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