linux/net/ipv6/ip6_gre.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *      GRE over IPv6 protocol decoder.
   4 *
   5 *      Authors: Dmitry Kozlov (xeb@mail.ru)
   6 */
   7
   8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   9
  10#include <linux/capability.h>
  11#include <linux/module.h>
  12#include <linux/types.h>
  13#include <linux/kernel.h>
  14#include <linux/slab.h>
  15#include <linux/uaccess.h>
  16#include <linux/skbuff.h>
  17#include <linux/netdevice.h>
  18#include <linux/in.h>
  19#include <linux/tcp.h>
  20#include <linux/udp.h>
  21#include <linux/if_arp.h>
  22#include <linux/init.h>
  23#include <linux/in6.h>
  24#include <linux/inetdevice.h>
  25#include <linux/igmp.h>
  26#include <linux/netfilter_ipv4.h>
  27#include <linux/etherdevice.h>
  28#include <linux/if_ether.h>
  29#include <linux/hash.h>
  30#include <linux/if_tunnel.h>
  31#include <linux/ip6_tunnel.h>
  32
  33#include <net/sock.h>
  34#include <net/ip.h>
  35#include <net/ip_tunnels.h>
  36#include <net/icmp.h>
  37#include <net/protocol.h>
  38#include <net/addrconf.h>
  39#include <net/arp.h>
  40#include <net/checksum.h>
  41#include <net/dsfield.h>
  42#include <net/inet_ecn.h>
  43#include <net/xfrm.h>
  44#include <net/net_namespace.h>
  45#include <net/netns/generic.h>
  46#include <net/rtnetlink.h>
  47
  48#include <net/ipv6.h>
  49#include <net/ip6_fib.h>
  50#include <net/ip6_route.h>
  51#include <net/ip6_tunnel.h>
  52#include <net/gre.h>
  53#include <net/erspan.h>
  54#include <net/dst_metadata.h>
  55
  56
  57static bool log_ecn_error = true;
  58module_param(log_ecn_error, bool, 0644);
  59MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  60
  61#define IP6_GRE_HASH_SIZE_SHIFT  5
  62#define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
  63
  64static unsigned int ip6gre_net_id __read_mostly;
  65struct ip6gre_net {
  66        struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
  67
  68        struct ip6_tnl __rcu *collect_md_tun;
  69        struct ip6_tnl __rcu *collect_md_tun_erspan;
  70        struct net_device *fb_tunnel_dev;
  71};
  72
  73static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
  74static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
  75static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
  76static int ip6gre_tunnel_init(struct net_device *dev);
  77static void ip6gre_tunnel_setup(struct net_device *dev);
  78static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
  79static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  80static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
  81
  82/* Tunnel hash table */
  83
  84/*
  85   4 hash tables:
  86
  87   3: (remote,local)
  88   2: (remote,*)
  89   1: (*,local)
  90   0: (*,*)
  91
  92   We require exact key match i.e. if a key is present in packet
  93   it will match only tunnel with the same key; if it is not present,
  94   it will match only keyless tunnel.
  95
  96   All keysless packets, if not matched configured keyless tunnels
  97   will match fallback tunnel.
  98 */
  99
 100#define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
 101static u32 HASH_ADDR(const struct in6_addr *addr)
 102{
 103        u32 hash = ipv6_addr_hash(addr);
 104
 105        return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
 106}
 107
 108#define tunnels_r_l     tunnels[3]
 109#define tunnels_r       tunnels[2]
 110#define tunnels_l       tunnels[1]
 111#define tunnels_wc      tunnels[0]
 112
 113/* Given src, dst and key, find appropriate for input tunnel. */
 114
 115static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
 116                const struct in6_addr *remote, const struct in6_addr *local,
 117                __be32 key, __be16 gre_proto)
 118{
 119        struct net *net = dev_net(dev);
 120        int link = dev->ifindex;
 121        unsigned int h0 = HASH_ADDR(remote);
 122        unsigned int h1 = HASH_KEY(key);
 123        struct ip6_tnl *t, *cand = NULL;
 124        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 125        int dev_type = (gre_proto == htons(ETH_P_TEB) ||
 126                        gre_proto == htons(ETH_P_ERSPAN) ||
 127                        gre_proto == htons(ETH_P_ERSPAN2)) ?
 128                       ARPHRD_ETHER : ARPHRD_IP6GRE;
 129        int score, cand_score = 4;
 130        struct net_device *ndev;
 131
 132        for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
 133                if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 134                    !ipv6_addr_equal(remote, &t->parms.raddr) ||
 135                    key != t->parms.i_key ||
 136                    !(t->dev->flags & IFF_UP))
 137                        continue;
 138
 139                if (t->dev->type != ARPHRD_IP6GRE &&
 140                    t->dev->type != dev_type)
 141                        continue;
 142
 143                score = 0;
 144                if (t->parms.link != link)
 145                        score |= 1;
 146                if (t->dev->type != dev_type)
 147                        score |= 2;
 148                if (score == 0)
 149                        return t;
 150
 151                if (score < cand_score) {
 152                        cand = t;
 153                        cand_score = score;
 154                }
 155        }
 156
 157        for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
 158                if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 159                    key != t->parms.i_key ||
 160                    !(t->dev->flags & IFF_UP))
 161                        continue;
 162
 163                if (t->dev->type != ARPHRD_IP6GRE &&
 164                    t->dev->type != dev_type)
 165                        continue;
 166
 167                score = 0;
 168                if (t->parms.link != link)
 169                        score |= 1;
 170                if (t->dev->type != dev_type)
 171                        score |= 2;
 172                if (score == 0)
 173                        return t;
 174
 175                if (score < cand_score) {
 176                        cand = t;
 177                        cand_score = score;
 178                }
 179        }
 180
 181        for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
 182                if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
 183                          (!ipv6_addr_equal(local, &t->parms.raddr) ||
 184                                 !ipv6_addr_is_multicast(local))) ||
 185                    key != t->parms.i_key ||
 186                    !(t->dev->flags & IFF_UP))
 187                        continue;
 188
 189                if (t->dev->type != ARPHRD_IP6GRE &&
 190                    t->dev->type != dev_type)
 191                        continue;
 192
 193                score = 0;
 194                if (t->parms.link != link)
 195                        score |= 1;
 196                if (t->dev->type != dev_type)
 197                        score |= 2;
 198                if (score == 0)
 199                        return t;
 200
 201                if (score < cand_score) {
 202                        cand = t;
 203                        cand_score = score;
 204                }
 205        }
 206
 207        for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
 208                if (t->parms.i_key != key ||
 209                    !(t->dev->flags & IFF_UP))
 210                        continue;
 211
 212                if (t->dev->type != ARPHRD_IP6GRE &&
 213                    t->dev->type != dev_type)
 214                        continue;
 215
 216                score = 0;
 217                if (t->parms.link != link)
 218                        score |= 1;
 219                if (t->dev->type != dev_type)
 220                        score |= 2;
 221                if (score == 0)
 222                        return t;
 223
 224                if (score < cand_score) {
 225                        cand = t;
 226                        cand_score = score;
 227                }
 228        }
 229
 230        if (cand)
 231                return cand;
 232
 233        if (gre_proto == htons(ETH_P_ERSPAN) ||
 234            gre_proto == htons(ETH_P_ERSPAN2))
 235                t = rcu_dereference(ign->collect_md_tun_erspan);
 236        else
 237                t = rcu_dereference(ign->collect_md_tun);
 238
 239        if (t && t->dev->flags & IFF_UP)
 240                return t;
 241
 242        ndev = READ_ONCE(ign->fb_tunnel_dev);
 243        if (ndev && ndev->flags & IFF_UP)
 244                return netdev_priv(ndev);
 245
 246        return NULL;
 247}
 248
 249static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
 250                const struct __ip6_tnl_parm *p)
 251{
 252        const struct in6_addr *remote = &p->raddr;
 253        const struct in6_addr *local = &p->laddr;
 254        unsigned int h = HASH_KEY(p->i_key);
 255        int prio = 0;
 256
 257        if (!ipv6_addr_any(local))
 258                prio |= 1;
 259        if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
 260                prio |= 2;
 261                h ^= HASH_ADDR(remote);
 262        }
 263
 264        return &ign->tunnels[prio][h];
 265}
 266
 267static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
 268{
 269        if (t->parms.collect_md)
 270                rcu_assign_pointer(ign->collect_md_tun, t);
 271}
 272
 273static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
 274{
 275        if (t->parms.collect_md)
 276                rcu_assign_pointer(ign->collect_md_tun_erspan, t);
 277}
 278
 279static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
 280{
 281        if (t->parms.collect_md)
 282                rcu_assign_pointer(ign->collect_md_tun, NULL);
 283}
 284
 285static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
 286                                       struct ip6_tnl *t)
 287{
 288        if (t->parms.collect_md)
 289                rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
 290}
 291
 292static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
 293                const struct ip6_tnl *t)
 294{
 295        return __ip6gre_bucket(ign, &t->parms);
 296}
 297
 298static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
 299{
 300        struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
 301
 302        rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 303        rcu_assign_pointer(*tp, t);
 304}
 305
 306static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
 307{
 308        struct ip6_tnl __rcu **tp;
 309        struct ip6_tnl *iter;
 310
 311        for (tp = ip6gre_bucket(ign, t);
 312             (iter = rtnl_dereference(*tp)) != NULL;
 313             tp = &iter->next) {
 314                if (t == iter) {
 315                        rcu_assign_pointer(*tp, t->next);
 316                        break;
 317                }
 318        }
 319}
 320
 321static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
 322                                           const struct __ip6_tnl_parm *parms,
 323                                           int type)
 324{
 325        const struct in6_addr *remote = &parms->raddr;
 326        const struct in6_addr *local = &parms->laddr;
 327        __be32 key = parms->i_key;
 328        int link = parms->link;
 329        struct ip6_tnl *t;
 330        struct ip6_tnl __rcu **tp;
 331        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 332
 333        for (tp = __ip6gre_bucket(ign, parms);
 334             (t = rtnl_dereference(*tp)) != NULL;
 335             tp = &t->next)
 336                if (ipv6_addr_equal(local, &t->parms.laddr) &&
 337                    ipv6_addr_equal(remote, &t->parms.raddr) &&
 338                    key == t->parms.i_key &&
 339                    link == t->parms.link &&
 340                    type == t->dev->type)
 341                        break;
 342
 343        return t;
 344}
 345
 346static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
 347                const struct __ip6_tnl_parm *parms, int create)
 348{
 349        struct ip6_tnl *t, *nt;
 350        struct net_device *dev;
 351        char name[IFNAMSIZ];
 352        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
 353
 354        t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
 355        if (t && create)
 356                return NULL;
 357        if (t || !create)
 358                return t;
 359
 360        if (parms->name[0]) {
 361                if (!dev_valid_name(parms->name))
 362                        return NULL;
 363                strlcpy(name, parms->name, IFNAMSIZ);
 364        } else {
 365                strcpy(name, "ip6gre%d");
 366        }
 367        dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 368                           ip6gre_tunnel_setup);
 369        if (!dev)
 370                return NULL;
 371
 372        dev_net_set(dev, net);
 373
 374        nt = netdev_priv(dev);
 375        nt->parms = *parms;
 376        dev->rtnl_link_ops = &ip6gre_link_ops;
 377
 378        nt->dev = dev;
 379        nt->net = dev_net(dev);
 380
 381        if (register_netdevice(dev) < 0)
 382                goto failed_free;
 383
 384        ip6gre_tnl_link_config(nt, 1);
 385
 386        /* Can use a lockless transmit, unless we generate output sequences */
 387        if (!(nt->parms.o_flags & TUNNEL_SEQ))
 388                dev->features |= NETIF_F_LLTX;
 389
 390        ip6gre_tunnel_link(ign, nt);
 391        return nt;
 392
 393failed_free:
 394        free_netdev(dev);
 395        return NULL;
 396}
 397
 398static void ip6erspan_tunnel_uninit(struct net_device *dev)
 399{
 400        struct ip6_tnl *t = netdev_priv(dev);
 401        struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 402
 403        ip6erspan_tunnel_unlink_md(ign, t);
 404        ip6gre_tunnel_unlink(ign, t);
 405        dst_cache_reset(&t->dst_cache);
 406        dev_put(dev);
 407}
 408
 409static void ip6gre_tunnel_uninit(struct net_device *dev)
 410{
 411        struct ip6_tnl *t = netdev_priv(dev);
 412        struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
 413
 414        ip6gre_tunnel_unlink_md(ign, t);
 415        ip6gre_tunnel_unlink(ign, t);
 416        if (ign->fb_tunnel_dev == dev)
 417                WRITE_ONCE(ign->fb_tunnel_dev, NULL);
 418        dst_cache_reset(&t->dst_cache);
 419        dev_put(dev);
 420}
 421
 422
 423static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 424                       u8 type, u8 code, int offset, __be32 info)
 425{
 426        struct net *net = dev_net(skb->dev);
 427        const struct ipv6hdr *ipv6h;
 428        struct tnl_ptk_info tpi;
 429        struct ip6_tnl *t;
 430
 431        if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
 432                             offset) < 0)
 433                return -EINVAL;
 434
 435        ipv6h = (const struct ipv6hdr *)skb->data;
 436        t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
 437                                 tpi.key, tpi.proto);
 438        if (!t)
 439                return -ENOENT;
 440
 441        switch (type) {
 442        case ICMPV6_DEST_UNREACH:
 443                net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 444                                    t->parms.name);
 445                if (code != ICMPV6_PORT_UNREACH)
 446                        break;
 447                return 0;
 448        case ICMPV6_TIME_EXCEED:
 449                if (code == ICMPV6_EXC_HOPLIMIT) {
 450                        net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 451                                            t->parms.name);
 452                        break;
 453                }
 454                return 0;
 455        case ICMPV6_PARAMPROB: {
 456                struct ipv6_tlv_tnl_enc_lim *tel;
 457                __u32 teli;
 458
 459                teli = 0;
 460                if (code == ICMPV6_HDR_FIELD)
 461                        teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 462
 463                if (teli && teli == be32_to_cpu(info) - 2) {
 464                        tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 465                        if (tel->encap_limit == 0) {
 466                                net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 467                                                    t->parms.name);
 468                        }
 469                } else {
 470                        net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 471                                            t->parms.name);
 472                }
 473                return 0;
 474        }
 475        case ICMPV6_PKT_TOOBIG:
 476                ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
 477                return 0;
 478        case NDISC_REDIRECT:
 479                ip6_redirect(skb, net, skb->dev->ifindex, 0,
 480                             sock_net_uid(net, NULL));
 481                return 0;
 482        }
 483
 484        if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
 485                t->err_count++;
 486        else
 487                t->err_count = 1;
 488        t->err_time = jiffies;
 489
 490        return 0;
 491}
 492
 493static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 494{
 495        const struct ipv6hdr *ipv6h;
 496        struct ip6_tnl *tunnel;
 497
 498        ipv6h = ipv6_hdr(skb);
 499        tunnel = ip6gre_tunnel_lookup(skb->dev,
 500                                      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 501                                      tpi->proto);
 502        if (tunnel) {
 503                if (tunnel->parms.collect_md) {
 504                        struct metadata_dst *tun_dst;
 505                        __be64 tun_id;
 506                        __be16 flags;
 507
 508                        flags = tpi->flags;
 509                        tun_id = key32_to_tunnel_id(tpi->key);
 510
 511                        tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
 512                        if (!tun_dst)
 513                                return PACKET_REJECT;
 514
 515                        ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
 516                } else {
 517                        ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 518                }
 519
 520                return PACKET_RCVD;
 521        }
 522
 523        return PACKET_REJECT;
 524}
 525
 526static int ip6erspan_rcv(struct sk_buff *skb,
 527                         struct tnl_ptk_info *tpi,
 528                         int gre_hdr_len)
 529{
 530        struct erspan_base_hdr *ershdr;
 531        const struct ipv6hdr *ipv6h;
 532        struct erspan_md2 *md2;
 533        struct ip6_tnl *tunnel;
 534        u8 ver;
 535
 536        ipv6h = ipv6_hdr(skb);
 537        ershdr = (struct erspan_base_hdr *)skb->data;
 538        ver = ershdr->ver;
 539
 540        tunnel = ip6gre_tunnel_lookup(skb->dev,
 541                                      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
 542                                      tpi->proto);
 543        if (tunnel) {
 544                int len = erspan_hdr_len(ver);
 545
 546                if (unlikely(!pskb_may_pull(skb, len)))
 547                        return PACKET_REJECT;
 548
 549                if (__iptunnel_pull_header(skb, len,
 550                                           htons(ETH_P_TEB),
 551                                           false, false) < 0)
 552                        return PACKET_REJECT;
 553
 554                if (tunnel->parms.collect_md) {
 555                        struct erspan_metadata *pkt_md, *md;
 556                        struct metadata_dst *tun_dst;
 557                        struct ip_tunnel_info *info;
 558                        unsigned char *gh;
 559                        __be64 tun_id;
 560                        __be16 flags;
 561
 562                        tpi->flags |= TUNNEL_KEY;
 563                        flags = tpi->flags;
 564                        tun_id = key32_to_tunnel_id(tpi->key);
 565
 566                        tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
 567                                                  sizeof(*md));
 568                        if (!tun_dst)
 569                                return PACKET_REJECT;
 570
 571                        /* skb can be uncloned in __iptunnel_pull_header, so
 572                         * old pkt_md is no longer valid and we need to reset
 573                         * it
 574                         */
 575                        gh = skb_network_header(skb) +
 576                             skb_network_header_len(skb);
 577                        pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
 578                                                            sizeof(*ershdr));
 579                        info = &tun_dst->u.tun_info;
 580                        md = ip_tunnel_info_opts(info);
 581                        md->version = ver;
 582                        md2 = &md->u.md2;
 583                        memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
 584                                                       ERSPAN_V2_MDSIZE);
 585                        info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
 586                        info->options_len = sizeof(*md);
 587
 588                        ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
 589
 590                } else {
 591                        ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 592                }
 593
 594                return PACKET_RCVD;
 595        }
 596
 597        return PACKET_REJECT;
 598}
 599
 600static int gre_rcv(struct sk_buff *skb)
 601{
 602        struct tnl_ptk_info tpi;
 603        bool csum_err = false;
 604        int hdr_len;
 605
 606        hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
 607        if (hdr_len < 0)
 608                goto drop;
 609
 610        if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
 611                goto drop;
 612
 613        if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
 614                     tpi.proto == htons(ETH_P_ERSPAN2))) {
 615                if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
 616                        return 0;
 617                goto out;
 618        }
 619
 620        if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
 621                return 0;
 622
 623out:
 624        icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 625drop:
 626        kfree_skb(skb);
 627        return 0;
 628}
 629
 630static int gre_handle_offloads(struct sk_buff *skb, bool csum)
 631{
 632        if (csum && skb_checksum_start(skb) < skb->data)
 633                return -EINVAL;
 634        return iptunnel_handle_offloads(skb,
 635                                        csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 636}
 637
 638static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
 639                                     struct net_device *dev,
 640                                     struct flowi6 *fl6, __u8 *dsfield,
 641                                     int *encap_limit)
 642{
 643        const struct iphdr *iph = ip_hdr(skb);
 644        struct ip6_tnl *t = netdev_priv(dev);
 645
 646        if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 647                *encap_limit = t->parms.encap_limit;
 648
 649        memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 650
 651        if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 652                *dsfield = ipv4_get_dsfield(iph);
 653        else
 654                *dsfield = ip6_tclass(t->parms.flowinfo);
 655
 656        if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 657                fl6->flowi6_mark = skb->mark;
 658        else
 659                fl6->flowi6_mark = t->parms.fwmark;
 660
 661        fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 662}
 663
 664static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
 665                                    struct net_device *dev,
 666                                    struct flowi6 *fl6, __u8 *dsfield,
 667                                    int *encap_limit)
 668{
 669        struct ipv6hdr *ipv6h;
 670        struct ip6_tnl *t = netdev_priv(dev);
 671        __u16 offset;
 672
 673        offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 674        /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 675        ipv6h = ipv6_hdr(skb);
 676
 677        if (offset > 0) {
 678                struct ipv6_tlv_tnl_enc_lim *tel;
 679
 680                tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 681                if (tel->encap_limit == 0) {
 682                        icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
 683                                        ICMPV6_HDR_FIELD, offset + 2);
 684                        return -1;
 685                }
 686                *encap_limit = tel->encap_limit - 1;
 687        } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
 688                *encap_limit = t->parms.encap_limit;
 689        }
 690
 691        memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 692
 693        if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 694                *dsfield = ipv6_get_dsfield(ipv6h);
 695        else
 696                *dsfield = ip6_tclass(t->parms.flowinfo);
 697
 698        if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 699                fl6->flowlabel |= ip6_flowlabel(ipv6h);
 700
 701        if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 702                fl6->flowi6_mark = skb->mark;
 703        else
 704                fl6->flowi6_mark = t->parms.fwmark;
 705
 706        fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 707
 708        return 0;
 709}
 710
 711static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
 712{
 713        struct ip_tunnel_info *tun_info;
 714
 715        tun_info = skb_tunnel_info(skb);
 716        if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
 717                return ERR_PTR(-EINVAL);
 718
 719        return tun_info;
 720}
 721
 722static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
 723                               struct net_device *dev, __u8 dsfield,
 724                               struct flowi6 *fl6, int encap_limit,
 725                               __u32 *pmtu, __be16 proto)
 726{
 727        struct ip6_tnl *tunnel = netdev_priv(dev);
 728        __be16 protocol;
 729
 730        if (dev->type == ARPHRD_ETHER)
 731                IPCB(skb)->flags = 0;
 732
 733        if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
 734                fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
 735        else
 736                fl6->daddr = tunnel->parms.raddr;
 737
 738        if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
 739                return -ENOMEM;
 740
 741        /* Push GRE header. */
 742        protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
 743
 744        if (tunnel->parms.collect_md) {
 745                struct ip_tunnel_info *tun_info;
 746                const struct ip_tunnel_key *key;
 747                __be16 flags;
 748
 749                tun_info = skb_tunnel_info_txcheck(skb);
 750                if (IS_ERR(tun_info) ||
 751                    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
 752                        return -EINVAL;
 753
 754                key = &tun_info->key;
 755                memset(fl6, 0, sizeof(*fl6));
 756                fl6->flowi6_proto = IPPROTO_GRE;
 757                fl6->daddr = key->u.ipv6.dst;
 758                fl6->flowlabel = key->label;
 759                fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 760
 761                dsfield = key->tos;
 762                flags = key->tun_flags &
 763                        (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
 764                tunnel->tun_hlen = gre_calc_hlen(flags);
 765
 766                gre_build_header(skb, tunnel->tun_hlen,
 767                                 flags, protocol,
 768                                 tunnel_id_to_key32(tun_info->key.tun_id),
 769                                 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
 770                                                      : 0);
 771
 772        } else {
 773                if (tunnel->parms.o_flags & TUNNEL_SEQ)
 774                        tunnel->o_seqno++;
 775
 776                gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
 777                                 protocol, tunnel->parms.o_key,
 778                                 htonl(tunnel->o_seqno));
 779        }
 780
 781        return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
 782                            NEXTHDR_GRE);
 783}
 784
 785static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 786{
 787        struct ip6_tnl *t = netdev_priv(dev);
 788        int encap_limit = -1;
 789        struct flowi6 fl6;
 790        __u8 dsfield = 0;
 791        __u32 mtu;
 792        int err;
 793
 794        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 795
 796        if (!t->parms.collect_md)
 797                prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
 798                                         &dsfield, &encap_limit);
 799
 800        err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 801        if (err)
 802                return -1;
 803
 804        err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 805                          skb->protocol);
 806        if (err != 0) {
 807                /* XXX: send ICMP error even if DF is not set. */
 808                if (err == -EMSGSIZE)
 809                        icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 810                                      htonl(mtu));
 811                return -1;
 812        }
 813
 814        return 0;
 815}
 816
 817static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 818{
 819        struct ip6_tnl *t = netdev_priv(dev);
 820        struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 821        int encap_limit = -1;
 822        struct flowi6 fl6;
 823        __u8 dsfield = 0;
 824        __u32 mtu;
 825        int err;
 826
 827        if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 828                return -1;
 829
 830        if (!t->parms.collect_md &&
 831            prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
 832                return -1;
 833
 834        if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
 835                return -1;
 836
 837        err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
 838                          &mtu, skb->protocol);
 839        if (err != 0) {
 840                if (err == -EMSGSIZE)
 841                        icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 842                return -1;
 843        }
 844
 845        return 0;
 846}
 847
 848/**
 849 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
 850 *   @t: the outgoing tunnel device
 851 *   @hdr: IPv6 header from the incoming packet
 852 *
 853 * Description:
 854 *   Avoid trivial tunneling loop by checking that tunnel exit-point
 855 *   doesn't match source of incoming packet.
 856 *
 857 * Return:
 858 *   1 if conflict,
 859 *   0 else
 860 **/
 861
 862static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 863        const struct ipv6hdr *hdr)
 864{
 865        return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 866}
 867
 868static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 869{
 870        struct ip6_tnl *t = netdev_priv(dev);
 871        int encap_limit = -1;
 872        struct flowi6 fl6;
 873        __u32 mtu;
 874        int err;
 875
 876        if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 877                encap_limit = t->parms.encap_limit;
 878
 879        if (!t->parms.collect_md)
 880                memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 881
 882        err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 883        if (err)
 884                return err;
 885
 886        err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
 887
 888        return err;
 889}
 890
 891static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 892        struct net_device *dev)
 893{
 894        struct ip6_tnl *t = netdev_priv(dev);
 895        struct net_device_stats *stats = &t->dev->stats;
 896        int ret;
 897
 898        if (!pskb_inet_may_pull(skb))
 899                goto tx_err;
 900
 901        if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 902                goto tx_err;
 903
 904        switch (skb->protocol) {
 905        case htons(ETH_P_IP):
 906                ret = ip6gre_xmit_ipv4(skb, dev);
 907                break;
 908        case htons(ETH_P_IPV6):
 909                ret = ip6gre_xmit_ipv6(skb, dev);
 910                break;
 911        default:
 912                ret = ip6gre_xmit_other(skb, dev);
 913                break;
 914        }
 915
 916        if (ret < 0)
 917                goto tx_err;
 918
 919        return NETDEV_TX_OK;
 920
 921tx_err:
 922        if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
 923                stats->tx_errors++;
 924        stats->tx_dropped++;
 925        kfree_skb(skb);
 926        return NETDEV_TX_OK;
 927}
 928
 929static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
 930                                         struct net_device *dev)
 931{
 932        struct ip_tunnel_info *tun_info = NULL;
 933        struct ip6_tnl *t = netdev_priv(dev);
 934        struct dst_entry *dst = skb_dst(skb);
 935        struct net_device_stats *stats;
 936        bool truncate = false;
 937        int encap_limit = -1;
 938        __u8 dsfield = false;
 939        struct flowi6 fl6;
 940        int err = -EINVAL;
 941        __be16 proto;
 942        __u32 mtu;
 943        int nhoff;
 944        int thoff;
 945
 946        if (!pskb_inet_may_pull(skb))
 947                goto tx_err;
 948
 949        if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 950                goto tx_err;
 951
 952        if (gre_handle_offloads(skb, false))
 953                goto tx_err;
 954
 955        if (skb->len > dev->mtu + dev->hard_header_len) {
 956                pskb_trim(skb, dev->mtu + dev->hard_header_len);
 957                truncate = true;
 958        }
 959
 960        nhoff = skb_network_header(skb) - skb_mac_header(skb);
 961        if (skb->protocol == htons(ETH_P_IP) &&
 962            (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
 963                truncate = true;
 964
 965        thoff = skb_transport_header(skb) - skb_mac_header(skb);
 966        if (skb->protocol == htons(ETH_P_IPV6) &&
 967            (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
 968                truncate = true;
 969
 970        if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
 971                goto tx_err;
 972
 973        t->parms.o_flags &= ~TUNNEL_KEY;
 974        IPCB(skb)->flags = 0;
 975
 976        /* For collect_md mode, derive fl6 from the tunnel key,
 977         * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
 978         */
 979        if (t->parms.collect_md) {
 980                const struct ip_tunnel_key *key;
 981                struct erspan_metadata *md;
 982                __be32 tun_id;
 983
 984                tun_info = skb_tunnel_info_txcheck(skb);
 985                if (IS_ERR(tun_info) ||
 986                    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
 987                        goto tx_err;
 988
 989                key = &tun_info->key;
 990                memset(&fl6, 0, sizeof(fl6));
 991                fl6.flowi6_proto = IPPROTO_GRE;
 992                fl6.daddr = key->u.ipv6.dst;
 993                fl6.flowlabel = key->label;
 994                fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 995
 996                dsfield = key->tos;
 997                if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
 998                        goto tx_err;
 999                if (tun_info->options_len < sizeof(*md))
1000                        goto tx_err;
1001                md = ip_tunnel_info_opts(tun_info);
1002
1003                tun_id = tunnel_id_to_key32(key->tun_id);
1004                if (md->version == 1) {
1005                        erspan_build_header(skb,
1006                                            ntohl(tun_id),
1007                                            ntohl(md->u.index), truncate,
1008                                            false);
1009                } else if (md->version == 2) {
1010                        erspan_build_header_v2(skb,
1011                                               ntohl(tun_id),
1012                                               md->u.md2.dir,
1013                                               get_hwid(&md->u.md2),
1014                                               truncate, false);
1015                } else {
1016                        goto tx_err;
1017                }
1018        } else {
1019                switch (skb->protocol) {
1020                case htons(ETH_P_IP):
1021                        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1022                        prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1023                                                 &dsfield, &encap_limit);
1024                        break;
1025                case htons(ETH_P_IPV6):
1026                        if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1027                                goto tx_err;
1028                        if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1029                                                     &dsfield, &encap_limit))
1030                                goto tx_err;
1031                        break;
1032                default:
1033                        memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1034                        break;
1035                }
1036
1037                if (t->parms.erspan_ver == 1)
1038                        erspan_build_header(skb, ntohl(t->parms.o_key),
1039                                            t->parms.index,
1040                                            truncate, false);
1041                else if (t->parms.erspan_ver == 2)
1042                        erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1043                                               t->parms.dir,
1044                                               t->parms.hwid,
1045                                               truncate, false);
1046                else
1047                        goto tx_err;
1048
1049                fl6.daddr = t->parms.raddr;
1050        }
1051
1052        /* Push GRE header. */
1053        proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1054                                           : htons(ETH_P_ERSPAN2);
1055        gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1056
1057        /* TooBig packet may have updated dst->dev's mtu */
1058        if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1059                dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1060
1061        err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1062                           NEXTHDR_GRE);
1063        if (err != 0) {
1064                /* XXX: send ICMP error even if DF is not set. */
1065                if (err == -EMSGSIZE) {
1066                        if (skb->protocol == htons(ETH_P_IP))
1067                                icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1068                                              ICMP_FRAG_NEEDED, htonl(mtu));
1069                        else
1070                                icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1071                }
1072
1073                goto tx_err;
1074        }
1075        return NETDEV_TX_OK;
1076
1077tx_err:
1078        stats = &t->dev->stats;
1079        if (!IS_ERR(tun_info))
1080                stats->tx_errors++;
1081        stats->tx_dropped++;
1082        kfree_skb(skb);
1083        return NETDEV_TX_OK;
1084}
1085
1086static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1087{
1088        struct net_device *dev = t->dev;
1089        struct __ip6_tnl_parm *p = &t->parms;
1090        struct flowi6 *fl6 = &t->fl.u.ip6;
1091
1092        if (dev->type != ARPHRD_ETHER) {
1093                memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1094                memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1095        }
1096
1097        /* Set up flowi template */
1098        fl6->saddr = p->laddr;
1099        fl6->daddr = p->raddr;
1100        fl6->flowi6_oif = p->link;
1101        fl6->flowlabel = 0;
1102        fl6->flowi6_proto = IPPROTO_GRE;
1103
1104        if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1105                fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1106        if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1107                fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1108
1109        p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1110        p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1111
1112        if (p->flags&IP6_TNL_F_CAP_XMIT &&
1113                        p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1114                dev->flags |= IFF_POINTOPOINT;
1115        else
1116                dev->flags &= ~IFF_POINTOPOINT;
1117}
1118
1119static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1120                                         int t_hlen)
1121{
1122        const struct __ip6_tnl_parm *p = &t->parms;
1123        struct net_device *dev = t->dev;
1124
1125        if (p->flags & IP6_TNL_F_CAP_XMIT) {
1126                int strict = (ipv6_addr_type(&p->raddr) &
1127                              (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1128
1129                struct rt6_info *rt = rt6_lookup(t->net,
1130                                                 &p->raddr, &p->laddr,
1131                                                 p->link, NULL, strict);
1132
1133                if (!rt)
1134                        return;
1135
1136                if (rt->dst.dev) {
1137                        unsigned short dst_len = rt->dst.dev->hard_header_len +
1138                                                 t_hlen;
1139
1140                        if (t->dev->header_ops)
1141                                dev->hard_header_len = dst_len;
1142                        else
1143                                dev->needed_headroom = dst_len;
1144
1145                        if (set_mtu) {
1146                                dev->mtu = rt->dst.dev->mtu - t_hlen;
1147                                if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1148                                        dev->mtu -= 8;
1149                                if (dev->type == ARPHRD_ETHER)
1150                                        dev->mtu -= ETH_HLEN;
1151
1152                                if (dev->mtu < IPV6_MIN_MTU)
1153                                        dev->mtu = IPV6_MIN_MTU;
1154                        }
1155                }
1156                ip6_rt_put(rt);
1157        }
1158}
1159
1160static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1161{
1162        int t_hlen;
1163
1164        tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1165        tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1166
1167        t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1168
1169        if (tunnel->dev->header_ops)
1170                tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1171        else
1172                tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1173
1174        return t_hlen;
1175}
1176
1177static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1178{
1179        ip6gre_tnl_link_config_common(t);
1180        ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1181}
1182
1183static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1184                                     const struct __ip6_tnl_parm *p)
1185{
1186        t->parms.laddr = p->laddr;
1187        t->parms.raddr = p->raddr;
1188        t->parms.flags = p->flags;
1189        t->parms.hop_limit = p->hop_limit;
1190        t->parms.encap_limit = p->encap_limit;
1191        t->parms.flowinfo = p->flowinfo;
1192        t->parms.link = p->link;
1193        t->parms.proto = p->proto;
1194        t->parms.i_key = p->i_key;
1195        t->parms.o_key = p->o_key;
1196        t->parms.i_flags = p->i_flags;
1197        t->parms.o_flags = p->o_flags;
1198        t->parms.fwmark = p->fwmark;
1199        t->parms.erspan_ver = p->erspan_ver;
1200        t->parms.index = p->index;
1201        t->parms.dir = p->dir;
1202        t->parms.hwid = p->hwid;
1203        dst_cache_reset(&t->dst_cache);
1204}
1205
1206static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1207                             int set_mtu)
1208{
1209        ip6gre_tnl_copy_tnl_parm(t, p);
1210        ip6gre_tnl_link_config(t, set_mtu);
1211        return 0;
1212}
1213
1214static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1215        const struct ip6_tnl_parm2 *u)
1216{
1217        p->laddr = u->laddr;
1218        p->raddr = u->raddr;
1219        p->flags = u->flags;
1220        p->hop_limit = u->hop_limit;
1221        p->encap_limit = u->encap_limit;
1222        p->flowinfo = u->flowinfo;
1223        p->link = u->link;
1224        p->i_key = u->i_key;
1225        p->o_key = u->o_key;
1226        p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1227        p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1228        memcpy(p->name, u->name, sizeof(u->name));
1229}
1230
1231static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1232        const struct __ip6_tnl_parm *p)
1233{
1234        u->proto = IPPROTO_GRE;
1235        u->laddr = p->laddr;
1236        u->raddr = p->raddr;
1237        u->flags = p->flags;
1238        u->hop_limit = p->hop_limit;
1239        u->encap_limit = p->encap_limit;
1240        u->flowinfo = p->flowinfo;
1241        u->link = p->link;
1242        u->i_key = p->i_key;
1243        u->o_key = p->o_key;
1244        u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1245        u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1246        memcpy(u->name, p->name, sizeof(u->name));
1247}
1248
1249static int ip6gre_tunnel_ioctl(struct net_device *dev,
1250        struct ifreq *ifr, int cmd)
1251{
1252        int err = 0;
1253        struct ip6_tnl_parm2 p;
1254        struct __ip6_tnl_parm p1;
1255        struct ip6_tnl *t = netdev_priv(dev);
1256        struct net *net = t->net;
1257        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1258
1259        memset(&p1, 0, sizeof(p1));
1260
1261        switch (cmd) {
1262        case SIOCGETTUNNEL:
1263                if (dev == ign->fb_tunnel_dev) {
1264                        if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1265                                err = -EFAULT;
1266                                break;
1267                        }
1268                        ip6gre_tnl_parm_from_user(&p1, &p);
1269                        t = ip6gre_tunnel_locate(net, &p1, 0);
1270                        if (!t)
1271                                t = netdev_priv(dev);
1272                }
1273                memset(&p, 0, sizeof(p));
1274                ip6gre_tnl_parm_to_user(&p, &t->parms);
1275                if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1276                        err = -EFAULT;
1277                break;
1278
1279        case SIOCADDTUNNEL:
1280        case SIOCCHGTUNNEL:
1281                err = -EPERM;
1282                if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1283                        goto done;
1284
1285                err = -EFAULT;
1286                if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1287                        goto done;
1288
1289                err = -EINVAL;
1290                if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1291                        goto done;
1292
1293                if (!(p.i_flags&GRE_KEY))
1294                        p.i_key = 0;
1295                if (!(p.o_flags&GRE_KEY))
1296                        p.o_key = 0;
1297
1298                ip6gre_tnl_parm_from_user(&p1, &p);
1299                t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1300
1301                if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1302                        if (t) {
1303                                if (t->dev != dev) {
1304                                        err = -EEXIST;
1305                                        break;
1306                                }
1307                        } else {
1308                                t = netdev_priv(dev);
1309
1310                                ip6gre_tunnel_unlink(ign, t);
1311                                synchronize_net();
1312                                ip6gre_tnl_change(t, &p1, 1);
1313                                ip6gre_tunnel_link(ign, t);
1314                                netdev_state_change(dev);
1315                        }
1316                }
1317
1318                if (t) {
1319                        err = 0;
1320
1321                        memset(&p, 0, sizeof(p));
1322                        ip6gre_tnl_parm_to_user(&p, &t->parms);
1323                        if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1324                                err = -EFAULT;
1325                } else
1326                        err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1327                break;
1328
1329        case SIOCDELTUNNEL:
1330                err = -EPERM;
1331                if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1332                        goto done;
1333
1334                if (dev == ign->fb_tunnel_dev) {
1335                        err = -EFAULT;
1336                        if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1337                                goto done;
1338                        err = -ENOENT;
1339                        ip6gre_tnl_parm_from_user(&p1, &p);
1340                        t = ip6gre_tunnel_locate(net, &p1, 0);
1341                        if (!t)
1342                                goto done;
1343                        err = -EPERM;
1344                        if (t == netdev_priv(ign->fb_tunnel_dev))
1345                                goto done;
1346                        dev = t->dev;
1347                }
1348                unregister_netdevice(dev);
1349                err = 0;
1350                break;
1351
1352        default:
1353                err = -EINVAL;
1354        }
1355
1356done:
1357        return err;
1358}
1359
1360static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1361                         unsigned short type, const void *daddr,
1362                         const void *saddr, unsigned int len)
1363{
1364        struct ip6_tnl *t = netdev_priv(dev);
1365        struct ipv6hdr *ipv6h;
1366        __be16 *p;
1367
1368        ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1369        ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1370                                                  t->fl.u.ip6.flowlabel,
1371                                                  true, &t->fl.u.ip6));
1372        ipv6h->hop_limit = t->parms.hop_limit;
1373        ipv6h->nexthdr = NEXTHDR_GRE;
1374        ipv6h->saddr = t->parms.laddr;
1375        ipv6h->daddr = t->parms.raddr;
1376
1377        p = (__be16 *)(ipv6h + 1);
1378        p[0] = t->parms.o_flags;
1379        p[1] = htons(type);
1380
1381        /*
1382         *      Set the source hardware address.
1383         */
1384
1385        if (saddr)
1386                memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1387        if (daddr)
1388                memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1389        if (!ipv6_addr_any(&ipv6h->daddr))
1390                return t->hlen;
1391
1392        return -t->hlen;
1393}
1394
1395static const struct header_ops ip6gre_header_ops = {
1396        .create = ip6gre_header,
1397};
1398
1399static const struct net_device_ops ip6gre_netdev_ops = {
1400        .ndo_init               = ip6gre_tunnel_init,
1401        .ndo_uninit             = ip6gre_tunnel_uninit,
1402        .ndo_start_xmit         = ip6gre_tunnel_xmit,
1403        .ndo_do_ioctl           = ip6gre_tunnel_ioctl,
1404        .ndo_change_mtu         = ip6_tnl_change_mtu,
1405        .ndo_get_stats64        = dev_get_tstats64,
1406        .ndo_get_iflink         = ip6_tnl_get_iflink,
1407};
1408
1409static void ip6gre_dev_free(struct net_device *dev)
1410{
1411        struct ip6_tnl *t = netdev_priv(dev);
1412
1413        gro_cells_destroy(&t->gro_cells);
1414        dst_cache_destroy(&t->dst_cache);
1415        free_percpu(dev->tstats);
1416}
1417
1418static void ip6gre_tunnel_setup(struct net_device *dev)
1419{
1420        dev->netdev_ops = &ip6gre_netdev_ops;
1421        dev->needs_free_netdev = true;
1422        dev->priv_destructor = ip6gre_dev_free;
1423
1424        dev->type = ARPHRD_IP6GRE;
1425
1426        dev->flags |= IFF_NOARP;
1427        dev->addr_len = sizeof(struct in6_addr);
1428        netif_keep_dst(dev);
1429        /* This perm addr will be used as interface identifier by IPv6 */
1430        dev->addr_assign_type = NET_ADDR_RANDOM;
1431        eth_random_addr(dev->perm_addr);
1432}
1433
1434#define GRE6_FEATURES (NETIF_F_SG |             \
1435                       NETIF_F_FRAGLIST |       \
1436                       NETIF_F_HIGHDMA |        \
1437                       NETIF_F_HW_CSUM)
1438
1439static void ip6gre_tnl_init_features(struct net_device *dev)
1440{
1441        struct ip6_tnl *nt = netdev_priv(dev);
1442
1443        dev->features           |= GRE6_FEATURES;
1444        dev->hw_features        |= GRE6_FEATURES;
1445
1446        if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1447                /* TCP offload with GRE SEQ is not supported, nor
1448                 * can we support 2 levels of outer headers requiring
1449                 * an update.
1450                 */
1451                if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1452                    nt->encap.type == TUNNEL_ENCAP_NONE) {
1453                        dev->features    |= NETIF_F_GSO_SOFTWARE;
1454                        dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1455                }
1456
1457                /* Can use a lockless transmit, unless we generate
1458                 * output sequences
1459                 */
1460                dev->features |= NETIF_F_LLTX;
1461        }
1462}
1463
1464static int ip6gre_tunnel_init_common(struct net_device *dev)
1465{
1466        struct ip6_tnl *tunnel;
1467        int ret;
1468        int t_hlen;
1469
1470        tunnel = netdev_priv(dev);
1471
1472        tunnel->dev = dev;
1473        tunnel->net = dev_net(dev);
1474        strcpy(tunnel->parms.name, dev->name);
1475
1476        dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1477        if (!dev->tstats)
1478                return -ENOMEM;
1479
1480        ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1481        if (ret)
1482                goto cleanup_alloc_pcpu_stats;
1483
1484        ret = gro_cells_init(&tunnel->gro_cells, dev);
1485        if (ret)
1486                goto cleanup_dst_cache_init;
1487
1488        t_hlen = ip6gre_calc_hlen(tunnel);
1489        dev->mtu = ETH_DATA_LEN - t_hlen;
1490        if (dev->type == ARPHRD_ETHER)
1491                dev->mtu -= ETH_HLEN;
1492        if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1493                dev->mtu -= 8;
1494
1495        if (tunnel->parms.collect_md) {
1496                netif_keep_dst(dev);
1497        }
1498        ip6gre_tnl_init_features(dev);
1499
1500        dev_hold(dev);
1501        return 0;
1502
1503cleanup_dst_cache_init:
1504        dst_cache_destroy(&tunnel->dst_cache);
1505cleanup_alloc_pcpu_stats:
1506        free_percpu(dev->tstats);
1507        dev->tstats = NULL;
1508        return ret;
1509}
1510
1511static int ip6gre_tunnel_init(struct net_device *dev)
1512{
1513        struct ip6_tnl *tunnel;
1514        int ret;
1515
1516        ret = ip6gre_tunnel_init_common(dev);
1517        if (ret)
1518                return ret;
1519
1520        tunnel = netdev_priv(dev);
1521
1522        if (tunnel->parms.collect_md)
1523                return 0;
1524
1525        memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1526        memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1527
1528        if (ipv6_addr_any(&tunnel->parms.raddr))
1529                dev->header_ops = &ip6gre_header_ops;
1530
1531        return 0;
1532}
1533
1534static void ip6gre_fb_tunnel_init(struct net_device *dev)
1535{
1536        struct ip6_tnl *tunnel = netdev_priv(dev);
1537
1538        tunnel->dev = dev;
1539        tunnel->net = dev_net(dev);
1540        strcpy(tunnel->parms.name, dev->name);
1541
1542        tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1543}
1544
1545static struct inet6_protocol ip6gre_protocol __read_mostly = {
1546        .handler     = gre_rcv,
1547        .err_handler = ip6gre_err,
1548        .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1549};
1550
1551static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1552{
1553        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1554        struct net_device *dev, *aux;
1555        int prio;
1556
1557        for_each_netdev_safe(net, dev, aux)
1558                if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1559                    dev->rtnl_link_ops == &ip6gre_tap_ops ||
1560                    dev->rtnl_link_ops == &ip6erspan_tap_ops)
1561                        unregister_netdevice_queue(dev, head);
1562
1563        for (prio = 0; prio < 4; prio++) {
1564                int h;
1565                for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1566                        struct ip6_tnl *t;
1567
1568                        t = rtnl_dereference(ign->tunnels[prio][h]);
1569
1570                        while (t) {
1571                                /* If dev is in the same netns, it has already
1572                                 * been added to the list by the previous loop.
1573                                 */
1574                                if (!net_eq(dev_net(t->dev), net))
1575                                        unregister_netdevice_queue(t->dev,
1576                                                                   head);
1577                                t = rtnl_dereference(t->next);
1578                        }
1579                }
1580        }
1581}
1582
1583static int __net_init ip6gre_init_net(struct net *net)
1584{
1585        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1586        struct net_device *ndev;
1587        int err;
1588
1589        if (!net_has_fallback_tunnels(net))
1590                return 0;
1591        ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1592                            NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1593        if (!ndev) {
1594                err = -ENOMEM;
1595                goto err_alloc_dev;
1596        }
1597        ign->fb_tunnel_dev = ndev;
1598        dev_net_set(ign->fb_tunnel_dev, net);
1599        /* FB netdevice is special: we have one, and only one per netns.
1600         * Allowing to move it to another netns is clearly unsafe.
1601         */
1602        ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1603
1604
1605        ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1606        ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1607
1608        err = register_netdev(ign->fb_tunnel_dev);
1609        if (err)
1610                goto err_reg_dev;
1611
1612        rcu_assign_pointer(ign->tunnels_wc[0],
1613                           netdev_priv(ign->fb_tunnel_dev));
1614        return 0;
1615
1616err_reg_dev:
1617        free_netdev(ndev);
1618err_alloc_dev:
1619        return err;
1620}
1621
1622static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1623{
1624        struct net *net;
1625        LIST_HEAD(list);
1626
1627        rtnl_lock();
1628        list_for_each_entry(net, net_list, exit_list)
1629                ip6gre_destroy_tunnels(net, &list);
1630        unregister_netdevice_many(&list);
1631        rtnl_unlock();
1632}
1633
1634static struct pernet_operations ip6gre_net_ops = {
1635        .init = ip6gre_init_net,
1636        .exit_batch = ip6gre_exit_batch_net,
1637        .id   = &ip6gre_net_id,
1638        .size = sizeof(struct ip6gre_net),
1639};
1640
1641static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1642                                  struct netlink_ext_ack *extack)
1643{
1644        __be16 flags;
1645
1646        if (!data)
1647                return 0;
1648
1649        flags = 0;
1650        if (data[IFLA_GRE_IFLAGS])
1651                flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1652        if (data[IFLA_GRE_OFLAGS])
1653                flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1654        if (flags & (GRE_VERSION|GRE_ROUTING))
1655                return -EINVAL;
1656
1657        return 0;
1658}
1659
1660static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1661                               struct netlink_ext_ack *extack)
1662{
1663        struct in6_addr daddr;
1664
1665        if (tb[IFLA_ADDRESS]) {
1666                if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1667                        return -EINVAL;
1668                if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1669                        return -EADDRNOTAVAIL;
1670        }
1671
1672        if (!data)
1673                goto out;
1674
1675        if (data[IFLA_GRE_REMOTE]) {
1676                daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1677                if (ipv6_addr_any(&daddr))
1678                        return -EINVAL;
1679        }
1680
1681out:
1682        return ip6gre_tunnel_validate(tb, data, extack);
1683}
1684
1685static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1686                                  struct netlink_ext_ack *extack)
1687{
1688        __be16 flags = 0;
1689        int ret, ver = 0;
1690
1691        if (!data)
1692                return 0;
1693
1694        ret = ip6gre_tap_validate(tb, data, extack);
1695        if (ret)
1696                return ret;
1697
1698        /* ERSPAN should only have GRE sequence and key flag */
1699        if (data[IFLA_GRE_OFLAGS])
1700                flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1701        if (data[IFLA_GRE_IFLAGS])
1702                flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1703        if (!data[IFLA_GRE_COLLECT_METADATA] &&
1704            flags != (GRE_SEQ | GRE_KEY))
1705                return -EINVAL;
1706
1707        /* ERSPAN Session ID only has 10-bit. Since we reuse
1708         * 32-bit key field as ID, check it's range.
1709         */
1710        if (data[IFLA_GRE_IKEY] &&
1711            (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1712                return -EINVAL;
1713
1714        if (data[IFLA_GRE_OKEY] &&
1715            (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1716                return -EINVAL;
1717
1718        if (data[IFLA_GRE_ERSPAN_VER]) {
1719                ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1720                if (ver != 1 && ver != 2)
1721                        return -EINVAL;
1722        }
1723
1724        if (ver == 1) {
1725                if (data[IFLA_GRE_ERSPAN_INDEX]) {
1726                        u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1727
1728                        if (index & ~INDEX_MASK)
1729                                return -EINVAL;
1730                }
1731        } else if (ver == 2) {
1732                if (data[IFLA_GRE_ERSPAN_DIR]) {
1733                        u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1734
1735                        if (dir & ~(DIR_MASK >> DIR_OFFSET))
1736                                return -EINVAL;
1737                }
1738
1739                if (data[IFLA_GRE_ERSPAN_HWID]) {
1740                        u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1741
1742                        if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1743                                return -EINVAL;
1744                }
1745        }
1746
1747        return 0;
1748}
1749
1750static void ip6erspan_set_version(struct nlattr *data[],
1751                                  struct __ip6_tnl_parm *parms)
1752{
1753        if (!data)
1754                return;
1755
1756        parms->erspan_ver = 1;
1757        if (data[IFLA_GRE_ERSPAN_VER])
1758                parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1759
1760        if (parms->erspan_ver == 1) {
1761                if (data[IFLA_GRE_ERSPAN_INDEX])
1762                        parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1763        } else if (parms->erspan_ver == 2) {
1764                if (data[IFLA_GRE_ERSPAN_DIR])
1765                        parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1766                if (data[IFLA_GRE_ERSPAN_HWID])
1767                        parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1768        }
1769}
1770
1771static void ip6gre_netlink_parms(struct nlattr *data[],
1772                                struct __ip6_tnl_parm *parms)
1773{
1774        memset(parms, 0, sizeof(*parms));
1775
1776        if (!data)
1777                return;
1778
1779        if (data[IFLA_GRE_LINK])
1780                parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1781
1782        if (data[IFLA_GRE_IFLAGS])
1783                parms->i_flags = gre_flags_to_tnl_flags(
1784                                nla_get_be16(data[IFLA_GRE_IFLAGS]));
1785
1786        if (data[IFLA_GRE_OFLAGS])
1787                parms->o_flags = gre_flags_to_tnl_flags(
1788                                nla_get_be16(data[IFLA_GRE_OFLAGS]));
1789
1790        if (data[IFLA_GRE_IKEY])
1791                parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1792
1793        if (data[IFLA_GRE_OKEY])
1794                parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1795
1796        if (data[IFLA_GRE_LOCAL])
1797                parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1798
1799        if (data[IFLA_GRE_REMOTE])
1800                parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1801
1802        if (data[IFLA_GRE_TTL])
1803                parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1804
1805        if (data[IFLA_GRE_ENCAP_LIMIT])
1806                parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1807
1808        if (data[IFLA_GRE_FLOWINFO])
1809                parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1810
1811        if (data[IFLA_GRE_FLAGS])
1812                parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1813
1814        if (data[IFLA_GRE_FWMARK])
1815                parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1816
1817        if (data[IFLA_GRE_COLLECT_METADATA])
1818                parms->collect_md = true;
1819}
1820
1821static int ip6gre_tap_init(struct net_device *dev)
1822{
1823        int ret;
1824
1825        ret = ip6gre_tunnel_init_common(dev);
1826        if (ret)
1827                return ret;
1828
1829        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1830
1831        return 0;
1832}
1833
1834static const struct net_device_ops ip6gre_tap_netdev_ops = {
1835        .ndo_init = ip6gre_tap_init,
1836        .ndo_uninit = ip6gre_tunnel_uninit,
1837        .ndo_start_xmit = ip6gre_tunnel_xmit,
1838        .ndo_set_mac_address = eth_mac_addr,
1839        .ndo_validate_addr = eth_validate_addr,
1840        .ndo_change_mtu = ip6_tnl_change_mtu,
1841        .ndo_get_stats64 = dev_get_tstats64,
1842        .ndo_get_iflink = ip6_tnl_get_iflink,
1843};
1844
1845static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1846{
1847        int t_hlen;
1848
1849        tunnel->tun_hlen = 8;
1850        tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1851                       erspan_hdr_len(tunnel->parms.erspan_ver);
1852
1853        t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1854        tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1855        return t_hlen;
1856}
1857
1858static int ip6erspan_tap_init(struct net_device *dev)
1859{
1860        struct ip6_tnl *tunnel;
1861        int t_hlen;
1862        int ret;
1863
1864        tunnel = netdev_priv(dev);
1865
1866        tunnel->dev = dev;
1867        tunnel->net = dev_net(dev);
1868        strcpy(tunnel->parms.name, dev->name);
1869
1870        dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1871        if (!dev->tstats)
1872                return -ENOMEM;
1873
1874        ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1875        if (ret)
1876                goto cleanup_alloc_pcpu_stats;
1877
1878        ret = gro_cells_init(&tunnel->gro_cells, dev);
1879        if (ret)
1880                goto cleanup_dst_cache_init;
1881
1882        t_hlen = ip6erspan_calc_hlen(tunnel);
1883        dev->mtu = ETH_DATA_LEN - t_hlen;
1884        if (dev->type == ARPHRD_ETHER)
1885                dev->mtu -= ETH_HLEN;
1886        if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1887                dev->mtu -= 8;
1888
1889        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1890        ip6erspan_tnl_link_config(tunnel, 1);
1891
1892        dev_hold(dev);
1893        return 0;
1894
1895cleanup_dst_cache_init:
1896        dst_cache_destroy(&tunnel->dst_cache);
1897cleanup_alloc_pcpu_stats:
1898        free_percpu(dev->tstats);
1899        dev->tstats = NULL;
1900        return ret;
1901}
1902
1903static const struct net_device_ops ip6erspan_netdev_ops = {
1904        .ndo_init =             ip6erspan_tap_init,
1905        .ndo_uninit =           ip6erspan_tunnel_uninit,
1906        .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1907        .ndo_set_mac_address =  eth_mac_addr,
1908        .ndo_validate_addr =    eth_validate_addr,
1909        .ndo_change_mtu =       ip6_tnl_change_mtu,
1910        .ndo_get_stats64 =      dev_get_tstats64,
1911        .ndo_get_iflink =       ip6_tnl_get_iflink,
1912};
1913
1914static void ip6gre_tap_setup(struct net_device *dev)
1915{
1916
1917        ether_setup(dev);
1918
1919        dev->max_mtu = 0;
1920        dev->netdev_ops = &ip6gre_tap_netdev_ops;
1921        dev->needs_free_netdev = true;
1922        dev->priv_destructor = ip6gre_dev_free;
1923
1924        dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1925        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1926        netif_keep_dst(dev);
1927}
1928
1929static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1930                                       struct ip_tunnel_encap *ipencap)
1931{
1932        bool ret = false;
1933
1934        memset(ipencap, 0, sizeof(*ipencap));
1935
1936        if (!data)
1937                return ret;
1938
1939        if (data[IFLA_GRE_ENCAP_TYPE]) {
1940                ret = true;
1941                ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1942        }
1943
1944        if (data[IFLA_GRE_ENCAP_FLAGS]) {
1945                ret = true;
1946                ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1947        }
1948
1949        if (data[IFLA_GRE_ENCAP_SPORT]) {
1950                ret = true;
1951                ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1952        }
1953
1954        if (data[IFLA_GRE_ENCAP_DPORT]) {
1955                ret = true;
1956                ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1957        }
1958
1959        return ret;
1960}
1961
1962static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1963                                 struct nlattr *tb[], struct nlattr *data[],
1964                                 struct netlink_ext_ack *extack)
1965{
1966        struct ip6_tnl *nt;
1967        struct ip_tunnel_encap ipencap;
1968        int err;
1969
1970        nt = netdev_priv(dev);
1971
1972        if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1973                int err = ip6_tnl_encap_setup(nt, &ipencap);
1974
1975                if (err < 0)
1976                        return err;
1977        }
1978
1979        if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1980                eth_hw_addr_random(dev);
1981
1982        nt->dev = dev;
1983        nt->net = dev_net(dev);
1984
1985        err = register_netdevice(dev);
1986        if (err)
1987                goto out;
1988
1989        if (tb[IFLA_MTU])
1990                ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1991
1992out:
1993        return err;
1994}
1995
1996static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1997                          struct nlattr *tb[], struct nlattr *data[],
1998                          struct netlink_ext_ack *extack)
1999{
2000        struct ip6_tnl *nt = netdev_priv(dev);
2001        struct net *net = dev_net(dev);
2002        struct ip6gre_net *ign;
2003        int err;
2004
2005        ip6gre_netlink_parms(data, &nt->parms);
2006        ign = net_generic(net, ip6gre_net_id);
2007
2008        if (nt->parms.collect_md) {
2009                if (rtnl_dereference(ign->collect_md_tun))
2010                        return -EEXIST;
2011        } else {
2012                if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2013                        return -EEXIST;
2014        }
2015
2016        err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2017        if (!err) {
2018                ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2019                ip6gre_tunnel_link_md(ign, nt);
2020                ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2021        }
2022        return err;
2023}
2024
2025static struct ip6_tnl *
2026ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2027                         struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2028                         struct netlink_ext_ack *extack)
2029{
2030        struct ip6_tnl *t, *nt = netdev_priv(dev);
2031        struct net *net = nt->net;
2032        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2033        struct ip_tunnel_encap ipencap;
2034
2035        if (dev == ign->fb_tunnel_dev)
2036                return ERR_PTR(-EINVAL);
2037
2038        if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2039                int err = ip6_tnl_encap_setup(nt, &ipencap);
2040
2041                if (err < 0)
2042                        return ERR_PTR(err);
2043        }
2044
2045        ip6gre_netlink_parms(data, p_p);
2046
2047        t = ip6gre_tunnel_locate(net, p_p, 0);
2048
2049        if (t) {
2050                if (t->dev != dev)
2051                        return ERR_PTR(-EEXIST);
2052        } else {
2053                t = nt;
2054        }
2055
2056        return t;
2057}
2058
2059static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2060                             struct nlattr *data[],
2061                             struct netlink_ext_ack *extack)
2062{
2063        struct ip6_tnl *t = netdev_priv(dev);
2064        struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2065        struct __ip6_tnl_parm p;
2066
2067        t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2068        if (IS_ERR(t))
2069                return PTR_ERR(t);
2070
2071        ip6gre_tunnel_unlink_md(ign, t);
2072        ip6gre_tunnel_unlink(ign, t);
2073        ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2074        ip6gre_tunnel_link_md(ign, t);
2075        ip6gre_tunnel_link(ign, t);
2076        return 0;
2077}
2078
2079static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2080{
2081        struct net *net = dev_net(dev);
2082        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2083
2084        if (dev != ign->fb_tunnel_dev)
2085                unregister_netdevice_queue(dev, head);
2086}
2087
2088static size_t ip6gre_get_size(const struct net_device *dev)
2089{
2090        return
2091                /* IFLA_GRE_LINK */
2092                nla_total_size(4) +
2093                /* IFLA_GRE_IFLAGS */
2094                nla_total_size(2) +
2095                /* IFLA_GRE_OFLAGS */
2096                nla_total_size(2) +
2097                /* IFLA_GRE_IKEY */
2098                nla_total_size(4) +
2099                /* IFLA_GRE_OKEY */
2100                nla_total_size(4) +
2101                /* IFLA_GRE_LOCAL */
2102                nla_total_size(sizeof(struct in6_addr)) +
2103                /* IFLA_GRE_REMOTE */
2104                nla_total_size(sizeof(struct in6_addr)) +
2105                /* IFLA_GRE_TTL */
2106                nla_total_size(1) +
2107                /* IFLA_GRE_ENCAP_LIMIT */
2108                nla_total_size(1) +
2109                /* IFLA_GRE_FLOWINFO */
2110                nla_total_size(4) +
2111                /* IFLA_GRE_FLAGS */
2112                nla_total_size(4) +
2113                /* IFLA_GRE_ENCAP_TYPE */
2114                nla_total_size(2) +
2115                /* IFLA_GRE_ENCAP_FLAGS */
2116                nla_total_size(2) +
2117                /* IFLA_GRE_ENCAP_SPORT */
2118                nla_total_size(2) +
2119                /* IFLA_GRE_ENCAP_DPORT */
2120                nla_total_size(2) +
2121                /* IFLA_GRE_COLLECT_METADATA */
2122                nla_total_size(0) +
2123                /* IFLA_GRE_FWMARK */
2124                nla_total_size(4) +
2125                /* IFLA_GRE_ERSPAN_INDEX */
2126                nla_total_size(4) +
2127                0;
2128}
2129
2130static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2131{
2132        struct ip6_tnl *t = netdev_priv(dev);
2133        struct __ip6_tnl_parm *p = &t->parms;
2134        __be16 o_flags = p->o_flags;
2135
2136        if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2137                if (!p->collect_md)
2138                        o_flags |= TUNNEL_KEY;
2139
2140                if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2141                        goto nla_put_failure;
2142
2143                if (p->erspan_ver == 1) {
2144                        if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2145                                goto nla_put_failure;
2146                } else {
2147                        if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2148                                goto nla_put_failure;
2149                        if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2150                                goto nla_put_failure;
2151                }
2152        }
2153
2154        if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2155            nla_put_be16(skb, IFLA_GRE_IFLAGS,
2156                         gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2157            nla_put_be16(skb, IFLA_GRE_OFLAGS,
2158                         gre_tnl_flags_to_gre_flags(o_flags)) ||
2159            nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2160            nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2161            nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2162            nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2163            nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2164            nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2165            nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2166            nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2167            nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2168                goto nla_put_failure;
2169
2170        if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2171                        t->encap.type) ||
2172            nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2173                         t->encap.sport) ||
2174            nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2175                         t->encap.dport) ||
2176            nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2177                        t->encap.flags))
2178                goto nla_put_failure;
2179
2180        if (p->collect_md) {
2181                if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2182                        goto nla_put_failure;
2183        }
2184
2185        return 0;
2186
2187nla_put_failure:
2188        return -EMSGSIZE;
2189}
2190
2191static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2192        [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2193        [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2194        [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2195        [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2196        [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2197        [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2198        [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2199        [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2200        [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2201        [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2202        [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2203        [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2204        [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2205        [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2206        [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2207        [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2208        [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2209        [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2210        [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2211        [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2212        [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2213};
2214
2215static void ip6erspan_tap_setup(struct net_device *dev)
2216{
2217        ether_setup(dev);
2218
2219        dev->max_mtu = 0;
2220        dev->netdev_ops = &ip6erspan_netdev_ops;
2221        dev->needs_free_netdev = true;
2222        dev->priv_destructor = ip6gre_dev_free;
2223
2224        dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2225        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2226        netif_keep_dst(dev);
2227}
2228
2229static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2230                             struct nlattr *tb[], struct nlattr *data[],
2231                             struct netlink_ext_ack *extack)
2232{
2233        struct ip6_tnl *nt = netdev_priv(dev);
2234        struct net *net = dev_net(dev);
2235        struct ip6gre_net *ign;
2236        int err;
2237
2238        ip6gre_netlink_parms(data, &nt->parms);
2239        ip6erspan_set_version(data, &nt->parms);
2240        ign = net_generic(net, ip6gre_net_id);
2241
2242        if (nt->parms.collect_md) {
2243                if (rtnl_dereference(ign->collect_md_tun_erspan))
2244                        return -EEXIST;
2245        } else {
2246                if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2247                        return -EEXIST;
2248        }
2249
2250        err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2251        if (!err) {
2252                ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2253                ip6erspan_tunnel_link_md(ign, nt);
2254                ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2255        }
2256        return err;
2257}
2258
2259static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2260{
2261        ip6gre_tnl_link_config_common(t);
2262        ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2263}
2264
2265static int ip6erspan_tnl_change(struct ip6_tnl *t,
2266                                const struct __ip6_tnl_parm *p, int set_mtu)
2267{
2268        ip6gre_tnl_copy_tnl_parm(t, p);
2269        ip6erspan_tnl_link_config(t, set_mtu);
2270        return 0;
2271}
2272
2273static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2274                                struct nlattr *data[],
2275                                struct netlink_ext_ack *extack)
2276{
2277        struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2278        struct __ip6_tnl_parm p;
2279        struct ip6_tnl *t;
2280
2281        t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2282        if (IS_ERR(t))
2283                return PTR_ERR(t);
2284
2285        ip6erspan_set_version(data, &p);
2286        ip6gre_tunnel_unlink_md(ign, t);
2287        ip6gre_tunnel_unlink(ign, t);
2288        ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2289        ip6erspan_tunnel_link_md(ign, t);
2290        ip6gre_tunnel_link(ign, t);
2291        return 0;
2292}
2293
2294static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2295        .kind           = "ip6gre",
2296        .maxtype        = IFLA_GRE_MAX,
2297        .policy         = ip6gre_policy,
2298        .priv_size      = sizeof(struct ip6_tnl),
2299        .setup          = ip6gre_tunnel_setup,
2300        .validate       = ip6gre_tunnel_validate,
2301        .newlink        = ip6gre_newlink,
2302        .changelink     = ip6gre_changelink,
2303        .dellink        = ip6gre_dellink,
2304        .get_size       = ip6gre_get_size,
2305        .fill_info      = ip6gre_fill_info,
2306        .get_link_net   = ip6_tnl_get_link_net,
2307};
2308
2309static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2310        .kind           = "ip6gretap",
2311        .maxtype        = IFLA_GRE_MAX,
2312        .policy         = ip6gre_policy,
2313        .priv_size      = sizeof(struct ip6_tnl),
2314        .setup          = ip6gre_tap_setup,
2315        .validate       = ip6gre_tap_validate,
2316        .newlink        = ip6gre_newlink,
2317        .changelink     = ip6gre_changelink,
2318        .get_size       = ip6gre_get_size,
2319        .fill_info      = ip6gre_fill_info,
2320        .get_link_net   = ip6_tnl_get_link_net,
2321};
2322
2323static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2324        .kind           = "ip6erspan",
2325        .maxtype        = IFLA_GRE_MAX,
2326        .policy         = ip6gre_policy,
2327        .priv_size      = sizeof(struct ip6_tnl),
2328        .setup          = ip6erspan_tap_setup,
2329        .validate       = ip6erspan_tap_validate,
2330        .newlink        = ip6erspan_newlink,
2331        .changelink     = ip6erspan_changelink,
2332        .get_size       = ip6gre_get_size,
2333        .fill_info      = ip6gre_fill_info,
2334        .get_link_net   = ip6_tnl_get_link_net,
2335};
2336
2337/*
2338 *      And now the modules code and kernel interface.
2339 */
2340
2341static int __init ip6gre_init(void)
2342{
2343        int err;
2344
2345        pr_info("GRE over IPv6 tunneling driver\n");
2346
2347        err = register_pernet_device(&ip6gre_net_ops);
2348        if (err < 0)
2349                return err;
2350
2351        err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2352        if (err < 0) {
2353                pr_info("%s: can't add protocol\n", __func__);
2354                goto add_proto_failed;
2355        }
2356
2357        err = rtnl_link_register(&ip6gre_link_ops);
2358        if (err < 0)
2359                goto rtnl_link_failed;
2360
2361        err = rtnl_link_register(&ip6gre_tap_ops);
2362        if (err < 0)
2363                goto tap_ops_failed;
2364
2365        err = rtnl_link_register(&ip6erspan_tap_ops);
2366        if (err < 0)
2367                goto erspan_link_failed;
2368
2369out:
2370        return err;
2371
2372erspan_link_failed:
2373        rtnl_link_unregister(&ip6gre_tap_ops);
2374tap_ops_failed:
2375        rtnl_link_unregister(&ip6gre_link_ops);
2376rtnl_link_failed:
2377        inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2378add_proto_failed:
2379        unregister_pernet_device(&ip6gre_net_ops);
2380        goto out;
2381}
2382
2383static void __exit ip6gre_fini(void)
2384{
2385        rtnl_link_unregister(&ip6gre_tap_ops);
2386        rtnl_link_unregister(&ip6gre_link_ops);
2387        rtnl_link_unregister(&ip6erspan_tap_ops);
2388        inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2389        unregister_pernet_device(&ip6gre_net_ops);
2390}
2391
2392module_init(ip6gre_init);
2393module_exit(ip6gre_fini);
2394MODULE_LICENSE("GPL");
2395MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2396MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2397MODULE_ALIAS_RTNL_LINK("ip6gre");
2398MODULE_ALIAS_RTNL_LINK("ip6gretap");
2399MODULE_ALIAS_RTNL_LINK("ip6erspan");
2400MODULE_ALIAS_NETDEV("ip6gre0");
2401