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         struct net *net = dev_net(skb->dev);
 373         const struct gre_base_hdr *greh;
 374         const struct ipv6hdr *ipv6h;
 375         int grehlen = sizeof(*greh);
 376         struct ip6_tnl *t;
 377         int key_off = 0;
 378         __be16 flags;
 379         __be32 key;
 380
 381         if (!pskb_may_pull(skb, offset + grehlen))
 382                 return;
 383         greh = (const struct gre_base_hdr *)(skb->data + offset);
 384         flags = greh->flags;
 385         if (flags & (GRE_VERSION | GRE_ROUTING))
 386                 return;
 387         if (flags & GRE_CSUM)
 388                 grehlen += 4;
 389         if (flags & GRE_KEY) {
 390                 key_off = grehlen + offset;
 391                 grehlen += 4;
 392         }
 393
 394         if (!pskb_may_pull(skb, offset + grehlen))
 395                 return;
 396         ipv6h = (const struct ipv6hdr *)skb->data;
 397         greh = (const struct gre_base_hdr *)(skb->data + offset);
 398         key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
 399
 400         t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 401                                  key, greh->protocol);
 402         if (!t)
 403                 return;
 404
 405         switch (type) {
 406                 struct ipv6_tlv_tnl_enc_lim *tel;
 407                 __u32 teli;
 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                 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
 439                 return;
 440         case NDISC_REDIRECT:
 441                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 442                              sock_net_uid(net, NULL));
 443                 return;
 444         }
 445
 446         if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 447                 t->err_count++;
 448         else
 449                 t->err_count = 1;
 450         t->err_time = jiffies;
 451 }
 452
 453 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 454 {
 455         const struct ipv6hdr *ipv6h;
 456         struct ip6_tnl *tunnel;
 457
 458         ipv6h = ipv6_hdr(skb);
 459         tunnel = ip6gre_tunnel_lookup(skb->dev,
 460                                       &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 461                                       tpi->proto);
 462         if (tunnel) {
 463                 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 464
 465                 return PACKET_RCVD;
 466         }
 467
 468         return PACKET_REJECT;
 469 }
 470
 471 static int gre_rcv(struct sk_buff *skb)
 472 {
 473         struct tnl_ptk_info tpi;
 474         bool csum_err = false;
 475         int hdr_len;
 476
 477         hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
 478         if (hdr_len < 0)
 479                 goto drop;
 480
 481         if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
 482                 goto drop;
 483
 484         if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
 485                 return 0;
 486
 487         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 488 drop:
 489         kfree_skb(skb);
 490         return 0;
 491 }
 492
 493 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
 494 {
 495         return iptunnel_handle_offloads(skb,
 496                                         csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 497 }
 498
 499 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
 500                                struct net_device *dev, __u8 dsfield,
 501                                struct flowi6 *fl6, int encap_limit,
 502                                __u32 *pmtu, __be16 proto)
 503 {
 504         struct ip6_tnl *tunnel = netdev_priv(dev);
 505         __be16 protocol;
 506
 507         if (dev->type == ARPHRD_ETHER)
 508                 IPCB(skb)->flags = 0;
 509
 510         if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
 511                 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
 512         else
 513                 fl6->daddr = tunnel->parms.raddr;
 514
 515         if (tunnel->parms.o_flags & TUNNEL_SEQ)
 516                 tunnel->o_seqno++;
 517
 518         /* Push GRE header. */
 519         protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
 520         gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
 521                          protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
 522
 523         return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
 524                             NEXTHDR_GRE);
 525 }
 526
 527 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 528 {
 529         struct ip6_tnl *t = netdev_priv(dev);
 530         const struct iphdr  *iph = ip_hdr(skb);
 531         int encap_limit = -1;
 532         struct flowi6 fl6;
 533         __u8 dsfield;
 534         __u32 mtu;
 535         int err;
 536
 537         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 538
 539         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 540                 encap_limit = t->parms.encap_limit;
 541
 542         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 543
 544         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 545                 dsfield = ipv4_get_dsfield(iph);
 546         else
 547                 dsfield = ip6_tclass(t->parms.flowinfo);
 548         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 549                 fl6.flowi6_mark = skb->mark;
 550         else
 551                 fl6.flowi6_mark = t->parms.fwmark;
 552
 553         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 554
 555         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 556         if (err)
 557                 return -1;
 558
 559         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 560                           skb->protocol);
 561         if (err != 0) {
 562                 /* XXX: send ICMP error even if DF is not set. */
 563                 if (err == -EMSGSIZE)
 564                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 565                                   htonl(mtu));
 566                 return -1;
 567         }
 568
 569         return 0;
 570 }
 571
 572 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 573 {
 574         struct ip6_tnl *t = netdev_priv(dev);
 575         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 576         int encap_limit = -1;
 577         __u16 offset;
 578         struct flowi6 fl6;
 579         __u8 dsfield;
 580         __u32 mtu;
 581         int err;
 582
 583         if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 584                 return -1;
 585
 586         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 587         /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 588         ipv6h = ipv6_hdr(skb);
 589
 590         if (offset > 0) {
 591                 struct ipv6_tlv_tnl_enc_lim *tel;
 592                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 593                 if (tel->encap_limit == 0) {
 594                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 595                                     ICMPV6_HDR_FIELD, offset + 2);
 596                         return -1;
 597                 }
 598                 encap_limit = tel->encap_limit - 1;
 599         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 600                 encap_limit = t->parms.encap_limit;
 601
 602         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 603
 604         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 605                 dsfield = ipv6_get_dsfield(ipv6h);
 606         else
 607                 dsfield = ip6_tclass(t->parms.flowinfo);
 608
 609         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 610                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
 611         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 612                 fl6.flowi6_mark = skb->mark;
 613         else
 614                 fl6.flowi6_mark = t->parms.fwmark;
 615
 616         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 617
 618         if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
 619                 return -1;
 620
 621         err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
 622                           &mtu, skb->protocol);
 623         if (err != 0) {
 624                 if (err == -EMSGSIZE)
 625                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 626                 return -1;
 627         }
 628
 629         return 0;
 630 }
 631
 632 /**
 633  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 634  *   @t: the outgoing tunnel device
 635  *   @hdr: IPv6 header from the incoming packet
 636  *
 637  * Description:
 638  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 639  *   doesn't match source of incoming packet.
 640  *
 641  * Return:
 642  *   1 if conflict,
 643  *   0 else
 644  **/
 645
 646 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 647         const struct ipv6hdr *hdr)
 648 {
 649         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 650 }
 651
 652 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 653 {
 654         struct ip6_tnl *t = netdev_priv(dev);
 655         int encap_limit = -1;
 656         struct flowi6 fl6;
 657         __u32 mtu;
 658         int err;
 659
 660         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 661                 encap_limit = t->parms.encap_limit;
 662
 663         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 664
 665         err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 666         if (err)
 667                 return err;
 668
 669         err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
 670
 671         return err;
 672 }
 673
 674 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 675         struct net_device *dev)
 676 {
 677         struct ip6_tnl *t = netdev_priv(dev);
 678         struct net_device_stats *stats = &t->dev->stats;
 679         int ret;
 680
 681         if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 682                 goto tx_err;
 683
 684         switch (skb->protocol) {
 685         case htons(ETH_P_IP):
 686                 ret = ip6gre_xmit_ipv4(skb, dev);
 687                 break;
 688         case htons(ETH_P_IPV6):
 689                 ret = ip6gre_xmit_ipv6(skb, dev);
 690                 break;
 691         default:
 692                 ret = ip6gre_xmit_other(skb, dev);
 693                 break;
 694         }
 695
 696         if (ret < 0)
 697                 goto tx_err;
 698
 699         return NETDEV_TX_OK;
 700
 701 tx_err:
 702         stats->tx_errors++;
 703         stats->tx_dropped++;
 704         kfree_skb(skb);
 705         return NETDEV_TX_OK;
 706 }
 707
 708 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
 709 {
 710         struct net_device *dev = t->dev;
 711         struct __ip6_tnl_parm *p = &t->parms;
 712         struct flowi6 *fl6 = &t->fl.u.ip6;
 713         int t_hlen;
 714
 715         if (dev->type != ARPHRD_ETHER) {
 716                 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 717                 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 718         }
 719
 720         /* Set up flowi template */
 721         fl6->saddr = p->laddr;
 722         fl6->daddr = p->raddr;
 723         fl6->flowi6_oif = p->link;
 724         fl6->flowlabel = 0;
 725         fl6->flowi6_proto = IPPROTO_GRE;
 726
 727         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 728                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 729         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 730                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 731
 732         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
 733         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
 734
 735         if (p->flags&IP6_TNL_F_CAP_XMIT &&
 736                         p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
 737                 dev->flags |= IFF_POINTOPOINT;
 738         else
 739                 dev->flags &= ~IFF_POINTOPOINT;
 740
 741         t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
 742
 743         t->hlen = t->encap_hlen + t->tun_hlen;
 744
 745         t_hlen = t->hlen + sizeof(struct ipv6hdr);
 746
 747         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 748                 int strict = (ipv6_addr_type(&p->raddr) &
 749                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 750
 751                 struct rt6_info *rt = rt6_lookup(t->net,
 752                                                  &p->raddr, &p->laddr,
 753                                                  p->link, strict);
 754
 755                 if (!rt)
 756                         return;
 757
 758                 if (rt->dst.dev) {
 759                         dev->hard_header_len = rt->dst.dev->hard_header_len +
 760                                                t_hlen;
 761
 762                         if (set_mtu) {
 763                                 dev->mtu = rt->dst.dev->mtu - t_hlen;
 764                                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 765                                         dev->mtu -= 8;
 766                                 if (dev->type == ARPHRD_ETHER)
 767                                         dev->mtu -= ETH_HLEN;
 768
 769                                 if (dev->mtu < IPV6_MIN_MTU)
 770                                         dev->mtu = IPV6_MIN_MTU;
 771                         }
 772                 }
 773                 ip6_rt_put(rt);
 774         }
 775 }
 776
 777 static int ip6gre_tnl_change(struct ip6_tnl *t,
 778         const struct __ip6_tnl_parm *p, int set_mtu)
 779 {
 780         t->parms.laddr = p->laddr;
 781         t->parms.raddr = p->raddr;
 782         t->parms.flags = p->flags;
 783         t->parms.hop_limit = p->hop_limit;
 784         t->parms.encap_limit = p->encap_limit;
 785         t->parms.flowinfo = p->flowinfo;
 786         t->parms.link = p->link;
 787         t->parms.proto = p->proto;
 788         t->parms.i_key = p->i_key;
 789         t->parms.o_key = p->o_key;
 790         t->parms.i_flags = p->i_flags;
 791         t->parms.o_flags = p->o_flags;
 792         t->parms.fwmark = p->fwmark;
 793         dst_cache_reset(&t->dst_cache);
 794         ip6gre_tnl_link_config(t, set_mtu);
 795         return 0;
 796 }
 797
 798 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
 799         const struct ip6_tnl_parm2 *u)
 800 {
 801         p->laddr = u->laddr;
 802         p->raddr = u->raddr;
 803         p->flags = u->flags;
 804         p->hop_limit = u->hop_limit;
 805         p->encap_limit = u->encap_limit;
 806         p->flowinfo = u->flowinfo;
 807         p->link = u->link;
 808         p->i_key = u->i_key;
 809         p->o_key = u->o_key;
 810         p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
 811         p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
 812         memcpy(p->name, u->name, sizeof(u->name));
 813 }
 814
 815 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
 816         const struct __ip6_tnl_parm *p)
 817 {
 818         u->proto = IPPROTO_GRE;
 819         u->laddr = p->laddr;
 820         u->raddr = p->raddr;
 821         u->flags = p->flags;
 822         u->hop_limit = p->hop_limit;
 823         u->encap_limit = p->encap_limit;
 824         u->flowinfo = p->flowinfo;
 825         u->link = p->link;
 826         u->i_key = p->i_key;
 827         u->o_key = p->o_key;
 828         u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
 829         u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
 830         memcpy(u->name, p->name, sizeof(u->name));
 831 }
 832
 833 static int ip6gre_tunnel_ioctl(struct net_device *dev,
 834         struct ifreq *ifr, int cmd)
 835 {
 836         int err = 0;
 837         struct ip6_tnl_parm2 p;
 838         struct __ip6_tnl_parm p1;
 839         struct ip6_tnl *t = netdev_priv(dev);
 840         struct net *net = t->net;
 841         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 842
 843         memset(&p1, 0, sizeof(p1));
 844
 845         switch (cmd) {
 846         case SIOCGETTUNNEL:
 847                 if (dev == ign->fb_tunnel_dev) {
 848                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 849                                 err = -EFAULT;
 850                                 break;
 851                         }
 852                         ip6gre_tnl_parm_from_user(&p1, &p);
 853                         t = ip6gre_tunnel_locate(net, &p1, 0);
 854                         if (!t)
 855                                 t = netdev_priv(dev);
 856                 }
 857                 memset(&p, 0, sizeof(p));
 858                 ip6gre_tnl_parm_to_user(&p, &t->parms);
 859                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 860                         err = -EFAULT;
 861                 break;
 862
 863         case SIOCADDTUNNEL:
 864         case SIOCCHGTUNNEL:
 865                 err = -EPERM;
 866                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 867                         goto done;
 868
 869                 err = -EFAULT;
 870                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 871                         goto done;
 872
 873                 err = -EINVAL;
 874                 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
 875                         goto done;
 876
 877                 if (!(p.i_flags&GRE_KEY))
 878                         p.i_key = 0;
 879                 if (!(p.o_flags&GRE_KEY))
 880                         p.o_key = 0;
 881
 882                 ip6gre_tnl_parm_from_user(&p1, &p);
 883                 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
 884
 885                 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 886                         if (t) {
 887                                 if (t->dev != dev) {
 888                                         err = -EEXIST;
 889                                         break;
 890                                 }
 891                         } else {
 892                                 t = netdev_priv(dev);
 893
 894                                 ip6gre_tunnel_unlink(ign, t);
 895                                 synchronize_net();
 896                                 ip6gre_tnl_change(t, &p1, 1);
 897                                 ip6gre_tunnel_link(ign, t);
 898                                 netdev_state_change(dev);
 899                         }
 900                 }
 901
 902                 if (t) {
 903                         err = 0;
 904
 905                         memset(&p, 0, sizeof(p));
 906                         ip6gre_tnl_parm_to_user(&p, &t->parms);
 907                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 908                                 err = -EFAULT;
 909                 } else
 910                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 911                 break;
 912
 913         case SIOCDELTUNNEL:
 914                 err = -EPERM;
 915                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 916                         goto done;
 917
 918                 if (dev == ign->fb_tunnel_dev) {
 919                         err = -EFAULT;
 920                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 921                                 goto done;
 922                         err = -ENOENT;
 923                         ip6gre_tnl_parm_from_user(&p1, &p);
 924                         t = ip6gre_tunnel_locate(net, &p1, 0);
 925                         if (!t)
 926                                 goto done;
 927                         err = -EPERM;
 928                         if (t == netdev_priv(ign->fb_tunnel_dev))
 929                                 goto done;
 930                         dev = t->dev;
 931                 }
 932                 unregister_netdevice(dev);
 933                 err = 0;
 934                 break;
 935
 936         default:
 937                 err = -EINVAL;
 938         }
 939
 940 done:
 941         return err;
 942 }
 943
 944 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
 945                          unsigned short type, const void *daddr,
 946                          const void *saddr, unsigned int len)
 947 {
 948         struct ip6_tnl *t = netdev_priv(dev);
 949         struct ipv6hdr *ipv6h;
 950         __be16 *p;
 951
 952         ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
 953         ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
 954                                                   t->fl.u.ip6.flowlabel,
 955                                                   true, &t->fl.u.ip6));
 956         ipv6h->hop_limit = t->parms.hop_limit;
 957         ipv6h->nexthdr = NEXTHDR_GRE;
 958         ipv6h->saddr = t->parms.laddr;
 959         ipv6h->daddr = t->parms.raddr;
 960
 961         p = (__be16 *)(ipv6h + 1);
 962         p[0] = t->parms.o_flags;
 963         p[1] = htons(type);
 964
 965         /*
 966          *      Set the source hardware address.
 967          */
 968
 969         if (saddr)
 970                 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
 971         if (daddr)
 972                 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
 973         if (!ipv6_addr_any(&ipv6h->daddr))
 974                 return t->hlen;
 975
 976         return -t->hlen;
 977 }
 978
 979 static const struct header_ops ip6gre_header_ops = {
 980         .create = ip6gre_header,
 981 };
 982
 983 static const struct net_device_ops ip6gre_netdev_ops = {
 984         .ndo_init               = ip6gre_tunnel_init,
 985         .ndo_uninit             = ip6gre_tunnel_uninit,
 986         .ndo_start_xmit         = ip6gre_tunnel_xmit,
 987         .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
 988         .ndo_change_mtu         = ip6_tnl_change_mtu,
 989         .ndo_get_stats64        = ip_tunnel_get_stats64,
 990         .ndo_get_iflink         = ip6_tnl_get_iflink,
 991 };
 992
 993 static void ip6gre_dev_free(struct net_device *dev)
 994 {
 995         struct ip6_tnl *t = netdev_priv(dev);
 996
 997         dst_cache_destroy(&t->dst_cache);
 998         free_percpu(dev->tstats);
 999 }
1000
1001 static void ip6gre_tunnel_setup(struct net_device *dev)
1002 {
1003         dev->netdev_ops = &ip6gre_netdev_ops;
1004         dev->needs_free_netdev = true;
1005         dev->priv_destructor = ip6gre_dev_free;
1006
1007         dev->type = ARPHRD_IP6GRE;
1008
1009         dev->flags |= IFF_NOARP;
1010         dev->addr_len = sizeof(struct in6_addr);
1011         netif_keep_dst(dev);
1012         /* This perm addr will be used as interface identifier by IPv6 */
1013         dev->addr_assign_type = NET_ADDR_RANDOM;
1014         eth_random_addr(dev->perm_addr);
1015 }
1016
1017 #define GRE6_FEATURES (NETIF_F_SG |             \
1018                        NETIF_F_FRAGLIST |       \
1019                        NETIF_F_HIGHDMA |        \
1020                        NETIF_F_HW_CSUM)
1021
1022 static void ip6gre_tnl_init_features(struct net_device *dev)
1023 {
1024         struct ip6_tnl *nt = netdev_priv(dev);
1025
1026         dev->features           |= GRE6_FEATURES;
1027         dev->hw_features        |= GRE6_FEATURES;
1028
1029         if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1030                 /* TCP offload with GRE SEQ is not supported, nor
1031                  * can we support 2 levels of outer headers requiring
1032                  * an update.
1033                  */
1034                 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1035                     nt->encap.type == TUNNEL_ENCAP_NONE) {
1036                         dev->features    |= NETIF_F_GSO_SOFTWARE;
1037                         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1038                 }
1039
1040                 /* Can use a lockless transmit, unless we generate
1041                  * output sequences
1042                  */
1043                 dev->features |= NETIF_F_LLTX;
1044         }
1045 }
1046
1047 static int ip6gre_tunnel_init_common(struct net_device *dev)
1048 {
1049         struct ip6_tnl *tunnel;
1050         int ret;
1051         int t_hlen;
1052
1053         tunnel = netdev_priv(dev);
1054
1055         tunnel->dev = dev;
1056         tunnel->net = dev_net(dev);
1057         strcpy(tunnel->parms.name, dev->name);
1058
1059         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1060         if (!dev->tstats)
1061                 return -ENOMEM;
1062
1063         ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1064         if (ret) {
1065                 free_percpu(dev->tstats);
1066                 dev->tstats = NULL;
1067                 return ret;
1068         }
1069
1070         tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1071         tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1072         t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1073
1074         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1075         dev->mtu = ETH_DATA_LEN - t_hlen;
1076         if (dev->type == ARPHRD_ETHER)
1077                 dev->mtu -= ETH_HLEN;
1078         if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1079                 dev->mtu -= 8;
1080
1081         ip6gre_tnl_init_features(dev);
1082
1083         return 0;
1084 }
1085
1086 static int ip6gre_tunnel_init(struct net_device *dev)
1087 {
1088         struct ip6_tnl *tunnel;
1089         int ret;
1090
1091         ret = ip6gre_tunnel_init_common(dev);
1092         if (ret)
1093                 return ret;
1094
1095         tunnel = netdev_priv(dev);
1096
1097         memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1098         memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1099
1100         if (ipv6_addr_any(&tunnel->parms.raddr))
1101                 dev->header_ops = &ip6gre_header_ops;
1102
1103         return 0;
1104 }
1105
1106 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1107 {
1108         struct ip6_tnl *tunnel = netdev_priv(dev);
1109
1110         tunnel->dev = dev;
1111         tunnel->net = dev_net(dev);
1112         strcpy(tunnel->parms.name, dev->name);
1113
1114         tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1115
1116         dev_hold(dev);
1117 }
1118
1119
1120 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1121         .handler     = gre_rcv,
1122         .err_handler = ip6gre_err,
1123         .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1124 };
1125
1126 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1127 {
1128         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1129         struct net_device *dev, *aux;
1130         int prio;
1131
1132         for_each_netdev_safe(net, dev, aux)
1133                 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1134                     dev->rtnl_link_ops == &ip6gre_tap_ops)
1135                         unregister_netdevice_queue(dev, head);
1136
1137         for (prio = 0; prio < 4; prio++) {
1138                 int h;
1139                 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1140                         struct ip6_tnl *t;
1141
1142                         t = rtnl_dereference(ign->tunnels[prio][h]);
1143
1144                         while (t) {
1145                                 /* If dev is in the same netns, it has already
1146                                  * been added to the list by the previous loop.
1147                                  */
1148                                 if (!net_eq(dev_net(t->dev), net))
1149                                         unregister_netdevice_queue(t->dev,
1150                                                                    head);
1151                                 t = rtnl_dereference(t->next);
1152                         }
1153                 }
1154         }
1155 }
1156
1157 static int __net_init ip6gre_init_net(struct net *net)
1158 {
1159         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1160         int err;
1161
1162         ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1163                                           NET_NAME_UNKNOWN,
1164                                           ip6gre_tunnel_setup);
1165         if (!ign->fb_tunnel_dev) {
1166                 err = -ENOMEM;
1167                 goto err_alloc_dev;
1168         }
1169         dev_net_set(ign->fb_tunnel_dev, net);
1170         /* FB netdevice is special: we have one, and only one per netns.
1171          * Allowing to move it to another netns is clearly unsafe.
1172          */
1173         ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1174
1175
1176         ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1177         ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1178
1179         err = register_netdev(ign->fb_tunnel_dev);
1180         if (err)
1181                 goto err_reg_dev;
1182
1183         rcu_assign_pointer(ign->tunnels_wc[0],
1184                            netdev_priv(ign->fb_tunnel_dev));
1185         return 0;
1186
1187 err_reg_dev:
1188         free_netdev(ign->fb_tunnel_dev);
1189 err_alloc_dev:
1190         return err;
1191 }
1192
1193 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1194 {
1195         struct net *net;
1196         LIST_HEAD(list);
1197
1198         rtnl_lock();
1199         list_for_each_entry(net, net_list, exit_list)
1200                 ip6gre_destroy_tunnels(net, &list);
1201         unregister_netdevice_many(&list);
1202         rtnl_unlock();
1203 }
1204
1205 static struct pernet_operations ip6gre_net_ops = {
1206         .init = ip6gre_init_net,
1207         .exit_batch = ip6gre_exit_batch_net,
1208         .id   = &ip6gre_net_id,
1209         .size = sizeof(struct ip6gre_net),
1210 };
1211
1212 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1213                                   struct netlink_ext_ack *extack)
1214 {
1215         __be16 flags;
1216
1217         if (!data)
1218                 return 0;
1219
1220         flags = 0;
1221         if (data[IFLA_GRE_IFLAGS])
1222                 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1223         if (data[IFLA_GRE_OFLAGS])
1224                 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1225         if (flags & (GRE_VERSION|GRE_ROUTING))
1226                 return -EINVAL;
1227
1228         return 0;
1229 }
1230
1231 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1232                                struct netlink_ext_ack *extack)
1233 {
1234         struct in6_addr daddr;
1235
1236         if (tb[IFLA_ADDRESS]) {
1237                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1238                         return -EINVAL;
1239                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1240                         return -EADDRNOTAVAIL;
1241         }
1242
1243         if (!data)
1244                 goto out;
1245
1246         if (data[IFLA_GRE_REMOTE]) {
1247                 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1248                 if (ipv6_addr_any(&daddr))
1249                         return -EINVAL;
1250         }
1251
1252 out:
1253         return ip6gre_tunnel_validate(tb, data, extack);
1254 }
1255
1256
1257 static void ip6gre_netlink_parms(struct nlattr *data[],
1258                                 struct __ip6_tnl_parm *parms)
1259 {
1260         memset(parms, 0, sizeof(*parms));
1261
1262         if (!data)
1263                 return;
1264
1265         if (data[IFLA_GRE_LINK])
1266                 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1267
1268         if (data[IFLA_GRE_IFLAGS])
1269                 parms->i_flags = gre_flags_to_tnl_flags(
1270                                 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1271
1272         if (data[IFLA_GRE_OFLAGS])
1273                 parms->o_flags = gre_flags_to_tnl_flags(
1274                                 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1275
1276         if (data[IFLA_GRE_IKEY])
1277                 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1278
1279         if (data[IFLA_GRE_OKEY])
1280                 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1281
1282         if (data[IFLA_GRE_LOCAL])
1283                 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1284
1285         if (data[IFLA_GRE_REMOTE])
1286                 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1287
1288         if (data[IFLA_GRE_TTL])
1289                 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1290
1291         if (data[IFLA_GRE_ENCAP_LIMIT])
1292                 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1293
1294         if (data[IFLA_GRE_FLOWINFO])
1295                 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1296
1297         if (data[IFLA_GRE_FLAGS])
1298                 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1299
1300         if (data[IFLA_GRE_FWMARK])
1301                 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1302 }
1303
1304 static int ip6gre_tap_init(struct net_device *dev)
1305 {
1306         struct ip6_tnl *tunnel;
1307         int ret;
1308
1309         ret = ip6gre_tunnel_init_common(dev);
1310         if (ret)
1311                 return ret;
1312
1313         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1314
1315         tunnel = netdev_priv(dev);
1316
1317         ip6gre_tnl_link_config(tunnel, 1);
1318
1319         return 0;
1320 }
1321
1322 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1323         .ndo_init = ip6gre_tap_init,
1324         .ndo_uninit = ip6gre_tunnel_uninit,
1325         .ndo_start_xmit = ip6gre_tunnel_xmit,
1326         .ndo_set_mac_address = eth_mac_addr,
1327         .ndo_validate_addr = eth_validate_addr,
1328         .ndo_change_mtu = ip6_tnl_change_mtu,
1329         .ndo_get_stats64 = ip_tunnel_get_stats64,
1330         .ndo_get_iflink = ip6_tnl_get_iflink,
1331 };
1332
1333 static void ip6gre_tap_setup(struct net_device *dev)
1334 {
1335
1336         ether_setup(dev);
1337
1338         dev->max_mtu = 0;
1339         dev->netdev_ops = &ip6gre_tap_netdev_ops;
1340         dev->needs_free_netdev = true;
1341         dev->priv_destructor = ip6gre_dev_free;
1342
1343         dev->features |= NETIF_F_NETNS_LOCAL;
1344         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1345         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1346         netif_keep_dst(dev);
1347 }
1348
1349 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1350                                        struct ip_tunnel_encap *ipencap)
1351 {
1352         bool ret = false;
1353
1354         memset(ipencap, 0, sizeof(*ipencap));
1355
1356         if (!data)
1357                 return ret;
1358
1359         if (data[IFLA_GRE_ENCAP_TYPE]) {
1360                 ret = true;
1361                 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1362         }
1363
1364         if (data[IFLA_GRE_ENCAP_FLAGS]) {
1365                 ret = true;
1366                 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1367         }
1368
1369         if (data[IFLA_GRE_ENCAP_SPORT]) {
1370                 ret = true;
1371                 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1372         }
1373
1374         if (data[IFLA_GRE_ENCAP_DPORT]) {
1375                 ret = true;
1376                 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1377         }
1378
1379         return ret;
1380 }
1381
1382 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1383                           struct nlattr *tb[], struct nlattr *data[],
1384                           struct netlink_ext_ack *extack)
1385 {
1386         struct ip6_tnl *nt;
1387         struct net *net = dev_net(dev);
1388         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1389         struct ip_tunnel_encap ipencap;
1390         int err;
1391
1392         nt = netdev_priv(dev);
1393
1394         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1395                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1396
1397                 if (err < 0)
1398                         return err;
1399         }
1400
1401         ip6gre_netlink_parms(data, &nt->parms);
1402
1403         if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1404                 return -EEXIST;
1405
1406         if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1407                 eth_hw_addr_random(dev);
1408
1409         nt->dev = dev;
1410         nt->net = dev_net(dev);
1411         ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1412
1413         err = register_netdevice(dev);
1414         if (err)
1415                 goto out;
1416
1417         dev_hold(dev);
1418         ip6gre_tunnel_link(ign, nt);
1419
1420 out:
1421         return err;
1422 }
1423
1424 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1425                              struct nlattr *data[],
1426                              struct netlink_ext_ack *extack)
1427 {
1428         struct ip6_tnl *t, *nt = netdev_priv(dev);
1429         struct net *net = nt->net;
1430         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1431         struct __ip6_tnl_parm p;
1432         struct ip_tunnel_encap ipencap;
1433
1434         if (dev == ign->fb_tunnel_dev)
1435                 return -EINVAL;
1436
1437         if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1438                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1439
1440                 if (err < 0)
1441                         return err;
1442         }
1443
1444         ip6gre_netlink_parms(data, &p);
1445
1446         t = ip6gre_tunnel_locate(net, &p, 0);
1447
1448         if (t) {
1449                 if (t->dev != dev)
1450                         return -EEXIST;
1451         } else {
1452                 t = nt;
1453         }
1454
1455         ip6gre_tunnel_unlink(ign, t);
1456         ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1457         ip6gre_tunnel_link(ign, t);
1458         return 0;
1459 }
1460
1461 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1462 {
1463         struct net *net = dev_net(dev);
1464         struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1465
1466         if (dev != ign->fb_tunnel_dev)
1467                 unregister_netdevice_queue(dev, head);
1468 }
1469
1470 static size_t ip6gre_get_size(const struct net_device *dev)
1471 {
1472         return
1473                 /* IFLA_GRE_LINK */
1474                 nla_total_size(4) +
1475                 /* IFLA_GRE_IFLAGS */
1476                 nla_total_size(2) +
1477                 /* IFLA_GRE_OFLAGS */
1478                 nla_total_size(2) +
1479                 /* IFLA_GRE_IKEY */
1480                 nla_total_size(4) +
1481                 /* IFLA_GRE_OKEY */
1482                 nla_total_size(4) +
1483                 /* IFLA_GRE_LOCAL */
1484                 nla_total_size(sizeof(struct in6_addr)) +
1485                 /* IFLA_GRE_REMOTE */
1486                 nla_total_size(sizeof(struct in6_addr)) +
1487                 /* IFLA_GRE_TTL */
1488                 nla_total_size(1) +
1489                 /* IFLA_GRE_ENCAP_LIMIT */
1490                 nla_total_size(1) +
1491                 /* IFLA_GRE_FLOWINFO */
1492                 nla_total_size(4) +
1493                 /* IFLA_GRE_FLAGS */
1494                 nla_total_size(4) +
1495                 /* IFLA_GRE_ENCAP_TYPE */
1496                 nla_total_size(2) +
1497                 /* IFLA_GRE_ENCAP_FLAGS */
1498                 nla_total_size(2) +
1499                 /* IFLA_GRE_ENCAP_SPORT */
1500                 nla_total_size(2) +
1501                 /* IFLA_GRE_ENCAP_DPORT */
1502                 nla_total_size(2) +
1503                 /* IFLA_GRE_FWMARK */
1504                 nla_total_size(4) +
1505                 0;
1506 }
1507
1508 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1509 {
1510         struct ip6_tnl *t = netdev_priv(dev);
1511         struct __ip6_tnl_parm *p = &t->parms;
1512
1513         if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1514             nla_put_be16(skb, IFLA_GRE_IFLAGS,
1515                          gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1516             nla_put_be16(skb, IFLA_GRE_OFLAGS,
1517                          gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1518             nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1519             nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1520             nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1521             nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1522             nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1523             nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1524             nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1525             nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1526             nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1527                 goto nla_put_failure;
1528
1529         if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1530                         t->encap.type) ||
1531             nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1532                          t->encap.sport) ||
1533             nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1534                          t->encap.dport) ||
1535             nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1536                         t->encap.flags))
1537                 goto nla_put_failure;
1538
1539         return 0;
1540
1541 nla_put_failure:
1542         return -EMSGSIZE;
1543 }
1544
1545 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1546         [IFLA_GRE_LINK]        = { .type = NLA_U32 },
1547         [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1548         [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1549         [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1550         [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1551         [IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1552         [IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1553         [IFLA_GRE_TTL]         = { .type = NLA_U8 },
1554         [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1555         [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1556         [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1557         [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
1558         [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
1559         [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
1560         [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
1561         [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
1562 };
1563
1564 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1565         .kind           = "ip6gre",
1566         .maxtype        = IFLA_GRE_MAX,
1567         .policy         = ip6gre_policy,
1568         .priv_size      = sizeof(struct ip6_tnl),
1569         .setup          = ip6gre_tunnel_setup,
1570         .validate       = ip6gre_tunnel_validate,
1571         .newlink        = ip6gre_newlink,
1572         .changelink     = ip6gre_changelink,
1573         .dellink        = ip6gre_dellink,
1574         .get_size       = ip6gre_get_size,
1575         .fill_info      = ip6gre_fill_info,
1576         .get_link_net   = ip6_tnl_get_link_net,
1577 };
1578
1579 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1580         .kind           = "ip6gretap",
1581         .maxtype        = IFLA_GRE_MAX,
1582         .policy         = ip6gre_policy,
1583         .priv_size      = sizeof(struct ip6_tnl),
1584         .setup          = ip6gre_tap_setup,
1585         .validate       = ip6gre_tap_validate,
1586         .newlink        = ip6gre_newlink,
1587         .changelink     = ip6gre_changelink,
1588         .get_size       = ip6gre_get_size,
1589         .fill_info      = ip6gre_fill_info,
1590         .get_link_net   = ip6_tnl_get_link_net,
1591 };
1592
1593 /*
1594  *      And now the modules code and kernel interface.
1595  */
1596
1597 static int __init ip6gre_init(void)
1598 {
1599         int err;
1600
1601         pr_info("GRE over IPv6 tunneling driver\n");
1602
1603         err = register_pernet_device(&ip6gre_net_ops);
1604         if (err < 0)
1605                 return err;
1606
1607         err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1608         if (err < 0) {
1609                 pr_info("%s: can't add protocol\n", __func__);
1610                 goto add_proto_failed;
1611         }
1612
1613         err = rtnl_link_register(&ip6gre_link_ops);
1614         if (err < 0)
1615                 goto rtnl_link_failed;
1616
1617         err = rtnl_link_register(&ip6gre_tap_ops);
1618         if (err < 0)
1619                 goto tap_ops_failed;
1620
1621 out:
1622         return err;
1623
1624 tap_ops_failed:
1625         rtnl_link_unregister(&ip6gre_link_ops);
1626 rtnl_link_failed:
1627         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1628 add_proto_failed:
1629         unregister_pernet_device(&ip6gre_net_ops);
1630         goto out;
1631 }
1632
1633 static void __exit ip6gre_fini(void)
1634 {
1635         rtnl_link_unregister(&ip6gre_tap_ops);
1636         rtnl_link_unregister(&ip6gre_link_ops);
1637         inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1638         unregister_pernet_device(&ip6gre_net_ops);
1639 }
1640
1641 module_init(ip6gre_init);
1642 module_exit(ip6gre_fini);
1643 MODULE_LICENSE("GPL");
1644 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1645 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1646 MODULE_ALIAS_RTNL_LINK("ip6gre");
1647 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1648 MODULE_ALIAS_NETDEV("ip6gre0");