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         skb->transport_header = skb->network_header;
 762
 763         proto = NEXTHDR_GRE;
 764         if (encap_limit >= 0) {
 765                 init_tel_txopt(&opt, encap_limit);
 766                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 767         }
 768
 769         skb_push(skb, gre_hlen);
 770         skb_reset_network_header(skb);
 771
 772         /*
 773          *      Push down and install the IP header.
 774          */
 775         ipv6h = ipv6_hdr(skb);
 776         *(__be32 *)ipv6h = fl6->flowlabel | htonl(0x60000000);
 777         dsfield = INET_ECN_encapsulate(0, dsfield);
 778         ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
 779         ipv6h->hop_limit = tunnel->parms.hop_limit;
 780         ipv6h->nexthdr = proto;
 781         ipv6h->saddr = fl6->saddr;
 782         ipv6h->daddr = fl6->daddr;
 783
 784         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 785         ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
 786                                    htons(ETH_P_TEB) : skb->protocol;
 787
 788         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 789                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 790
 791                 if (tunnel->parms.o_flags&GRE_SEQ) {
 792                         ++tunnel->o_seqno;
 793                         *ptr = htonl(tunnel->o_seqno);
 794                         ptr--;
 795                 }
 796                 if (tunnel->parms.o_flags&GRE_KEY) {
 797                         *ptr = tunnel->parms.o_key;
 798                         ptr--;
 799                 }
 800                 if (tunnel->parms.o_flags&GRE_CSUM) {
 801                         *ptr = 0;
 802                         *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
 803                                 skb->len - sizeof(struct ipv6hdr));
 804                 }
 805         }
 806
 807         nf_reset(skb);
 808         pkt_len = skb->len;
 809         err = ip6_local_out(skb);
 810
 811         if (net_xmit_eval(err) == 0) {
 812                 struct pcpu_tstats *tstats = this_cpu_ptr(tunnel->dev->tstats);
 813
 814                 tstats->tx_bytes += pkt_len;
 815                 tstats->tx_packets++;
 816         } else {
 817                 stats->tx_errors++;
 818                 stats->tx_aborted_errors++;
 819         }
 820
 821         if (ndst)
 822                 ip6_tnl_dst_store(tunnel, ndst);
 823
 824         return 0;
 825 tx_err_link_failure:
 826         stats->tx_carrier_errors++;
 827         dst_link_failure(skb);
 828 tx_err_dst_release:
 829         dst_release(ndst);
 830         return err;
 831 }
 832
 833 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 834 {
 835         struct ip6_tnl *t = netdev_priv(dev);
 836         const struct iphdr  *iph = ip_hdr(skb);
 837         int encap_limit = -1;
 838         struct flowi6 fl6;
 839         __u8 dsfield;
 840         __u32 mtu;
 841         int err;
 842
 843         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 844                 encap_limit = t->parms.encap_limit;
 845
 846         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 847         fl6.flowi6_proto = IPPROTO_IPIP;
 848
 849         dsfield = ipv4_get_dsfield(iph);
 850
 851         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 852                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 853                                           & IPV6_TCLASS_MASK;
 854         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 855                 fl6.flowi6_mark = skb->mark;
 856
 857         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 858         if (err != 0) {
 859                 /* XXX: send ICMP error even if DF is not set. */
 860                 if (err == -EMSGSIZE)
 861                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 862                                   htonl(mtu));
 863                 return -1;
 864         }
 865
 866         return 0;
 867 }
 868
 869 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 870 {
 871         struct ip6_tnl *t = netdev_priv(dev);
 872         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 873         int encap_limit = -1;
 874         __u16 offset;
 875         struct flowi6 fl6;
 876         __u8 dsfield;
 877         __u32 mtu;
 878         int err;
 879
 880         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 881                 return -1;
 882
 883         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 884         if (offset > 0) {
 885                 struct ipv6_tlv_tnl_enc_lim *tel;
 886                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 887                 if (tel->encap_limit == 0) {
 888                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 889                                     ICMPV6_HDR_FIELD, offset + 2);
 890                         return -1;
 891                 }
 892                 encap_limit = tel->encap_limit - 1;
 893         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 894                 encap_limit = t->parms.encap_limit;
 895
 896         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 897         fl6.flowi6_proto = IPPROTO_IPV6;
 898
 899         dsfield = ipv6_get_dsfield(ipv6h);
 900         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 901                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 902         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 903                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 904         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 905                 fl6.flowi6_mark = skb->mark;
 906
 907         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 908         if (err != 0) {
 909                 if (err == -EMSGSIZE)
 910                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 911                 return -1;
 912         }
 913
 914         return 0;
 915 }
 916
 917 /**
 918  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 919  *   @t: the outgoing tunnel device
 920  *   @hdr: IPv6 header from the incoming packet
 921  *
 922  * Description:
 923  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 924  *   doesn't match source of incoming packet.
 925  *
 926  * Return:
 927  *   1 if conflict,
 928  *   0 else
 929  **/
 930
 931 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 932         const struct ipv6hdr *hdr)
 933 {
 934         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 935 }
 936
 937 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 938 {
 939         struct ip6_tnl *t = netdev_priv(dev);
 940         int encap_limit = -1;
 941         struct flowi6 fl6;
 942         __u32 mtu;
 943         int err;
 944
 945         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 946                 encap_limit = t->parms.encap_limit;
 947
 948         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 949         fl6.flowi6_proto = skb->protocol;
 950
 951         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 952
 953         return err;
 954 }
 955
 956 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 957         struct net_device *dev)
 958 {
 959         struct ip6_tnl *t = netdev_priv(dev);
 960         struct net_device_stats *stats = &t->dev->stats;
 961         int ret;
 962
 963         if (!ip6_tnl_xmit_ctl(t))
 964                 return -1;
 965
 966         switch (skb->protocol) {
 967         case htons(ETH_P_IP):
 968                 ret = ip6gre_xmit_ipv4(skb, dev);
 969                 break;
 970         case htons(ETH_P_IPV6):
 971                 ret = ip6gre_xmit_ipv6(skb, dev);
 972                 break;
 973         default:
 974                 ret = ip6gre_xmit_other(skb, dev);
 975                 break;
 976         }
 977
 978         if (ret < 0)
 979                 goto tx_err;
 980
 981         return NETDEV_TX_OK;
 982
 983 tx_err:
 984         stats->tx_errors++;
 985         stats->tx_dropped++;
 986         kfree_skb(skb);
 987         return NETDEV_TX_OK;
 988 }
 989
 990 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 991 {
 992         struct net_device *dev = t->dev;
 993         struct __ip6_tnl_parm *p = &t->parms;
 994         struct flowi6 *fl6 = &t->fl.u.ip6;
 995         int addend = sizeof(struct ipv6hdr) + 4;
 996
 997         if (dev->type != ARPHRD_ETHER) {
 998                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 999                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1000         }
1001
1002         /* Set up flowi template */
1003         fl6->saddr = p->laddr;
1004         fl6->daddr = p->raddr;
1005         fl6->flowi6_oif = p->link;
1006         fl6->flowlabel = 0;
1007
1008         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1009                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1010         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1011                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1012
1013         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1014         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1015
1016         if (p->flags&IP6_TNL_F_CAP_XMIT &&
1017                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1018                 dev->flags |= IFF_POINTOPOINT;
1019         else
1020                 dev->flags &= ~IFF_POINTOPOINT;
1021
1022         dev->iflink = p->link;
1023
1024         /* Precalculate GRE options length */
1025         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
1026                 if (t->parms.o_flags&GRE_CSUM)
1027                         addend += 4;
1028                 if (t->parms.o_flags&GRE_KEY)
1029                         addend += 4;
1030                 if (t->parms.o_flags&GRE_SEQ)
1031                         addend += 4;
1032         }
1033
1034         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1035                 int strict = (ipv6_addr_type(&p->raddr) &
1036                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1037
1038                 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1039                                                  &p->raddr, &p->laddr,
1040                                                  p->link, strict);
1041
1042                 if (rt == NULL)
1043                         return;
1044
1045                 if (rt->dst.dev) {
1046                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
1047
1048                         if (set_mtu) {
1049                                 dev->mtu = rt->dst.dev->mtu - addend;
1050                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1051                                         dev->mtu -= 8;
1052
1053                                 if (dev->mtu < IPV6_MIN_MTU)
1054                                         dev->mtu = IPV6_MIN_MTU;
1055                         }
1056                 }
1057                 ip6_rt_put(rt);
1058         }
1059
1060         t->hlen = addend;
1061 }
1062
1063 static int ip6gre_tnl_change(struct ip6_tnl *t,
1064         const struct __ip6_tnl_parm *p, int set_mtu)
1065 {
1066         t->parms.laddr = p->laddr;
1067         t->parms.raddr = p->raddr;
1068         t->parms.flags = p->flags;
1069         t->parms.hop_limit = p->hop_limit;
1070         t->parms.encap_limit = p->encap_limit;
1071         t->parms.flowinfo = p->flowinfo;
1072         t->parms.link = p->link;
1073         t->parms.proto = p->proto;
1074         t->parms.i_key = p->i_key;
1075         t->parms.o_key = p->o_key;
1076         t->parms.i_flags = p->i_flags;
1077         t->parms.o_flags = p->o_flags;
1078         ip6_tnl_dst_reset(t);
1079         ip6gre_tnl_link_config(t, set_mtu);
1080         return 0;
1081 }
1082
1083 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1084         const struct ip6_tnl_parm2 *u)
1085 {
1086         p->laddr = u->laddr;
1087         p->raddr = u->raddr;
1088         p->flags = u->flags;
1089         p->hop_limit = u->hop_limit;
1090         p->encap_limit = u->encap_limit;
1091         p->flowinfo = u->flowinfo;
1092         p->link = u->link;
1093         p->i_key = u->i_key;
1094         p->o_key = u->o_key;
1095         p->i_flags = u->i_flags;
1096         p->o_flags = u->o_flags;
1097         memcpy(p->name, u->name, sizeof(u->name));
1098 }
1099
1100 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1101         const struct __ip6_tnl_parm *p)
1102 {
1103         u->proto = IPPROTO_GRE;
1104         u->laddr = p->laddr;
1105         u->raddr = p->raddr;
1106         u->flags = p->flags;
1107         u->hop_limit = p->hop_limit;
1108         u->encap_limit = p->encap_limit;
1109         u->flowinfo = p->flowinfo;
1110         u->link = p->link;
1111         u->i_key = p->i_key;
1112         u->o_key = p->o_key;
1113         u->i_flags = p->i_flags;
1114         u->o_flags = p->o_flags;
1115         memcpy(u->name, p->name, sizeof(u->name));
1116 }
1117
1118 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1119         struct ifreq *ifr, int cmd)
1120 {
1121         int err = 0;
1122         struct ip6_tnl_parm2 p;
1123         struct __ip6_tnl_parm p1;
1124         struct ip6_tnl *t;
1125         struct net *net = dev_net(dev);
1126         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1127
1128         switch (cmd) {
1129         case SIOCGETTUNNEL:
1130                 t = NULL;
1131                 if (dev == ign->fb_tunnel_dev) {
1132                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1133                                 err = -EFAULT;
1134                                 break;
1135                         }
1136                         ip6gre_tnl_parm_from_user(&p1, &p);
1137                         t = ip6gre_tunnel_locate(net, &p1, 0);
1138                 }
1139                 if (t == NULL)
1140                         t = netdev_priv(dev);
1141                 ip6gre_tnl_parm_to_user(&p, &t->parms);
1142                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1143                         err = -EFAULT;
1144                 break;
1145
1146         case SIOCADDTUNNEL:
1147         case SIOCCHGTUNNEL:
1148                 err = -EPERM;
1149                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1150                         goto done;
1151
1152                 err = -EFAULT;
1153                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1154                         goto done;
1155
1156                 err = -EINVAL;
1157                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1158                         goto done;
1159
1160                 if (!(p.i_flags&GRE_KEY))
1161                         p.i_key = 0;
1162                 if (!(p.o_flags&GRE_KEY))
1163                         p.o_key = 0;
1164
1165                 ip6gre_tnl_parm_from_user(&p1, &p);
1166                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1167
1168                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1169                         if (t != NULL) {
1170                                 if (t->dev != dev) {
1171                                         err = -EEXIST;
1172                                         break;
1173                                 }
1174                         } else {
1175                                 t = netdev_priv(dev);
1176
1177                                 ip6gre_tunnel_unlink(ign, t);
1178                                 synchronize_net();
1179                                 ip6gre_tnl_change(t, &p1, 1);
1180                                 ip6gre_tunnel_link(ign, t);
1181                                 netdev_state_change(dev);
1182                         }
1183                 }
1184
1185                 if (t) {
1186                         err = 0;
1187
1188                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1189                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1190                                 err = -EFAULT;
1191                 } else
1192                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1193                 break;
1194
1195         case SIOCDELTUNNEL:
1196                 err = -EPERM;
1197                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1198                         goto done;
1199
1200                 if (dev == ign->fb_tunnel_dev) {
1201                         err = -EFAULT;
1202                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1203                                 goto done;
1204                         err = -ENOENT;
1205                         ip6gre_tnl_parm_from_user(&p1, &p);
1206                         t = ip6gre_tunnel_locate(net, &p1, 0);
1207                         if (t == NULL)
1208                                 goto done;
1209                         err = -EPERM;
1210                         if (t == netdev_priv(ign->fb_tunnel_dev))
1211                                 goto done;
1212                         dev = t->dev;
1213                 }
1214                 unregister_netdevice(dev);
1215                 err = 0;
1216                 break;
1217
1218         default:
1219                 err = -EINVAL;
1220         }
1221
1222 done:
1223         return err;
1224 }
1225
1226 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1227 {
1228         struct ip6_tnl *tunnel = netdev_priv(dev);
1229         if (new_mtu < 68 ||
1230             new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1231                 return -EINVAL;
1232         dev->mtu = new_mtu;
1233         return 0;
1234 }
1235
1236 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1237                         unsigned short type,
1238                         const void *daddr, const void *saddr, unsigned int len)
1239 {
1240         struct ip6_tnl *t = netdev_priv(dev);
1241         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1242         __be16 *p = (__be16 *)(ipv6h+1);
1243
1244         *(__be32 *)ipv6h = t->fl.u.ip6.flowlabel | htonl(0x60000000);
1245         ipv6h->hop_limit = t->parms.hop_limit;
1246         ipv6h->nexthdr = NEXTHDR_GRE;
1247         ipv6h->saddr = t->parms.laddr;
1248         ipv6h->daddr = t->parms.raddr;
1249
1250         p[0]            = t->parms.o_flags;
1251         p[1]            = htons(type);
1252
1253         /*
1254          *      Set the source hardware address.
1255          */
1256
1257         if (saddr)
1258                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1259         if (daddr)
1260                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1261         if (!ipv6_addr_any(&ipv6h->daddr))
1262                 return t->hlen;
1263
1264         return -t->hlen;
1265 }
1266
1267 static const struct header_ops ip6gre_header_ops = {
1268         .create = ip6gre_header,
1269 };
1270
1271 static const struct net_device_ops ip6gre_netdev_ops = {
1272         .ndo_init               = ip6gre_tunnel_init,
1273         .ndo_uninit             = ip6gre_tunnel_uninit,
1274         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1275         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1276         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1277         .ndo_get_stats64        = ip6gre_get_stats64,
1278 };
1279
1280 static void ip6gre_dev_free(struct net_device *dev)
1281 {
1282         free_percpu(dev->tstats);
1283         free_netdev(dev);
1284 }
1285
1286 static void ip6gre_tunnel_setup(struct net_device *dev)
1287 {
1288         struct ip6_tnl *t;
1289
1290         dev->netdev_ops = &ip6gre_netdev_ops;
1291         dev->destructor = ip6gre_dev_free;
1292
1293         dev->type = ARPHRD_IP6GRE;
1294         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1295         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1296         t = netdev_priv(dev);
1297         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1298                 dev->mtu -= 8;
1299         dev->flags |= IFF_NOARP;
1300         dev->iflink = 0;
1301         dev->addr_len = sizeof(struct in6_addr);
1302         dev->features |= NETIF_F_NETNS_LOCAL;
1303         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1304 }
1305
1306 static int ip6gre_tunnel_init(struct net_device *dev)
1307 {
1308         struct ip6_tnl *tunnel;
1309
1310         tunnel = netdev_priv(dev);
1311
1312         tunnel->dev = dev;
1313         strcpy(tunnel->parms.name, dev->name);
1314
1315         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1316         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1317
1318         if (ipv6_addr_any(&tunnel->parms.raddr))
1319                 dev->header_ops = &ip6gre_header_ops;
1320
1321         dev->tstats = alloc_percpu(struct pcpu_tstats);
1322         if (!dev->tstats)
1323                 return -ENOMEM;
1324
1325         return 0;
1326 }
1327
1328 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1329 {
1330         struct ip6_tnl *tunnel = netdev_priv(dev);
1331
1332         tunnel->dev = dev;
1333         strcpy(tunnel->parms.name, dev->name);
1334
1335         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1336
1337         dev_hold(dev);
1338 }
1339
1340
1341 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1342         .handler     = ip6gre_rcv,
1343         .err_handler = ip6gre_err,
1344         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1345 };
1346
1347 static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1348         struct list_head *head)
1349 {
1350         int prio;
1351
1352         for (prio = 0; prio < 4; prio++) {
1353                 int h;
1354                 for (h = 0; h < HASH_SIZE; h++) {
1355                         struct ip6_tnl *t;
1356
1357                         t = rtnl_dereference(ign->tunnels[prio][h]);
1358
1359                         while (t != NULL) {
1360                                 unregister_netdevice_queue(t->dev, head);
1361                                 t = rtnl_dereference(t->next);
1362                         }
1363                 }
1364         }
1365 }
1366
1367 static int __net_init ip6gre_init_net(struct net *net)
1368 {
1369         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1370         int err;
1371
1372         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1373                                            ip6gre_tunnel_setup);
1374         if (!ign->fb_tunnel_dev) {
1375                 err = -ENOMEM;
1376                 goto err_alloc_dev;
1377         }
1378         dev_net_set(ign->fb_tunnel_dev, net);
1379
1380         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1381         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1382
1383         err = register_netdev(ign->fb_tunnel_dev);
1384         if (err)
1385                 goto err_reg_dev;
1386
1387         rcu_assign_pointer(ign->tunnels_wc[0],
1388                            netdev_priv(ign->fb_tunnel_dev));
1389         return 0;
1390
1391 err_reg_dev:
1392         ip6gre_dev_free(ign->fb_tunnel_dev);
1393 err_alloc_dev:
1394         return err;
1395 }
1396
1397 static void __net_exit ip6gre_exit_net(struct net *net)
1398 {
1399         struct ip6gre_net *ign;
1400         LIST_HEAD(list);
1401
1402         ign = net_generic(net, ip6gre_net_id);
1403         rtnl_lock();
1404         ip6gre_destroy_tunnels(ign, &list);
1405         unregister_netdevice_many(&list);
1406         rtnl_unlock();
1407 }
1408
1409 static struct pernet_operations ip6gre_net_ops = {
1410         .init = ip6gre_init_net,
1411         .exit = ip6gre_exit_net,
1412         .id   = &ip6gre_net_id,
1413         .size = sizeof(struct ip6gre_net),
1414 };
1415
1416 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1417 {
1418         __be16 flags;
1419
1420         if (!data)
1421                 return 0;
1422
1423         flags = 0;
1424         if (data[IFLA_GRE_IFLAGS])
1425                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1426         if (data[IFLA_GRE_OFLAGS])
1427                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1428         if (flags & (GRE_VERSION|GRE_ROUTING))
1429                 return -EINVAL;
1430
1431         return 0;
1432 }
1433
1434 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1435 {
1436         struct in6_addr daddr;
1437
1438         if (tb[IFLA_ADDRESS]) {
1439                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1440                         return -EINVAL;
1441                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1442                         return -EADDRNOTAVAIL;
1443         }
1444
1445         if (!data)
1446                 goto out;
1447
1448         if (data[IFLA_GRE_REMOTE]) {
1449                 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1450                 if (ipv6_addr_any(&daddr))
1451                         return -EINVAL;
1452         }
1453
1454 out:
1455         return ip6gre_tunnel_validate(tb, data);
1456 }
1457
1458
1459 static void ip6gre_netlink_parms(struct nlattr *data[],
1460                                 struct __ip6_tnl_parm *parms)
1461 {
1462         memset(parms, 0, sizeof(*parms));
1463
1464         if (!data)
1465                 return;
1466
1467         if (data[IFLA_GRE_LINK])
1468                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1469
1470         if (data[IFLA_GRE_IFLAGS])
1471                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1472
1473         if (data[IFLA_GRE_OFLAGS])
1474                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1475
1476         if (data[IFLA_GRE_IKEY])
1477                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1478
1479         if (data[IFLA_GRE_OKEY])
1480                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1481
1482         if (data[IFLA_GRE_LOCAL])
1483                 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1484
1485         if (data[IFLA_GRE_REMOTE])
1486                 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1487
1488         if (data[IFLA_GRE_TTL])
1489                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1490
1491         if (data[IFLA_GRE_ENCAP_LIMIT])
1492                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1493
1494         if (data[IFLA_GRE_FLOWINFO])
1495                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1496
1497         if (data[IFLA_GRE_FLAGS])
1498                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1499 }
1500
1501 static int ip6gre_tap_init(struct net_device *dev)
1502 {
1503         struct ip6_tnl *tunnel;
1504
1505         tunnel = netdev_priv(dev);
1506
1507         tunnel->dev = dev;
1508         strcpy(tunnel->parms.name, dev->name);
1509
1510         ip6gre_tnl_link_config(tunnel, 1);
1511
1512         dev->tstats = alloc_percpu(struct pcpu_tstats);
1513         if (!dev->tstats)
1514                 return -ENOMEM;
1515
1516         return 0;
1517 }
1518
1519 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1520         .ndo_init = ip6gre_tap_init,
1521         .ndo_uninit = ip6gre_tunnel_uninit,
1522         .ndo_start_xmit = ip6gre_tunnel_xmit,
1523         .ndo_set_mac_address = eth_mac_addr,
1524         .ndo_validate_addr = eth_validate_addr,
1525         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1526         .ndo_get_stats64 = ip6gre_get_stats64,
1527 };
1528
1529 static void ip6gre_tap_setup(struct net_device *dev)
1530 {
1531
1532         ether_setup(dev);
1533
1534         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1535         dev->destructor = ip6gre_dev_free;
1536
1537         dev->iflink = 0;
1538         dev->features |= NETIF_F_NETNS_LOCAL;
1539 }
1540
1541 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1542         struct nlattr *tb[], struct nlattr *data[])
1543 {
1544         struct ip6_tnl *nt;
1545         struct net *net = dev_net(dev);
1546         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1547         int err;
1548
1549         nt = netdev_priv(dev);
1550         ip6gre_netlink_parms(data, &nt->parms);
1551
1552         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1553                 return -EEXIST;
1554
1555         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1556                 eth_hw_addr_random(dev);
1557
1558         nt->dev = dev;
1559         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1560
1561         /* Can use a lockless transmit, unless we generate output sequences */
1562         if (!(nt->parms.o_flags & GRE_SEQ))
1563                 dev->features |= NETIF_F_LLTX;
1564
1565         err = register_netdevice(dev);
1566         if (err)
1567                 goto out;
1568
1569         dev_hold(dev);
1570         ip6gre_tunnel_link(ign, nt);
1571
1572 out:
1573         return err;
1574 }
1575
1576 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1577                             struct nlattr *data[])
1578 {
1579         struct ip6_tnl *t, *nt;
1580         struct net *net = dev_net(dev);
1581         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1582         struct __ip6_tnl_parm p;
1583
1584         if (dev == ign->fb_tunnel_dev)
1585                 return -EINVAL;
1586
1587         nt = netdev_priv(dev);
1588         ip6gre_netlink_parms(data, &p);
1589
1590         t = ip6gre_tunnel_locate(net, &p, 0);
1591
1592         if (t) {
1593                 if (t->dev != dev)
1594                         return -EEXIST;
1595         } else {
1596                 t = nt;
1597
1598                 ip6gre_tunnel_unlink(ign, t);
1599                 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1600                 ip6gre_tunnel_link(ign, t);
1601                 netdev_state_change(dev);
1602         }
1603
1604         return 0;
1605 }
1606
1607 static size_t ip6gre_get_size(const struct net_device *dev)
1608 {
1609         return
1610                 /* IFLA_GRE_LINK */
1611                 nla_total_size(4) +
1612                 /* IFLA_GRE_IFLAGS */
1613                 nla_total_size(2) +
1614                 /* IFLA_GRE_OFLAGS */
1615                 nla_total_size(2) +
1616                 /* IFLA_GRE_IKEY */
1617                 nla_total_size(4) +
1618                 /* IFLA_GRE_OKEY */
1619                 nla_total_size(4) +
1620                 /* IFLA_GRE_LOCAL */
1621                 nla_total_size(sizeof(struct in6_addr)) +
1622                 /* IFLA_GRE_REMOTE */
1623                 nla_total_size(sizeof(struct in6_addr)) +
1624                 /* IFLA_GRE_TTL */
1625                 nla_total_size(1) +
1626                 /* IFLA_GRE_TOS */
1627                 nla_total_size(1) +
1628                 /* IFLA_GRE_ENCAP_LIMIT */
1629                 nla_total_size(1) +
1630                 /* IFLA_GRE_FLOWINFO */
1631                 nla_total_size(4) +
1632                 /* IFLA_GRE_FLAGS */
1633                 nla_total_size(4) +
1634                 0;
1635 }
1636
1637 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1638 {
1639         struct ip6_tnl *t = netdev_priv(dev);
1640         struct __ip6_tnl_parm *p = &t->parms;
1641
1642         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1643             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1644             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1645             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1646             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1647             nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1648             nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1649             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1650             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1651             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1652             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1653             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1654                 goto nla_put_failure;
1655         return 0;
1656
1657 nla_put_failure:
1658         return -EMSGSIZE;
1659 }
1660
1661 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1662         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1663         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1664         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1665         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1666         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1667         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1668         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1669         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1670         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1671         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1672         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1673 };
1674
1675 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1676         .kind           = "ip6gre",
1677         .maxtype        = IFLA_GRE_MAX,
1678         .policy         = ip6gre_policy,
1679         .priv_size      = sizeof(struct ip6_tnl),
1680         .setup          = ip6gre_tunnel_setup,
1681         .validate       = ip6gre_tunnel_validate,
1682         .newlink        = ip6gre_newlink,
1683         .changelink     = ip6gre_changelink,
1684         .get_size       = ip6gre_get_size,
1685         .fill_info      = ip6gre_fill_info,
1686 };
1687
1688 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1689         .kind           = "ip6gretap",
1690         .maxtype        = IFLA_GRE_MAX,
1691         .policy         = ip6gre_policy,
1692         .priv_size      = sizeof(struct ip6_tnl),
1693         .setup          = ip6gre_tap_setup,
1694         .validate       = ip6gre_tap_validate,
1695         .newlink        = ip6gre_newlink,
1696         .changelink     = ip6gre_changelink,
1697         .get_size       = ip6gre_get_size,
1698         .fill_info      = ip6gre_fill_info,
1699 };
1700
1701 /*
1702  *      And now the modules code and kernel interface.
1703  */
1704
1705 static int __init ip6gre_init(void)
1706 {
1707         int err;
1708
1709         pr_info("GRE over IPv6 tunneling driver\n");
1710
1711         err = register_pernet_device(&ip6gre_net_ops);
1712         if (err < 0)
1713                 return err;
1714
1715         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1716         if (err < 0) {
1717                 pr_info("%s: can't add protocol\n", __func__);
1718                 goto add_proto_failed;
1719         }
1720
1721         err = rtnl_link_register(&ip6gre_link_ops);
1722         if (err < 0)
1723                 goto rtnl_link_failed;
1724
1725         err = rtnl_link_register(&ip6gre_tap_ops);
1726         if (err < 0)
1727                 goto tap_ops_failed;
1728
1729 out:
1730         return err;
1731
1732 tap_ops_failed:
1733         rtnl_link_unregister(&ip6gre_link_ops);
1734 rtnl_link_failed:
1735         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1736 add_proto_failed:
1737         unregister_pernet_device(&ip6gre_net_ops);
1738         goto out;
1739 }
1740
1741 static void __exit ip6gre_fini(void)
1742 {
1743         rtnl_link_unregister(&ip6gre_tap_ops);
1744         rtnl_link_unregister(&ip6gre_link_ops);
1745         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1746         unregister_pernet_device(&ip6gre_net_ops);
1747 }
1748
1749 module_init(ip6gre_init);
1750 module_exit(ip6gre_fini);
1751 MODULE_LICENSE("GPL");
1752 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1753 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1754 MODULE_ALIAS_RTNL_LINK("ip6gre");
1755 MODULE_ALIAS_NETDEV("ip6gre0");