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 IP6_GRE_HASH_SIZE_SHIFT  5
  65 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
  66
  67 static unsigned int ip6gre_net_id __read_mostly;
  68 struct ip6gre_net {
  69         struct ip6_tnl __rcu *tunnels[4][IP6_GRE_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))&(IP6_GRE_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, IP6_GRE_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 & TUNNEL_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 gre_base_hdr *greh;
 373         const struct ipv6hdr *ipv6h;
 374         int grehlen = sizeof(*greh);
 375         struct ip6_tnl *t;
 376         int key_off = 0;
 377         __be16 flags;
 378         __be32 key;
 379
 380         if (!pskb_may_pull(skb, offset + grehlen))
 381                 return;
 382         greh = (const struct gre_base_hdr *)(skb->data + offset);
 383         flags = greh->flags;
 384         if (flags & (GRE_VERSION | GRE_ROUTING))
 385                 return;
 386         if (flags & GRE_CSUM)
 387                 grehlen += 4;
 388         if (flags & GRE_KEY) {
 389                 key_off = grehlen + offset;
 390                 grehlen += 4;
 391         }
 392
 393         if (!pskb_may_pull(skb, offset + grehlen))
 394                 return;
 395         ipv6h = (const struct ipv6hdr *)skb->data;
 396         greh = (const struct gre_base_hdr *)(skb->data + offset);
 397         key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
 398
 399         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 400                                  key, greh->protocol);
 401         if (!t)
 402                 return;
 403
 404         switch (type) {
 405                 __u32 teli;
 406                 struct ipv6_tlv_tnl_enc_lim *tel;
 407                 __u32 mtu;
 408         case ICMPV6_DEST_UNREACH:
 409                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 410                                     t->parms.name);
 411                 break;
 412         case ICMPV6_TIME_EXCEED:
 413                 if (code == ICMPV6_EXC_HOPLIMIT) {
 414                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 415                                             t->parms.name);
 416                 }
 417                 break;
 418         case ICMPV6_PARAMPROB:
 419                 teli = 0;
 420                 if (code == ICMPV6_HDR_FIELD)
 421                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 422
 423                 if (teli && teli == be32_to_cpu(info) - 2) {
 424                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 425                         if (tel->encap_limit == 0) {
 426                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 427                                                     t->parms.name);
 428                         }
 429                 } else {
 430                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 431                                             t->parms.name);
 432                 }
 433                 break;
 434         case ICMPV6_PKT_TOOBIG:
 435                 mtu = be32_to_cpu(info) - offset;
 436                 if (mtu < IPV6_MIN_MTU)
 437                         mtu = IPV6_MIN_MTU;
 438                 t->dev->mtu = mtu;
 439                 break;
 440         }
 441
 442         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 443                 t->err_count++;
 444         else
 445                 t->err_count = 1;
 446         t->err_time = jiffies;
 447 }
 448
 449 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 450 {
 451         const struct ipv6hdr *ipv6h;
 452         struct ip6_tnl *tunnel;
 453
 454         ipv6h = ipv6_hdr(skb);
 455         tunnel = ip6gre_tunnel_lookup(skb->dev,
 456                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 457                                       tpi->proto);
 458         if (tunnel) {
 459                 ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
 460
 461                 return PACKET_RCVD;
 462         }
 463
 464         return PACKET_REJECT;
 465 }
 466
 467 static int gre_rcv(struct sk_buff *skb)
 468 {
 469         struct tnl_ptk_info tpi;
 470         bool csum_err = false;
 471         int hdr_len;
 472
 473         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
 474         if (hdr_len < 0)
 475                 goto drop;
 476
 477         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
 478                 goto drop;
 479
 480         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
 481                 return 0;
 482
 483         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 484 drop:
 485         kfree_skb(skb);
 486         return 0;
 487 }
 488
 489 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
 490 {
 491         return iptunnel_handle_offloads(skb,
 492                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 493 }
 494
 495 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
 496                                struct net_device *dev, __u8 dsfield,
 497                                struct flowi6 *fl6, int encap_limit,
 498                                __u32 *pmtu, __be16 proto)
 499 {
 500         struct ip6_tnl *tunnel = netdev_priv(dev);
 501         __be16 protocol = (dev->type == ARPHRD_ETHER) ?
 502                           htons(ETH_P_TEB) : proto;
 503
 504         if (dev->type == ARPHRD_ETHER)
 505                 IPCB(skb)->flags = 0;
 506
 507         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
 508                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
 509         else
 510                 fl6->daddr = tunnel->parms.raddr;
 511
 512         if (tunnel->parms.o_flags & TUNNEL_SEQ)
 513                 tunnel->o_seqno++;
 514
 515         /* Push GRE header. */
 516         gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
 517                          protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
 518
 519         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
 520                             NEXTHDR_GRE);
 521 }
 522
 523 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 524 {
 525         struct ip6_tnl *t = netdev_priv(dev);
 526         const struct iphdr  *iph = ip_hdr(skb);
 527         int encap_limit = -1;
 528         struct flowi6 fl6;
 529         __u8 dsfield;
 530         __u32 mtu;
 531         int err;
 532
 533         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 534
 535         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 536                 encap_limit = t->parms.encap_limit;
 537
 538         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 539
 540         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 541                 dsfield = ipv4_get_dsfield(iph);
 542         else
 543                 dsfield = ip6_tclass(t->parms.flowinfo);
 544         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 545                 fl6.flowi6_mark = skb->mark;
 546         else
 547                 fl6.flowi6_mark = t->parms.fwmark;
 548
 549         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 550
 551         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 552         if (err)
 553                 return -1;
 554
 555         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 556                           skb->protocol);
 557         if (err != 0) {
 558                 /* XXX: send ICMP error even if DF is not set. */
 559                 if (err == -EMSGSIZE)
 560                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 561                                   htonl(mtu));
 562                 return -1;
 563         }
 564
 565         return 0;
 566 }
 567
 568 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 569 {
 570         struct ip6_tnl *t = netdev_priv(dev);
 571         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 572         int encap_limit = -1;
 573         __u16 offset;
 574         struct flowi6 fl6;
 575         __u8 dsfield;
 576         __u32 mtu;
 577         int err;
 578
 579         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 580                 return -1;
 581
 582         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 583         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 584         ipv6h = ipv6_hdr(skb);
 585
 586         if (offset > 0) {
 587                 struct ipv6_tlv_tnl_enc_lim *tel;
 588                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 589                 if (tel->encap_limit == 0) {
 590                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 591                                     ICMPV6_HDR_FIELD, offset + 2);
 592                         return -1;
 593                 }
 594                 encap_limit = tel->encap_limit - 1;
 595         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 596                 encap_limit = t->parms.encap_limit;
 597
 598         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 599
 600         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 601                 dsfield = ipv6_get_dsfield(ipv6h);
 602         else
 603                 dsfield = ip6_tclass(t->parms.flowinfo);
 604
 605         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 606                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
 607         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 608                 fl6.flowi6_mark = skb->mark;
 609         else
 610                 fl6.flowi6_mark = t->parms.fwmark;
 611
 612         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 613
 614         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
 615                 return -1;
 616
 617         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
 618                           &mtu, skb->protocol);
 619         if (err != 0) {
 620                 if (err == -EMSGSIZE)
 621                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 622                 return -1;
 623         }
 624
 625         return 0;
 626 }
 627
 628 /**
 629  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 630  *   @t: the outgoing tunnel device
 631  *   @hdr: IPv6 header from the incoming packet
 632  *
 633  * Description:
 634  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 635  *   doesn't match source of incoming packet.
 636  *
 637  * Return:
 638  *   1 if conflict,
 639  *   0 else
 640  **/
 641
 642 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 643         const struct ipv6hdr *hdr)
 644 {
 645         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 646 }
 647
 648 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 649 {
 650         struct ip6_tnl *t = netdev_priv(dev);
 651         int encap_limit = -1;
 652         struct flowi6 fl6;
 653         __u32 mtu;
 654         int err;
 655
 656         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 657                 encap_limit = t->parms.encap_limit;
 658
 659         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 660
 661         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 662         if (err)
 663                 return err;
 664
 665         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
 666
 667         return err;
 668 }
 669
 670 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 671         struct net_device *dev)
 672 {
 673         struct ip6_tnl *t = netdev_priv(dev);
 674         struct net_device_stats *stats = &t->dev->stats;
 675         int ret;
 676
 677         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 678                 goto tx_err;
 679
 680         switch (skb->protocol) {
 681         case htons(ETH_P_IP):
 682                 ret = ip6gre_xmit_ipv4(skb, dev);
 683                 break;
 684         case htons(ETH_P_IPV6):
 685                 ret = ip6gre_xmit_ipv6(skb, dev);
 686                 break;
 687         default:
 688                 ret = ip6gre_xmit_other(skb, dev);
 689                 break;
 690         }
 691
 692         if (ret < 0)
 693                 goto tx_err;
 694
 695         return NETDEV_TX_OK;
 696
 697 tx_err:
 698         stats->tx_errors++;
 699         stats->tx_dropped++;
 700         kfree_skb(skb);
 701         return NETDEV_TX_OK;
 702 }
 703
 704 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 705 {
 706         struct net_device *dev = t->dev;
 707         struct __ip6_tnl_parm *p = &t->parms;
 708         struct flowi6 *fl6 = &t->fl.u.ip6;
 709         int t_hlen;
 710
 711         if (dev->type != ARPHRD_ETHER) {
 712                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 713                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 714         }
 715
 716         /* Set up flowi template */
 717         fl6->saddr = p->laddr;
 718         fl6->daddr = p->raddr;
 719         fl6->flowi6_oif = p->link;
 720         fl6->flowlabel = 0;
 721         fl6->flowi6_proto = IPPROTO_GRE;
 722
 723         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 724                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 725         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 726                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 727
 728         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 729         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 730
 731         if (p->flags&IP6_TNL_F_CAP_XMIT &&
 732                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
 733                 dev->flags |= IFF_POINTOPOINT;
 734         else
 735                 dev->flags &= ~IFF_POINTOPOINT;
 736
 737         t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
 738
 739         t->hlen = t->encap_hlen + t->tun_hlen;
 740
 741         t_hlen = t->hlen + sizeof(struct ipv6hdr);
 742
 743         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 744                 int strict = (ipv6_addr_type(&p->raddr) &
 745                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 746
 747                 struct rt6_info *rt = rt6_lookup(t->net,
 748                                                  &p->raddr, &p->laddr,
 749                                                  p->link, strict);
 750
 751                 if (!rt)
 752                         return;
 753
 754                 if (rt->dst.dev) {
 755                         dev->hard_header_len = rt->dst.dev->hard_header_len +
 756                                                t_hlen;
 757
 758                         if (set_mtu) {
 759                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
 760                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 761                                         dev->mtu -= 8;
 762                                 if (dev->type == ARPHRD_ETHER)
 763                                         dev->mtu -= ETH_HLEN;
 764
 765                                 if (dev->mtu < IPV6_MIN_MTU)
 766                                         dev->mtu = IPV6_MIN_MTU;
 767                         }
 768                 }
 769                 ip6_rt_put(rt);
 770         }
 771 }
 772
 773 static int ip6gre_tnl_change(struct ip6_tnl *t,
 774         const struct __ip6_tnl_parm *p, int set_mtu)
 775 {
 776         t->parms.laddr = p->laddr;
 777         t->parms.raddr = p->raddr;
 778         t->parms.flags = p->flags;
 779         t->parms.hop_limit = p->hop_limit;
 780         t->parms.encap_limit = p->encap_limit;
 781         t->parms.flowinfo = p->flowinfo;
 782         t->parms.link = p->link;
 783         t->parms.proto = p->proto;
 784         t->parms.i_key = p->i_key;
 785         t->parms.o_key = p->o_key;
 786         t->parms.i_flags = p->i_flags;
 787         t->parms.o_flags = p->o_flags;
 788         t->parms.fwmark = p->fwmark;
 789         dst_cache_reset(&t->dst_cache);
 790         ip6gre_tnl_link_config(t, set_mtu);
 791         return 0;
 792 }
 793
 794 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
 795         const struct ip6_tnl_parm2 *u)
 796 {
 797         p->laddr = u->laddr;
 798         p->raddr = u->raddr;
 799         p->flags = u->flags;
 800         p->hop_limit = u->hop_limit;
 801         p->encap_limit = u->encap_limit;
 802         p->flowinfo = u->flowinfo;
 803         p->link = u->link;
 804         p->i_key = u->i_key;
 805         p->o_key = u->o_key;
 806         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
 807         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
 808         memcpy(p->name, u->name, sizeof(u->name));
 809 }
 810
 811 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
 812         const struct __ip6_tnl_parm *p)
 813 {
 814         u->proto = IPPROTO_GRE;
 815         u->laddr = p->laddr;
 816         u->raddr = p->raddr;
 817         u->flags = p->flags;
 818         u->hop_limit = p->hop_limit;
 819         u->encap_limit = p->encap_limit;
 820         u->flowinfo = p->flowinfo;
 821         u->link = p->link;
 822         u->i_key = p->i_key;
 823         u->o_key = p->o_key;
 824         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
 825         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
 826         memcpy(u->name, p->name, sizeof(u->name));
 827 }
 828
 829 static int ip6gre_tunnel_ioctl(struct net_device *dev,
 830         struct ifreq *ifr, int cmd)
 831 {
 832         int err = 0;
 833         struct ip6_tnl_parm2 p;
 834         struct __ip6_tnl_parm p1;
 835         struct ip6_tnl *t = netdev_priv(dev);
 836         struct net *net = t->net;
 837         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 838
 839         memset(&p1, 0, sizeof(p1));
 840
 841         switch (cmd) {
 842         case SIOCGETTUNNEL:
 843                 if (dev == ign->fb_tunnel_dev) {
 844                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 845                                 err = -EFAULT;
 846                                 break;
 847                         }
 848                         ip6gre_tnl_parm_from_user(&p1, &p);
 849                         t = ip6gre_tunnel_locate(net, &p1, 0);
 850                         if (!t)
 851                                 t = netdev_priv(dev);
 852                 }
 853                 memset(&p, 0, sizeof(p));
 854                 ip6gre_tnl_parm_to_user(&p, &t->parms);
 855                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 856                         err = -EFAULT;
 857                 break;
 858
 859         case SIOCADDTUNNEL:
 860         case SIOCCHGTUNNEL:
 861                 err = -EPERM;
 862                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 863                         goto done;
 864
 865                 err = -EFAULT;
 866                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 867                         goto done;
 868
 869                 err = -EINVAL;
 870                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
 871                         goto done;
 872
 873                 if (!(p.i_flags&GRE_KEY))
 874                         p.i_key = 0;
 875                 if (!(p.o_flags&GRE_KEY))
 876                         p.o_key = 0;
 877
 878                 ip6gre_tnl_parm_from_user(&p1, &p);
 879                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
 880
 881                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 882                         if (t) {
 883                                 if (t->dev != dev) {
 884                                         err = -EEXIST;
 885                                         break;
 886                                 }
 887                         } else {
 888                                 t = netdev_priv(dev);
 889
 890                                 ip6gre_tunnel_unlink(ign, t);
 891                                 synchronize_net();
 892                                 ip6gre_tnl_change(t, &p1, 1);
 893                                 ip6gre_tunnel_link(ign, t);
 894                                 netdev_state_change(dev);
 895                         }
 896                 }
 897
 898                 if (t) {
 899                         err = 0;
 900
 901                         memset(&p, 0, sizeof(p));
 902                         ip6gre_tnl_parm_to_user(&p, &t->parms);
 903                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 904                                 err = -EFAULT;
 905                 } else
 906                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 907                 break;
 908
 909         case SIOCDELTUNNEL:
 910                 err = -EPERM;
 911                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 912                         goto done;
 913
 914                 if (dev == ign->fb_tunnel_dev) {
 915                         err = -EFAULT;
 916                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 917                                 goto done;
 918                         err = -ENOENT;
 919                         ip6gre_tnl_parm_from_user(&p1, &p);
 920                         t = ip6gre_tunnel_locate(net, &p1, 0);
 921                         if (!t)
 922                                 goto done;
 923                         err = -EPERM;
 924                         if (t == netdev_priv(ign->fb_tunnel_dev))
 925                                 goto done;
 926                         dev = t->dev;
 927                 }
 928                 unregister_netdevice(dev);
 929                 err = 0;
 930                 break;
 931
 932         default:
 933                 err = -EINVAL;
 934         }
 935
 936 done:
 937         return err;
 938 }
 939
 940 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
 941                         unsigned short type,
 942                         const void *daddr, const void *saddr, unsigned int len)
 943 {
 944         struct ip6_tnl *t = netdev_priv(dev);
 945         struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
 946         __be16 *p = (__be16 *)(ipv6h+1);
 947
 948         ip6_flow_hdr(ipv6h, 0,
 949                      ip6_make_flowlabel(dev_net(dev), skb,
 950                                         t->fl.u.ip6.flowlabel, true,
 951                                         &t->fl.u.ip6));
 952         ipv6h->hop_limit = t->parms.hop_limit;
 953         ipv6h->nexthdr = NEXTHDR_GRE;
 954         ipv6h->saddr = t->parms.laddr;
 955         ipv6h->daddr = t->parms.raddr;
 956
 957         p[0]            = t->parms.o_flags;
 958         p[1]            = htons(type);
 959
 960         /*
 961          *      Set the source hardware address.
 962          */
 963
 964         if (saddr)
 965                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
 966         if (daddr)
 967                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
 968         if (!ipv6_addr_any(&ipv6h->daddr))
 969                 return t->hlen;
 970
 971         return -t->hlen;
 972 }
 973
 974 static const struct header_ops ip6gre_header_ops = {
 975         .create = ip6gre_header,
 976 };
 977
 978 static const struct net_device_ops ip6gre_netdev_ops = {
 979         .ndo_init               = ip6gre_tunnel_init,
 980         .ndo_uninit             = ip6gre_tunnel_uninit,
 981         .ndo_start_xmit         = ip6gre_tunnel_xmit,
 982         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
 983         .ndo_change_mtu         = ip6_tnl_change_mtu,
 984         .ndo_get_stats64        = ip_tunnel_get_stats64,
 985         .ndo_get_iflink         = ip6_tnl_get_iflink,
 986 };
 987
 988 static void ip6gre_dev_free(struct net_device *dev)
 989 {
 990         struct ip6_tnl *t = netdev_priv(dev);
 991
 992         dst_cache_destroy(&t->dst_cache);
 993         free_percpu(dev->tstats);
 994 }
 995
 996 static void ip6gre_tunnel_setup(struct net_device *dev)
 997 {
 998         dev->netdev_ops = &ip6gre_netdev_ops;
 999         dev->needs_free_netdev = true;
1000         dev->priv_destructor = ip6gre_dev_free;
1001
1002         dev->type = ARPHRD_IP6GRE;
1003
1004         dev->flags |= IFF_NOARP;
1005         dev->addr_len = sizeof(struct in6_addr);
1006         netif_keep_dst(dev);
1007         /* This perm addr will be used as interface identifier by IPv6 */
1008         dev->addr_assign_type = NET_ADDR_RANDOM;
1009         eth_random_addr(dev->perm_addr);
1010 }
1011
1012 static int ip6gre_tunnel_init_common(struct net_device *dev)
1013 {
1014         struct ip6_tnl *tunnel;
1015         int ret;
1016         int t_hlen;
1017
1018         tunnel = netdev_priv(dev);
1019
1020         tunnel->dev = dev;
1021         tunnel->net = dev_net(dev);
1022         strcpy(tunnel->parms.name, dev->name);
1023
1024         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1025         if (!dev->tstats)
1026                 return -ENOMEM;
1027
1028         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1029         if (ret) {
1030                 free_percpu(dev->tstats);
1031                 dev->tstats = NULL;
1032                 return ret;
1033         }
1034
1035         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1036         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1037         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1038
1039         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1040         dev->mtu = ETH_DATA_LEN - t_hlen;
1041         if (dev->type == ARPHRD_ETHER)
1042                 dev->mtu -= ETH_HLEN;
1043         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1044                 dev->mtu -= 8;
1045
1046         return 0;
1047 }
1048
1049 static int ip6gre_tunnel_init(struct net_device *dev)
1050 {
1051         struct ip6_tnl *tunnel;
1052         int ret;
1053
1054         ret = ip6gre_tunnel_init_common(dev);
1055         if (ret)
1056                 return ret;
1057
1058         tunnel = netdev_priv(dev);
1059
1060         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1061         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1062
1063         if (ipv6_addr_any(&tunnel->parms.raddr))
1064                 dev->header_ops = &ip6gre_header_ops;
1065
1066         return 0;
1067 }
1068
1069 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1070 {
1071         struct ip6_tnl *tunnel = netdev_priv(dev);
1072
1073         tunnel->dev = dev;
1074         tunnel->net = dev_net(dev);
1075         strcpy(tunnel->parms.name, dev->name);
1076
1077         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1078
1079         dev_hold(dev);
1080 }
1081
1082
1083 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1084         .handler     = gre_rcv,
1085         .err_handler = ip6gre_err,
1086         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1087 };
1088
1089 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1090 {
1091         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1092         struct net_device *dev, *aux;
1093         int prio;
1094
1095         for_each_netdev_safe(net, dev, aux)
1096                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1097                     dev->rtnl_link_ops == &ip6gre_tap_ops)
1098                         unregister_netdevice_queue(dev, head);
1099
1100         for (prio = 0; prio < 4; prio++) {
1101                 int h;
1102                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1103                         struct ip6_tnl *t;
1104
1105                         t = rtnl_dereference(ign->tunnels[prio][h]);
1106
1107                         while (t) {
1108                                 /* If dev is in the same netns, it has already
1109                                  * been added to the list by the previous loop.
1110                                  */
1111                                 if (!net_eq(dev_net(t->dev), net))
1112                                         unregister_netdevice_queue(t->dev,
1113                                                                    head);
1114                                 t = rtnl_dereference(t->next);
1115                         }
1116                 }
1117         }
1118 }
1119
1120 static int __net_init ip6gre_init_net(struct net *net)
1121 {
1122         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1123         int err;
1124
1125         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1126                                           NET_NAME_UNKNOWN,
1127                                           ip6gre_tunnel_setup);
1128         if (!ign->fb_tunnel_dev) {
1129                 err = -ENOMEM;
1130                 goto err_alloc_dev;
1131         }
1132         dev_net_set(ign->fb_tunnel_dev, net);
1133         /* FB netdevice is special: we have one, and only one per netns.
1134          * Allowing to move it to another netns is clearly unsafe.
1135          */
1136         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1137
1138
1139         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1140         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1141
1142         err = register_netdev(ign->fb_tunnel_dev);
1143         if (err)
1144                 goto err_reg_dev;
1145
1146         rcu_assign_pointer(ign->tunnels_wc[0],
1147                            netdev_priv(ign->fb_tunnel_dev));
1148         return 0;
1149
1150 err_reg_dev:
1151         free_netdev(ign->fb_tunnel_dev);
1152 err_alloc_dev:
1153         return err;
1154 }
1155
1156 static void __net_exit ip6gre_exit_net(struct net *net)
1157 {
1158         LIST_HEAD(list);
1159
1160         rtnl_lock();
1161         ip6gre_destroy_tunnels(net, &list);
1162         unregister_netdevice_many(&list);
1163         rtnl_unlock();
1164 }
1165
1166 static struct pernet_operations ip6gre_net_ops = {
1167         .init = ip6gre_init_net,
1168         .exit = ip6gre_exit_net,
1169         .id   = &ip6gre_net_id,
1170         .size = sizeof(struct ip6gre_net),
1171 };
1172
1173 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1174 {
1175         __be16 flags;
1176
1177         if (!data)
1178                 return 0;
1179
1180         flags = 0;
1181         if (data[IFLA_GRE_IFLAGS])
1182                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1183         if (data[IFLA_GRE_OFLAGS])
1184                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1185         if (flags & (GRE_VERSION|GRE_ROUTING))
1186                 return -EINVAL;
1187
1188         return 0;
1189 }
1190
1191 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1192 {
1193         struct in6_addr daddr;
1194
1195         if (tb[IFLA_ADDRESS]) {
1196                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1197                         return -EINVAL;
1198                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1199                         return -EADDRNOTAVAIL;
1200         }
1201
1202         if (!data)
1203                 goto out;
1204
1205         if (data[IFLA_GRE_REMOTE]) {
1206                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1207                 if (ipv6_addr_any(&daddr))
1208                         return -EINVAL;
1209         }
1210
1211 out:
1212         return ip6gre_tunnel_validate(tb, data);
1213 }
1214
1215
1216 static void ip6gre_netlink_parms(struct nlattr *data[],
1217                                 struct __ip6_tnl_parm *parms)
1218 {
1219         memset(parms, 0, sizeof(*parms));
1220
1221         if (!data)
1222                 return;
1223
1224         if (data[IFLA_GRE_LINK])
1225                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1226
1227         if (data[IFLA_GRE_IFLAGS])
1228                 parms->i_flags = gre_flags_to_tnl_flags(
1229                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1230
1231         if (data[IFLA_GRE_OFLAGS])
1232                 parms->o_flags = gre_flags_to_tnl_flags(
1233                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1234
1235         if (data[IFLA_GRE_IKEY])
1236                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1237
1238         if (data[IFLA_GRE_OKEY])
1239                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1240
1241         if (data[IFLA_GRE_LOCAL])
1242                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1243
1244         if (data[IFLA_GRE_REMOTE])
1245                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1246
1247         if (data[IFLA_GRE_TTL])
1248                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1249
1250         if (data[IFLA_GRE_ENCAP_LIMIT])
1251                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1252
1253         if (data[IFLA_GRE_FLOWINFO])
1254                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1255
1256         if (data[IFLA_GRE_FLAGS])
1257                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1258
1259         if (data[IFLA_GRE_FWMARK])
1260                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1261 }
1262
1263 static int ip6gre_tap_init(struct net_device *dev)
1264 {
1265         struct ip6_tnl *tunnel;
1266         int ret;
1267
1268         ret = ip6gre_tunnel_init_common(dev);
1269         if (ret)
1270                 return ret;
1271
1272         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1273
1274         tunnel = netdev_priv(dev);
1275
1276         ip6gre_tnl_link_config(tunnel, 1);
1277
1278         return 0;
1279 }
1280
1281 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1282         .ndo_init = ip6gre_tap_init,
1283         .ndo_uninit = ip6gre_tunnel_uninit,
1284         .ndo_start_xmit = ip6gre_tunnel_xmit,
1285         .ndo_set_mac_address = eth_mac_addr,
1286         .ndo_validate_addr = eth_validate_addr,
1287         .ndo_change_mtu = ip6_tnl_change_mtu,
1288         .ndo_get_stats64 = ip_tunnel_get_stats64,
1289         .ndo_get_iflink = ip6_tnl_get_iflink,
1290 };
1291
1292 #define GRE6_FEATURES (NETIF_F_SG |             \
1293                        NETIF_F_FRAGLIST |       \
1294                        NETIF_F_HIGHDMA |                \
1295                        NETIF_F_HW_CSUM)
1296
1297 static void ip6gre_tap_setup(struct net_device *dev)
1298 {
1299
1300         ether_setup(dev);
1301
1302         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1303         dev->needs_free_netdev = true;
1304         dev->priv_destructor = ip6gre_dev_free;
1305
1306         dev->features |= NETIF_F_NETNS_LOCAL;
1307         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1308         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1309 }
1310
1311 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1312                                        struct ip_tunnel_encap *ipencap)
1313 {
1314         bool ret = false;
1315
1316         memset(ipencap, 0, sizeof(*ipencap));
1317
1318         if (!data)
1319                 return ret;
1320
1321         if (data[IFLA_GRE_ENCAP_TYPE]) {
1322                 ret = true;
1323                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1324         }
1325
1326         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1327                 ret = true;
1328                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1329         }
1330
1331         if (data[IFLA_GRE_ENCAP_SPORT]) {
1332                 ret = true;
1333                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1334         }
1335
1336         if (data[IFLA_GRE_ENCAP_DPORT]) {
1337                 ret = true;
1338                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1339         }
1340
1341         return ret;
1342 }
1343
1344 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1345         struct nlattr *tb[], struct nlattr *data[])
1346 {
1347         struct ip6_tnl *nt;
1348         struct net *net = dev_net(dev);
1349         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1350         struct ip_tunnel_encap ipencap;
1351         int err;
1352
1353         nt = netdev_priv(dev);
1354
1355         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1356                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1357
1358                 if (err < 0)
1359                         return err;
1360         }
1361
1362         ip6gre_netlink_parms(data, &nt->parms);
1363
1364         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1365                 return -EEXIST;
1366
1367         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1368                 eth_hw_addr_random(dev);
1369
1370         nt->dev = dev;
1371         nt->net = dev_net(dev);
1372         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1373
1374         dev->features           |= GRE6_FEATURES;
1375         dev->hw_features        |= GRE6_FEATURES;
1376
1377         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1378                 /* TCP offload with GRE SEQ is not supported, nor
1379                  * can we support 2 levels of outer headers requiring
1380                  * an update.
1381                  */
1382                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1383                     (nt->encap.type == TUNNEL_ENCAP_NONE)) {
1384                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1385                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1386                 }
1387
1388                 /* Can use a lockless transmit, unless we generate
1389                  * output sequences
1390                  */
1391                 dev->features |= NETIF_F_LLTX;
1392         }
1393
1394         err = register_netdevice(dev);
1395         if (err)
1396                 goto out;
1397
1398         dev_hold(dev);
1399         ip6gre_tunnel_link(ign, nt);
1400
1401 out:
1402         return err;
1403 }
1404
1405 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1406                             struct nlattr *data[])
1407 {
1408         struct ip6_tnl *t, *nt = netdev_priv(dev);
1409         struct net *net = nt->net;
1410         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1411         struct __ip6_tnl_parm p;
1412         struct ip_tunnel_encap ipencap;
1413
1414         if (dev == ign->fb_tunnel_dev)
1415                 return -EINVAL;
1416
1417         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1418                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1419
1420                 if (err < 0)
1421                         return err;
1422         }
1423
1424         ip6gre_netlink_parms(data, &p);
1425
1426         t = ip6gre_tunnel_locate(net, &p, 0);
1427
1428         if (t) {
1429                 if (t->dev != dev)
1430                         return -EEXIST;
1431         } else {
1432                 t = nt;
1433         }
1434
1435         ip6gre_tunnel_unlink(ign, t);
1436         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1437         ip6gre_tunnel_link(ign, t);
1438         return 0;
1439 }
1440
1441 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1442 {
1443         struct net *net = dev_net(dev);
1444         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1445
1446         if (dev != ign->fb_tunnel_dev)
1447                 unregister_netdevice_queue(dev, head);
1448 }
1449
1450 static size_t ip6gre_get_size(const struct net_device *dev)
1451 {
1452         return
1453                 /* IFLA_GRE_LINK */
1454                 nla_total_size(4) +
1455                 /* IFLA_GRE_IFLAGS */
1456                 nla_total_size(2) +
1457                 /* IFLA_GRE_OFLAGS */
1458                 nla_total_size(2) +
1459                 /* IFLA_GRE_IKEY */
1460                 nla_total_size(4) +
1461                 /* IFLA_GRE_OKEY */
1462                 nla_total_size(4) +
1463                 /* IFLA_GRE_LOCAL */
1464                 nla_total_size(sizeof(struct in6_addr)) +
1465                 /* IFLA_GRE_REMOTE */
1466                 nla_total_size(sizeof(struct in6_addr)) +
1467                 /* IFLA_GRE_TTL */
1468                 nla_total_size(1) +
1469                 /* IFLA_GRE_ENCAP_LIMIT */
1470                 nla_total_size(1) +
1471                 /* IFLA_GRE_FLOWINFO */
1472                 nla_total_size(4) +
1473                 /* IFLA_GRE_FLAGS */
1474                 nla_total_size(4) +
1475                 /* IFLA_GRE_ENCAP_TYPE */
1476                 nla_total_size(2) +
1477                 /* IFLA_GRE_ENCAP_FLAGS */
1478                 nla_total_size(2) +
1479                 /* IFLA_GRE_ENCAP_SPORT */
1480                 nla_total_size(2) +
1481                 /* IFLA_GRE_ENCAP_DPORT */
1482                 nla_total_size(2) +
1483                 /* IFLA_GRE_FWMARK */
1484                 nla_total_size(4) +
1485                 0;
1486 }
1487
1488 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1489 {
1490         struct ip6_tnl *t = netdev_priv(dev);
1491         struct __ip6_tnl_parm *p = &t->parms;
1492
1493         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1494             nla_put_be16(skb, IFLA_GRE_IFLAGS,
1495                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1496             nla_put_be16(skb, IFLA_GRE_OFLAGS,
1497                          gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1498             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1499             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1500             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1501             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1502             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1503             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1504             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1505             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1506             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1507                 goto nla_put_failure;
1508
1509         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1510                         t->encap.type) ||
1511             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1512                          t->encap.sport) ||
1513             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1514                          t->encap.dport) ||
1515             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1516                         t->encap.flags))
1517                 goto nla_put_failure;
1518
1519         return 0;
1520
1521 nla_put_failure:
1522         return -EMSGSIZE;
1523 }
1524
1525 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1526         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1527         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1528         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1529         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1530         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1531         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1532         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1533         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1534         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1535         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1536         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1537         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
1538         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
1539         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
1540         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
1541         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
1542 };
1543
1544 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1545         .kind           = "ip6gre",
1546         .maxtype        = IFLA_GRE_MAX,
1547         .policy         = ip6gre_policy,
1548         .priv_size      = sizeof(struct ip6_tnl),
1549         .setup          = ip6gre_tunnel_setup,
1550         .validate       = ip6gre_tunnel_validate,
1551         .newlink        = ip6gre_newlink,
1552         .changelink     = ip6gre_changelink,
1553         .dellink        = ip6gre_dellink,
1554         .get_size       = ip6gre_get_size,
1555         .fill_info      = ip6gre_fill_info,
1556         .get_link_net   = ip6_tnl_get_link_net,
1557 };
1558
1559 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1560         .kind           = "ip6gretap",
1561         .maxtype        = IFLA_GRE_MAX,
1562         .policy         = ip6gre_policy,
1563         .priv_size      = sizeof(struct ip6_tnl),
1564         .setup          = ip6gre_tap_setup,
1565         .validate       = ip6gre_tap_validate,
1566         .newlink        = ip6gre_newlink,
1567         .changelink     = ip6gre_changelink,
1568         .get_size       = ip6gre_get_size,
1569         .fill_info      = ip6gre_fill_info,
1570         .get_link_net   = ip6_tnl_get_link_net,
1571 };
1572
1573 /*
1574  *      And now the modules code and kernel interface.
1575  */
1576
1577 static int __init ip6gre_init(void)
1578 {
1579         int err;
1580
1581         pr_info("GRE over IPv6 tunneling driver\n");
1582
1583         err = register_pernet_device(&ip6gre_net_ops);
1584         if (err < 0)
1585                 return err;
1586
1587         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1588         if (err < 0) {
1589                 pr_info("%s: can't add protocol\n", __func__);
1590                 goto add_proto_failed;
1591         }
1592
1593         err = rtnl_link_register(&ip6gre_link_ops);
1594         if (err < 0)
1595                 goto rtnl_link_failed;
1596
1597         err = rtnl_link_register(&ip6gre_tap_ops);
1598         if (err < 0)
1599                 goto tap_ops_failed;
1600
1601 out:
1602         return err;
1603
1604 tap_ops_failed:
1605         rtnl_link_unregister(&ip6gre_link_ops);
1606 rtnl_link_failed:
1607         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1608 add_proto_failed:
1609         unregister_pernet_device(&ip6gre_net_ops);
1610         goto out;
1611 }
1612
1613 static void __exit ip6gre_fini(void)
1614 {
1615         rtnl_link_unregister(&ip6gre_tap_ops);
1616         rtnl_link_unregister(&ip6gre_link_ops);
1617         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1618         unregister_pernet_device(&ip6gre_net_ops);
1619 }
1620
1621 module_init(ip6gre_init);
1622 module_exit(ip6gre_fini);
1623 MODULE_LICENSE("GPL");
1624 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1625 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1626 MODULE_ALIAS_RTNL_LINK("ip6gre");
1627 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1628 MODULE_ALIAS_NETDEV("ip6gre0");