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        return iptunnel_handle_offloads(skb,
 633                                        csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
 634}
 635
 636static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
 637                                     struct net_device *dev,
 638                                     struct flowi6 *fl6, __u8 *dsfield,
 639                                     int *encap_limit)
 640{
 641        const struct iphdr *iph = ip_hdr(skb);
 642        struct ip6_tnl *t = netdev_priv(dev);
 643
 644        if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 645                *encap_limit = t->parms.encap_limit;
 646
 647        memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 648
 649        if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 650                *dsfield = ipv4_get_dsfield(iph);
 651        else
 652                *dsfield = ip6_tclass(t->parms.flowinfo);
 653
 654        if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 655                fl6->flowi6_mark = skb->mark;
 656        else
 657                fl6->flowi6_mark = t->parms.fwmark;
 658
 659        fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 660}
 661
 662static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
 663                                    struct net_device *dev,
 664                                    struct flowi6 *fl6, __u8 *dsfield,
 665                                    int *encap_limit)
 666{
 667        struct ipv6hdr *ipv6h;
 668        struct ip6_tnl *t = netdev_priv(dev);
 669        __u16 offset;
 670
 671        offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
 672        /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
 673        ipv6h = ipv6_hdr(skb);
 674
 675        if (offset > 0) {
 676                struct ipv6_tlv_tnl_enc_lim *tel;
 677
 678                tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 679                if (tel->encap_limit == 0) {
 680                        icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
 681                                        ICMPV6_HDR_FIELD, offset + 2);
 682                        return -1;
 683                }
 684                *encap_limit = tel->encap_limit - 1;
 685        } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
 686                *encap_limit = t->parms.encap_limit;
 687        }
 688
 689        memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
 690
 691        if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
 692                *dsfield = ipv6_get_dsfield(ipv6h);
 693        else
 694                *dsfield = ip6_tclass(t->parms.flowinfo);
 695
 696        if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
 697                fl6->flowlabel |= ip6_flowlabel(ipv6h);
 698
 699        if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
 700                fl6->flowi6_mark = skb->mark;
 701        else
 702                fl6->flowi6_mark = t->parms.fwmark;
 703
 704        fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 705
 706        return 0;
 707}
 708
 709static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
 710{
 711        struct ip_tunnel_info *tun_info;
 712
 713        tun_info = skb_tunnel_info(skb);
 714        if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
 715                return ERR_PTR(-EINVAL);
 716
 717        return tun_info;
 718}
 719
 720static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
 721                               struct net_device *dev, __u8 dsfield,
 722                               struct flowi6 *fl6, int encap_limit,
 723                               __u32 *pmtu, __be16 proto)
 724{
 725        struct ip6_tnl *tunnel = netdev_priv(dev);
 726        __be16 protocol;
 727
 728        if (dev->type == ARPHRD_ETHER)
 729                IPCB(skb)->flags = 0;
 730
 731        if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
 732                fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
 733        else
 734                fl6->daddr = tunnel->parms.raddr;
 735
 736        if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
 737                return -ENOMEM;
 738
 739        /* Push GRE header. */
 740        protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
 741
 742        if (tunnel->parms.collect_md) {
 743                struct ip_tunnel_info *tun_info;
 744                const struct ip_tunnel_key *key;
 745                __be16 flags;
 746
 747                tun_info = skb_tunnel_info_txcheck(skb);
 748                if (IS_ERR(tun_info) ||
 749                    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
 750                        return -EINVAL;
 751
 752                key = &tun_info->key;
 753                memset(fl6, 0, sizeof(*fl6));
 754                fl6->flowi6_proto = IPPROTO_GRE;
 755                fl6->daddr = key->u.ipv6.dst;
 756                fl6->flowlabel = key->label;
 757                fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 758
 759                dsfield = key->tos;
 760                flags = key->tun_flags &
 761                        (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
 762                tunnel->tun_hlen = gre_calc_hlen(flags);
 763
 764                gre_build_header(skb, tunnel->tun_hlen,
 765                                 flags, protocol,
 766                                 tunnel_id_to_key32(tun_info->key.tun_id),
 767                                 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
 768                                                      : 0);
 769
 770        } else {
 771                if (tunnel->parms.o_flags & TUNNEL_SEQ)
 772                        tunnel->o_seqno++;
 773
 774                gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
 775                                 protocol, tunnel->parms.o_key,
 776                                 htonl(tunnel->o_seqno));
 777        }
 778
 779        return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
 780                            NEXTHDR_GRE);
 781}
 782
 783static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
 784{
 785        struct ip6_tnl *t = netdev_priv(dev);
 786        int encap_limit = -1;
 787        struct flowi6 fl6;
 788        __u8 dsfield = 0;
 789        __u32 mtu;
 790        int err;
 791
 792        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 793
 794        if (!t->parms.collect_md)
 795                prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
 796                                         &dsfield, &encap_limit);
 797
 798        err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 799        if (err)
 800                return -1;
 801
 802        err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
 803                          skb->protocol);
 804        if (err != 0) {
 805                /* XXX: send ICMP error even if DF is not set. */
 806                if (err == -EMSGSIZE)
 807                        icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 808                                      htonl(mtu));
 809                return -1;
 810        }
 811
 812        return 0;
 813}
 814
 815static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
 816{
 817        struct ip6_tnl *t = netdev_priv(dev);
 818        struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 819        int encap_limit = -1;
 820        struct flowi6 fl6;
 821        __u8 dsfield = 0;
 822        __u32 mtu;
 823        int err;
 824
 825        if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
 826                return -1;
 827
 828        if (!t->parms.collect_md &&
 829            prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
 830                return -1;
 831
 832        if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
 833                return -1;
 834
 835        err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
 836                          &mtu, skb->protocol);
 837        if (err != 0) {
 838                if (err == -EMSGSIZE)
 839                        icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 840                return -1;
 841        }
 842
 843        return 0;
 844}
 845
 846/**
 847 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
 848 *   @t: the outgoing tunnel device
 849 *   @hdr: IPv6 header from the incoming packet
 850 *
 851 * Description:
 852 *   Avoid trivial tunneling loop by checking that tunnel exit-point
 853 *   doesn't match source of incoming packet.
 854 *
 855 * Return:
 856 *   1 if conflict,
 857 *   0 else
 858 **/
 859
 860static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
 861        const struct ipv6hdr *hdr)
 862{
 863        return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 864}
 865
 866static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
 867{
 868        struct ip6_tnl *t = netdev_priv(dev);
 869        int encap_limit = -1;
 870        struct flowi6 fl6;
 871        __u32 mtu;
 872        int err;
 873
 874        if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 875                encap_limit = t->parms.encap_limit;
 876
 877        if (!t->parms.collect_md)
 878                memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
 879
 880        err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
 881        if (err)
 882                return err;
 883
 884        err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
 885
 886        return err;
 887}
 888
 889static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
 890        struct net_device *dev)
 891{
 892        struct ip6_tnl *t = netdev_priv(dev);
 893        struct net_device_stats *stats = &t->dev->stats;
 894        int ret;
 895
 896        if (!pskb_inet_may_pull(skb))
 897                goto tx_err;
 898
 899        if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 900                goto tx_err;
 901
 902        switch (skb->protocol) {
 903        case htons(ETH_P_IP):
 904                ret = ip6gre_xmit_ipv4(skb, dev);
 905                break;
 906        case htons(ETH_P_IPV6):
 907                ret = ip6gre_xmit_ipv6(skb, dev);
 908                break;
 909        default:
 910                ret = ip6gre_xmit_other(skb, dev);
 911                break;
 912        }
 913
 914        if (ret < 0)
 915                goto tx_err;
 916
 917        return NETDEV_TX_OK;
 918
 919tx_err:
 920        if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
 921                stats->tx_errors++;
 922        stats->tx_dropped++;
 923        kfree_skb(skb);
 924        return NETDEV_TX_OK;
 925}
 926
 927static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
 928                                         struct net_device *dev)
 929{
 930        struct ip_tunnel_info *tun_info = NULL;
 931        struct ip6_tnl *t = netdev_priv(dev);
 932        struct dst_entry *dst = skb_dst(skb);
 933        struct net_device_stats *stats;
 934        bool truncate = false;
 935        int encap_limit = -1;
 936        __u8 dsfield = false;
 937        struct flowi6 fl6;
 938        int err = -EINVAL;
 939        __be16 proto;
 940        __u32 mtu;
 941        int nhoff;
 942        int thoff;
 943
 944        if (!pskb_inet_may_pull(skb))
 945                goto tx_err;
 946
 947        if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
 948                goto tx_err;
 949
 950        if (gre_handle_offloads(skb, false))
 951                goto tx_err;
 952
 953        if (skb->len > dev->mtu + dev->hard_header_len) {
 954                pskb_trim(skb, dev->mtu + dev->hard_header_len);
 955                truncate = true;
 956        }
 957
 958        nhoff = skb_network_header(skb) - skb_mac_header(skb);
 959        if (skb->protocol == htons(ETH_P_IP) &&
 960            (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
 961                truncate = true;
 962
 963        thoff = skb_transport_header(skb) - skb_mac_header(skb);
 964        if (skb->protocol == htons(ETH_P_IPV6) &&
 965            (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
 966                truncate = true;
 967
 968        if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
 969                goto tx_err;
 970
 971        t->parms.o_flags &= ~TUNNEL_KEY;
 972        IPCB(skb)->flags = 0;
 973
 974        /* For collect_md mode, derive fl6 from the tunnel key,
 975         * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
 976         */
 977        if (t->parms.collect_md) {
 978                const struct ip_tunnel_key *key;
 979                struct erspan_metadata *md;
 980                __be32 tun_id;
 981
 982                tun_info = skb_tunnel_info_txcheck(skb);
 983                if (IS_ERR(tun_info) ||
 984                    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
 985                        goto tx_err;
 986
 987                key = &tun_info->key;
 988                memset(&fl6, 0, sizeof(fl6));
 989                fl6.flowi6_proto = IPPROTO_GRE;
 990                fl6.daddr = key->u.ipv6.dst;
 991                fl6.flowlabel = key->label;
 992                fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
 993
 994                dsfield = key->tos;
 995                if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
 996                        goto tx_err;
 997                if (tun_info->options_len < sizeof(*md))
 998                        goto tx_err;
 999                md = ip_tunnel_info_opts(tun_info);
1000
1001                tun_id = tunnel_id_to_key32(key->tun_id);
1002                if (md->version == 1) {
1003                        erspan_build_header(skb,
1004                                            ntohl(tun_id),
1005                                            ntohl(md->u.index), truncate,
1006                                            false);
1007                } else if (md->version == 2) {
1008                        erspan_build_header_v2(skb,
1009                                               ntohl(tun_id),
1010                                               md->u.md2.dir,
1011                                               get_hwid(&md->u.md2),
1012                                               truncate, false);
1013                } else {
1014                        goto tx_err;
1015                }
1016        } else {
1017                switch (skb->protocol) {
1018                case htons(ETH_P_IP):
1019                        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1020                        prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1021                                                 &dsfield, &encap_limit);
1022                        break;
1023                case htons(ETH_P_IPV6):
1024                        if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1025                                goto tx_err;
1026                        if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1027                                                     &dsfield, &encap_limit))
1028                                goto tx_err;
1029                        break;
1030                default:
1031                        memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1032                        break;
1033                }
1034
1035                if (t->parms.erspan_ver == 1)
1036                        erspan_build_header(skb, ntohl(t->parms.o_key),
1037                                            t->parms.index,
1038                                            truncate, false);
1039                else if (t->parms.erspan_ver == 2)
1040                        erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1041                                               t->parms.dir,
1042                                               t->parms.hwid,
1043                                               truncate, false);
1044                else
1045                        goto tx_err;
1046
1047                fl6.daddr = t->parms.raddr;
1048        }
1049
1050        /* Push GRE header. */
1051        proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1052                                           : htons(ETH_P_ERSPAN2);
1053        gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++));
1054
1055        /* TooBig packet may have updated dst->dev's mtu */
1056        if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1057                dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1058
1059        err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1060                           NEXTHDR_GRE);
1061        if (err != 0) {
1062                /* XXX: send ICMP error even if DF is not set. */
1063                if (err == -EMSGSIZE) {
1064                        if (skb->protocol == htons(ETH_P_IP))
1065                                icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1066                                              ICMP_FRAG_NEEDED, htonl(mtu));
1067                        else
1068                                icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1069                }
1070
1071                goto tx_err;
1072        }
1073        return NETDEV_TX_OK;
1074
1075tx_err:
1076        stats = &t->dev->stats;
1077        if (!IS_ERR(tun_info))
1078                stats->tx_errors++;
1079        stats->tx_dropped++;
1080        kfree_skb(skb);
1081        return NETDEV_TX_OK;
1082}
1083
1084static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1085{
1086        struct net_device *dev = t->dev;
1087        struct __ip6_tnl_parm *p = &t->parms;
1088        struct flowi6 *fl6 = &t->fl.u.ip6;
1089
1090        if (dev->type != ARPHRD_ETHER) {
1091                memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1092                memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1093        }
1094
1095        /* Set up flowi template */
1096        fl6->saddr = p->laddr;
1097        fl6->daddr = p->raddr;
1098        fl6->flowi6_oif = p->link;
1099        fl6->flowlabel = 0;
1100        fl6->flowi6_proto = IPPROTO_GRE;
1101
1102        if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1103                fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1104        if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1105                fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1106
1107        p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1108        p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1109
1110        if (p->flags&IP6_TNL_F_CAP_XMIT &&
1111                        p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1112                dev->flags |= IFF_POINTOPOINT;
1113        else
1114                dev->flags &= ~IFF_POINTOPOINT;
1115}
1116
1117static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1118                                         int t_hlen)
1119{
1120        const struct __ip6_tnl_parm *p = &t->parms;
1121        struct net_device *dev = t->dev;
1122
1123        if (p->flags & IP6_TNL_F_CAP_XMIT) {
1124                int strict = (ipv6_addr_type(&p->raddr) &
1125                              (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1126
1127                struct rt6_info *rt = rt6_lookup(t->net,
1128                                                 &p->raddr, &p->laddr,
1129                                                 p->link, NULL, strict);
1130
1131                if (!rt)
1132                        return;
1133
1134                if (rt->dst.dev) {
1135                        unsigned short dst_len = rt->dst.dev->hard_header_len +
1136                                                 t_hlen;
1137
1138                        if (t->dev->header_ops)
1139                                dev->hard_header_len = dst_len;
1140                        else
1141                                dev->needed_headroom = dst_len;
1142
1143                        if (set_mtu) {
1144                                dev->mtu = rt->dst.dev->mtu - t_hlen;
1145                                if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1146                                        dev->mtu -= 8;
1147                                if (dev->type == ARPHRD_ETHER)
1148                                        dev->mtu -= ETH_HLEN;
1149
1150                                if (dev->mtu < IPV6_MIN_MTU)
1151                                        dev->mtu = IPV6_MIN_MTU;
1152                        }
1153                }
1154                ip6_rt_put(rt);
1155        }
1156}
1157
1158static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1159{
1160        int t_hlen;
1161
1162        tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1163        tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1164
1165        t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1166
1167        if (tunnel->dev->header_ops)
1168                tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1169        else
1170                tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1171
1172        return t_hlen;
1173}
1174
1175static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1176{
1177        ip6gre_tnl_link_config_common(t);
1178        ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1179}
1180
1181static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1182                                     const struct __ip6_tnl_parm *p)
1183{
1184        t->parms.laddr = p->laddr;
1185        t->parms.raddr = p->raddr;
1186        t->parms.flags = p->flags;
1187        t->parms.hop_limit = p->hop_limit;
1188        t->parms.encap_limit = p->encap_limit;
1189        t->parms.flowinfo = p->flowinfo;
1190        t->parms.link = p->link;
1191        t->parms.proto = p->proto;
1192        t->parms.i_key = p->i_key;
1193        t->parms.o_key = p->o_key;
1194        t->parms.i_flags = p->i_flags;
1195        t->parms.o_flags = p->o_flags;
1196        t->parms.fwmark = p->fwmark;
1197        t->parms.erspan_ver = p->erspan_ver;
1198        t->parms.index = p->index;
1199        t->parms.dir = p->dir;
1200        t->parms.hwid = p->hwid;
1201        dst_cache_reset(&t->dst_cache);
1202}
1203
1204static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1205                             int set_mtu)
1206{
1207        ip6gre_tnl_copy_tnl_parm(t, p);
1208        ip6gre_tnl_link_config(t, set_mtu);
1209        return 0;
1210}
1211
1212static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1213        const struct ip6_tnl_parm2 *u)
1214{
1215        p->laddr = u->laddr;
1216        p->raddr = u->raddr;
1217        p->flags = u->flags;
1218        p->hop_limit = u->hop_limit;
1219        p->encap_limit = u->encap_limit;
1220        p->flowinfo = u->flowinfo;
1221        p->link = u->link;
1222        p->i_key = u->i_key;
1223        p->o_key = u->o_key;
1224        p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1225        p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1226        memcpy(p->name, u->name, sizeof(u->name));
1227}
1228
1229static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1230        const struct __ip6_tnl_parm *p)
1231{
1232        u->proto = IPPROTO_GRE;
1233        u->laddr = p->laddr;
1234        u->raddr = p->raddr;
1235        u->flags = p->flags;
1236        u->hop_limit = p->hop_limit;
1237        u->encap_limit = p->encap_limit;
1238        u->flowinfo = p->flowinfo;
1239        u->link = p->link;
1240        u->i_key = p->i_key;
1241        u->o_key = p->o_key;
1242        u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1243        u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1244        memcpy(u->name, p->name, sizeof(u->name));
1245}
1246
1247static int ip6gre_tunnel_siocdevprivate(struct net_device *dev,
1248                                        struct ifreq *ifr, void __user *data,
1249                                        int cmd)
1250{
1251        int err = 0;
1252        struct ip6_tnl_parm2 p;
1253        struct __ip6_tnl_parm p1;
1254        struct ip6_tnl *t = netdev_priv(dev);
1255        struct net *net = t->net;
1256        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1257
1258        memset(&p1, 0, sizeof(p1));
1259
1260        switch (cmd) {
1261        case SIOCGETTUNNEL:
1262                if (dev == ign->fb_tunnel_dev) {
1263                        if (copy_from_user(&p, data, sizeof(p))) {
1264                                err = -EFAULT;
1265                                break;
1266                        }
1267                        ip6gre_tnl_parm_from_user(&p1, &p);
1268                        t = ip6gre_tunnel_locate(net, &p1, 0);
1269                        if (!t)
1270                                t = netdev_priv(dev);
1271                }
1272                memset(&p, 0, sizeof(p));
1273                ip6gre_tnl_parm_to_user(&p, &t->parms);
1274                if (copy_to_user(data, &p, sizeof(p)))
1275                        err = -EFAULT;
1276                break;
1277
1278        case SIOCADDTUNNEL:
1279        case SIOCCHGTUNNEL:
1280                err = -EPERM;
1281                if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1282                        goto done;
1283
1284                err = -EFAULT;
1285                if (copy_from_user(&p, data, sizeof(p)))
1286                        goto done;
1287
1288                err = -EINVAL;
1289                if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1290                        goto done;
1291
1292                if (!(p.i_flags&GRE_KEY))
1293                        p.i_key = 0;
1294                if (!(p.o_flags&GRE_KEY))
1295                        p.o_key = 0;
1296
1297                ip6gre_tnl_parm_from_user(&p1, &p);
1298                t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1299
1300                if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1301                        if (t) {
1302                                if (t->dev != dev) {
1303                                        err = -EEXIST;
1304                                        break;
1305                                }
1306                        } else {
1307                                t = netdev_priv(dev);
1308
1309                                ip6gre_tunnel_unlink(ign, t);
1310                                synchronize_net();
1311                                ip6gre_tnl_change(t, &p1, 1);
1312                                ip6gre_tunnel_link(ign, t);
1313                                netdev_state_change(dev);
1314                        }
1315                }
1316
1317                if (t) {
1318                        err = 0;
1319
1320                        memset(&p, 0, sizeof(p));
1321                        ip6gre_tnl_parm_to_user(&p, &t->parms);
1322                        if (copy_to_user(data, &p, sizeof(p)))
1323                                err = -EFAULT;
1324                } else
1325                        err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1326                break;
1327
1328        case SIOCDELTUNNEL:
1329                err = -EPERM;
1330                if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1331                        goto done;
1332
1333                if (dev == ign->fb_tunnel_dev) {
1334                        err = -EFAULT;
1335                        if (copy_from_user(&p, data, sizeof(p)))
1336                                goto done;
1337                        err = -ENOENT;
1338                        ip6gre_tnl_parm_from_user(&p1, &p);
1339                        t = ip6gre_tunnel_locate(net, &p1, 0);
1340                        if (!t)
1341                                goto done;
1342                        err = -EPERM;
1343                        if (t == netdev_priv(ign->fb_tunnel_dev))
1344                                goto done;
1345                        dev = t->dev;
1346                }
1347                unregister_netdevice(dev);
1348                err = 0;
1349                break;
1350
1351        default:
1352                err = -EINVAL;
1353        }
1354
1355done:
1356        return err;
1357}
1358
1359static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1360                         unsigned short type, const void *daddr,
1361                         const void *saddr, unsigned int len)
1362{
1363        struct ip6_tnl *t = netdev_priv(dev);
1364        struct ipv6hdr *ipv6h;
1365        __be16 *p;
1366
1367        ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1368        ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1369                                                  t->fl.u.ip6.flowlabel,
1370                                                  true, &t->fl.u.ip6));
1371        ipv6h->hop_limit = t->parms.hop_limit;
1372        ipv6h->nexthdr = NEXTHDR_GRE;
1373        ipv6h->saddr = t->parms.laddr;
1374        ipv6h->daddr = t->parms.raddr;
1375
1376        p = (__be16 *)(ipv6h + 1);
1377        p[0] = t->parms.o_flags;
1378        p[1] = htons(type);
1379
1380        /*
1381         *      Set the source hardware address.
1382         */
1383
1384        if (saddr)
1385                memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1386        if (daddr)
1387                memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1388        if (!ipv6_addr_any(&ipv6h->daddr))
1389                return t->hlen;
1390
1391        return -t->hlen;
1392}
1393
1394static const struct header_ops ip6gre_header_ops = {
1395        .create = ip6gre_header,
1396};
1397
1398static const struct net_device_ops ip6gre_netdev_ops = {
1399        .ndo_init               = ip6gre_tunnel_init,
1400        .ndo_uninit             = ip6gre_tunnel_uninit,
1401        .ndo_start_xmit         = ip6gre_tunnel_xmit,
1402        .ndo_siocdevprivate     = ip6gre_tunnel_siocdevprivate,
1403        .ndo_change_mtu         = ip6_tnl_change_mtu,
1404        .ndo_get_stats64        = dev_get_tstats64,
1405        .ndo_get_iflink         = ip6_tnl_get_iflink,
1406};
1407
1408static void ip6gre_dev_free(struct net_device *dev)
1409{
1410        struct ip6_tnl *t = netdev_priv(dev);
1411
1412        gro_cells_destroy(&t->gro_cells);
1413        dst_cache_destroy(&t->dst_cache);
1414        free_percpu(dev->tstats);
1415}
1416
1417static void ip6gre_tunnel_setup(struct net_device *dev)
1418{
1419        dev->netdev_ops = &ip6gre_netdev_ops;
1420        dev->needs_free_netdev = true;
1421        dev->priv_destructor = ip6gre_dev_free;
1422
1423        dev->type = ARPHRD_IP6GRE;
1424
1425        dev->flags |= IFF_NOARP;
1426        dev->addr_len = sizeof(struct in6_addr);
1427        netif_keep_dst(dev);
1428        /* This perm addr will be used as interface identifier by IPv6 */
1429        dev->addr_assign_type = NET_ADDR_RANDOM;
1430        eth_random_addr(dev->perm_addr);
1431}
1432
1433#define GRE6_FEATURES (NETIF_F_SG |             \
1434                       NETIF_F_FRAGLIST |       \
1435                       NETIF_F_HIGHDMA |        \
1436                       NETIF_F_HW_CSUM)
1437
1438static void ip6gre_tnl_init_features(struct net_device *dev)
1439{
1440        struct ip6_tnl *nt = netdev_priv(dev);
1441
1442        dev->features           |= GRE6_FEATURES;
1443        dev->hw_features        |= GRE6_FEATURES;
1444
1445        if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1446                /* TCP offload with GRE SEQ is not supported, nor
1447                 * can we support 2 levels of outer headers requiring
1448                 * an update.
1449                 */
1450                if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1451                    nt->encap.type == TUNNEL_ENCAP_NONE) {
1452                        dev->features    |= NETIF_F_GSO_SOFTWARE;
1453                        dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1454                }
1455
1456                /* Can use a lockless transmit, unless we generate
1457                 * output sequences
1458                 */
1459                dev->features |= NETIF_F_LLTX;
1460        }
1461}
1462
1463static int ip6gre_tunnel_init_common(struct net_device *dev)
1464{
1465        struct ip6_tnl *tunnel;
1466        int ret;
1467        int t_hlen;
1468
1469        tunnel = netdev_priv(dev);
1470
1471        tunnel->dev = dev;
1472        tunnel->net = dev_net(dev);
1473        strcpy(tunnel->parms.name, dev->name);
1474
1475        dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1476        if (!dev->tstats)
1477                return -ENOMEM;
1478
1479        ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1480        if (ret)
1481                goto cleanup_alloc_pcpu_stats;
1482
1483        ret = gro_cells_init(&tunnel->gro_cells, dev);
1484        if (ret)
1485                goto cleanup_dst_cache_init;
1486
1487        t_hlen = ip6gre_calc_hlen(tunnel);
1488        dev->mtu = ETH_DATA_LEN - t_hlen;
1489        if (dev->type == ARPHRD_ETHER)
1490                dev->mtu -= ETH_HLEN;
1491        if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1492                dev->mtu -= 8;
1493
1494        if (tunnel->parms.collect_md) {
1495                netif_keep_dst(dev);
1496        }
1497        ip6gre_tnl_init_features(dev);
1498
1499        dev_hold(dev);
1500        return 0;
1501
1502cleanup_dst_cache_init:
1503        dst_cache_destroy(&tunnel->dst_cache);
1504cleanup_alloc_pcpu_stats:
1505        free_percpu(dev->tstats);
1506        dev->tstats = NULL;
1507        return ret;
1508}
1509
1510static int ip6gre_tunnel_init(struct net_device *dev)
1511{
1512        struct ip6_tnl *tunnel;
1513        int ret;
1514
1515        ret = ip6gre_tunnel_init_common(dev);
1516        if (ret)
1517                return ret;
1518
1519        tunnel = netdev_priv(dev);
1520
1521        if (tunnel->parms.collect_md)
1522                return 0;
1523
1524        memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1525        memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1526
1527        if (ipv6_addr_any(&tunnel->parms.raddr))
1528                dev->header_ops = &ip6gre_header_ops;
1529
1530        return 0;
1531}
1532
1533static void ip6gre_fb_tunnel_init(struct net_device *dev)
1534{
1535        struct ip6_tnl *tunnel = netdev_priv(dev);
1536
1537        tunnel->dev = dev;
1538        tunnel->net = dev_net(dev);
1539        strcpy(tunnel->parms.name, dev->name);
1540
1541        tunnel->hlen            = sizeof(struct ipv6hdr) + 4;
1542}
1543
1544static struct inet6_protocol ip6gre_protocol __read_mostly = {
1545        .handler     = gre_rcv,
1546        .err_handler = ip6gre_err,
1547        .flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1548};
1549
1550static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1551{
1552        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1553        struct net_device *dev, *aux;
1554        int prio;
1555
1556        for_each_netdev_safe(net, dev, aux)
1557                if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1558                    dev->rtnl_link_ops == &ip6gre_tap_ops ||
1559                    dev->rtnl_link_ops == &ip6erspan_tap_ops)
1560                        unregister_netdevice_queue(dev, head);
1561
1562        for (prio = 0; prio < 4; prio++) {
1563                int h;
1564                for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1565                        struct ip6_tnl *t;
1566
1567                        t = rtnl_dereference(ign->tunnels[prio][h]);
1568
1569                        while (t) {
1570                                /* If dev is in the same netns, it has already
1571                                 * been added to the list by the previous loop.
1572                                 */
1573                                if (!net_eq(dev_net(t->dev), net))
1574                                        unregister_netdevice_queue(t->dev,
1575                                                                   head);
1576                                t = rtnl_dereference(t->next);
1577                        }
1578                }
1579        }
1580}
1581
1582static int __net_init ip6gre_init_net(struct net *net)
1583{
1584        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1585        struct net_device *ndev;
1586        int err;
1587
1588        if (!net_has_fallback_tunnels(net))
1589                return 0;
1590        ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1591                            NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1592        if (!ndev) {
1593                err = -ENOMEM;
1594                goto err_alloc_dev;
1595        }
1596        ign->fb_tunnel_dev = ndev;
1597        dev_net_set(ign->fb_tunnel_dev, net);
1598        /* FB netdevice is special: we have one, and only one per netns.
1599         * Allowing to move it to another netns is clearly unsafe.
1600         */
1601        ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1602
1603
1604        ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1605        ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1606
1607        err = register_netdev(ign->fb_tunnel_dev);
1608        if (err)
1609                goto err_reg_dev;
1610
1611        rcu_assign_pointer(ign->tunnels_wc[0],
1612                           netdev_priv(ign->fb_tunnel_dev));
1613        return 0;
1614
1615err_reg_dev:
1616        free_netdev(ndev);
1617err_alloc_dev:
1618        return err;
1619}
1620
1621static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1622{
1623        struct net *net;
1624        LIST_HEAD(list);
1625
1626        rtnl_lock();
1627        list_for_each_entry(net, net_list, exit_list)
1628                ip6gre_destroy_tunnels(net, &list);
1629        unregister_netdevice_many(&list);
1630        rtnl_unlock();
1631}
1632
1633static struct pernet_operations ip6gre_net_ops = {
1634        .init = ip6gre_init_net,
1635        .exit_batch = ip6gre_exit_batch_net,
1636        .id   = &ip6gre_net_id,
1637        .size = sizeof(struct ip6gre_net),
1638};
1639
1640static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1641                                  struct netlink_ext_ack *extack)
1642{
1643        __be16 flags;
1644
1645        if (!data)
1646                return 0;
1647
1648        flags = 0;
1649        if (data[IFLA_GRE_IFLAGS])
1650                flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1651        if (data[IFLA_GRE_OFLAGS])
1652                flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1653        if (flags & (GRE_VERSION|GRE_ROUTING))
1654                return -EINVAL;
1655
1656        return 0;
1657}
1658
1659static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1660                               struct netlink_ext_ack *extack)
1661{
1662        struct in6_addr daddr;
1663
1664        if (tb[IFLA_ADDRESS]) {
1665                if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1666                        return -EINVAL;
1667                if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1668                        return -EADDRNOTAVAIL;
1669        }
1670
1671        if (!data)
1672                goto out;
1673
1674        if (data[IFLA_GRE_REMOTE]) {
1675                daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1676                if (ipv6_addr_any(&daddr))
1677                        return -EINVAL;
1678        }
1679
1680out:
1681        return ip6gre_tunnel_validate(tb, data, extack);
1682}
1683
1684static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1685                                  struct netlink_ext_ack *extack)
1686{
1687        __be16 flags = 0;
1688        int ret, ver = 0;
1689
1690        if (!data)
1691                return 0;
1692
1693        ret = ip6gre_tap_validate(tb, data, extack);
1694        if (ret)
1695                return ret;
1696
1697        /* ERSPAN should only have GRE sequence and key flag */
1698        if (data[IFLA_GRE_OFLAGS])
1699                flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1700        if (data[IFLA_GRE_IFLAGS])
1701                flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1702        if (!data[IFLA_GRE_COLLECT_METADATA] &&
1703            flags != (GRE_SEQ | GRE_KEY))
1704                return -EINVAL;
1705
1706        /* ERSPAN Session ID only has 10-bit. Since we reuse
1707         * 32-bit key field as ID, check it's range.
1708         */
1709        if (data[IFLA_GRE_IKEY] &&
1710            (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1711                return -EINVAL;
1712
1713        if (data[IFLA_GRE_OKEY] &&
1714            (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1715                return -EINVAL;
1716
1717        if (data[IFLA_GRE_ERSPAN_VER]) {
1718                ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1719                if (ver != 1 && ver != 2)
1720                        return -EINVAL;
1721        }
1722
1723        if (ver == 1) {
1724                if (data[IFLA_GRE_ERSPAN_INDEX]) {
1725                        u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1726
1727                        if (index & ~INDEX_MASK)
1728                                return -EINVAL;
1729                }
1730        } else if (ver == 2) {
1731                if (data[IFLA_GRE_ERSPAN_DIR]) {
1732                        u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1733
1734                        if (dir & ~(DIR_MASK >> DIR_OFFSET))
1735                                return -EINVAL;
1736                }
1737
1738                if (data[IFLA_GRE_ERSPAN_HWID]) {
1739                        u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1740
1741                        if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1742                                return -EINVAL;
1743                }
1744        }
1745
1746        return 0;
1747}
1748
1749static void ip6erspan_set_version(struct nlattr *data[],
1750                                  struct __ip6_tnl_parm *parms)
1751{
1752        if (!data)
1753                return;
1754
1755        parms->erspan_ver = 1;
1756        if (data[IFLA_GRE_ERSPAN_VER])
1757                parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1758
1759        if (parms->erspan_ver == 1) {
1760                if (data[IFLA_GRE_ERSPAN_INDEX])
1761                        parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1762        } else if (parms->erspan_ver == 2) {
1763                if (data[IFLA_GRE_ERSPAN_DIR])
1764                        parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1765                if (data[IFLA_GRE_ERSPAN_HWID])
1766                        parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1767        }
1768}
1769
1770static void ip6gre_netlink_parms(struct nlattr *data[],
1771                                struct __ip6_tnl_parm *parms)
1772{
1773        memset(parms, 0, sizeof(*parms));
1774
1775        if (!data)
1776                return;
1777
1778        if (data[IFLA_GRE_LINK])
1779                parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1780
1781        if (data[IFLA_GRE_IFLAGS])
1782                parms->i_flags = gre_flags_to_tnl_flags(
1783                                nla_get_be16(data[IFLA_GRE_IFLAGS]));
1784
1785        if (data[IFLA_GRE_OFLAGS])
1786                parms->o_flags = gre_flags_to_tnl_flags(
1787                                nla_get_be16(data[IFLA_GRE_OFLAGS]));
1788
1789        if (data[IFLA_GRE_IKEY])
1790                parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1791
1792        if (data[IFLA_GRE_OKEY])
1793                parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1794
1795        if (data[IFLA_GRE_LOCAL])
1796                parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1797
1798        if (data[IFLA_GRE_REMOTE])
1799                parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1800
1801        if (data[IFLA_GRE_TTL])
1802                parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1803
1804        if (data[IFLA_GRE_ENCAP_LIMIT])
1805                parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1806
1807        if (data[IFLA_GRE_FLOWINFO])
1808                parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1809
1810        if (data[IFLA_GRE_FLAGS])
1811                parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1812
1813        if (data[IFLA_GRE_FWMARK])
1814                parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1815
1816        if (data[IFLA_GRE_COLLECT_METADATA])
1817                parms->collect_md = true;
1818}
1819
1820static int ip6gre_tap_init(struct net_device *dev)
1821{
1822        int ret;
1823
1824        ret = ip6gre_tunnel_init_common(dev);
1825        if (ret)
1826                return ret;
1827
1828        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1829
1830        return 0;
1831}
1832
1833static const struct net_device_ops ip6gre_tap_netdev_ops = {
1834        .ndo_init = ip6gre_tap_init,
1835        .ndo_uninit = ip6gre_tunnel_uninit,
1836        .ndo_start_xmit = ip6gre_tunnel_xmit,
1837        .ndo_set_mac_address = eth_mac_addr,
1838        .ndo_validate_addr = eth_validate_addr,
1839        .ndo_change_mtu = ip6_tnl_change_mtu,
1840        .ndo_get_stats64 = dev_get_tstats64,
1841        .ndo_get_iflink = ip6_tnl_get_iflink,
1842};
1843
1844static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1845{
1846        int t_hlen;
1847
1848        tunnel->tun_hlen = 8;
1849        tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1850                       erspan_hdr_len(tunnel->parms.erspan_ver);
1851
1852        t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1853        tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1854        return t_hlen;
1855}
1856
1857static int ip6erspan_tap_init(struct net_device *dev)
1858{
1859        struct ip6_tnl *tunnel;
1860        int t_hlen;
1861        int ret;
1862
1863        tunnel = netdev_priv(dev);
1864
1865        tunnel->dev = dev;
1866        tunnel->net = dev_net(dev);
1867        strcpy(tunnel->parms.name, dev->name);
1868
1869        dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1870        if (!dev->tstats)
1871                return -ENOMEM;
1872
1873        ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1874        if (ret)
1875                goto cleanup_alloc_pcpu_stats;
1876
1877        ret = gro_cells_init(&tunnel->gro_cells, dev);
1878        if (ret)
1879                goto cleanup_dst_cache_init;
1880
1881        t_hlen = ip6erspan_calc_hlen(tunnel);
1882        dev->mtu = ETH_DATA_LEN - t_hlen;
1883        if (dev->type == ARPHRD_ETHER)
1884                dev->mtu -= ETH_HLEN;
1885        if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1886                dev->mtu -= 8;
1887
1888        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1889        ip6erspan_tnl_link_config(tunnel, 1);
1890
1891        dev_hold(dev);
1892        return 0;
1893
1894cleanup_dst_cache_init:
1895        dst_cache_destroy(&tunnel->dst_cache);
1896cleanup_alloc_pcpu_stats:
1897        free_percpu(dev->tstats);
1898        dev->tstats = NULL;
1899        return ret;
1900}
1901
1902static const struct net_device_ops ip6erspan_netdev_ops = {
1903        .ndo_init =             ip6erspan_tap_init,
1904        .ndo_uninit =           ip6erspan_tunnel_uninit,
1905        .ndo_start_xmit =       ip6erspan_tunnel_xmit,
1906        .ndo_set_mac_address =  eth_mac_addr,
1907        .ndo_validate_addr =    eth_validate_addr,
1908        .ndo_change_mtu =       ip6_tnl_change_mtu,
1909        .ndo_get_stats64 =      dev_get_tstats64,
1910        .ndo_get_iflink =       ip6_tnl_get_iflink,
1911};
1912
1913static void ip6gre_tap_setup(struct net_device *dev)
1914{
1915
1916        ether_setup(dev);
1917
1918        dev->max_mtu = 0;
1919        dev->netdev_ops = &ip6gre_tap_netdev_ops;
1920        dev->needs_free_netdev = true;
1921        dev->priv_destructor = ip6gre_dev_free;
1922
1923        dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1924        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1925        netif_keep_dst(dev);
1926}
1927
1928static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1929                                       struct ip_tunnel_encap *ipencap)
1930{
1931        bool ret = false;
1932
1933        memset(ipencap, 0, sizeof(*ipencap));
1934
1935        if (!data)
1936                return ret;
1937
1938        if (data[IFLA_GRE_ENCAP_TYPE]) {
1939                ret = true;
1940                ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1941        }
1942
1943        if (data[IFLA_GRE_ENCAP_FLAGS]) {
1944                ret = true;
1945                ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1946        }
1947
1948        if (data[IFLA_GRE_ENCAP_SPORT]) {
1949                ret = true;
1950                ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1951        }
1952
1953        if (data[IFLA_GRE_ENCAP_DPORT]) {
1954                ret = true;
1955                ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1956        }
1957
1958        return ret;
1959}
1960
1961static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1962                                 struct nlattr *tb[], struct nlattr *data[],
1963                                 struct netlink_ext_ack *extack)
1964{
1965        struct ip6_tnl *nt;
1966        struct ip_tunnel_encap ipencap;
1967        int err;
1968
1969        nt = netdev_priv(dev);
1970
1971        if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1972                int err = ip6_tnl_encap_setup(nt, &ipencap);
1973
1974                if (err < 0)
1975                        return err;
1976        }
1977
1978        if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1979                eth_hw_addr_random(dev);
1980
1981        nt->dev = dev;
1982        nt->net = dev_net(dev);
1983
1984        err = register_netdevice(dev);
1985        if (err)
1986                goto out;
1987
1988        if (tb[IFLA_MTU])
1989                ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1990
1991out:
1992        return err;
1993}
1994
1995static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1996                          struct nlattr *tb[], struct nlattr *data[],
1997                          struct netlink_ext_ack *extack)
1998{
1999        struct ip6_tnl *nt = netdev_priv(dev);
2000        struct net *net = dev_net(dev);
2001        struct ip6gre_net *ign;
2002        int err;
2003
2004        ip6gre_netlink_parms(data, &nt->parms);
2005        ign = net_generic(net, ip6gre_net_id);
2006
2007        if (nt->parms.collect_md) {
2008                if (rtnl_dereference(ign->collect_md_tun))
2009                        return -EEXIST;
2010        } else {
2011                if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2012                        return -EEXIST;
2013        }
2014
2015        err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2016        if (!err) {
2017                ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2018                ip6gre_tunnel_link_md(ign, nt);
2019                ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2020        }
2021        return err;
2022}
2023
2024static struct ip6_tnl *
2025ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2026                         struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2027                         struct netlink_ext_ack *extack)
2028{
2029        struct ip6_tnl *t, *nt = netdev_priv(dev);
2030        struct net *net = nt->net;
2031        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2032        struct ip_tunnel_encap ipencap;
2033
2034        if (dev == ign->fb_tunnel_dev)
2035                return ERR_PTR(-EINVAL);
2036
2037        if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2038                int err = ip6_tnl_encap_setup(nt, &ipencap);
2039
2040                if (err < 0)
2041                        return ERR_PTR(err);
2042        }
2043
2044        ip6gre_netlink_parms(data, p_p);
2045
2046        t = ip6gre_tunnel_locate(net, p_p, 0);
2047
2048        if (t) {
2049                if (t->dev != dev)
2050                        return ERR_PTR(-EEXIST);
2051        } else {
2052                t = nt;
2053        }
2054
2055        return t;
2056}
2057
2058static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2059                             struct nlattr *data[],
2060                             struct netlink_ext_ack *extack)
2061{
2062        struct ip6_tnl *t = netdev_priv(dev);
2063        struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2064        struct __ip6_tnl_parm p;
2065
2066        t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2067        if (IS_ERR(t))
2068                return PTR_ERR(t);
2069
2070        ip6gre_tunnel_unlink_md(ign, t);
2071        ip6gre_tunnel_unlink(ign, t);
2072        ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2073        ip6gre_tunnel_link_md(ign, t);
2074        ip6gre_tunnel_link(ign, t);
2075        return 0;
2076}
2077
2078static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2079{
2080        struct net *net = dev_net(dev);
2081        struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2082
2083        if (dev != ign->fb_tunnel_dev)
2084                unregister_netdevice_queue(dev, head);
2085}
2086
2087static size_t ip6gre_get_size(const struct net_device *dev)
2088{
2089        return
2090                /* IFLA_GRE_LINK */
2091                nla_total_size(4) +
2092                /* IFLA_GRE_IFLAGS */
2093                nla_total_size(2) +
2094                /* IFLA_GRE_OFLAGS */
2095                nla_total_size(2) +
2096                /* IFLA_GRE_IKEY */
2097                nla_total_size(4) +
2098                /* IFLA_GRE_OKEY */
2099                nla_total_size(4) +
2100                /* IFLA_GRE_LOCAL */
2101                nla_total_size(sizeof(struct in6_addr)) +
2102                /* IFLA_GRE_REMOTE */
2103                nla_total_size(sizeof(struct in6_addr)) +
2104                /* IFLA_GRE_TTL */
2105                nla_total_size(1) +
2106                /* IFLA_GRE_ENCAP_LIMIT */
2107                nla_total_size(1) +
2108                /* IFLA_GRE_FLOWINFO */
2109                nla_total_size(4) +
2110                /* IFLA_GRE_FLAGS */
2111                nla_total_size(4) +
2112                /* IFLA_GRE_ENCAP_TYPE */
2113                nla_total_size(2) +
2114                /* IFLA_GRE_ENCAP_FLAGS */
2115                nla_total_size(2) +
2116                /* IFLA_GRE_ENCAP_SPORT */
2117                nla_total_size(2) +
2118                /* IFLA_GRE_ENCAP_DPORT */
2119                nla_total_size(2) +
2120                /* IFLA_GRE_COLLECT_METADATA */
2121                nla_total_size(0) +
2122                /* IFLA_GRE_FWMARK */
2123                nla_total_size(4) +
2124                /* IFLA_GRE_ERSPAN_INDEX */
2125                nla_total_size(4) +
2126                0;
2127}
2128
2129static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2130{
2131        struct ip6_tnl *t = netdev_priv(dev);
2132        struct __ip6_tnl_parm *p = &t->parms;
2133        __be16 o_flags = p->o_flags;
2134
2135        if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2136                if (!p->collect_md)
2137                        o_flags |= TUNNEL_KEY;
2138
2139                if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2140                        goto nla_put_failure;
2141
2142                if (p->erspan_ver == 1) {
2143                        if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2144                                goto nla_put_failure;
2145                } else {
2146                        if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2147                                goto nla_put_failure;
2148                        if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2149                                goto nla_put_failure;
2150                }
2151        }
2152
2153        if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2154            nla_put_be16(skb, IFLA_GRE_IFLAGS,
2155                         gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2156            nla_put_be16(skb, IFLA_GRE_OFLAGS,
2157                         gre_tnl_flags_to_gre_flags(o_flags)) ||
2158            nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2159            nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2160            nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2161            nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2162            nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2163            nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2164            nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2165            nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2166            nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2167                goto nla_put_failure;
2168
2169        if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2170                        t->encap.type) ||
2171            nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2172                         t->encap.sport) ||
2173            nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2174                         t->encap.dport) ||
2175            nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2176                        t->encap.flags))
2177                goto nla_put_failure;
2178
2179        if (p->collect_md) {
2180                if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2181                        goto nla_put_failure;
2182        }
2183
2184        return 0;
2185
2186nla_put_failure:
2187        return -EMSGSIZE;
2188}
2189
2190static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2191        [IFLA_GRE_LINK]        = { .type = NLA_U32 },
2192        [IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2193        [IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2194        [IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2195        [IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2196        [IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2197        [IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2198        [IFLA_GRE_TTL]         = { .type = NLA_U8 },
2199        [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2200        [IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2201        [IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2202        [IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2203        [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2204        [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2205        [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2206        [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2207        [IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2208        [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2209        [IFLA_GRE_ERSPAN_VER]   = { .type = NLA_U8 },
2210        [IFLA_GRE_ERSPAN_DIR]   = { .type = NLA_U8 },
2211        [IFLA_GRE_ERSPAN_HWID]  = { .type = NLA_U16 },
2212};
2213
2214static void ip6erspan_tap_setup(struct net_device *dev)
2215{
2216        ether_setup(dev);
2217
2218        dev->max_mtu = 0;
2219        dev->netdev_ops = &ip6erspan_netdev_ops;
2220        dev->needs_free_netdev = true;
2221        dev->priv_destructor = ip6gre_dev_free;
2222
2223        dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2224        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2225        netif_keep_dst(dev);
2226}
2227
2228static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2229                             struct nlattr *tb[], struct nlattr *data[],
2230                             struct netlink_ext_ack *extack)
2231{
2232        struct ip6_tnl *nt = netdev_priv(dev);
2233        struct net *net = dev_net(dev);
2234        struct ip6gre_net *ign;
2235        int err;
2236
2237        ip6gre_netlink_parms(data, &nt->parms);
2238        ip6erspan_set_version(data, &nt->parms);
2239        ign = net_generic(net, ip6gre_net_id);
2240
2241        if (nt->parms.collect_md) {
2242                if (rtnl_dereference(ign->collect_md_tun_erspan))
2243                        return -EEXIST;
2244        } else {
2245                if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2246                        return -EEXIST;
2247        }
2248
2249        err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2250        if (!err) {
2251                ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2252                ip6erspan_tunnel_link_md(ign, nt);
2253                ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2254        }
2255        return err;
2256}
2257
2258static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2259{
2260        ip6gre_tnl_link_config_common(t);
2261        ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2262}
2263
2264static int ip6erspan_tnl_change(struct ip6_tnl *t,
2265                                const struct __ip6_tnl_parm *p, int set_mtu)
2266{
2267        ip6gre_tnl_copy_tnl_parm(t, p);
2268        ip6erspan_tnl_link_config(t, set_mtu);
2269        return 0;
2270}
2271
2272static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2273                                struct nlattr *data[],
2274                                struct netlink_ext_ack *extack)
2275{
2276        struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2277        struct __ip6_tnl_parm p;
2278        struct ip6_tnl *t;
2279
2280        t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2281        if (IS_ERR(t))
2282                return PTR_ERR(t);
2283
2284        ip6erspan_set_version(data, &p);
2285        ip6gre_tunnel_unlink_md(ign, t);
2286        ip6gre_tunnel_unlink(ign, t);
2287        ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2288        ip6erspan_tunnel_link_md(ign, t);
2289        ip6gre_tunnel_link(ign, t);
2290        return 0;
2291}
2292
2293static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2294        .kind           = "ip6gre",
2295        .maxtype        = IFLA_GRE_MAX,
2296        .policy         = ip6gre_policy,
2297        .priv_size      = sizeof(struct ip6_tnl),
2298        .setup          = ip6gre_tunnel_setup,
2299        .validate       = ip6gre_tunnel_validate,
2300        .newlink        = ip6gre_newlink,
2301        .changelink     = ip6gre_changelink,
2302        .dellink        = ip6gre_dellink,
2303        .get_size       = ip6gre_get_size,
2304        .fill_info      = ip6gre_fill_info,
2305        .get_link_net   = ip6_tnl_get_link_net,
2306};
2307
2308static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2309        .kind           = "ip6gretap",
2310        .maxtype        = IFLA_GRE_MAX,
2311        .policy         = ip6gre_policy,
2312        .priv_size      = sizeof(struct ip6_tnl),
2313        .setup          = ip6gre_tap_setup,
2314        .validate       = ip6gre_tap_validate,
2315        .newlink        = ip6gre_newlink,
2316        .changelink     = ip6gre_changelink,
2317        .get_size       = ip6gre_get_size,
2318        .fill_info      = ip6gre_fill_info,
2319        .get_link_net   = ip6_tnl_get_link_net,
2320};
2321
2322static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2323        .kind           = "ip6erspan",
2324        .maxtype        = IFLA_GRE_MAX,
2325        .policy         = ip6gre_policy,
2326        .priv_size      = sizeof(struct ip6_tnl),
2327        .setup          = ip6erspan_tap_setup,
2328        .validate       = ip6erspan_tap_validate,
2329        .newlink        = ip6erspan_newlink,
2330        .changelink     = ip6erspan_changelink,
2331        .get_size       = ip6gre_get_size,
2332        .fill_info      = ip6gre_fill_info,
2333        .get_link_net   = ip6_tnl_get_link_net,
2334};
2335
2336/*
2337 *      And now the modules code and kernel interface.
2338 */
2339
2340static int __init ip6gre_init(void)
2341{
2342        int err;
2343
2344        pr_info("GRE over IPv6 tunneling driver\n");
2345
2346        err = register_pernet_device(&ip6gre_net_ops);
2347        if (err < 0)
2348                return err;
2349
2350        err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2351        if (err < 0) {
2352                pr_info("%s: can't add protocol\n", __func__);
2353                goto add_proto_failed;
2354        }
2355
2356        err = rtnl_link_register(&ip6gre_link_ops);
2357        if (err < 0)
2358                goto rtnl_link_failed;
2359
2360        err = rtnl_link_register(&ip6gre_tap_ops);
2361        if (err < 0)
2362                goto tap_ops_failed;
2363
2364        err = rtnl_link_register(&ip6erspan_tap_ops);
2365        if (err < 0)
2366                goto erspan_link_failed;
2367
2368out:
2369        return err;
2370
2371erspan_link_failed:
2372        rtnl_link_unregister(&ip6gre_tap_ops);
2373tap_ops_failed:
2374        rtnl_link_unregister(&ip6gre_link_ops);
2375rtnl_link_failed:
2376        inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2377add_proto_failed:
2378        unregister_pernet_device(&ip6gre_net_ops);
2379        goto out;
2380}
2381
2382static void __exit ip6gre_fini(void)
2383{
2384        rtnl_link_unregister(&ip6gre_tap_ops);
2385        rtnl_link_unregister(&ip6gre_link_ops);
2386        rtnl_link_unregister(&ip6erspan_tap_ops);
2387        inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2388        unregister_pernet_device(&ip6gre_net_ops);
2389}
2390
2391module_init(ip6gre_init);
2392module_exit(ip6gre_fini);
2393MODULE_LICENSE("GPL");
2394MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2395MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2396MODULE_ALIAS_RTNL_LINK("ip6gre");
2397MODULE_ALIAS_RTNL_LINK("ip6gretap");
2398MODULE_ALIAS_RTNL_LINK("ip6erspan");
2399MODULE_ALIAS_NETDEV("ip6gre0");
2400