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/init.h>
  28 #include <linux/in6.h>
  29 #include <linux/inetdevice.h>
  30 #include <linux/igmp.h>
  31 #include <linux/netfilter_ipv4.h>
  32 #include <linux/etherdevice.h>
  33 #include <linux/if_ether.h>
  34 #include <linux/hash.h>
  35 #include <linux/if_tunnel.h>
  36 #include <linux/ip6_tunnel.h>
  37
  38 #include <net/sock.h>
  39 #include <net/ip.h>
  40 #include <net/ip_tunnels.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 #include <net/gre.h>
  58
  59
  60 static bool log_ecn_error = true;
  61 module_param(log_ecn_error, bool, 0644);
  62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  63
  64 #define HASH_SIZE_SHIFT  5
  65 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  66
  67 static int ip6gre_net_id __read_mostly;
  68 struct ip6gre_net {
  69         struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
  70
  71         struct net_device *fb_tunnel_dev;
  72 };
  73
  74 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  75 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
  76 static int ip6gre_tunnel_init(struct net_device *dev);
  77 static void ip6gre_tunnel_setup(struct net_device *dev);
  78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  80
  81 /* Tunnel hash table */
  82
  83 /*
  84    4 hash tables:
  85
  86    3: (remote,local)
  87    2: (remote,*)
  88    1: (*,local)
  89    0: (*,*)
  90
  91    We require exact key match i.e. if a key is present in packet
  92    it will match only tunnel with the same key; if it is not present,
  93    it will match only keyless tunnel.
  94
  95    All keysless packets, if not matched configured keyless tunnels
  96    will match fallback tunnel.
  97  */
  98
  99 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
 100 static u32 HASH_ADDR(const struct in6_addr *addr)
 101 {
 102         u32 hash = ipv6_addr_hash(addr);
 103
 104         return hash_32(hash, HASH_SIZE_SHIFT);
 105 }
 106
 107 #define tunnels_r_l     tunnels[3]
 108 #define tunnels_r       tunnels[2]
 109 #define tunnels_l       tunnels[1]
 110 #define tunnels_wc      tunnels[0]
 111
 112 /* Given src, dst and key, find appropriate for input tunnel. */
 113
 114 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 115                 const struct in6_addr *remote, const struct in6_addr *local,
 116                 __be32 key, __be16 gre_proto)
 117 {
 118         struct net *net = dev_net(dev);
 119         int link = dev->ifindex;
 120         unsigned int h0 = HASH_ADDR(remote);
 121         unsigned int h1 = HASH_KEY(key);
 122         struct ip6_tnl *t, *cand = NULL;
 123         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 124         int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
 125                        ARPHRD_ETHER : ARPHRD_IP6GRE;
 126         int score, cand_score = 4;
 127
 128         for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 129                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 130                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 131                     key != t->parms.i_key ||
 132                     !(t->dev->flags & IFF_UP))
 133                         continue;
 134
 135                 if (t->dev->type != ARPHRD_IP6GRE &&
 136                     t->dev->type != dev_type)
 137                         continue;
 138
 139                 score = 0;
 140                 if (t->parms.link != link)
 141                         score |= 1;
 142                 if (t->dev->type != dev_type)
 143                         score |= 2;
 144                 if (score == 0)
 145                         return t;
 146
 147                 if (score < cand_score) {
 148                         cand = t;
 149                         cand_score = score;
 150                 }
 151         }
 152
 153         for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 154                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 155                     key != t->parms.i_key ||
 156                     !(t->dev->flags & IFF_UP))
 157                         continue;
 158
 159                 if (t->dev->type != ARPHRD_IP6GRE &&
 160                     t->dev->type != dev_type)
 161                         continue;
 162
 163                 score = 0;
 164                 if (t->parms.link != link)
 165                         score |= 1;
 166                 if (t->dev->type != dev_type)
 167                         score |= 2;
 168                 if (score == 0)
 169                         return t;
 170
 171                 if (score < cand_score) {
 172                         cand = t;
 173                         cand_score = score;
 174                 }
 175         }
 176
 177         for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 178                 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 179                           (!ipv6_addr_equal(local, &t->parms.raddr) ||
 180                                  !ipv6_addr_is_multicast(local))) ||
 181                     key != t->parms.i_key ||
 182                     !(t->dev->flags & IFF_UP))
 183                         continue;
 184
 185                 if (t->dev->type != ARPHRD_IP6GRE &&
 186                     t->dev->type != dev_type)
 187                         continue;
 188
 189                 score = 0;
 190                 if (t->parms.link != link)
 191                         score |= 1;
 192                 if (t->dev->type != dev_type)
 193                         score |= 2;
 194                 if (score == 0)
 195                         return t;
 196
 197                 if (score < cand_score) {
 198                         cand = t;
 199                         cand_score = score;
 200                 }
 201         }
 202
 203         for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 204                 if (t->parms.i_key != key ||
 205                     !(t->dev->flags & IFF_UP))
 206                         continue;
 207
 208                 if (t->dev->type != ARPHRD_IP6GRE &&
 209                     t->dev->type != dev_type)
 210                         continue;
 211
 212                 score = 0;
 213                 if (t->parms.link != link)
 214                         score |= 1;
 215                 if (t->dev->type != dev_type)
 216                         score |= 2;
 217                 if (score == 0)
 218                         return t;
 219
 220                 if (score < cand_score) {
 221                         cand = t;
 222                         cand_score = score;
 223                 }
 224         }
 225
 226         if (cand)
 227                 return cand;
 228
 229         dev = ign->fb_tunnel_dev;
 230         if (dev->flags & IFF_UP)
 231                 return netdev_priv(dev);
 232
 233         return NULL;
 234 }
 235
 236 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 237                 const struct __ip6_tnl_parm *p)
 238 {
 239         const struct in6_addr *remote = &p->raddr;
 240         const struct in6_addr *local = &p->laddr;
 241         unsigned int h = HASH_KEY(p->i_key);
 242         int prio = 0;
 243
 244         if (!ipv6_addr_any(local))
 245                 prio |= 1;
 246         if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 247                 prio |= 2;
 248                 h ^= HASH_ADDR(remote);
 249         }
 250
 251         return &ign->tunnels[prio][h];
 252 }
 253
 254 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 255                 const struct ip6_tnl *t)
 256 {
 257         return __ip6gre_bucket(ign, &t->parms);
 258 }
 259
 260 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 261 {
 262         struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 263
 264         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 265         rcu_assign_pointer(*tp, t);
 266 }
 267
 268 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 269 {
 270         struct ip6_tnl __rcu **tp;
 271         struct ip6_tnl *iter;
 272
 273         for (tp = ip6gre_bucket(ign, t);
 274              (iter = rtnl_dereference(*tp)) != NULL;
 275              tp = &iter->next) {
 276                 if (t == iter) {
 277                         rcu_assign_pointer(*tp, t->next);
 278                         break;
 279                 }
 280         }
 281 }
 282
 283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 284                                            const struct __ip6_tnl_parm *parms,
 285                                            int type)
 286 {
 287         const struct in6_addr *remote = &parms->raddr;
 288         const struct in6_addr *local = &parms->laddr;
 289         __be32 key = parms->i_key;
 290         int link = parms->link;
 291         struct ip6_tnl *t;
 292         struct ip6_tnl __rcu **tp;
 293         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 294
 295         for (tp = __ip6gre_bucket(ign, parms);
 296              (t = rtnl_dereference(*tp)) != NULL;
 297              tp = &t->next)
 298                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 299                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 300                     key == t->parms.i_key &&
 301                     link == t->parms.link &&
 302                     type == t->dev->type)
 303                         break;
 304
 305         return t;
 306 }
 307
 308 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 309                 const struct __ip6_tnl_parm *parms, int create)
 310 {
 311         struct ip6_tnl *t, *nt;
 312         struct net_device *dev;
 313         char name[IFNAMSIZ];
 314         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 315
 316         t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 317         if (t && create)
 318                 return NULL;
 319         if (t || !create)
 320                 return t;
 321
 322         if (parms->name[0])
 323                 strlcpy(name, parms->name, IFNAMSIZ);
 324         else
 325                 strcpy(name, "ip6gre%d");
 326
 327         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 328                            ip6gre_tunnel_setup);
 329         if (!dev)
 330                 return NULL;
 331
 332         dev_net_set(dev, net);
 333
 334         nt = netdev_priv(dev);
 335         nt->parms = *parms;
 336         dev->rtnl_link_ops = &ip6gre_link_ops;
 337
 338         nt->dev = dev;
 339         nt->net = dev_net(dev);
 340         ip6gre_tnl_link_config(nt, 1);
 341
 342         if (register_netdevice(dev) < 0)
 343                 goto failed_free;
 344
 345         /* Can use a lockless transmit, unless we generate output sequences */
 346         if (!(nt->parms.o_flags & GRE_SEQ))
 347                 dev->features |= NETIF_F_LLTX;
 348
 349         dev_hold(dev);
 350         ip6gre_tunnel_link(ign, nt);
 351         return nt;
 352
 353 failed_free:
 354         free_netdev(dev);
 355         return NULL;
 356 }
 357
 358 static void ip6gre_tunnel_uninit(struct net_device *dev)
 359 {
 360         struct ip6_tnl *t = netdev_priv(dev);
 361         struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 362
 363         ip6gre_tunnel_unlink(ign, t);
 364         dst_cache_reset(&t->dst_cache);
 365         dev_put(dev);
 366 }
 367
 368
 369 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 370                 u8 type, u8 code, int offset, __be32 info)
 371 {
 372         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 373         __be16 *p = (__be16 *)(skb->data + offset);
 374         int grehlen = offset + 4;
 375         struct ip6_tnl *t;
 376         __be16 flags;
 377
 378         flags = p[0];
 379         if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
 380                 if (flags&(GRE_VERSION|GRE_ROUTING))
 381                         return;
 382                 if (flags&GRE_KEY) {
 383                         grehlen += 4;
 384                         if (flags&GRE_CSUM)
 385                                 grehlen += 4;
 386                 }
 387         }
 388
 389         /* If only 8 bytes returned, keyed message will be dropped here */
 390         if (!pskb_may_pull(skb, grehlen))
 391                 return;
 392         ipv6h = (const struct ipv6hdr *)skb->data;
 393         p = (__be16 *)(skb->data + offset);
 394
 395         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 396                                 flags & GRE_KEY ?
 397                                 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
 398                                 p[1]);
 399         if (!t)
 400                 return;
 401
 402         switch (type) {
 403                 __u32 teli;
 404                 struct ipv6_tlv_tnl_enc_lim *tel;
 405                 __u32 mtu;
 406         case ICMPV6_DEST_UNREACH:
 407                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 408                                     t->parms.name);
 409                 break;
 410         case ICMPV6_TIME_EXCEED:
 411                 if (code == ICMPV6_EXC_HOPLIMIT) {
 412                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 413                                             t->parms.name);
 414                 }
 415                 break;
 416         case ICMPV6_PARAMPROB:
 417                 teli = 0;
 418                 if (code == ICMPV6_HDR_FIELD)
 419                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 420
 421                 if (teli && teli == be32_to_cpu(info) - 2) {
 422                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 423                         if (tel->encap_limit == 0) {
 424                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 425                                                     t->parms.name);
 426                         }
 427                 } else {
 428                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 429                                             t->parms.name);
 430                 }
 431                 break;
 432         case ICMPV6_PKT_TOOBIG:
 433                 mtu = be32_to_cpu(info) - offset;
 434                 if (mtu < IPV6_MIN_MTU)
 435                         mtu = IPV6_MIN_MTU;
 436                 t->dev->mtu = mtu;
 437                 break;
 438         }
 439
 440         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 441                 t->err_count++;
 442         else
 443                 t->err_count = 1;
 444         t->err_time = jiffies;
 445 }
 446
 447 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 448 {
 449         const struct ipv6hdr *ipv6h;
 450         struct ip6_tnl *tunnel;
 451
 452         ipv6h = ipv6_hdr(skb);
 453         tunnel = ip6gre_tunnel_lookup(skb->dev,
 454                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 455                                       tpi->proto);
 456         if (tunnel) {
 457                 ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
 458
 459                 return PACKET_RCVD;
 460         }
 461
 462         return PACKET_REJECT;
 463 }
 464
 465 static int gre_rcv(struct sk_buff *skb)
 466 {
 467         struct tnl_ptk_info tpi;
 468         bool csum_err = false;
 469         int hdr_len;
 470
 471         if (gre_parse_header(skb, &tpi, &csum_err, &hdr_len) < 0)
 472                 goto drop;
 473
 474         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
 475                 goto drop;
 476
 477         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
 478                 return 0;
 479
 480         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 481 drop:
 482         kfree_skb(skb);
 483         return 0;
 484 }
 485
 486 struct ipv6_tel_txoption {
 487         struct ipv6_txoptions ops;
 488         __u8 dst_opt[8];
 489 };
 490
 491 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 492 {
 493         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 494
 495         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 496         opt->dst_opt[3] = 1;
 497         opt->dst_opt[4] = encap_limit;
 498         opt->dst_opt[5] = IPV6_TLV_PADN;
 499         opt->dst_opt[6] = 1;
 500
 501         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 502         opt->ops.opt_nflen = 8;
 503 }
 504
 505 static __sum16 gre6_checksum(struct sk_buff *skb)
 506 {
 507         __wsum csum;
 508
 509         if (skb->ip_summed == CHECKSUM_PARTIAL)
 510                 csum = lco_csum(skb);
 511         else
 512                 csum = skb_checksum(skb, sizeof(struct ipv6hdr),
 513                                     skb->len - sizeof(struct ipv6hdr), 0);
 514         return csum_fold(csum);
 515 }
 516
 517 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 518                          struct net_device *dev,
 519                          __u8 dsfield,
 520                          struct flowi6 *fl6,
 521                          int encap_limit,
 522                          __u32 *pmtu)
 523 {
 524         struct ip6_tnl *tunnel = netdev_priv(dev);
 525         struct net *net = tunnel->net;
 526         struct net_device *tdev;    /* Device to other host */
 527         struct ipv6hdr  *ipv6h;     /* Our new IP header */
 528         unsigned int min_headroom = 0; /* The extra header space needed */
 529         int    gre_hlen;
 530         struct ipv6_tel_txoption opt;
 531         int    mtu;
 532         struct dst_entry *dst = NULL, *ndst = NULL;
 533         struct net_device_stats *stats = &tunnel->dev->stats;
 534         int err = -1;
 535         u8 proto;
 536         __be16 protocol;
 537
 538         if (dev->type == ARPHRD_ETHER)
 539                 IPCB(skb)->flags = 0;
 540
 541         if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
 542                 gre_hlen = 0;
 543                 ipv6h = (struct ipv6hdr *)skb->data;
 544                 fl6->daddr = ipv6h->daddr;
 545         } else {
 546                 gre_hlen = tunnel->hlen;
 547                 fl6->daddr = tunnel->parms.raddr;
 548         }
 549
 550         if (!fl6->flowi6_mark)
 551                 dst = dst_cache_get(&tunnel->dst_cache);
 552
 553         if (!dst) {
 554                 dst = ip6_route_output(net, NULL, fl6);
 555
 556                 if (dst->error)
 557                         goto tx_err_link_failure;
 558                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
 559                 if (IS_ERR(dst)) {
 560                         err = PTR_ERR(dst);
 561                         dst = NULL;
 562                         goto tx_err_link_failure;
 563                 }
 564                 ndst = dst;
 565         }
 566
 567         tdev = dst->dev;
 568
 569         if (tdev == dev) {
 570                 stats->collisions++;
 571                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 572                                      tunnel->parms.name);
 573                 goto tx_err_dst_release;
 574         }
 575
 576         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 577         if (encap_limit >= 0) {
 578                 min_headroom += 8;
 579                 mtu -= 8;
 580         }
 581         if (mtu < IPV6_MIN_MTU)
 582                 mtu = IPV6_MIN_MTU;
 583         if (skb_dst(skb))
 584                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 585         if (skb->len > mtu && !skb_is_gso(skb)) {
 586                 *pmtu = mtu;
 587                 err = -EMSGSIZE;
 588                 goto tx_err_dst_release;
 589         }
 590
 591         if (tunnel->err_count > 0) {
 592                 if (time_before(jiffies,
 593                                 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
 594                         tunnel->err_count--;
 595
 596                         dst_link_failure(skb);
 597                 } else
 598                         tunnel->err_count = 0;
 599         }
 600
 601         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
 602
 603         min_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
 604
 605         if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
 606                 int head_delta = SKB_DATA_ALIGN(min_headroom -
 607                                                 skb_headroom(skb) +
 608                                                 16);
 609
 610                 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
 611                                        0, GFP_ATOMIC);
 612                 if (min_headroom > dev->needed_headroom)
 613                         dev->needed_headroom = min_headroom;
 614                 if (unlikely(err))
 615                         goto tx_err_dst_release;
 616         }
 617
 618         if (!fl6->flowi6_mark && ndst)
 619                 dst_cache_set_ip6(&tunnel->dst_cache, ndst, &fl6->saddr);
 620         skb_dst_set(skb, dst);
 621
 622         proto = NEXTHDR_GRE;
 623         if (encap_limit >= 0) {
 624                 init_tel_txopt(&opt, encap_limit);
 625                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 626         }
 627
 628         err = iptunnel_handle_offloads(skb,
 629                                        (tunnel->parms.o_flags & GRE_CSUM) ?
 630                                        SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 631         if (err)
 632                 goto tx_err_dst_release;
 633
 634         skb_push(skb, gre_hlen);
 635         skb_reset_network_header(skb);
 636         skb_set_transport_header(skb, sizeof(*ipv6h));
 637
 638         /*
 639          *      Push down and install the IP header.
 640          */
 641         ipv6h = ipv6_hdr(skb);
 642         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
 643                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
 644         ipv6h->hop_limit = tunnel->parms.hop_limit;
 645         ipv6h->nexthdr = proto;
 646         ipv6h->saddr = fl6->saddr;
 647         ipv6h->daddr = fl6->daddr;
 648
 649         ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
 650         protocol = (dev->type == ARPHRD_ETHER) ?
 651                     htons(ETH_P_TEB) : skb->protocol;
 652         ((__be16 *)(ipv6h + 1))[1] = protocol;
 653
 654         if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
 655                 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
 656
 657                 if (tunnel->parms.o_flags&GRE_SEQ) {
 658                         ++tunnel->o_seqno;
 659                         *ptr = htonl(tunnel->o_seqno);
 660                         ptr--;
 661                 }
 662                 if (tunnel->parms.o_flags&GRE_KEY) {
 663                         *ptr = tunnel->parms.o_key;
 664                         ptr--;
 665                 }
 666                 if ((tunnel->parms.o_flags & GRE_CSUM) &&
 667                     !(skb_shinfo(skb)->gso_type &
 668                       (SKB_GSO_GRE | SKB_GSO_GRE_CSUM))) {
 669                         *ptr = 0;
 670                         *(__sum16 *)ptr = gre6_checksum(skb);
 671                 }
 672         }
 673
 674         skb_set_inner_protocol(skb, protocol);
 675
 676         ip6tunnel_xmit(NULL, skb, dev);
 677         return 0;
 678 tx_err_link_failure:
 679         stats->tx_carrier_errors++;
 680         dst_link_failure(skb);
 681 tx_err_dst_release:
 682         dst_release(dst);
 683         return err;
 684 }
 685
 686 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 687 {
 688         struct ip6_tnl *t = netdev_priv(dev);
 689         const struct iphdr  *iph = ip_hdr(skb);
 690         int encap_limit = -1;
 691         struct flowi6 fl6;
 692         __u8 dsfield;
 693         __u32 mtu;
 694         int err;
 695
 696         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 697
 698         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 699                 encap_limit = t->parms.encap_limit;
 700
 701         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 702         fl6.flowi6_proto = IPPROTO_GRE;
 703
 704         dsfield = ipv4_get_dsfield(iph);
 705
 706         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 707                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 708                                           & IPV6_TCLASS_MASK;
 709         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 710                 fl6.flowi6_mark = skb->mark;
 711
 712         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 713         if (err != 0) {
 714                 /* XXX: send ICMP error even if DF is not set. */
 715                 if (err == -EMSGSIZE)
 716                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 717                                   htonl(mtu));
 718                 return -1;
 719         }
 720
 721         return 0;
 722 }
 723
 724 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 725 {
 726         struct ip6_tnl *t = netdev_priv(dev);
 727         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 728         int encap_limit = -1;
 729         __u16 offset;
 730         struct flowi6 fl6;
 731         __u8 dsfield;
 732         __u32 mtu;
 733         int err;
 734
 735         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 736                 return -1;
 737
 738         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 739         if (offset > 0) {
 740                 struct ipv6_tlv_tnl_enc_lim *tel;
 741                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 742                 if (tel->encap_limit == 0) {
 743                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 744                                     ICMPV6_HDR_FIELD, offset + 2);
 745                         return -1;
 746                 }
 747                 encap_limit = tel->encap_limit - 1;
 748         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 749                 encap_limit = t->parms.encap_limit;
 750
 751         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 752         fl6.flowi6_proto = IPPROTO_GRE;
 753
 754         dsfield = ipv6_get_dsfield(ipv6h);
 755         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 756                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 757         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 758                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
 759         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 760                 fl6.flowi6_mark = skb->mark;
 761
 762         err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
 763         if (err != 0) {
 764                 if (err == -EMSGSIZE)
 765                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 766                 return -1;
 767         }
 768
 769         return 0;
 770 }
 771
 772 /**
 773  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 774  *   @t: the outgoing tunnel device
 775  *   @hdr: IPv6 header from the incoming packet
 776  *
 777  * Description:
 778  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 779  *   doesn't match source of incoming packet.
 780  *
 781  * Return:
 782  *   1 if conflict,
 783  *   0 else
 784  **/
 785
 786 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 787         const struct ipv6hdr *hdr)
 788 {
 789         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 790 }
 791
 792 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 793 {
 794         struct ip6_tnl *t = netdev_priv(dev);
 795         int encap_limit = -1;
 796         struct flowi6 fl6;
 797         __u32 mtu;
 798         int err;
 799
 800         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 801                 encap_limit = t->parms.encap_limit;
 802
 803         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 804         fl6.flowi6_proto = skb->protocol;
 805
 806         err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
 807
 808         return err;
 809 }
 810
 811 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 812         struct net_device *dev)
 813 {
 814         struct ip6_tnl *t = netdev_priv(dev);
 815         struct net_device_stats *stats = &t->dev->stats;
 816         int ret;
 817
 818         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 819                 goto tx_err;
 820
 821         switch (skb->protocol) {
 822         case htons(ETH_P_IP):
 823                 ret = ip6gre_xmit_ipv4(skb, dev);
 824                 break;
 825         case htons(ETH_P_IPV6):
 826                 ret = ip6gre_xmit_ipv6(skb, dev);
 827                 break;
 828         default:
 829                 ret = ip6gre_xmit_other(skb, dev);
 830                 break;
 831         }
 832
 833         if (ret < 0)
 834                 goto tx_err;
 835
 836         return NETDEV_TX_OK;
 837
 838 tx_err:
 839         stats->tx_errors++;
 840         stats->tx_dropped++;
 841         kfree_skb(skb);
 842         return NETDEV_TX_OK;
 843 }
 844
 845 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 846 {
 847         struct net_device *dev = t->dev;
 848         struct __ip6_tnl_parm *p = &t->parms;
 849         struct flowi6 *fl6 = &t->fl.u.ip6;
 850         int addend = sizeof(struct ipv6hdr) + 4;
 851
 852         if (dev->type != ARPHRD_ETHER) {
 853                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 854                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 855         }
 856
 857         /* Set up flowi template */
 858         fl6->saddr = p->laddr;
 859         fl6->daddr = p->raddr;
 860         fl6->flowi6_oif = p->link;
 861         fl6->flowlabel = 0;
 862
 863         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 864                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 865         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 866                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 867
 868         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 869         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 870
 871         if (p->flags&IP6_TNL_F_CAP_XMIT &&
 872                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
 873                 dev->flags |= IFF_POINTOPOINT;
 874         else
 875                 dev->flags &= ~IFF_POINTOPOINT;
 876
 877         /* Precalculate GRE options length */
 878         if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
 879                 if (t->parms.o_flags&GRE_CSUM)
 880                         addend += 4;
 881                 if (t->parms.o_flags&GRE_KEY)
 882                         addend += 4;
 883                 if (t->parms.o_flags&GRE_SEQ)
 884                         addend += 4;
 885         }
 886         t->hlen = addend;
 887
 888         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 889                 int strict = (ipv6_addr_type(&p->raddr) &
 890                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 891
 892                 struct rt6_info *rt = rt6_lookup(t->net,
 893                                                  &p->raddr, &p->laddr,
 894                                                  p->link, strict);
 895
 896                 if (!rt)
 897                         return;
 898
 899                 if (rt->dst.dev) {
 900                         dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
 901
 902                         if (set_mtu) {
 903                                 dev->mtu = rt->dst.dev->mtu - addend;
 904                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 905                                         dev->mtu -= 8;
 906                                 if (dev->type == ARPHRD_ETHER)
 907                                         dev->mtu -= ETH_HLEN;
 908
 909                                 if (dev->mtu < IPV6_MIN_MTU)
 910                                         dev->mtu = IPV6_MIN_MTU;
 911                         }
 912                 }
 913                 ip6_rt_put(rt);
 914         }
 915 }
 916
 917 static int ip6gre_tnl_change(struct ip6_tnl *t,
 918         const struct __ip6_tnl_parm *p, int set_mtu)
 919 {
 920         t->parms.laddr = p->laddr;
 921         t->parms.raddr = p->raddr;
 922         t->parms.flags = p->flags;
 923         t->parms.hop_limit = p->hop_limit;
 924         t->parms.encap_limit = p->encap_limit;
 925         t->parms.flowinfo = p->flowinfo;
 926         t->parms.link = p->link;
 927         t->parms.proto = p->proto;
 928         t->parms.i_key = p->i_key;
 929         t->parms.o_key = p->o_key;
 930         t->parms.i_flags = p->i_flags;
 931         t->parms.o_flags = p->o_flags;
 932         dst_cache_reset(&t->dst_cache);
 933         ip6gre_tnl_link_config(t, set_mtu);
 934         return 0;
 935 }
 936
 937 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
 938         const struct ip6_tnl_parm2 *u)
 939 {
 940         p->laddr = u->laddr;
 941         p->raddr = u->raddr;
 942         p->flags = u->flags;
 943         p->hop_limit = u->hop_limit;
 944         p->encap_limit = u->encap_limit;
 945         p->flowinfo = u->flowinfo;
 946         p->link = u->link;
 947         p->i_key = u->i_key;
 948         p->o_key = u->o_key;
 949         p->i_flags = u->i_flags;
 950         p->o_flags = u->o_flags;
 951         memcpy(p->name, u->name, sizeof(u->name));
 952 }
 953
 954 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
 955         const struct __ip6_tnl_parm *p)
 956 {
 957         u->proto = IPPROTO_GRE;
 958         u->laddr = p->laddr;
 959         u->raddr = p->raddr;
 960         u->flags = p->flags;
 961         u->hop_limit = p->hop_limit;
 962         u->encap_limit = p->encap_limit;
 963         u->flowinfo = p->flowinfo;
 964         u->link = p->link;
 965         u->i_key = p->i_key;
 966         u->o_key = p->o_key;
 967         u->i_flags = p->i_flags;
 968         u->o_flags = p->o_flags;
 969         memcpy(u->name, p->name, sizeof(u->name));
 970 }
 971
 972 static int ip6gre_tunnel_ioctl(struct net_device *dev,
 973         struct ifreq *ifr, int cmd)
 974 {
 975         int err = 0;
 976         struct ip6_tnl_parm2 p;
 977         struct __ip6_tnl_parm p1;
 978         struct ip6_tnl *t = netdev_priv(dev);
 979         struct net *net = t->net;
 980         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 981
 982         memset(&p1, 0, sizeof(p1));
 983
 984         switch (cmd) {
 985         case SIOCGETTUNNEL:
 986                 if (dev == ign->fb_tunnel_dev) {
 987                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 988                                 err = -EFAULT;
 989                                 break;
 990                         }
 991                         ip6gre_tnl_parm_from_user(&p1, &p);
 992                         t = ip6gre_tunnel_locate(net, &p1, 0);
 993                         if (!t)
 994                                 t = netdev_priv(dev);
 995                 }
 996                 memset(&p, 0, sizeof(p));
 997                 ip6gre_tnl_parm_to_user(&p, &t->parms);
 998                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 999                         err = -EFAULT;
1000                 break;
1001
1002         case SIOCADDTUNNEL:
1003         case SIOCCHGTUNNEL:
1004                 err = -EPERM;
1005                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1006                         goto done;
1007
1008                 err = -EFAULT;
1009                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1010                         goto done;
1011
1012                 err = -EINVAL;
1013                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1014                         goto done;
1015
1016                 if (!(p.i_flags&GRE_KEY))
1017                         p.i_key = 0;
1018                 if (!(p.o_flags&GRE_KEY))
1019                         p.o_key = 0;
1020
1021                 ip6gre_tnl_parm_from_user(&p1, &p);
1022                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1023
1024                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1025                         if (t) {
1026                                 if (t->dev != dev) {
1027                                         err = -EEXIST;
1028                                         break;
1029                                 }
1030                         } else {
1031                                 t = netdev_priv(dev);
1032
1033                                 ip6gre_tunnel_unlink(ign, t);
1034                                 synchronize_net();
1035                                 ip6gre_tnl_change(t, &p1, 1);
1036                                 ip6gre_tunnel_link(ign, t);
1037                                 netdev_state_change(dev);
1038                         }
1039                 }
1040
1041                 if (t) {
1042                         err = 0;
1043
1044                         memset(&p, 0, sizeof(p));
1045                         ip6gre_tnl_parm_to_user(&p, &t->parms);
1046                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1047                                 err = -EFAULT;
1048                 } else
1049                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1050                 break;
1051
1052         case SIOCDELTUNNEL:
1053                 err = -EPERM;
1054                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1055                         goto done;
1056
1057                 if (dev == ign->fb_tunnel_dev) {
1058                         err = -EFAULT;
1059                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1060                                 goto done;
1061                         err = -ENOENT;
1062                         ip6gre_tnl_parm_from_user(&p1, &p);
1063                         t = ip6gre_tunnel_locate(net, &p1, 0);
1064                         if (!t)
1065                                 goto done;
1066                         err = -EPERM;
1067                         if (t == netdev_priv(ign->fb_tunnel_dev))
1068                                 goto done;
1069                         dev = t->dev;
1070                 }
1071                 unregister_netdevice(dev);
1072                 err = 0;
1073                 break;
1074
1075         default:
1076                 err = -EINVAL;
1077         }
1078
1079 done:
1080         return err;
1081 }
1082
1083 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1084 {
1085         if (new_mtu < 68 ||
1086             new_mtu > 0xFFF8 - dev->hard_header_len)
1087                 return -EINVAL;
1088         dev->mtu = new_mtu;
1089         return 0;
1090 }
1091
1092 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1093                         unsigned short type,
1094                         const void *daddr, const void *saddr, unsigned int len)
1095 {
1096         struct ip6_tnl *t = netdev_priv(dev);
1097         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1098         __be16 *p = (__be16 *)(ipv6h+1);
1099
1100         ip6_flow_hdr(ipv6h, 0,
1101                      ip6_make_flowlabel(dev_net(dev), skb,
1102                                         t->fl.u.ip6.flowlabel, true,
1103                                         &t->fl.u.ip6));
1104         ipv6h->hop_limit = t->parms.hop_limit;
1105         ipv6h->nexthdr = NEXTHDR_GRE;
1106         ipv6h->saddr = t->parms.laddr;
1107         ipv6h->daddr = t->parms.raddr;
1108
1109         p[0]            = t->parms.o_flags;
1110         p[1]            = htons(type);
1111
1112         /*
1113          *      Set the source hardware address.
1114          */
1115
1116         if (saddr)
1117                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1118         if (daddr)
1119                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1120         if (!ipv6_addr_any(&ipv6h->daddr))
1121                 return t->hlen;
1122
1123         return -t->hlen;
1124 }
1125
1126 static const struct header_ops ip6gre_header_ops = {
1127         .create = ip6gre_header,
1128 };
1129
1130 static const struct net_device_ops ip6gre_netdev_ops = {
1131         .ndo_init               = ip6gre_tunnel_init,
1132         .ndo_uninit             = ip6gre_tunnel_uninit,
1133         .ndo_start_xmit         = ip6gre_tunnel_xmit,
1134         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1135         .ndo_change_mtu         = ip6gre_tunnel_change_mtu,
1136         .ndo_get_stats64        = ip_tunnel_get_stats64,
1137         .ndo_get_iflink         = ip6_tnl_get_iflink,
1138 };
1139
1140 static void ip6gre_dev_free(struct net_device *dev)
1141 {
1142         struct ip6_tnl *t = netdev_priv(dev);
1143
1144         dst_cache_destroy(&t->dst_cache);
1145         free_percpu(dev->tstats);
1146         free_netdev(dev);
1147 }
1148
1149 static void ip6gre_tunnel_setup(struct net_device *dev)
1150 {
1151         struct ip6_tnl *t;
1152
1153         dev->netdev_ops = &ip6gre_netdev_ops;
1154         dev->destructor = ip6gre_dev_free;
1155
1156         dev->type = ARPHRD_IP6GRE;
1157         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1158         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1159         t = netdev_priv(dev);
1160         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1161                 dev->mtu -= 8;
1162         dev->flags |= IFF_NOARP;
1163         dev->addr_len = sizeof(struct in6_addr);
1164         netif_keep_dst(dev);
1165 }
1166
1167 static int ip6gre_tunnel_init_common(struct net_device *dev)
1168 {
1169         struct ip6_tnl *tunnel;
1170         int ret;
1171
1172         tunnel = netdev_priv(dev);
1173
1174         tunnel->dev = dev;
1175         tunnel->net = dev_net(dev);
1176         strcpy(tunnel->parms.name, dev->name);
1177
1178         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1179         if (!dev->tstats)
1180                 return -ENOMEM;
1181
1182         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1183         if (ret) {
1184                 free_percpu(dev->tstats);
1185                 dev->tstats = NULL;
1186                 return ret;
1187         }
1188
1189         return 0;
1190 }
1191
1192 static int ip6gre_tunnel_init(struct net_device *dev)
1193 {
1194         struct ip6_tnl *tunnel;
1195         int ret;
1196
1197         ret = ip6gre_tunnel_init_common(dev);
1198         if (ret)
1199                 return ret;
1200
1201         tunnel = netdev_priv(dev);
1202
1203         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1204         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1205
1206         if (ipv6_addr_any(&tunnel->parms.raddr))
1207                 dev->header_ops = &ip6gre_header_ops;
1208
1209         return 0;
1210 }
1211
1212 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1213 {
1214         struct ip6_tnl *tunnel = netdev_priv(dev);
1215
1216         tunnel->dev = dev;
1217         tunnel->net = dev_net(dev);
1218         strcpy(tunnel->parms.name, dev->name);
1219
1220         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1221
1222         dev_hold(dev);
1223 }
1224
1225
1226 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1227         .handler     = gre_rcv,
1228         .err_handler = ip6gre_err,
1229         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1230 };
1231
1232 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1233 {
1234         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1235         struct net_device *dev, *aux;
1236         int prio;
1237
1238         for_each_netdev_safe(net, dev, aux)
1239                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1240                     dev->rtnl_link_ops == &ip6gre_tap_ops)
1241                         unregister_netdevice_queue(dev, head);
1242
1243         for (prio = 0; prio < 4; prio++) {
1244                 int h;
1245                 for (h = 0; h < HASH_SIZE; h++) {
1246                         struct ip6_tnl *t;
1247
1248                         t = rtnl_dereference(ign->tunnels[prio][h]);
1249
1250                         while (t) {
1251                                 /* If dev is in the same netns, it has already
1252                                  * been added to the list by the previous loop.
1253                                  */
1254                                 if (!net_eq(dev_net(t->dev), net))
1255                                         unregister_netdevice_queue(t->dev,
1256                                                                    head);
1257                                 t = rtnl_dereference(t->next);
1258                         }
1259                 }
1260         }
1261 }
1262
1263 static int __net_init ip6gre_init_net(struct net *net)
1264 {
1265         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1266         int err;
1267
1268         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1269                                           NET_NAME_UNKNOWN,
1270                                           ip6gre_tunnel_setup);
1271         if (!ign->fb_tunnel_dev) {
1272                 err = -ENOMEM;
1273                 goto err_alloc_dev;
1274         }
1275         dev_net_set(ign->fb_tunnel_dev, net);
1276         /* FB netdevice is special: we have one, and only one per netns.
1277          * Allowing to move it to another netns is clearly unsafe.
1278          */
1279         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1280
1281
1282         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1283         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1284
1285         err = register_netdev(ign->fb_tunnel_dev);
1286         if (err)
1287                 goto err_reg_dev;
1288
1289         rcu_assign_pointer(ign->tunnels_wc[0],
1290                            netdev_priv(ign->fb_tunnel_dev));
1291         return 0;
1292
1293 err_reg_dev:
1294         ip6gre_dev_free(ign->fb_tunnel_dev);
1295 err_alloc_dev:
1296         return err;
1297 }
1298
1299 static void __net_exit ip6gre_exit_net(struct net *net)
1300 {
1301         LIST_HEAD(list);
1302
1303         rtnl_lock();
1304         ip6gre_destroy_tunnels(net, &list);
1305         unregister_netdevice_many(&list);
1306         rtnl_unlock();
1307 }
1308
1309 static struct pernet_operations ip6gre_net_ops = {
1310         .init = ip6gre_init_net,
1311         .exit = ip6gre_exit_net,
1312         .id   = &ip6gre_net_id,
1313         .size = sizeof(struct ip6gre_net),
1314 };
1315
1316 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1317 {
1318         __be16 flags;
1319
1320         if (!data)
1321                 return 0;
1322
1323         flags = 0;
1324         if (data[IFLA_GRE_IFLAGS])
1325                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1326         if (data[IFLA_GRE_OFLAGS])
1327                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1328         if (flags & (GRE_VERSION|GRE_ROUTING))
1329                 return -EINVAL;
1330
1331         return 0;
1332 }
1333
1334 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1335 {
1336         struct in6_addr daddr;
1337
1338         if (tb[IFLA_ADDRESS]) {
1339                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1340                         return -EINVAL;
1341                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1342                         return -EADDRNOTAVAIL;
1343         }
1344
1345         if (!data)
1346                 goto out;
1347
1348         if (data[IFLA_GRE_REMOTE]) {
1349                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1350                 if (ipv6_addr_any(&daddr))
1351                         return -EINVAL;
1352         }
1353
1354 out:
1355         return ip6gre_tunnel_validate(tb, data);
1356 }
1357
1358
1359 static void ip6gre_netlink_parms(struct nlattr *data[],
1360                                 struct __ip6_tnl_parm *parms)
1361 {
1362         memset(parms, 0, sizeof(*parms));
1363
1364         if (!data)
1365                 return;
1366
1367         if (data[IFLA_GRE_LINK])
1368                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1369
1370         if (data[IFLA_GRE_IFLAGS])
1371                 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1372
1373         if (data[IFLA_GRE_OFLAGS])
1374                 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1375
1376         if (data[IFLA_GRE_IKEY])
1377                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1378
1379         if (data[IFLA_GRE_OKEY])
1380                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1381
1382         if (data[IFLA_GRE_LOCAL])
1383                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1384
1385         if (data[IFLA_GRE_REMOTE])
1386                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1387
1388         if (data[IFLA_GRE_TTL])
1389                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1390
1391         if (data[IFLA_GRE_ENCAP_LIMIT])
1392                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1393
1394         if (data[IFLA_GRE_FLOWINFO])
1395                 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1396
1397         if (data[IFLA_GRE_FLAGS])
1398                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1399 }
1400
1401 static int ip6gre_tap_init(struct net_device *dev)
1402 {
1403         struct ip6_tnl *tunnel;
1404         int ret;
1405
1406         ret = ip6gre_tunnel_init_common(dev);
1407         if (ret)
1408                 return ret;
1409
1410         tunnel = netdev_priv(dev);
1411
1412         ip6gre_tnl_link_config(tunnel, 1);
1413
1414         return 0;
1415 }
1416
1417 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1418         .ndo_init = ip6gre_tap_init,
1419         .ndo_uninit = ip6gre_tunnel_uninit,
1420         .ndo_start_xmit = ip6gre_tunnel_xmit,
1421         .ndo_set_mac_address = eth_mac_addr,
1422         .ndo_validate_addr = eth_validate_addr,
1423         .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1424         .ndo_get_stats64 = ip_tunnel_get_stats64,
1425         .ndo_get_iflink = ip6_tnl_get_iflink,
1426 };
1427
1428 #define GRE6_FEATURES (NETIF_F_SG |             \
1429                        NETIF_F_FRAGLIST |       \
1430                        NETIF_F_HIGHDMA |                \
1431                        NETIF_F_HW_CSUM)
1432
1433 static void ip6gre_tap_setup(struct net_device *dev)
1434 {
1435
1436         ether_setup(dev);
1437
1438         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1439         dev->destructor = ip6gre_dev_free;
1440
1441         dev->features |= NETIF_F_NETNS_LOCAL;
1442         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1443 }
1444
1445 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1446         struct nlattr *tb[], struct nlattr *data[])
1447 {
1448         struct ip6_tnl *nt;
1449         struct net *net = dev_net(dev);
1450         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1451         int err;
1452
1453         nt = netdev_priv(dev);
1454         ip6gre_netlink_parms(data, &nt->parms);
1455
1456         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1457                 return -EEXIST;
1458
1459         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1460                 eth_hw_addr_random(dev);
1461
1462         nt->dev = dev;
1463         nt->net = dev_net(dev);
1464         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1465
1466         dev->features           |= GRE6_FEATURES;
1467         dev->hw_features        |= GRE6_FEATURES;
1468
1469         if (!(nt->parms.o_flags & GRE_SEQ)) {
1470                 /* TCP segmentation offload is not supported when we
1471                  * generate output sequences.
1472                  */
1473                 dev->features    |= NETIF_F_GSO_SOFTWARE;
1474                 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1475
1476                 /* Can use a lockless transmit, unless we generate
1477                  * output sequences
1478                  */
1479                 dev->features |= NETIF_F_LLTX;
1480         }
1481
1482         err = register_netdevice(dev);
1483         if (err)
1484                 goto out;
1485
1486         dev_hold(dev);
1487         ip6gre_tunnel_link(ign, nt);
1488
1489 out:
1490         return err;
1491 }
1492
1493 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1494                             struct nlattr *data[])
1495 {
1496         struct ip6_tnl *t, *nt = netdev_priv(dev);
1497         struct net *net = nt->net;
1498         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1499         struct __ip6_tnl_parm p;
1500
1501         if (dev == ign->fb_tunnel_dev)
1502                 return -EINVAL;
1503
1504         ip6gre_netlink_parms(data, &p);
1505
1506         t = ip6gre_tunnel_locate(net, &p, 0);
1507
1508         if (t) {
1509                 if (t->dev != dev)
1510                         return -EEXIST;
1511         } else {
1512                 t = nt;
1513         }
1514
1515         ip6gre_tunnel_unlink(ign, t);
1516         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1517         ip6gre_tunnel_link(ign, t);
1518         return 0;
1519 }
1520
1521 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1522 {
1523         struct net *net = dev_net(dev);
1524         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1525
1526         if (dev != ign->fb_tunnel_dev)
1527                 unregister_netdevice_queue(dev, head);
1528 }
1529
1530 static size_t ip6gre_get_size(const struct net_device *dev)
1531 {
1532         return
1533                 /* IFLA_GRE_LINK */
1534                 nla_total_size(4) +
1535                 /* IFLA_GRE_IFLAGS */
1536                 nla_total_size(2) +
1537                 /* IFLA_GRE_OFLAGS */
1538                 nla_total_size(2) +
1539                 /* IFLA_GRE_IKEY */
1540                 nla_total_size(4) +
1541                 /* IFLA_GRE_OKEY */
1542                 nla_total_size(4) +
1543                 /* IFLA_GRE_LOCAL */
1544                 nla_total_size(sizeof(struct in6_addr)) +
1545                 /* IFLA_GRE_REMOTE */
1546                 nla_total_size(sizeof(struct in6_addr)) +
1547                 /* IFLA_GRE_TTL */
1548                 nla_total_size(1) +
1549                 /* IFLA_GRE_TOS */
1550                 nla_total_size(1) +
1551                 /* IFLA_GRE_ENCAP_LIMIT */
1552                 nla_total_size(1) +
1553                 /* IFLA_GRE_FLOWINFO */
1554                 nla_total_size(4) +
1555                 /* IFLA_GRE_FLAGS */
1556                 nla_total_size(4) +
1557                 0;
1558 }
1559
1560 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1561 {
1562         struct ip6_tnl *t = netdev_priv(dev);
1563         struct __ip6_tnl_parm *p = &t->parms;
1564
1565         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1566             nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1567             nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1568             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1569             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1570             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1571             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1572             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1573             /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1574             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1575             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1576             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1577                 goto nla_put_failure;
1578         return 0;
1579
1580 nla_put_failure:
1581         return -EMSGSIZE;
1582 }
1583
1584 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1585         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1586         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1587         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1588         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1589         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1590         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1591         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1592         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1593         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1594         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1595         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1596 };
1597
1598 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1599         .kind           = "ip6gre",
1600         .maxtype        = IFLA_GRE_MAX,
1601         .policy         = ip6gre_policy,
1602         .priv_size      = sizeof(struct ip6_tnl),
1603         .setup          = ip6gre_tunnel_setup,
1604         .validate       = ip6gre_tunnel_validate,
1605         .newlink        = ip6gre_newlink,
1606         .changelink     = ip6gre_changelink,
1607         .dellink        = ip6gre_dellink,
1608         .get_size       = ip6gre_get_size,
1609         .fill_info      = ip6gre_fill_info,
1610         .get_link_net   = ip6_tnl_get_link_net,
1611 };
1612
1613 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1614         .kind           = "ip6gretap",
1615         .maxtype        = IFLA_GRE_MAX,
1616         .policy         = ip6gre_policy,
1617         .priv_size      = sizeof(struct ip6_tnl),
1618         .setup          = ip6gre_tap_setup,
1619         .validate       = ip6gre_tap_validate,
1620         .newlink        = ip6gre_newlink,
1621         .changelink     = ip6gre_changelink,
1622         .get_size       = ip6gre_get_size,
1623         .fill_info      = ip6gre_fill_info,
1624         .get_link_net   = ip6_tnl_get_link_net,
1625 };
1626
1627 /*
1628  *      And now the modules code and kernel interface.
1629  */
1630
1631 static int __init ip6gre_init(void)
1632 {
1633         int err;
1634
1635         pr_info("GRE over IPv6 tunneling driver\n");
1636
1637         err = register_pernet_device(&ip6gre_net_ops);
1638         if (err < 0)
1639                 return err;
1640
1641         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1642         if (err < 0) {
1643                 pr_info("%s: can't add protocol\n", __func__);
1644                 goto add_proto_failed;
1645         }
1646
1647         err = rtnl_link_register(&ip6gre_link_ops);
1648         if (err < 0)
1649                 goto rtnl_link_failed;
1650
1651         err = rtnl_link_register(&ip6gre_tap_ops);
1652         if (err < 0)
1653                 goto tap_ops_failed;
1654
1655 out:
1656         return err;
1657
1658 tap_ops_failed:
1659         rtnl_link_unregister(&ip6gre_link_ops);
1660 rtnl_link_failed:
1661         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1662 add_proto_failed:
1663         unregister_pernet_device(&ip6gre_net_ops);
1664         goto out;
1665 }
1666
1667 static void __exit ip6gre_fini(void)
1668 {
1669         rtnl_link_unregister(&ip6gre_tap_ops);
1670         rtnl_link_unregister(&ip6gre_link_ops);
1671         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1672         unregister_pernet_device(&ip6gre_net_ops);
1673 }
1674
1675 module_init(ip6gre_init);
1676 module_exit(ip6gre_fini);
1677 MODULE_LICENSE("GPL");
1678 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1679 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1680 MODULE_ALIAS_RTNL_LINK("ip6gre");
1681 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1682 MODULE_ALIAS_NETDEV("ip6gre0");