINT;
1036
1037         dev->iflink = p->link;
1038
1039         /* Precalculate GRE options length */
1040         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
1041                 if (t->parms.o_flags&GRE_CSUM)
1042                         addend += 4;
1043                 if (t->parms.o_flags&GRE_KEY)
1044                         addend += 4;
1045                 if (t->parms.o_flags&GRE_SEQ)
1046                         addend += 4;
1047         }
1048
1049         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1050                 int strict = (ipv6_addr_type(&p->raddr) &
1051                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1052
1053                 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1054                                                  &p->raddr, &p->laddr,
1055                                                  p->link, strict);
1056
1057                 if (rt == NULL)
1058                         return;
1059
1060                 if (rt->dst.dev) {
1061                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
1062
1063                         if (set_mtu) {
1064                                 dev->mtu = rt->dst.dev->mtu - addend;
1065                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1066                                         dev->mtu -= 8;
1067
1068                                 if (dev->mtu < IPV6_MIN_MTU)
1069                                         dev->mtu = IPV6_MIN_MTU;
1070                         }
1071                 }
1072                 dst_release(&rt->dst);
1073         }
1074
1075         t->hlen = addend;
1076 }
1077
1078 static int ip6gre_tnl_change(struct ip6_tnl *t,
1079         const struct __ip6_tnl_parm *p, int set_mtu)
1080 {
1081         t->parms.laddr = p->laddr;
1082         t->parms.raddr = p->raddr;
1083         t->parms.flags = p->flags;
1084         t->parms.hop_limit = p->hop_limit;
1085         t->parms.encap_limit = p->encap_limit;
1086         t->parms.flowinfo = p->flowinfo;
1087         t->parms.link = p->link;
1088         t->parms.proto = p->proto;
1089         t->parms.i_key = p->i_key;
1090         t->parms.o_key = p->o_key;
1091         t->parms.i_flags = p->i_flags;
1092         t->parms.o_flags = p->o_flags;
1093         ip6_tnl_dst_reset(t);
1094         ip6gre_tnl_link_config(t, set_mtu);
1095         return 0;
1096 }
1097
1098 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1099         const struct ip6_tnl_parm2 *u)
1100 {
1101         p->laddr = u->laddr;
1102         p->raddr = u->raddr;
1103         p->flags = u->flags;
1104         p->hop_limit = u->hop_limit;
1105         p->encap_limit = u->encap_limit;
1106         p->flowinfo = u->flowinfo;
1107         p->link = u->link;
1108         p->i_key = u->i_key;
1109         p->o_key = u->o_key;
1110         p->i_flags = u->i_flags;
1111         p->o_flags = u->o_flags;
1112         memcpy(p->name, u->name, sizeof(u->name));
1113 }
1114
1115 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1116         const struct __ip6_tnl_parm *p)
1117 {
1118         u->proto = IPPROTO_GRE;
1119         u->laddr = p->laddr;
1120         u->raddr = p->raddr;
1121         u->flags = p->flags;
1122         u->hop_limit = p->hop_limit;
1123         u->encap_limit = p->encap_limit;
1124         u->flowinfo = p->flowinfo;
1125         u->link = p->link;
1126         u->i_key = p->i_key;
1127         u->o_key = p->o_key;
1128         u->i_flags = p->i_flags;
1129         u->o_flags = p->o_flags;
1130         memcpy(u->name, p->name, sizeof(u->name));
1131 }
1132
1133 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1134         struct ifreq *ifr, int cmd)
1135 {
1136         int err = 0;
1137         struct ip6_tnl_parm2 p;
1138         struct __ip6_tnl_parm p1;
1139         struct ip6_tnl *t;
1140         struct net *net = dev_net(dev);
1141         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1142
1143         switch (cmd) {
1144         case SIOCGETTUNNEL:
1145                 t = NULL;
1146                 if (dev == ign->fb_tunnel_dev) {
1147                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1148                                 err = -EFAULT;
1149                                 break;
1150                         }
1151                         ip6gre_tnl_parm_from_user(&p1, &p);
1152                         t = ip6gre_tunnel_locate(net, &p1, 0);
1153                 }
1154                 if (t == NULL)
1155                         t = netdev_priv(dev);
1156                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1157                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1158                         err = -EFAULT;
1159                 break;
1160
1161         case SIOCADDTUNNEL:
1162         case SIOCCHGTUNNEL:
1163                 err = -EPERM;
1164                 if (!capable(CAP_NET_ADMIN))
1165                         goto done;
1166
1167                 err = -EFAULT;
1168                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1169                         goto done;
1170
1171                 err = -EINVAL;
1172                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1173                         goto done;
1174
1175                 if (!(p.i_flags&GRE_KEY))
1176                         p.i_key = 0;
1177                 if (!(p.o_flags&GRE_KEY))
1178                         p.o_key = 0;
1179
1180                 ip6gre_tnl_parm_from_user(&p1, &p);
1181                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1182
1183                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1184                         if (t != NULL) {
1185                                 if (t->dev != dev) {
1186                                         err = -EEXIST;
1187                                         break;
1188                                 }
1189                         } else {
1190                                 t = netdev_priv(dev);
1191
1192                                 ip6gre_tunnel_unlink(ign, t);
1193                                 synchronize_net();
1194                                 ip6gre_tnl_change(t, &p1, 1);
1195                                 ip6gre_tunnel_link(ign, t);
1196                                 netdev_state_change(dev);
1197                         }
1198                 }
1199
1200                 if (t) {
1201                         err = 0;
1202
1203                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1204                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1205                                 err = -EFAULT;
1206                 } else
1207                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1208                 break;
1209
1210         case SIOCDELTUNNEL:
1211                 err = -EPERM;
1212                 if (!capable(CAP_NET_ADMIN))
1213                         goto done;
1214
1215                 if (dev == ign->fb_tunnel_dev) {
1216                         err = -EFAULT;
1217                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1218                                 goto done;
1219                         err = -ENOENT;
1220                         ip6gre_tnl_parm_from_user(&p1, &p);
1221                         t = ip6gre_tunnel_locate(net, &p1, 0);
1222                         if (t == NULL)
1223                                 goto done;
1224                         err = -EPERM;
1225                         if (t == netdev_priv(ign->fb_tunnel_dev))
1226                                 goto done;
1227                         dev = t->dev;
1228                 }
1229                 unregister_netdevice(dev);
1230                 err = 0;
1231                 break;
1232
1233         default:
1234                 err = -EINVAL;
1235         }
1236
1237 done:
1238         return err;
1239 }
1240
1241 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1242 {
1243         struct ip6_tnl *tunnel = netdev_priv(dev);
1244         if (new_mtu < 68 ||
1245             new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1246                 return -EINVAL;
1247         dev->mtu = new_mtu;
1248         return 0;
1249 }
1250
1251 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1252                         unsigned short type,
1253                         const void *daddr, const void *saddr, unsigned int len)
1254 {
1255         struct ip6_tnl *t = netdev_priv(dev);
1256         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1257         __be16 *p = (__be16 *)(ipv6h+1);
1258
1259         *(__be32 *)ipv6h = t->fl.u.ip6.flowlabel | htonl(0x60000000);
1260         ipv6h->hop_limit = t->parms.hop_limit;
1261         ipv6h->nexthdr = NEXTHDR_GRE;
1262         ipv6h->saddr = t->parms.laddr;
1263         ipv6h->daddr = t->parms.raddr;
1264
1265         p[0]            = t->parms.o_flags;
1266         p[1]            = htons(type);
1267
1268         /*
1269          *      Set the source hardware address.
1270          */
1271
1272         if (saddr)
1273                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1274         if (daddr)
1275                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1276         if (!ipv6_addr_any(&ipv6h->daddr))
1277                 return t->hlen;
1278
1279         return -t->hlen;
1280 }
1281
1282 static int ip6gre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
1283 {
1284         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb_mac_header(skb);
1285         memcpy(haddr, &ipv6h->saddr, sizeof(struct in6_addr));
1286         return sizeof(struct in6_addr);
1287 }
1288
1289 static const struct header_ops ip6gre_header_ops = {
1290         .create = ip6gre_header,
1291         .parse  = ip6gre_header_parse,
1292 };
1293
1294 static const struct net_device_ops ip6gre_netdev_ops = {
1295         .ndo_init               = ip6gre_tunnel_init,
1296         .ndo_uninit             = ip6gre_tunnel_uninit,
1297         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1298         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1299         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1300         .ndo_get_stats64        = ip6gre_get_stats64,
1301 };
1302
1303 static void ip6gre_dev_free(struct net_device *dev)
1304 {
1305         free_percpu(dev->tstats);
1306         free_netdev(dev);
1307 }
1308
1309 static void ip6gre_tunnel_setup(struct net_device *dev)
1310 {
1311         struct ip6_tnl *t;
1312
1313         dev->netdev_ops = &ip6gre_netdev_ops;
1314         dev->destructor = ip6gre_dev_free;
1315
1316         dev->type = ARPHRD_IP6GRE;
1317         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1318         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1319         t = netdev_priv(dev);
1320         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1321                 dev->mtu -= 8;
1322         dev->flags |= IFF_NOARP;
1323         dev->iflink = 0;
1324         dev->addr_len = sizeof(struct in6_addr);
1325         dev->features |= NETIF_F_NETNS_LOCAL;
1326         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1327 }
1328
1329 static int ip6gre_tunnel_init(struct net_device *dev)
1330 {
1331         struct ip6_tnl *tunnel;
1332
1333         tunnel = netdev_priv(dev);
1334
1335         tunnel->dev = dev;
1336         strcpy(tunnel->parms.name, dev->name);
1337
1338         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1339         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1340
1341         if (ipv6_addr_any(&tunnel->parms.raddr))
1342                 dev->header_ops = &ip6gre_header_ops;
1343
1344         dev->tstats = alloc_percpu(struct pcpu_tstats);
1345         if (!dev->tstats)
1346                 return -ENOMEM;
1347
1348         return 0;
1349 }
1350
1351 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1352 {
1353         struct ip6_tnl *tunnel = netdev_priv(dev);
1354
1355         tunnel->dev = dev;
1356         strcpy(tunnel->parms.name, dev->name);
1357
1358         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1359
1360         dev_hold(dev);
1361 }
1362
1363
1364 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1365         .handler     = ip6gre_rcv,
1366         .err_handler = ip6gre_err,
1367         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1368 };
1369
1370 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1371         struct list_head *head)
1372 {
1373         int prio;
1374
1375         for (prio = 0; prio < 4; prio++) {
1376                 int h;
1377                 for (h = 0; h < HASH_SIZE; h++) {
1378                         struct ip6_tnl *t;
1379
1380                         t = rtnl_dereference(ign->tunnels[prio][h]);
1381
1382                         while (t != NULL) {
1383                                 unregister_netdevice_queue(t->dev, head);
1384                                 t = rtnl_dereference(t->next);
1385                         }
1386                 }
1387         }
1388 }
1389
1390 static int __net_init ip6gre_init_net(struct net *net)
1391 {
1392         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1393         int err;
1394
1395         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1396                                            ip6gre_tunnel_setup);
1397         if (!ign->fb_tunnel_dev) {
1398                 err = -ENOMEM;
1399                 goto err_alloc_dev;
1400         }
1401         dev_net_set(ign->fb_tunnel_dev, net);
1402
1403         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1404         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1405
1406         err = register_netdev(ign->fb_tunnel_dev);
1407         if (err)
1408                 goto err_reg_dev;
1409
1410         rcu_assign_pointer(ign->tunnels_wc[0],
1411                            netdev_priv(ign->fb_tunnel_dev));
1412         return 0;
1413
1414 err_reg_dev:
1415         ip6gre_dev_free(ign->fb_tunnel_dev);
1416 err_alloc_dev:
1417         return err;
1418 }
1419
1420 static void __net_exit ip6gre_exit_net(struct net *net)
1421 {
1422         struct ip6gre_net *ign;
1423         LIST_HEAD(list);
1424
1425         ign = net_generic(net, ip6gre_net_id);
1426         rtnl_lock();
1427         ip6gre_destroy_tunnels(ign, &list);
1428         unregister_netdevice_many(&list);
1429         rtnl_unlock();
1430 }
1431
1432 static struct pernet_operations ip6gre_net_ops = {
1433         .init = ip6gre_init_net,
1434         .exit = ip6gre_exit_net,
1435         .id   = &ip6gre_net_id,
1436         .size = sizeof(struct ip6gre_net),
1437 };
1438
1439 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1440 {
1441         __be16 flags;
1442
1443         if (!data)
1444                 return 0;
1445
1446         flags = 0;
1447         if (data[IFLA_GRE_IFLAGS])
1448                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1449         if (data[IFLA_GRE_OFLAGS])
1450                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1451         if (flags & (GRE_VERSION|GRE_ROUTING))
1452                 return -EINVAL;
1453
1454         return 0;
1455 }
1456
1457 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1458 {
1459         struct in6_addr daddr;
1460
1461         if (tb[IFLA_ADDRESS]) {
1462                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1463                         return -EINVAL;
1464                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1465                         return -EADDRNOTAVAIL;
1466         }
1467
1468         if (!data)
1469                 goto out;
1470
1471         if (data[IFLA_GRE_REMOTE]) {
1472                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1473                 if (ipv6_addr_any(&daddr))
1474                         return -EINVAL;
1475         }
1476
1477 out:
1478         return ip6gre_tunnel_validate(tb, data);
1479 }
1480
1481
1482 static void ip6gre_netlink_parms(struct nlattr *data[],
1483                                 struct __ip6_tnl_parm *parms)
1484 {
1485         memset(parms, 0, sizeof(*parms));
1486
1487         if (!data)
1488                 return;
1489
1490         if (data[IFLA_GRE_LINK])
1491                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1492
1493         if (data[IFLA_GRE_IFLAGS])
1494                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1495
1496         if (data[IFLA_GRE_OFLAGS])
1497                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1498
1499         if (data[IFLA_GRE_IKEY])
1500                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1501
1502         if (data[IFLA_GRE_OKEY])
1503                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1504
1505         if (data[IFLA_GRE_LOCAL])
1506                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1507
1508         if (data[IFLA_GRE_REMOTE])
1509                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1510
1511         if (data[IFLA_GRE_TTL])
1512                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1513
1514         if (data[IFLA_GRE_ENCAP_LIMIT])
1515                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1516
1517         if (data[IFLA_GRE_FLOWINFO])
1518                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1519
1520         if (data[IFLA_GRE_FLAGS])
1521                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1522 }
1523
1524 static int ip6gre_tap_init(struct net_device *dev)
1525 {
1526         struct ip6_tnl *tunnel;
1527
1528         tunnel = netdev_priv(dev);
1529
1530         tunnel->dev = dev;
1531         strcpy(tunnel->parms.name, dev->name);
1532
1533         ip6gre_tnl_link_config(tunnel, 1);
1534
1535         dev->tstats = alloc_percpu(struct pcpu_tstats);
1536         if (!dev->tstats)
1537                 return -ENOMEM;
1538
1539         return 0;
1540 }
1541
1542 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1543         .ndo_init = ip6gre_tap_init,
1544         .ndo_uninit = ip6gre_tunnel_uninit,
1545         .ndo_start_xmit = ip6gre_tunnel_xmit,
1546         .ndo_set_mac_address = eth_mac_addr,
1547         .ndo_validate_addr = eth_validate_addr,
1548         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1549         .ndo_get_stats64 = ip6gre_get_stats64,
1550 };
1551
1552 static void ip6gre_tap_setup(struct net_device *dev)
1553 {
1554
1555         ether_setup(dev);
1556
1557         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1558         dev->destructor = ip6gre_dev_free;
1559
1560         dev->iflink = 0;
1561         dev->features |= NETIF_F_NETNS_LOCAL;
1562 }
1563
1564 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1565         struct nlattr *tb[], struct nlattr *data[])
1566 {
1567         struct ip6_tnl *nt;
1568         struct net *net = dev_net(dev);
1569         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1570         int err;
1571
1572         nt = netdev_priv(dev);
1573         ip6gre_netlink_parms(data, &nt->parms);
1574
1575         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1576                 return -EEXIST;
1577
1578         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1579                 eth_hw_addr_random(dev);
1580
1581         nt->dev = dev;
1582         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1583
1584         /* Can use a lockless transmit, unless we generate output sequences */
1585         if (!(nt->parms.o_flags & GRE_SEQ))
1586                 dev->features |= NETIF_F_LLTX;
1587
1588         err = register_netdevice(dev);
1589         if (err)
1590                 goto out;
1591
1592         dev_hold(dev);
1593         ip6gre_tunnel_link(ign, nt);
1594
1595 out:
1596         return err;
1597 }
1598
1599 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1600                             struct nlattr *data[])
1601 {
1602         struct ip6_tnl *t, *nt;
1603         struct net *net = dev_net(dev);
1604         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1605         struct __ip6_tnl_parm p;
1606
1607         if (dev == ign->fb_tunnel_dev)
1608                 return -EINVAL;
1609
1610         nt = netdev_priv(dev);
1611         ip6gre_netlink_parms(data, &p);
1612
1613         t = ip6gre_tunnel_locate(net, &p, 0);
1614
1615         if (t) {
1616                 if (t->dev != dev)
1617                         return -EEXIST;
1618         } else {
1619                 t = nt;
1620
1621                 ip6gre_tunnel_unlink(ign, t);
1622                 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1623                 ip6gre_tunnel_link(ign, t);
1624                 netdev_state_change(dev);
1625         }
1626
1627         return 0;
1628 }
1629
1630 static size_t ip6gre_get_size(const struct net_device *dev)
1631 {
1632         return
1633                 /* IFLA_GRE_LINK */
1634                 nla_total_size(4) +
1635                 /* IFLA_GRE_IFLAGS */
1636                 nla_total_size(2) +
1637                 /* IFLA_GRE_OFLAGS */
1638                 nla_total_size(2) +
1639                 /* IFLA_GRE_IKEY */
1640                 nla_total_size(4) +
1641                 /* IFLA_GRE_OKEY */
1642                 nla_total_size(4) +
1643                 /* IFLA_GRE_LOCAL */
1644                 nla_total_size(4) +
1645                 /* IFLA_GRE_REMOTE */
1646                 nla_total_size(4) +
1647                 /* IFLA_GRE_TTL */
1648                 nla_total_size(1) +
1649                 /* IFLA_GRE_TOS */
1650                 nla_total_size(1) +
1651                 /* IFLA_GRE_ENCAP_LIMIT */
1652                 nla_total_size(1) +
1653                 /* IFLA_GRE_FLOWINFO */
1654                 nla_total_size(4) +
1655                 /* IFLA_GRE_FLAGS */
1656                 nla_total_size(4) +
1657                 0;
1658 }
1659
1660 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1661 {
1662         struct ip6_tnl *t = netdev_priv(dev);
1663         struct __ip6_tnl_parm *p = &t->parms;
1664
1665         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1666             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1667             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1668             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1669             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1670             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->raddr) ||
1671             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->laddr) ||
1672             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1673             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1674             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1675             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1676             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1677                 goto nla_put_failure;
1678         return 0;
1679
1680 nla_put_failure:
1681         return -EMSGSIZE;
1682 }
1683
1684 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1685         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1686         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1687         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1688         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1689         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1690         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1691         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1692         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1693         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1694         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1695         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1696 };
1697
1698 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1699         .kind           = "ip6gre",
1700         .maxtype        = IFLA_GRE_MAX,
1701         .policy         = ip6gre_policy,
1702         .priv_size      = sizeof(struct ip6_tnl),
1703         .setup          = ip6gre_tunnel_setup,
1704         .validate       = ip6gre_tunnel_validate,
1705         .newlink        = ip6gre_newlink,
1706         .changelink     = ip6gre_changelink,
1707         .get_size       = ip6gre_get_size,
1708         .fill_info      = ip6gre_fill_info,
1709 };
1710
1711 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1712         .kind           = "ip6gretap",
1713         .maxtype        = IFLA_GRE_MAX,
1714         .policy         = ip6gre_policy,
1715         .priv_size      = sizeof(struct ip6_tnl),
1716         .setup          = ip6gre_tap_setup,
1717         .validate       = ip6gre_tap_validate,
1718         .newlink        = ip6gre_newlink,
1719         .changelink     = ip6gre_changelink,
1720         .get_size       = ip6gre_get_size,
1721         .fill_info      = ip6gre_fill_info,
1722 };
1723
1724 /*
1725  *      And now the modules code and kernel interface.
1726  */
1727
1728 static int __init ip6gre_init(void)
1729 {
1730         int err;
1731
1732         pr_info("GRE over IPv6 tunneling driver\n");
1733
1734         err = register_pernet_device(&ip6gre_net_ops);
1735         if (err < 0)
1736                 return err;
1737
1738         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1739         if (err < 0) {
1740                 pr_info("%s: can't add protocol\n", __func__);
1741                 goto add_proto_failed;
1742         }
1743
1744         err = rtnl_link_register(&ip6gre_link_ops);
1745         if (err < 0)
1746                 goto rtnl_link_failed;
1747
1748         err = rtnl_link_register(&ip6gre_tap_ops);
1749         if (err < 0)
1750                 goto tap_ops_failed;
1751
1752 out:
1753         return err;
1754
1755 tap_ops_failed:
1756         rtnl_link_unregister(&ip6gre_link_ops);
1757 rtnl_link_failed:
1758         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1759 add_proto_failed:
1760         unregister_pernet_device(&ip6gre_net_ops);
1761         goto out;
1762 }
1763
1764 static void __exit ip6gre_fini(void)
1765 {
1766         rtnl_link_unregister(&ip6gre_tap_ops);
1767         rtnl_link_unregister(&ip6gre_link_ops);
1768         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1769         unregister_pernet_device(&ip6gre_net_ops);
1770 }
1771
1772 module_init(ip6gre_init);
1773 module_exit(ip6gre_fini);
1774 MODULE_LICENSE("GPL");
1775 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1776 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1777 MODULE_ALIAS_RTNL_LINK("ip6gre");
1778 MODULE_ALIAS_NETDEV("ip6gre0");