net/ipv6/ip6_gre.c

   1 /*
   2  *      GRE over IPv6 protocol decoder.
   3  *
   4  *      Authors: Dmitry Kozlov ([email protected])
   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                 if (code != ICMPV6_PORT_UNREACH)
 412                         break;
 413                 return;
 414         case ICMPV6_TIME_EXCEED:
 415                 if (code == ICMPV6_EXC_HOPLIMIT) {
 416                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 417                                             t->parms.name);
 418                         break;
 419                 }
 420                 return;
 421         case ICMPV6_PARAMPROB:
 422                 teli = 0;
 423                 if (code == ICMPV6_HDR_FIELD)
 424                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 425
 426                 if (teli && teli == be32_to_cpu(info) - 2) {
 427                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 428                         if (tel->encap_limit == 0) {
 429                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 430                                                     t->parms.name);
 431                         }
 432                 } else {
 433                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 434                                             t->parms.name);
 435                 }
 436                 return;
 437         case ICMPV6_PKT_TOOBIG:
 438                 mtu = be32_to_cpu(info) - offset - t->tun_hlen;
 439                 if (t->dev->type == ARPHRD_ETHER)
 440                         mtu -= ETH_HLEN;
 441                 if (mtu < IPV6_MIN_MTU)
 442                         mtu = IPV6_MIN_MTU;
 443                 t->dev->mtu = mtu;
 444                 return;
 445         }
 446
 447         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 448                 t->err_count++;
 449         else
 450                 t->err_count = 1;
 451         t->err_time = jiffies;
 452 }
 453
 454 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 455 {
 456         const struct ipv6hdr *ipv6h;
 457         struct ip6_tnl *tunnel;
 458
 459         ipv6h = ipv6_hdr(skb);
 460         tunnel = ip6gre_tunnel_lookup(skb->dev,
 461                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 462                                       tpi->proto);
 463         if (tunnel) {
 464                 ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
 465
 466                 return PACKET_RCVD;
 467         }
 468
 469         return PACKET_REJECT;
 470 }
 471
 472 static int gre_rcv(struct sk_buff *skb)
 473 {
 474         struct tnl_ptk_info tpi;
 475         bool csum_err = false;
 476         int hdr_len;
 477
 478         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
 479         if (hdr_len < 0)
 480                 goto drop;
 481
 482         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
 483                 goto drop;
 484
 485         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
 486                 return 0;
 487
 488         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 489 drop:
 490         kfree_skb(skb);
 491         return 0;
 492 }
 493
 494 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
 495 {
 496         return iptunnel_handle_offloads(skb,
 497                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 498 }
 499
 500 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
 501                                struct net_device *dev, __u8 dsfield,
 502                                struct flowi6 *fl6, int encap_limit,
 503                                __u32 *pmtu, __be16 proto)
 504 {
 505         struct ip6_tnl *tunnel = netdev_priv(dev);
 506         struct dst_entry *dst = skb_dst(skb);
 507         __be16 protocol;
 508
 509         if (dev->type == ARPHRD_ETHER)
 510                 IPCB(skb)->flags = 0;
 511
 512         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
 513                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
 514         else
 515                 fl6->daddr = tunnel->parms.raddr;
 516
 517         if (tunnel->parms.o_flags & TUNNEL_SEQ)
 518                 tunnel->o_seqno++;
 519
 520         /* Push GRE header. */
 521         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
 522         gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
 523                          protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
 524
 525         /* TooBig packet may have updated dst->dev's mtu */
 526         if (dst && dst_mtu(dst) > dst->dev->mtu)
 527                 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
 528
 529         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
 530                             NEXTHDR_GRE);
 531 }
 532
 533 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 534 {
 535         struct ip6_tnl *t = netdev_priv(dev);
 536         const struct iphdr  *iph = ip_hdr(skb);
 537         int encap_limit = -1;
 538         struct flowi6 fl6;
 539         __u8 dsfield;
 540         __u32 mtu;
 541         int err;
 542
 543         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 544
 545         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 546                 encap_limit = t->parms.encap_limit;
 547
 548         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 549
 550         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 551                 dsfield = ipv4_get_dsfield(iph);
 552         else
 553                 dsfield = ip6_tclass(t->parms.flowinfo);
 554         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 555                 fl6.flowi6_mark = skb->mark;
 556         else
 557                 fl6.flowi6_mark = t->parms.fwmark;
 558
 559         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 560
 561         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 562         if (err)
 563                 return -1;
 564
 565         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 566                           skb->protocol);
 567         if (err != 0) {
 568                 /* XXX: send ICMP error even if DF is not set. */
 569                 if (err == -EMSGSIZE)
 570                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 571                                   htonl(mtu));
 572                 return -1;
 573         }
 574
 575         return 0;
 576 }
 577
 578 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 579 {
 580         struct ip6_tnl *t = netdev_priv(dev);
 581         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 582         int encap_limit = -1;
 583         __u16 offset;
 584         struct flowi6 fl6;
 585         __u8 dsfield;
 586         __u32 mtu;
 587         int err;
 588
 589         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 590                 return -1;
 591
 592         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 593         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 594         ipv6h = ipv6_hdr(skb);
 595
 596         if (offset > 0) {
 597                 struct ipv6_tlv_tnl_enc_lim *tel;
 598                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 599                 if (tel->encap_limit == 0) {
 600                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 601                                     ICMPV6_HDR_FIELD, offset + 2);
 602                         return -1;
 603                 }
 604                 encap_limit = tel->encap_limit - 1;
 605         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 606                 encap_limit = t->parms.encap_limit;
 607
 608         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 609
 610         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 611                 dsfield = ipv6_get_dsfield(ipv6h);
 612         else
 613                 dsfield = ip6_tclass(t->parms.flowinfo);
 614
 615         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 616                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
 617         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 618                 fl6.flowi6_mark = skb->mark;
 619         else
 620                 fl6.flowi6_mark = t->parms.fwmark;
 621
 622         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 623
 624         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
 625                 return -1;
 626
 627         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
 628                           &mtu, skb->protocol);
 629         if (err != 0) {
 630                 if (err == -EMSGSIZE)
 631                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 632                 return -1;
 633         }
 634
 635         return 0;
 636 }
 637
 638 /**
 639  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 640  *   @t: the outgoing tunnel device
 641  *   @hdr: IPv6 header from the incoming packet
 642  *
 643  * Description:
 644  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 645  *   doesn't match source of incoming packet.
 646  *
 647  * Return:
 648  *   1 if conflict,
 649  *   0 else
 650  **/
 651
 652 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 653         const struct ipv6hdr *hdr)
 654 {
 655         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 656 }
 657
 658 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 659 {
 660         struct ip6_tnl *t = netdev_priv(dev);
 661         int encap_limit = -1;
 662         struct flowi6 fl6;
 663         __u32 mtu;
 664         int err;
 665
 666         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 667                 encap_limit = t->parms.encap_limit;
 668
 669         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 670
 671         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 672         if (err)
 673                 return err;
 674
 675         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
 676
 677         return err;
 678 }
 679
 680 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 681         struct net_device *dev)
 682 {
 683         struct ip6_tnl *t = netdev_priv(dev);
 684         struct net_device_stats *stats = &t->dev->stats;
 685         int ret;
 686
 687         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 688                 goto tx_err;
 689
 690         switch (skb->protocol) {
 691         case htons(ETH_P_IP):
 692                 ret = ip6gre_xmit_ipv4(skb, dev);
 693                 break;
 694         case htons(ETH_P_IPV6):
 695                 ret = ip6gre_xmit_ipv6(skb, dev);
 696                 break;
 697         default:
 698                 ret = ip6gre_xmit_other(skb, dev);
 699                 break;
 700         }
 701
 702         if (ret < 0)
 703                 goto tx_err;
 704
 705         return NETDEV_TX_OK;
 706
 707 tx_err:
 708         stats->tx_errors++;
 709         stats->tx_dropped++;
 710         kfree_skb(skb);
 711         return NETDEV_TX_OK;
 712 }
 713
 714 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 715 {
 716         struct net_device *dev = t->dev;
 717         struct __ip6_tnl_parm *p = &t->parms;
 718         struct flowi6 *fl6 = &t->fl.u.ip6;
 719         int t_hlen;
 720
 721         if (dev->type != ARPHRD_ETHER) {
 722                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 723                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 724         }
 725
 726         /* Set up flowi template */
 727         fl6->saddr = p->laddr;
 728         fl6->daddr = p->raddr;
 729         fl6->flowi6_oif = p->link;
 730         fl6->flowlabel = 0;
 731         fl6->flowi6_proto = IPPROTO_GRE;
 732
 733         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 734                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 735         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 736                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 737
 738         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 739         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 740
 741         if (p->flags&IP6_TNL_F_CAP_XMIT &&
 742                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
 743                 dev->flags |= IFF_POINTOPOINT;
 744         else
 745                 dev->flags &= ~IFF_POINTOPOINT;
 746
 747         t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
 748
 749         t->hlen = t->encap_hlen + t->tun_hlen;
 750
 751         t_hlen = t->hlen + sizeof(struct ipv6hdr);
 752
 753         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 754                 int strict = (ipv6_addr_type(&p->raddr) &
 755                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 756
 757                 struct rt6_info *rt = rt6_lookup(t->net,
 758                                                  &p->raddr, &p->laddr,
 759                                                  p->link, strict);
 760
 761                 if (!rt)
 762                         return;
 763
 764                 if (rt->dst.dev) {
 765                         dev->hard_header_len = rt->dst.dev->hard_header_len +
 766                                                t_hlen;
 767
 768                         if (set_mtu) {
 769                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
 770                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 771                                         dev->mtu -= 8;
 772                                 if (dev->type == ARPHRD_ETHER)
 773                                         dev->mtu -= ETH_HLEN;
 774
 775                                 if (dev->mtu < IPV6_MIN_MTU)
 776                                         dev->mtu = IPV6_MIN_MTU;
 777                         }
 778                 }
 779                 ip6_rt_put(rt);
 780         }
 781 }
 782
 783 static int ip6gre_tnl_change(struct ip6_tnl *t,
 784         const struct __ip6_tnl_parm *p, int set_mtu)
 785 {
 786         t->parms.laddr = p->laddr;
 787         t->parms.raddr = p->raddr;
 788         t->parms.flags = p->flags;
 789         t->parms.hop_limit = p->hop_limit;
 790         t->parms.encap_limit = p->encap_limit;
 791         t->parms.flowinfo = p->flowinfo;
 792         t->parms.link = p->link;
 793         t->parms.proto = p->proto;
 794         t->parms.i_key = p->i_key;
 795         t->parms.o_key = p->o_key;
 796         t->parms.i_flags = p->i_flags;
 797         t->parms.o_flags = p->o_flags;
 798         t->parms.fwmark = p->fwmark;
 799         dst_cache_reset(&t->dst_cache);
 800         ip6gre_tnl_link_config(t, set_mtu);
 801         return 0;
 802 }
 803
 804 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
 805         const struct ip6_tnl_parm2 *u)
 806 {
 807         p->laddr = u->laddr;
 808         p->raddr = u->raddr;
 809         p->flags = u->flags;
 810         p->hop_limit = u->hop_limit;
 811         p->encap_limit = u->encap_limit;
 812         p->flowinfo = u->flowinfo;
 813         p->link = u->link;
 814         p->i_key = u->i_key;
 815         p->o_key = u->o_key;
 816         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
 817         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
 818         memcpy(p->name, u->name, sizeof(u->name));
 819 }
 820
 821 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
 822         const struct __ip6_tnl_parm *p)
 823 {
 824         u->proto = IPPROTO_GRE;
 825         u->laddr = p->laddr;
 826         u->raddr = p->raddr;
 827         u->flags = p->flags;
 828         u->hop_limit = p->hop_limit;
 829         u->encap_limit = p->encap_limit;
 830         u->flowinfo = p->flowinfo;
 831         u->link = p->link;
 832         u->i_key = p->i_key;
 833         u->o_key = p->o_key;
 834         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
 835         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
 836         memcpy(u->name, p->name, sizeof(u->name));
 837 }
 838
 839 static int ip6gre_tunnel_ioctl(struct net_device *dev,
 840         struct ifreq *ifr, int cmd)
 841 {
 842         int err = 0;
 843         struct ip6_tnl_parm2 p;
 844         struct __ip6_tnl_parm p1;
 845         struct ip6_tnl *t = netdev_priv(dev);
 846         struct net *net = t->net;
 847         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 848
 849         memset(&p1, 0, sizeof(p1));
 850
 851         switch (cmd) {
 852         case SIOCGETTUNNEL:
 853                 if (dev == ign->fb_tunnel_dev) {
 854                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 855                                 err = -EFAULT;
 856                                 break;
 857                         }
 858                         ip6gre_tnl_parm_from_user(&p1, &p);
 859                         t = ip6gre_tunnel_locate(net, &p1, 0);
 860                         if (!t)
 861                                 t = netdev_priv(dev);
 862                 }
 863                 memset(&p, 0, sizeof(p));
 864                 ip6gre_tnl_parm_to_user(&p, &t->parms);
 865                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 866                         err = -EFAULT;
 867                 break;
 868
 869         case SIOCADDTUNNEL:
 870         case SIOCCHGTUNNEL:
 871                 err = -EPERM;
 872                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 873                         goto done;
 874
 875                 err = -EFAULT;
 876                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 877                         goto done;
 878
 879                 err = -EINVAL;
 880                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
 881                         goto done;
 882
 883                 if (!(p.i_flags&GRE_KEY))
 884                         p.i_key = 0;
 885                 if (!(p.o_flags&GRE_KEY))
 886                         p.o_key = 0;
 887
 888                 ip6gre_tnl_parm_from_user(&p1, &p);
 889                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
 890
 891                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 892                         if (t) {
 893                                 if (t->dev != dev) {
 894                                         err = -EEXIST;
 895                                         break;
 896                                 }
 897                         } else {
 898                                 t = netdev_priv(dev);
 899
 900                                 ip6gre_tunnel_unlink(ign, t);
 901                                 synchronize_net();
 902                                 ip6gre_tnl_change(t, &p1, 1);
 903                                 ip6gre_tunnel_link(ign, t);
 904                                 netdev_state_change(dev);
 905                         }
 906                 }
 907
 908                 if (t) {
 909                         err = 0;
 910
 911                         memset(&p, 0, sizeof(p));
 912                         ip6gre_tnl_parm_to_user(&p, &t->parms);
 913                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 914                                 err = -EFAULT;
 915                 } else
 916                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 917                 break;
 918
 919         case SIOCDELTUNNEL:
 920                 err = -EPERM;
 921                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 922                         goto done;
 923
 924                 if (dev == ign->fb_tunnel_dev) {
 925                         err = -EFAULT;
 926                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 927                                 goto done;
 928                         err = -ENOENT;
 929                         ip6gre_tnl_parm_from_user(&p1, &p);
 930                         t = ip6gre_tunnel_locate(net, &p1, 0);
 931                         if (!t)
 932                                 goto done;
 933                         err = -EPERM;
 934                         if (t == netdev_priv(ign->fb_tunnel_dev))
 935                                 goto done;
 936                         dev = t->dev;
 937                 }
 938                 unregister_netdevice(dev);
 939                 err = 0;
 940                 break;
 941
 942         default:
 943                 err = -EINVAL;
 944         }
 945
 946 done:
 947         return err;
 948 }
 949
 950 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
 951                          unsigned short type, const void *daddr,
 952                          const void *saddr, unsigned int len)
 953 {
 954         struct ip6_tnl *t = netdev_priv(dev);
 955         struct ipv6hdr *ipv6h;
 956         __be16 *p;
 957
 958         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
 959         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
 960                                                   t->fl.u.ip6.flowlabel,
 961                                                   true, &t->fl.u.ip6));
 962         ipv6h->hop_limit = t->parms.hop_limit;
 963         ipv6h->nexthdr = NEXTHDR_GRE;
 964         ipv6h->saddr = t->parms.laddr;
 965         ipv6h->daddr = t->parms.raddr;
 966
 967         p = (__be16 *)(ipv6h + 1);
 968         p[0] = t->parms.o_flags;
 969         p[1] = htons(type);
 970
 971         /*
 972          *      Set the source hardware address.
 973          */
 974
 975         if (saddr)
 976                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
 977         if (daddr)
 978                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
 979         if (!ipv6_addr_any(&ipv6h->daddr))
 980                 return t->hlen;
 981
 982         return -t->hlen;
 983 }
 984
 985 static const struct header_ops ip6gre_header_ops = {
 986         .create = ip6gre_header,
 987 };
 988
 989 static const struct net_device_ops ip6gre_netdev_ops = {
 990         .ndo_init               = ip6gre_tunnel_init,
 991         .ndo_uninit             = ip6gre_tunnel_uninit,
 992         .ndo_start_xmit         = ip6gre_tunnel_xmit,
 993         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
 994         .ndo_change_mtu         = ip6_tnl_change_mtu,
 995         .ndo_get_stats64        = ip_tunnel_get_stats64,
 996         .ndo_get_iflink         = ip6_tnl_get_iflink,
 997 };
 998
 999 static void ip6gre_dev_free(struct net_device *dev)
1000 {
1001         struct ip6_tnl *t = netdev_priv(dev);
1002
1003         dst_cache_destroy(&t->dst_cache);
1004         free_percpu(dev->tstats);
1005 }
1006
1007 static void ip6gre_tunnel_setup(struct net_device *dev)
1008 {
1009         dev->netdev_ops = &ip6gre_netdev_ops;
1010         dev->needs_free_netdev = true;
1011         dev->priv_destructor = ip6gre_dev_free;
1012
1013         dev->type = ARPHRD_IP6GRE;
1014
1015         dev->flags |= IFF_NOARP;
1016         dev->addr_len = sizeof(struct in6_addr);
1017         netif_keep_dst(dev);
1018         /* This perm addr will be used as interface identifier by IPv6 */
1019         dev->addr_assign_type = NET_ADDR_RANDOM;
1020         eth_random_addr(dev->perm_addr);
1021 }
1022
1023 static int ip6gre_tunnel_init_common(struct net_device *dev)
1024 {
1025         struct ip6_tnl *tunnel;
1026         int ret;
1027         int t_hlen;
1028
1029         tunnel = netdev_priv(dev);
1030
1031         tunnel->dev = dev;
1032         tunnel->net = dev_net(dev);
1033         strcpy(tunnel->parms.name, dev->name);
1034
1035         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1036         if (!dev->tstats)
1037                 return -ENOMEM;
1038
1039         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1040         if (ret) {
1041                 free_percpu(dev->tstats);
1042                 dev->tstats = NULL;
1043                 return ret;
1044         }
1045
1046         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1047         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1048         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1049
1050         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1051         dev->mtu = ETH_DATA_LEN - t_hlen;
1052         if (dev->type == ARPHRD_ETHER)
1053                 dev->mtu -= ETH_HLEN;
1054         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1055                 dev->mtu -= 8;
1056
1057         return 0;
1058 }
1059
1060 static int ip6gre_tunnel_init(struct net_device *dev)
1061 {
1062         struct ip6_tnl *tunnel;
1063         int ret;
1064
1065         ret = ip6gre_tunnel_init_common(dev);
1066         if (ret)
1067                 return ret;
1068
1069         tunnel = netdev_priv(dev);
1070
1071         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1072         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1073
1074         if (ipv6_addr_any(&tunnel->parms.raddr))
1075                 dev->header_ops = &ip6gre_header_ops;
1076
1077         return 0;
1078 }
1079
1080 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1081 {
1082         struct ip6_tnl *tunnel = netdev_priv(dev);
1083
1084         tunnel->dev = dev;
1085         tunnel->net = dev_net(dev);
1086         strcpy(tunnel->parms.name, dev->name);
1087
1088         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1089
1090         dev_hold(dev);
1091 }
1092
1093
1094 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1095         .handler     = gre_rcv,
1096         .err_handler = ip6gre_err,
1097         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1098 };
1099
1100 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1101 {
1102         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1103         struct net_device *dev, *aux;
1104         int prio;
1105
1106         for_each_netdev_safe(net, dev, aux)
1107                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1108                     dev->rtnl_link_ops == &ip6gre_tap_ops)
1109                         unregister_netdevice_queue(dev, head);
1110
1111         for (prio = 0; prio < 4; prio++) {
1112                 int h;
1113                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1114                         struct ip6_tnl *t;
1115
1116                         t = rtnl_dereference(ign->tunnels[prio][h]);
1117
1118                         while (t) {
1119                                 /* If dev is in the same netns, it has already
1120                                  * been added to the list by the previous loop.
1121                                  */
1122                                 if (!net_eq(dev_net(t->dev), net))
1123                                         unregister_netdevice_queue(t->dev,
1124                                                                    head);
1125                                 t = rtnl_dereference(t->next);
1126                         }
1127                 }
1128         }
1129 }
1130
1131 static int __net_init ip6gre_init_net(struct net *net)
1132 {
1133         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1134         int err;
1135
1136         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1137                                           NET_NAME_UNKNOWN,
1138                                           ip6gre_tunnel_setup);
1139         if (!ign->fb_tunnel_dev) {
1140                 err = -ENOMEM;
1141                 goto err_alloc_dev;
1142         }
1143         dev_net_set(ign->fb_tunnel_dev, net);
1144         /* FB netdevice is special: we have one, and only one per netns.
1145          * Allowing to move it to another netns is clearly unsafe.
1146          */
1147         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1148
1149
1150         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1151         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1152
1153         err = register_netdev(ign->fb_tunnel_dev);
1154         if (err)
1155                 goto err_reg_dev;
1156
1157         rcu_assign_pointer(ign->tunnels_wc[0],
1158                            netdev_priv(ign->fb_tunnel_dev));
1159         return 0;
1160
1161 err_reg_dev:
1162         free_netdev(ign->fb_tunnel_dev);
1163 err_alloc_dev:
1164         return err;
1165 }
1166
1167 static void __net_exit ip6gre_exit_net(struct net *net)
1168 {
1169         LIST_HEAD(list);
1170
1171         rtnl_lock();
1172         ip6gre_destroy_tunnels(net, &list);
1173         unregister_netdevice_many(&list);
1174         rtnl_unlock();
1175 }
1176
1177 static struct pernet_operations ip6gre_net_ops = {
1178         .init = ip6gre_init_net,
1179         .exit = ip6gre_exit_net,
1180         .id   = &ip6gre_net_id,
1181         .size = sizeof(struct ip6gre_net),
1182 };
1183
1184 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1185                                   struct netlink_ext_ack *extack)
1186 {
1187         __be16 flags;
1188
1189         if (!data)
1190                 return 0;
1191
1192         flags = 0;
1193         if (data[IFLA_GRE_IFLAGS])
1194                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1195         if (data[IFLA_GRE_OFLAGS])
1196                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1197         if (flags & (GRE_VERSION|GRE_ROUTING))
1198                 return -EINVAL;
1199
1200         return 0;
1201 }
1202
1203 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1204                                struct netlink_ext_ack *extack)
1205 {
1206         struct in6_addr daddr;
1207
1208         if (tb[IFLA_ADDRESS]) {
1209                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1210                         return -EINVAL;
1211                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1212                         return -EADDRNOTAVAIL;
1213         }
1214
1215         if (!data)
1216                 goto out;
1217
1218         if (data[IFLA_GRE_REMOTE]) {
1219                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1220                 if (ipv6_addr_any(&daddr))
1221                         return -EINVAL;
1222         }
1223
1224 out:
1225         return ip6gre_tunnel_validate(tb, data, extack);
1226 }
1227
1228
1229 static void ip6gre_netlink_parms(struct nlattr *data[],
1230                                 struct __ip6_tnl_parm *parms)
1231 {
1232         memset(parms, 0, sizeof(*parms));
1233
1234         if (!data)
1235                 return;
1236
1237         if (data[IFLA_GRE_LINK])
1238                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1239
1240         if (data[IFLA_GRE_IFLAGS])
1241                 parms->i_flags = gre_flags_to_tnl_flags(
1242                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1243
1244         if (data[IFLA_GRE_OFLAGS])
1245                 parms->o_flags = gre_flags_to_tnl_flags(
1246                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1247
1248         if (data[IFLA_GRE_IKEY])
1249                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1250
1251         if (data[IFLA_GRE_OKEY])
1252                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1253
1254         if (data[IFLA_GRE_LOCAL])
1255                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1256
1257         if (data[IFLA_GRE_REMOTE])
1258                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1259
1260         if (data[IFLA_GRE_TTL])
1261                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1262
1263         if (data[IFLA_GRE_ENCAP_LIMIT])
1264                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1265
1266         if (data[IFLA_GRE_FLOWINFO])
1267                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1268
1269         if (data[IFLA_GRE_FLAGS])
1270                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1271
1272         if (data[IFLA_GRE_FWMARK])
1273                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1274 }
1275
1276 static int ip6gre_tap_init(struct net_device *dev)
1277 {
1278         struct ip6_tnl *tunnel;
1279         int ret;
1280
1281         ret = ip6gre_tunnel_init_common(dev);
1282         if (ret)
1283                 return ret;
1284
1285         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1286
1287         tunnel = netdev_priv(dev);
1288
1289         ip6gre_tnl_link_config(tunnel, 1);
1290
1291         return 0;
1292 }
1293
1294 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1295         .ndo_init = ip6gre_tap_init,
1296         .ndo_uninit = ip6gre_tunnel_uninit,
1297         .ndo_start_xmit = ip6gre_tunnel_xmit,
1298         .ndo_set_mac_address = eth_mac_addr,
1299         .ndo_validate_addr = eth_validate_addr,
1300         .ndo_change_mtu = ip6_tnl_change_mtu,
1301         .ndo_get_stats64 = ip_tunnel_get_stats64,
1302         .ndo_get_iflink = ip6_tnl_get_iflink,
1303 };
1304
1305 #define GRE6_FEATURES (NETIF_F_SG |             \
1306                        NETIF_F_FRAGLIST |       \
1307                        NETIF_F_HIGHDMA |                \
1308                        NETIF_F_HW_CSUM)
1309
1310 static void ip6gre_tap_setup(struct net_device *dev)
1311 {
1312
1313         ether_setup(dev);
1314
1315         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1316         dev->needs_free_netdev = true;
1317         dev->priv_destructor = ip6gre_dev_free;
1318
1319         dev->features |= NETIF_F_NETNS_LOCAL;
1320         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1321         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1322         netif_keep_dst(dev);
1323 }
1324
1325 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1326                                        struct ip_tunnel_encap *ipencap)
1327 {
1328         bool ret = false;
1329
1330         memset(ipencap, 0, sizeof(*ipencap));
1331
1332         if (!data)
1333                 return ret;
1334
1335         if (data[IFLA_GRE_ENCAP_TYPE]) {
1336                 ret = true;
1337                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1338         }
1339
1340         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1341                 ret = true;
1342                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1343         }
1344
1345         if (data[IFLA_GRE_ENCAP_SPORT]) {
1346                 ret = true;
1347                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1348         }
1349
1350         if (data[IFLA_GRE_ENCAP_DPORT]) {
1351                 ret = true;
1352                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1353         }
1354
1355         return ret;
1356 }
1357
1358 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1359                           struct nlattr *tb[], struct nlattr *data[],
1360                           struct netlink_ext_ack *extack)
1361 {
1362         struct ip6_tnl *nt;
1363         struct net *net = dev_net(dev);
1364         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1365         struct ip_tunnel_encap ipencap;
1366         int err;
1367
1368         nt = netdev_priv(dev);
1369
1370         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1371                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1372
1373                 if (err < 0)
1374                         return err;
1375         }
1376
1377         ip6gre_netlink_parms(data, &nt->parms);
1378
1379         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1380                 return -EEXIST;
1381
1382         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1383                 eth_hw_addr_random(dev);
1384
1385         nt->dev = dev;
1386         nt->net = dev_net(dev);
1387         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1388
1389         dev->features           |= GRE6_FEATURES;
1390         dev->hw_features        |= GRE6_FEATURES;
1391
1392         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1393                 /* TCP offload with GRE SEQ is not supported, nor
1394                  * can we support 2 levels of outer headers requiring
1395                  * an update.
1396                  */
1397                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1398                     (nt->encap.type == TUNNEL_ENCAP_NONE)) {
1399                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1400                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1401                 }
1402
1403                 /* Can use a lockless transmit, unless we generate
1404                  * output sequences
1405                  */
1406                 dev->features |= NETIF_F_LLTX;
1407         }
1408
1409         err = register_netdevice(dev);
1410         if (err)
1411                 goto out;
1412
1413         dev_hold(dev);
1414         ip6gre_tunnel_link(ign, nt);
1415
1416 out:
1417         return err;
1418 }
1419
1420 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1421                              struct nlattr *data[],
1422                              struct netlink_ext_ack *extack)
1423 {
1424         struct ip6_tnl *t, *nt = netdev_priv(dev);
1425         struct net *net = nt->net;
1426         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1427         struct __ip6_tnl_parm p;
1428         struct ip_tunnel_encap ipencap;
1429
1430         if (dev == ign->fb_tunnel_dev)
1431                 return -EINVAL;
1432
1433         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1434                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1435
1436                 if (err < 0)
1437                         return err;
1438         }
1439
1440         ip6gre_netlink_parms(data, &p);
1441
1442         t = ip6gre_tunnel_locate(net, &p, 0);
1443
1444         if (t) {
1445                 if (t->dev != dev)
1446                         return -EEXIST;
1447         } else {
1448                 t = nt;
1449         }
1450
1451         ip6gre_tunnel_unlink(ign, t);
1452         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1453         ip6gre_tunnel_link(ign, t);
1454         return 0;
1455 }
1456
1457 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1458 {
1459         struct net *net = dev_net(dev);
1460         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1461
1462         if (dev != ign->fb_tunnel_dev)
1463                 unregister_netdevice_queue(dev, head);
1464 }
1465
1466 static size_t ip6gre_get_size(const struct net_device *dev)
1467 {
1468         return
1469                 /* IFLA_GRE_LINK */
1470                 nla_total_size(4) +
1471                 /* IFLA_GRE_IFLAGS */
1472                 nla_total_size(2) +
1473                 /* IFLA_GRE_OFLAGS */
1474                 nla_total_size(2) +
1475                 /* IFLA_GRE_IKEY */
1476                 nla_total_size(4) +
1477                 /* IFLA_GRE_OKEY */
1478                 nla_total_size(4) +
1479                 /* IFLA_GRE_LOCAL */
1480                 nla_total_size(sizeof(struct in6_addr)) +
1481                 /* IFLA_GRE_REMOTE */
1482                 nla_total_size(sizeof(struct in6_addr)) +
1483                 /* IFLA_GRE_TTL */
1484                 nla_total_size(1) +
1485                 /* IFLA_GRE_ENCAP_LIMIT */
1486                 nla_total_size(1) +
1487                 /* IFLA_GRE_FLOWINFO */
1488                 nla_total_size(4) +
1489                 /* IFLA_GRE_FLAGS */
1490                 nla_total_size(4) +
1491                 /* IFLA_GRE_ENCAP_TYPE */
1492                 nla_total_size(2) +
1493                 /* IFLA_GRE_ENCAP_FLAGS */
1494                 nla_total_size(2) +
1495                 /* IFLA_GRE_ENCAP_SPORT */
1496                 nla_total_size(2) +
1497                 /* IFLA_GRE_ENCAP_DPORT */
1498                 nla_total_size(2) +
1499                 /* IFLA_GRE_FWMARK */
1500                 nla_total_size(4) +
1501                 0;
1502 }
1503
1504 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1505 {
1506         struct ip6_tnl *t = netdev_priv(dev);
1507         struct __ip6_tnl_parm *p = &t->parms;
1508
1509         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1510             nla_put_be16(skb, IFLA_GRE_IFLAGS,
1511                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1512             nla_put_be16(skb, IFLA_GRE_OFLAGS,
1513                          gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1514             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1515             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1516             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1517             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1518             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1519             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1520             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1521             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1522             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1523                 goto nla_put_failure;
1524
1525         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1526                         t->encap.type) ||
1527             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1528                          t->encap.sport) ||
1529             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1530                          t->encap.dport) ||
1531             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1532                         t->encap.flags))
1533                 goto nla_put_failure;
1534
1535         return 0;
1536
1537 nla_put_failure:
1538         return -EMSGSIZE;
1539 }
1540
1541 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1542         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1543         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1544         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1545         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1546         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1547         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1548         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1549         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1550         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1551         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1552         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1553         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
1554         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
1555         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
1556         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
1557         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
1558 };
1559
1560 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1561         .kind           = "ip6gre",
1562         .maxtype        = IFLA_GRE_MAX,
1563         .policy         = ip6gre_policy,
1564         .priv_size      = sizeof(struct ip6_tnl),
1565         .setup          = ip6gre_tunnel_setup,
1566         .validate       = ip6gre_tunnel_validate,
1567         .newlink        = ip6gre_newlink,
1568         .changelink     = ip6gre_changelink,
1569         .dellink        = ip6gre_dellink,
1570         .get_size       = ip6gre_get_size,
1571         .fill_info      = ip6gre_fill_info,
1572         .get_link_net   = ip6_tnl_get_link_net,
1573 };
1574
1575 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1576         .kind           = "ip6gretap",
1577         .maxtype        = IFLA_GRE_MAX,
1578         .policy         = ip6gre_policy,
1579         .priv_size      = sizeof(struct ip6_tnl),
1580         .setup          = ip6gre_tap_setup,
1581         .validate       = ip6gre_tap_validate,
1582         .newlink        = ip6gre_newlink,
1583         .changelink     = ip6gre_changelink,
1584         .get_size       = ip6gre_get_size,
1585         .fill_info      = ip6gre_fill_info,
1586         .get_link_net   = ip6_tnl_get_link_net,
1587 };
1588
1589 /*
1590  *      And now the modules code and kernel interface.
1591  */
1592
1593 static int __init ip6gre_init(void)
1594 {
1595         int err;
1596
1597         pr_info("GRE over IPv6 tunneling driver\n");
1598
1599         err = register_pernet_device(&ip6gre_net_ops);
1600         if (err < 0)
1601                 return err;
1602
1603         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1604         if (err < 0) {
1605                 pr_info("%s: can't add protocol\n", __func__);
1606                 goto add_proto_failed;
1607         }
1608
1609         err = rtnl_link_register(&ip6gre_link_ops);
1610         if (err < 0)
1611                 goto rtnl_link_failed;
1612
1613         err = rtnl_link_register(&ip6gre_tap_ops);
1614         if (err < 0)
1615                 goto tap_ops_failed;
1616
1617 out:
1618         return err;
1619
1620 tap_ops_failed:
1621         rtnl_link_unregister(&ip6gre_link_ops);
1622 rtnl_link_failed:
1623         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1624 add_proto_failed:
1625         unregister_pernet_device(&ip6gre_net_ops);
1626         goto out;
1627 }
1628
1629 static void __exit ip6gre_fini(void)
1630 {
1631         rtnl_link_unregister(&ip6gre_tap_ops);
1632         rtnl_link_unregister(&ip6gre_link_ops);
1633         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1634         unregister_pernet_device(&ip6gre_net_ops);
1635 }
1636
1637 module_init(ip6gre_init);
1638 module_exit(ip6gre_fini);
1639 MODULE_LICENSE("GPL");
1640 MODULE_AUTHOR("D. Kozlov ([email protected])");
1641 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1642 MODULE_ALIAS_RTNL_LINK("ip6gre");
1643 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1644 MODULE_ALIAS_NETDEV("ip6gre0");