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         int pkt_len;
 671         struct sk_buff *new_skb;
 672
 673         if (dev->type == ARPHRD_ETHER)
 674                 IPCB(skb)->flags = 0;
 675
 676         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 677                 gre_hlen = 0;
 678                 ipv6h = (struct ipv6hdr *)skb->data;
 679                 fl6->daddr = ipv6h->daddr;
 680         } else {
 681                 gre_hlen = tunnel->hlen;
 682                 fl6->daddr = tunnel->parms.raddr;
 683         }
 684
 685         if (!fl6->flowi6_mark)
 686                 dst = ip6_tnl_dst_check(tunnel);
 687
 688         if (!dst) {
 689                 ndst = ip6_route_output(net, NULL, fl6);
 690
 691                 if (ndst->error)
 692                         goto tx_err_link_failure;
 693                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
 694                 if (IS_ERR(ndst)) {
 695                         err = PTR_ERR(ndst);
 696                         ndst = NULL;
 697                         goto tx_err_link_failure;
 698                 }
 699                 dst = ndst;
 700         }
 701
 702         tdev = dst->dev;
 703
 704         if (tdev == dev) {
 705                 stats->collisions++;
 706                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 707                                      tunnel->parms.name);
 708                 goto tx_err_dst_release;
 709         }
 710
 711         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 712         if (encap_limit >= 0) {
 713                 max_headroom += 8;
 714                 mtu -= 8;
 715         }
 716         if (mtu < IPV6_MIN_MTU)
 717                 mtu = IPV6_MIN_MTU;
 718         if (skb_dst(skb))
 719                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 720         if (skb->len > mtu) {
 721                 *pmtu = mtu;
 722                 err = -EMSGSIZE;
 723                 goto tx_err_dst_release;
 724         }
 725
 726         if (tunnel->err_count > 0) {
 727                 if (time_before(jiffies,
 728                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 729                         tunnel->err_count--;
 730
 731                         dst_link_failure(skb);
 732                 } else
 733                         tunnel->err_count = 0;
 734         }
 735
 736         max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 737
 738         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 739             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 740                 new_skb = skb_realloc_headroom(skb, max_headroom);
 741                 if (max_headroom > dev->needed_headroom)
 742                         dev->needed_headroom = max_headroom;
 743                 if (!new_skb)
 744                         goto tx_err_dst_release;
 745
 746                 if (skb->sk)
 747                         skb_set_owner_w(new_skb, skb->sk);
 748                 consume_skb(skb);
 749                 skb = new_skb;
 750         }
 751
 752         skb_dst_drop(skb);
 753
 754         if (fl6->flowi6_mark) {
 755                 skb_dst_set(skb, dst);
 756                 ndst = NULL;
 757         } else {
 758                 skb_dst_set_noref(skb, dst);
 759         }
 760
 761         proto = NEXTHDR_GRE;
 762         if (encap_limit >= 0) {
 763                 init_tel_txopt(&opt, encap_limit);
 764                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 765         }
 766
 767         skb_push(skb, gre_hlen);
 768         skb_reset_network_header(skb);
 769         skb_set_transport_header(skb, sizeof(*ipv6h));
 770
 771         /*
 772          *      Push down and install the IP header.
 773          */
 774         ipv6h = ipv6_hdr(skb);
 775         *(__be32 *)ipv6h = fl6->flowlabel | htonl(0x60000000);
 776         dsfield = INET_ECN_encapsulate(0, dsfield);
 777         ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
 778         ipv6h->hop_limit = tunnel->parms.hop_limit;
 779         ipv6h->nexthdr = proto;
 780         ipv6h->saddr = fl6->saddr;
 781         ipv6h->daddr = fl6->daddr;
 782
 783         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 784         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
 785                                    htons(ETH_P_TEB) : skb->protocol;
 786
 787         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 788                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 789
 790                 if (tunnel->parms.o_flags&GRE_SEQ) {
 791                         ++tunnel->o_seqno;
 792                         *ptr = htonl(tunnel->o_seqno);
 793                         ptr--;
 794                 }
 795                 if (tunnel->parms.o_flags&GRE_KEY) {
 796                         *ptr = tunnel->parms.o_key;
 797                         ptr--;
 798                 }
 799                 if (tunnel->parms.o_flags&GRE_CSUM) {
 800                         *ptr = 0;
 801                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
 802                                 skb->len - sizeof(struct ipv6hdr));
 803                 }
 804         }
 805
 806         nf_reset(skb);
 807         pkt_len = skb->len;
 808         err = ip6_local_out(skb);
 809
 810         if (net_xmit_eval(err) == 0) {
 811                 struct pcpu_tstats *tstats = this_cpu_ptr(tunnel->dev->tstats);
 812
 813                 tstats->tx_bytes += pkt_len;
 814                 tstats->tx_packets++;
 815         } else {
 816                 stats->tx_errors++;
 817                 stats->tx_aborted_errors++;
 818         }
 819
 820         if (ndst)
 821                 ip6_tnl_dst_store(tunnel, ndst);
 822
 823         return 0;
 824 tx_err_link_failure:
 825         stats->tx_carrier_errors++;
 826         dst_link_failure(skb);
 827 tx_err_dst_release:
 828         dst_release(ndst);
 829         return err;
 830 }
 831
 832 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 833 {
 834         struct ip6_tnl *t = netdev_priv(dev);
 835         const struct iphdr  *iph = ip_hdr(skb);
 836         int encap_limit = -1;
 837         struct flowi6 fl6;
 838         __u8 dsfield;
 839         __u32 mtu;
 840         int err;
 841
 842         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 843                 encap_limit = t->parms.encap_limit;
 844
 845         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 846         fl6.flowi6_proto = IPPROTO_IPIP;
 847
 848         dsfield = ipv4_get_dsfield(iph);
 849
 850         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 851                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 852                                           & IPV6_TCLASS_MASK;
 853         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 854                 fl6.flowi6_mark = skb->mark;
 855
 856         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 857         if (err != 0) {
 858                 /* XXX: send ICMP error even if DF is not set. */
 859                 if (err == -EMSGSIZE)
 860                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 861                                   htonl(mtu));
 862                 return -1;
 863         }
 864
 865         return 0;
 866 }
 867
 868 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 869 {
 870         struct ip6_tnl *t = netdev_priv(dev);
 871         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 872         int encap_limit = -1;
 873         __u16 offset;
 874         struct flowi6 fl6;
 875         __u8 dsfield;
 876         __u32 mtu;
 877         int err;
 878
 879         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 880                 return -1;
 881
 882         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 883         if (offset > 0) {
 884                 struct ipv6_tlv_tnl_enc_lim *tel;
 885                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 886                 if (tel->encap_limit == 0) {
 887                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 888                                     ICMPV6_HDR_FIELD, offset + 2);
 889                         return -1;
 890                 }
 891                 encap_limit = tel->encap_limit - 1;
 892         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 893                 encap_limit = t->parms.encap_limit;
 894
 895         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 896         fl6.flowi6_proto = IPPROTO_IPV6;
 897
 898         dsfield = ipv6_get_dsfield(ipv6h);
 899         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 900                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 901         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 902                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 903         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 904                 fl6.flowi6_mark = skb->mark;
 905
 906         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 907         if (err != 0) {
 908                 if (err == -EMSGSIZE)
 909                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 910                 return -1;
 911         }
 912
 913         return 0;
 914 }
 915
 916 /**
 917  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 918  *   @t: the outgoing tunnel device
 919  *   @hdr: IPv6 header from the incoming packet
 920  *
 921  * Description:
 922  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 923  *   doesn't match source of incoming packet.
 924  *
 925  * Return:
 926  *   1 if conflict,
 927  *   0 else
 928  **/
 929
 930 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 931         const struct ipv6hdr *hdr)
 932 {
 933         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 934 }
 935
 936 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 937 {
 938         struct ip6_tnl *t = netdev_priv(dev);
 939         int encap_limit = -1;
 940         struct flowi6 fl6;
 941         __u32 mtu;
 942         int err;
 943
 944         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 945                 encap_limit = t->parms.encap_limit;
 946
 947         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 948         fl6.flowi6_proto = skb->protocol;
 949
 950         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 951
 952         return err;
 953 }
 954
 955 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 956         struct net_device *dev)
 957 {
 958         struct ip6_tnl *t = netdev_priv(dev);
 959         struct net_device_stats *stats = &t->dev->stats;
 960         int ret;
 961
 962         if (!ip6_tnl_xmit_ctl(t))
 963                 goto tx_err;
 964
 965         switch (skb->protocol) {
 966         case htons(ETH_P_IP):
 967                 ret = ip6gre_xmit_ipv4(skb, dev);
 968                 break;
 969         case htons(ETH_P_IPV6):
 970                 ret = ip6gre_xmit_ipv6(skb, dev);
 971                 break;
 972         default:
 973                 ret = ip6gre_xmit_other(skb, dev);
 974                 break;
 975         }
 976
 977         if (ret < 0)
 978                 goto tx_err;
 979
 980         return NETDEV_TX_OK;
 981
 982 tx_err:
 983         stats->tx_errors++;
 984         stats->tx_dropped++;
 985         kfree_skb(skb);
 986         return NETDEV_TX_OK;
 987 }
 988
 989 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 990 {
 991         struct net_device *dev = t->dev;
 992         struct __ip6_tnl_parm *p = &t->parms;
 993         struct flowi6 *fl6 = &t->fl.u.ip6;
 994         int addend = sizeof(struct ipv6hdr) + 4;
 995
 996         if (dev->type != ARPHRD_ETHER) {
 997                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 998                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 999         }
1000
1001         /* Set up flowi template */
1002         fl6->saddr = p->laddr;
1003         fl6->daddr = p->raddr;
1004         fl6->flowi6_oif = p->link;
1005         fl6->flowlabel = 0;
1006
1007         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1008                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1009         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1010                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1011
1012         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1013         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1014
1015         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1016                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1017                 dev->flags |= IFF_POINTOPOINT;
1018         else
1019                 dev->flags &= ~IFF_POINTOPOINT;
1020
1021         dev->iflink = p->link;
1022
1023         /* Precalculate GRE options length */
1024         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
1025                 if (t->parms.o_flags&GRE_CSUM)
1026                         addend += 4;
1027                 if (t->parms.o_flags&GRE_KEY)
1028                         addend += 4;
1029                 if (t->parms.o_flags&GRE_SEQ)
1030                         addend += 4;
1031         }
1032
1033         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1034                 int strict = (ipv6_addr_type(&p->raddr) &
1035                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1036
1037                 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1038                                                  &p->raddr, &p->laddr,
1039                                                  p->link, strict);
1040
1041                 if (rt == NULL)
1042                         return;
1043
1044                 if (rt->dst.dev) {
1045                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
1046
1047                         if (set_mtu) {
1048                                 dev->mtu = rt->dst.dev->mtu - addend;
1049                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1050                                         dev->mtu -= 8;
1051
1052                                 if (dev->mtu < IPV6_MIN_MTU)
1053                                         dev->mtu = IPV6_MIN_MTU;
1054                         }
1055                 }
1056                 ip6_rt_put(rt);
1057         }
1058
1059         t->hlen = addend;
1060 }
1061
1062 static int ip6gre_tnl_change(struct ip6_tnl *t,
1063         const struct __ip6_tnl_parm *p, int set_mtu)
1064 {
1065         t->parms.laddr = p->laddr;
1066         t->parms.raddr = p->raddr;
1067         t->parms.flags = p->flags;
1068         t->parms.hop_limit = p->hop_limit;
1069         t->parms.encap_limit = p->encap_limit;
1070         t->parms.flowinfo = p->flowinfo;
1071         t->parms.link = p->link;
1072         t->parms.proto = p->proto;
1073         t->parms.i_key = p->i_key;
1074         t->parms.o_key = p->o_key;
1075         t->parms.i_flags = p->i_flags;
1076         t->parms.o_flags = p->o_flags;
1077         ip6_tnl_dst_reset(t);
1078         ip6gre_tnl_link_config(t, set_mtu);
1079         return 0;
1080 }
1081
1082 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1083         const struct ip6_tnl_parm2 *u)
1084 {
1085         p->laddr = u->laddr;
1086         p->raddr = u->raddr;
1087         p->flags = u->flags;
1088         p->hop_limit = u->hop_limit;
1089         p->encap_limit = u->encap_limit;
1090         p->flowinfo = u->flowinfo;
1091         p->link = u->link;
1092         p->i_key = u->i_key;
1093         p->o_key = u->o_key;
1094         p->i_flags = u->i_flags;
1095         p->o_flags = u->o_flags;
1096         memcpy(p->name, u->name, sizeof(u->name));
1097 }
1098
1099 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1100         const struct __ip6_tnl_parm *p)
1101 {
1102         u->proto = IPPROTO_GRE;
1103         u->laddr = p->laddr;
1104         u->raddr = p->raddr;
1105         u->flags = p->flags;
1106         u->hop_limit = p->hop_limit;
1107         u->encap_limit = p->encap_limit;
1108         u->flowinfo = p->flowinfo;
1109         u->link = p->link;
1110         u->i_key = p->i_key;
1111         u->o_key = p->o_key;
1112         u->i_flags = p->i_flags;
1113         u->o_flags = p->o_flags;
1114         memcpy(u->name, p->name, sizeof(u->name));
1115 }
1116
1117 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1118         struct ifreq *ifr, int cmd)
1119 {
1120         int err = 0;
1121         struct ip6_tnl_parm2 p;
1122         struct __ip6_tnl_parm p1;
1123         struct ip6_tnl *t;
1124         struct net *net = dev_net(dev);
1125         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1126
1127         switch (cmd) {
1128         case SIOCGETTUNNEL:
1129                 t = NULL;
1130                 if (dev == ign->fb_tunnel_dev) {
1131                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1132                                 err = -EFAULT;
1133                                 break;
1134                         }
1135                         ip6gre_tnl_parm_from_user(&p1, &p);
1136                         t = ip6gre_tunnel_locate(net, &p1, 0);
1137                 }
1138                 if (t == NULL)
1139                         t = netdev_priv(dev);
1140                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1141                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1142                         err = -EFAULT;
1143                 break;
1144
1145         case SIOCADDTUNNEL:
1146         case SIOCCHGTUNNEL:
1147                 err = -EPERM;
1148                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1149                         goto done;
1150
1151                 err = -EFAULT;
1152                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1153                         goto done;
1154
1155                 err = -EINVAL;
1156                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1157                         goto done;
1158
1159                 if (!(p.i_flags&GRE_KEY))
1160                         p.i_key = 0;
1161                 if (!(p.o_flags&GRE_KEY))
1162                         p.o_key = 0;
1163
1164                 ip6gre_tnl_parm_from_user(&p1, &p);
1165                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1166
1167                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1168                         if (t != NULL) {
1169                                 if (t->dev != dev) {
1170                                         err = -EEXIST;
1171                                         break;
1172                                 }
1173                         } else {
1174                                 t = netdev_priv(dev);
1175
1176                                 ip6gre_tunnel_unlink(ign, t);
1177                                 synchronize_net();
1178                                 ip6gre_tnl_change(t, &p1, 1);
1179                                 ip6gre_tunnel_link(ign, t);
1180                                 netdev_state_change(dev);
1181                         }
1182                 }
1183
1184                 if (t) {
1185                         err = 0;
1186
1187                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1188                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1189                                 err = -EFAULT;
1190                 } else
1191                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1192                 break;
1193
1194         case SIOCDELTUNNEL:
1195                 err = -EPERM;
1196                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1197                         goto done;
1198
1199                 if (dev == ign->fb_tunnel_dev) {
1200                         err = -EFAULT;
1201                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1202                                 goto done;
1203                         err = -ENOENT;
1204                         ip6gre_tnl_parm_from_user(&p1, &p);
1205                         t = ip6gre_tunnel_locate(net, &p1, 0);
1206                         if (t == NULL)
1207                                 goto done;
1208                         err = -EPERM;
1209                         if (t == netdev_priv(ign->fb_tunnel_dev))
1210                                 goto done;
1211                         dev = t->dev;
1212                 }
1213                 unregister_netdevice(dev);
1214                 err = 0;
1215                 break;
1216
1217         default:
1218                 err = -EINVAL;
1219         }
1220
1221 done:
1222         return err;
1223 }
1224
1225 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1226 {
1227         struct ip6_tnl *tunnel = netdev_priv(dev);
1228         if (new_mtu < 68 ||
1229             new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1230                 return -EINVAL;
1231         dev->mtu = new_mtu;
1232         return 0;
1233 }
1234
1235 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1236                         unsigned short type,
1237                         const void *daddr, const void *saddr, unsigned int len)
1238 {
1239         struct ip6_tnl *t = netdev_priv(dev);
1240         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1241         __be16 *p = (__be16 *)(ipv6h+1);
1242
1243         *(__be32 *)ipv6h = t->fl.u.ip6.flowlabel | htonl(0x60000000);
1244         ipv6h->hop_limit = t->parms.hop_limit;
1245         ipv6h->nexthdr = NEXTHDR_GRE;
1246         ipv6h->saddr = t->parms.laddr;
1247         ipv6h->daddr = t->parms.raddr;
1248
1249         p[0]            = t->parms.o_flags;
1250         p[1]            = htons(type);
1251
1252         /*
1253          *      Set the source hardware address.
1254          */
1255
1256         if (saddr)
1257                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1258         if (daddr)
1259                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1260         if (!ipv6_addr_any(&ipv6h->daddr))
1261                 return t->hlen;
1262
1263         return -t->hlen;
1264 }
1265
1266 static const struct header_ops ip6gre_header_ops = {
1267         .create = ip6gre_header,
1268 };
1269
1270 static const struct net_device_ops ip6gre_netdev_ops = {
1271         .ndo_init               = ip6gre_tunnel_init,
1272         .ndo_uninit             = ip6gre_tunnel_uninit,
1273         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1274         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1275         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1276         .ndo_get_stats64        = ip6gre_get_stats64,
1277 };
1278
1279 static void ip6gre_dev_free(struct net_device *dev)
1280 {
1281         free_percpu(dev->tstats);
1282         free_netdev(dev);
1283 }
1284
1285 static void ip6gre_tunnel_setup(struct net_device *dev)
1286 {
1287         struct ip6_tnl *t;
1288
1289         dev->netdev_ops = &ip6gre_netdev_ops;
1290         dev->destructor = ip6gre_dev_free;
1291
1292         dev->type = ARPHRD_IP6GRE;
1293         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1294         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1295         t = netdev_priv(dev);
1296         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1297                 dev->mtu -= 8;
1298         dev->flags |= IFF_NOARP;
1299         dev->iflink = 0;
1300         dev->addr_len = sizeof(struct in6_addr);
1301         dev->features |= NETIF_F_NETNS_LOCAL;
1302         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1303 }
1304
1305 static int ip6gre_tunnel_init(struct net_device *dev)
1306 {
1307         struct ip6_tnl *tunnel;
1308
1309         tunnel = netdev_priv(dev);
1310
1311         tunnel->dev = dev;
1312         strcpy(tunnel->parms.name, dev->name);
1313
1314         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1315         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1316
1317         if (ipv6_addr_any(&tunnel->parms.raddr))
1318                 dev->header_ops = &ip6gre_header_ops;
1319
1320         dev->tstats = alloc_percpu(struct pcpu_tstats);
1321         if (!dev->tstats)
1322                 return -ENOMEM;
1323
1324         return 0;
1325 }
1326
1327 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1328 {
1329         struct ip6_tnl *tunnel = netdev_priv(dev);
1330
1331         tunnel->dev = dev;
1332         strcpy(tunnel->parms.name, dev->name);
1333
1334         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1335
1336         dev_hold(dev);
1337 }
1338
1339
1340 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1341         .handler     = ip6gre_rcv,
1342         .err_handler = ip6gre_err,
1343         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1344 };
1345
1346 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1347         struct list_head *head)
1348 {
1349         int prio;
1350
1351         for (prio = 0; prio < 4; prio++) {
1352                 int h;
1353                 for (h = 0; h < HASH_SIZE; h++) {
1354                         struct ip6_tnl *t;
1355
1356                         t = rtnl_dereference(ign->tunnels[prio][h]);
1357
1358                         while (t != NULL) {
1359                                 unregister_netdevice_queue(t->dev, head);
1360                                 t = rtnl_dereference(t->next);
1361                         }
1362                 }
1363         }
1364 }
1365
1366 static int __net_init ip6gre_init_net(struct net *net)
1367 {
1368         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1369         int err;
1370
1371         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1372                                            ip6gre_tunnel_setup);
1373         if (!ign->fb_tunnel_dev) {
1374                 err = -ENOMEM;
1375                 goto err_alloc_dev;
1376         }
1377         dev_net_set(ign->fb_tunnel_dev, net);
1378
1379         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1380         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1381
1382         err = register_netdev(ign->fb_tunnel_dev);
1383         if (err)
1384                 goto err_reg_dev;
1385
1386         rcu_assign_pointer(ign->tunnels_wc[0],
1387                            netdev_priv(ign->fb_tunnel_dev));
1388         return 0;
1389
1390 err_reg_dev:
1391         ip6gre_dev_free(ign->fb_tunnel_dev);
1392 err_alloc_dev:
1393         return err;
1394 }
1395
1396 static void __net_exit ip6gre_exit_net(struct net *net)
1397 {
1398         struct ip6gre_net *ign;
1399         LIST_HEAD(list);
1400
1401         ign = net_generic(net, ip6gre_net_id);
1402         rtnl_lock();
1403         ip6gre_destroy_tunnels(ign, &list);
1404         unregister_netdevice_many(&list);
1405         rtnl_unlock();
1406 }
1407
1408 static struct pernet_operations ip6gre_net_ops = {
1409         .init = ip6gre_init_net,
1410         .exit = ip6gre_exit_net,
1411         .id   = &ip6gre_net_id,
1412         .size = sizeof(struct ip6gre_net),
1413 };
1414
1415 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1416 {
1417         __be16 flags;
1418
1419         if (!data)
1420                 return 0;
1421
1422         flags = 0;
1423         if (data[IFLA_GRE_IFLAGS])
1424                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1425         if (data[IFLA_GRE_OFLAGS])
1426                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1427         if (flags & (GRE_VERSION|GRE_ROUTING))
1428                 return -EINVAL;
1429
1430         return 0;
1431 }
1432
1433 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1434 {
1435         struct in6_addr daddr;
1436
1437         if (tb[IFLA_ADDRESS]) {
1438                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1439                         return -EINVAL;
1440                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1441                         return -EADDRNOTAVAIL;
1442         }
1443
1444         if (!data)
1445                 goto out;
1446
1447         if (data[IFLA_GRE_REMOTE]) {
1448                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1449                 if (ipv6_addr_any(&daddr))
1450                         return -EINVAL;
1451         }
1452
1453 out:
1454         return ip6gre_tunnel_validate(tb, data);
1455 }
1456
1457
1458 static void ip6gre_netlink_parms(struct nlattr *data[],
1459                                 struct __ip6_tnl_parm *parms)
1460 {
1461         memset(parms, 0, sizeof(*parms));
1462
1463         if (!data)
1464                 return;
1465
1466         if (data[IFLA_GRE_LINK])
1467                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1468
1469         if (data[IFLA_GRE_IFLAGS])
1470                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1471
1472         if (data[IFLA_GRE_OFLAGS])
1473                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1474
1475         if (data[IFLA_GRE_IKEY])
1476                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1477
1478         if (data[IFLA_GRE_OKEY])
1479                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1480
1481         if (data[IFLA_GRE_LOCAL])
1482                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1483
1484         if (data[IFLA_GRE_REMOTE])
1485                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1486
1487         if (data[IFLA_GRE_TTL])
1488                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1489
1490         if (data[IFLA_GRE_ENCAP_LIMIT])
1491                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1492
1493         if (data[IFLA_GRE_FLOWINFO])
1494                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1495
1496         if (data[IFLA_GRE_FLAGS])
1497                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1498 }
1499
1500 static int ip6gre_tap_init(struct net_device *dev)
1501 {
1502         struct ip6_tnl *tunnel;
1503
1504         tunnel = netdev_priv(dev);
1505
1506         tunnel->dev = dev;
1507         strcpy(tunnel->parms.name, dev->name);
1508
1509         ip6gre_tnl_link_config(tunnel, 1);
1510
1511         dev->tstats = alloc_percpu(struct pcpu_tstats);
1512         if (!dev->tstats)
1513                 return -ENOMEM;
1514
1515         return 0;
1516 }
1517
1518 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1519         .ndo_init = ip6gre_tap_init,
1520         .ndo_uninit = ip6gre_tunnel_uninit,
1521         .ndo_start_xmit = ip6gre_tunnel_xmit,
1522         .ndo_set_mac_address = eth_mac_addr,
1523         .ndo_validate_addr = eth_validate_addr,
1524         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1525         .ndo_get_stats64 = ip6gre_get_stats64,
1526 };
1527
1528 static void ip6gre_tap_setup(struct net_device *dev)
1529 {
1530
1531         ether_setup(dev);
1532
1533         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1534         dev->destructor = ip6gre_dev_free;
1535
1536         dev->iflink = 0;
1537         dev->features |= NETIF_F_NETNS_LOCAL;
1538 }
1539
1540 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1541         struct nlattr *tb[], struct nlattr *data[])
1542 {
1543         struct ip6_tnl *nt;
1544         struct net *net = dev_net(dev);
1545         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1546         int err;
1547
1548         nt = netdev_priv(dev);
1549         ip6gre_netlink_parms(data, &nt->parms);
1550
1551         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1552                 return -EEXIST;
1553
1554         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1555                 eth_hw_addr_random(dev);
1556
1557         nt->dev = dev;
1558         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1559
1560         /* Can use a lockless transmit, unless we generate output sequences */
1561         if (!(nt->parms.o_flags & GRE_SEQ))
1562                 dev->features |= NETIF_F_LLTX;
1563
1564         err = register_netdevice(dev);
1565         if (err)
1566                 goto out;
1567
1568         dev_hold(dev);
1569         ip6gre_tunnel_link(ign, nt);
1570
1571 out:
1572         return err;
1573 }
1574
1575 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1576                             struct nlattr *data[])
1577 {
1578         struct ip6_tnl *t, *nt;
1579         struct net *net = dev_net(dev);
1580         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1581         struct __ip6_tnl_parm p;
1582
1583         if (dev == ign->fb_tunnel_dev)
1584                 return -EINVAL;
1585
1586         nt = netdev_priv(dev);
1587         ip6gre_netlink_parms(data, &p);
1588
1589         t = ip6gre_tunnel_locate(net, &p, 0);
1590
1591         if (t) {
1592                 if (t->dev != dev)
1593                         return -EEXIST;
1594         } else {
1595                 t = nt;
1596
1597                 ip6gre_tunnel_unlink(ign, t);
1598                 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1599                 ip6gre_tunnel_link(ign, t);
1600                 netdev_state_change(dev);
1601         }
1602
1603         return 0;
1604 }
1605
1606 static size_t ip6gre_get_size(const struct net_device *dev)
1607 {
1608         return
1609                 /* IFLA_GRE_LINK */
1610                 nla_total_size(4) +
1611                 /* IFLA_GRE_IFLAGS */
1612                 nla_total_size(2) +
1613                 /* IFLA_GRE_OFLAGS */
1614                 nla_total_size(2) +
1615                 /* IFLA_GRE_IKEY */
1616                 nla_total_size(4) +
1617                 /* IFLA_GRE_OKEY */
1618                 nla_total_size(4) +
1619                 /* IFLA_GRE_LOCAL */
1620                 nla_total_size(sizeof(struct in6_addr)) +
1621                 /* IFLA_GRE_REMOTE */
1622                 nla_total_size(sizeof(struct in6_addr)) +
1623                 /* IFLA_GRE_TTL */
1624                 nla_total_size(1) +
1625                 /* IFLA_GRE_TOS */
1626                 nla_total_size(1) +
1627                 /* IFLA_GRE_ENCAP_LIMIT */
1628                 nla_total_size(1) +
1629                 /* IFLA_GRE_FLOWINFO */
1630                 nla_total_size(4) +
1631                 /* IFLA_GRE_FLAGS */
1632                 nla_total_size(4) +
1633                 0;
1634 }
1635
1636 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1637 {
1638         struct ip6_tnl *t = netdev_priv(dev);
1639         struct __ip6_tnl_parm *p = &t->parms;
1640
1641         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1642             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1643             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1644             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1645             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1646             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1647             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1648             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1649             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1650             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1651             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1652             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1653                 goto nla_put_failure;
1654         return 0;
1655
1656 nla_put_failure:
1657         return -EMSGSIZE;
1658 }
1659
1660 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1661         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1662         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1663         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1664         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1665         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1666         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1667         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1668         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1669         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1670         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1671         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1672 };
1673
1674 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1675         .kind           = "ip6gre",
1676         .maxtype        = IFLA_GRE_MAX,
1677         .policy         = ip6gre_policy,
1678         .priv_size      = sizeof(struct ip6_tnl),
1679         .setup          = ip6gre_tunnel_setup,
1680         .validate       = ip6gre_tunnel_validate,
1681         .newlink        = ip6gre_newlink,
1682         .changelink     = ip6gre_changelink,
1683         .get_size       = ip6gre_get_size,
1684         .fill_info      = ip6gre_fill_info,
1685 };
1686
1687 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1688         .kind           = "ip6gretap",
1689         .maxtype        = IFLA_GRE_MAX,
1690         .policy         = ip6gre_policy,
1691         .priv_size      = sizeof(struct ip6_tnl),
1692         .setup          = ip6gre_tap_setup,
1693         .validate       = ip6gre_tap_validate,
1694         .newlink        = ip6gre_newlink,
1695         .changelink     = ip6gre_changelink,
1696         .get_size       = ip6gre_get_size,
1697         .fill_info      = ip6gre_fill_info,
1698 };
1699
1700 /*
1701  *      And now the modules code and kernel interface.
1702  */
1703
1704 static int __init ip6gre_init(void)
1705 {
1706         int err;
1707
1708         pr_info("GRE over IPv6 tunneling driver\n");
1709
1710         err = register_pernet_device(&ip6gre_net_ops);
1711         if (err < 0)
1712                 return err;
1713
1714         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1715         if (err < 0) {
1716                 pr_info("%s: can't add protocol\n", __func__);
1717                 goto add_proto_failed;
1718         }
1719
1720         err = rtnl_link_register(&ip6gre_link_ops);
1721         if (err < 0)
1722                 goto rtnl_link_failed;
1723
1724         err = rtnl_link_register(&ip6gre_tap_ops);
1725         if (err < 0)
1726                 goto tap_ops_failed;
1727
1728 out:
1729         return err;
1730
1731 tap_ops_failed:
1732         rtnl_link_unregister(&ip6gre_link_ops);
1733 rtnl_link_failed:
1734         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1735 add_proto_failed:
1736         unregister_pernet_device(&ip6gre_net_ops);
1737         goto out;
1738 }
1739
1740 static void __exit ip6gre_fini(void)
1741 {
1742         rtnl_link_unregister(&ip6gre_tap_ops);
1743         rtnl_link_unregister(&ip6gre_link_ops);
1744         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1745         unregister_pernet_device(&ip6gre_net_ops);
1746 }
1747
1748 module_init(ip6gre_init);
1749 module_exit(ip6gre_fini);
1750 MODULE_LICENSE("GPL");
1751 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1752 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1753 MODULE_ALIAS_RTNL_LINK("ip6gre");
1754 MODULE_ALIAS_NETDEV("ip6gre0");