net/netfilter/nf_flow_table_ip.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 #include <linux/kernel.h>
   3 #include <linux/init.h>
   4 #include <linux/module.h>
   5 #include <linux/netfilter.h>
   6 #include <linux/rhashtable.h>
   7 #include <linux/ip.h>
   8 #include <linux/ipv6.h>
   9 #include <linux/netdevice.h>
  10 #include <linux/if_ether.h>
  11 #include <net/ip.h>
  12 #include <net/ipv6.h>
  13 #include <net/ip6_route.h>
  14 #include <net/neighbour.h>
  15 #include <net/netfilter/nf_flow_table.h>
  16 #include <net/netfilter/nf_conntrack_acct.h>
  17 /* For layer 4 checksum field offset. */
  18 #include <linux/tcp.h>
  19 #include <linux/udp.h>
  20
  21 static int nf_flow_state_check(struct flow_offload *flow, int proto,
  22                                struct sk_buff *skb, unsigned int thoff)
  23 {
  24         struct tcphdr *tcph;
  25
  26         if (proto != IPPROTO_TCP)
  27                 return 0;
  28
  29         tcph = (void *)(skb_network_header(skb) + thoff);
  30         if (unlikely(tcph->fin || tcph->rst)) {
  31                 flow_offload_teardown(flow);
  32                 return -1;
  33         }
  34
  35         return 0;
  36 }
  37
  38 static void nf_flow_nat_ip_tcp(struct sk_buff *skb, unsigned int thoff,
  39                                __be32 addr, __be32 new_addr)
  40 {
  41         struct tcphdr *tcph;
  42
  43         tcph = (void *)(skb_network_header(skb) + thoff);
  44         inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, true);
  45 }
  46
  47 static void nf_flow_nat_ip_udp(struct sk_buff *skb, unsigned int thoff,
  48                                __be32 addr, __be32 new_addr)
  49 {
  50         struct udphdr *udph;
  51
  52         udph = (void *)(skb_network_header(skb) + thoff);
  53         if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
  54                 inet_proto_csum_replace4(&udph->check, skb, addr,
  55                                          new_addr, true);
  56                 if (!udph->check)
  57                         udph->check = CSUM_MANGLED_0;
  58         }
  59 }
  60
  61 static void nf_flow_nat_ip_l4proto(struct sk_buff *skb, struct iphdr *iph,
  62                                    unsigned int thoff, __be32 addr,
  63                                    __be32 new_addr)
  64 {
  65         switch (iph->protocol) {
  66         case IPPROTO_TCP:
  67                 nf_flow_nat_ip_tcp(skb, thoff, addr, new_addr);
  68                 break;
  69         case IPPROTO_UDP:
  70                 nf_flow_nat_ip_udp(skb, thoff, addr, new_addr);
  71                 break;
  72         }
  73 }
  74
  75 static void nf_flow_snat_ip(const struct flow_offload *flow,
  76                             struct sk_buff *skb, struct iphdr *iph,
  77                             unsigned int thoff, enum flow_offload_tuple_dir dir)
  78 {
  79         __be32 addr, new_addr;
  80
  81         switch (dir) {
  82         case FLOW_OFFLOAD_DIR_ORIGINAL:
  83                 addr = iph->saddr;
  84                 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_v4.s_addr;
  85                 iph->saddr = new_addr;
  86                 break;
  87         case FLOW_OFFLOAD_DIR_REPLY:
  88                 addr = iph->daddr;
  89                 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_v4.s_addr;
  90                 iph->daddr = new_addr;
  91                 break;
  92         }
  93         csum_replace4(&iph->check, addr, new_addr);
  94
  95         nf_flow_nat_ip_l4proto(skb, iph, thoff, addr, new_addr);
  96 }
  97
  98 static void nf_flow_dnat_ip(const struct flow_offload *flow,
  99                             struct sk_buff *skb, struct iphdr *iph,
 100                             unsigned int thoff, enum flow_offload_tuple_dir dir)
 101 {
 102         __be32 addr, new_addr;
 103
 104         switch (dir) {
 105         case FLOW_OFFLOAD_DIR_ORIGINAL:
 106                 addr = iph->daddr;
 107                 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_v4.s_addr;
 108                 iph->daddr = new_addr;
 109                 break;
 110         case FLOW_OFFLOAD_DIR_REPLY:
 111                 addr = iph->saddr;
 112                 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_v4.s_addr;
 113                 iph->saddr = new_addr;
 114                 break;
 115         }
 116         csum_replace4(&iph->check, addr, new_addr);
 117
 118         nf_flow_nat_ip_l4proto(skb, iph, thoff, addr, new_addr);
 119 }
 120
 121 static void nf_flow_nat_ip(const struct flow_offload *flow, struct sk_buff *skb,
 122                           unsigned int thoff, enum flow_offload_tuple_dir dir,
 123                           struct iphdr *iph)
 124 {
 125         if (test_bit(NF_FLOW_SNAT, &flow->flags)) {
 126                 nf_flow_snat_port(flow, skb, thoff, iph->protocol, dir);
 127                 nf_flow_snat_ip(flow, skb, iph, thoff, dir);
 128         }
 129         if (test_bit(NF_FLOW_DNAT, &flow->flags)) {
 130                 nf_flow_dnat_port(flow, skb, thoff, iph->protocol, dir);
 131                 nf_flow_dnat_ip(flow, skb, iph, thoff, dir);
 132         }
 133 }
 134
 135 static bool ip_has_options(unsigned int thoff)
 136 {
 137         return thoff != sizeof(struct iphdr);
 138 }
 139
 140 static void nf_flow_tuple_encap(struct sk_buff *skb,
 141                                 struct flow_offload_tuple *tuple)
 142 {
 143         struct vlan_ethhdr *veth;
 144         struct pppoe_hdr *phdr;
 145         int i = 0;
 146
 147         if (skb_vlan_tag_present(skb)) {
 148                 tuple->encap[i].id = skb_vlan_tag_get(skb);
 149                 tuple->encap[i].proto = skb->vlan_proto;
 150                 i++;
 151         }
 152         switch (skb->protocol) {
 153         case htons(ETH_P_8021Q):
 154                 veth = (struct vlan_ethhdr *)skb_mac_header(skb);
 155                 tuple->encap[i].id = ntohs(veth->h_vlan_TCI);
 156                 tuple->encap[i].proto = skb->protocol;
 157                 break;
 158         case htons(ETH_P_PPP_SES):
 159                 phdr = (struct pppoe_hdr *)skb_network_header(skb);
 160                 tuple->encap[i].id = ntohs(phdr->sid);
 161                 tuple->encap[i].proto = skb->protocol;
 162                 break;
 163         }
 164 }
 165
 166 static int nf_flow_tuple_ip(struct sk_buff *skb, const struct net_device *dev,
 167                             struct flow_offload_tuple *tuple, u32 *hdrsize,
 168                             u32 offset)
 169 {
 170         struct flow_ports *ports;
 171         unsigned int thoff;
 172         struct iphdr *iph;
 173         u8 ipproto;
 174
 175         if (!pskb_may_pull(skb, sizeof(*iph) + offset))
 176                 return -1;
 177
 178         iph = (struct iphdr *)(skb_network_header(skb) + offset);
 179         thoff = (iph->ihl * 4);
 180
 181         if (ip_is_fragment(iph) ||
 182             unlikely(ip_has_options(thoff)))
 183                 return -1;
 184
 185         thoff += offset;
 186
 187         ipproto = iph->protocol;
 188         switch (ipproto) {
 189         case IPPROTO_TCP:
 190                 *hdrsize = sizeof(struct tcphdr);
 191                 break;
 192         case IPPROTO_UDP:
 193                 *hdrsize = sizeof(struct udphdr);
 194                 break;
 195 #ifdef CONFIG_NF_CT_PROTO_GRE
 196         case IPPROTO_GRE:
 197                 *hdrsize = sizeof(struct gre_base_hdr);
 198                 break;
 199 #endif
 200         default:
 201                 return -1;
 202         }
 203
 204         if (iph->ttl <= 1)
 205                 return -1;
 206
 207         if (!pskb_may_pull(skb, thoff + *hdrsize))
 208                 return -1;
 209
 210         switch (ipproto) {
 211         case IPPROTO_TCP:
 212         case IPPROTO_UDP:
 213                 ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
 214                 tuple->src_port         = ports->source;
 215                 tuple->dst_port         = ports->dest;
 216                 break;
 217         case IPPROTO_GRE: {
 218                 struct gre_base_hdr *greh;
 219
 220                 greh = (struct gre_base_hdr *)(skb_network_header(skb) + thoff);
 221                 if ((greh->flags & GRE_VERSION) != GRE_VERSION_0)
 222                         return -1;
 223                 break;
 224         }
 225         }
 226
 227         iph = (struct iphdr *)(skb_network_header(skb) + offset);
 228
 229         tuple->src_v4.s_addr    = iph->saddr;
 230         tuple->dst_v4.s_addr    = iph->daddr;
 231         tuple->l3proto          = AF_INET;
 232         tuple->l4proto          = ipproto;
 233         tuple->iifidx           = dev->ifindex;
 234         nf_flow_tuple_encap(skb, tuple);
 235
 236         return 0;
 237 }
 238
 239 /* Based on ip_exceeds_mtu(). */
 240 static bool nf_flow_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
 241 {
 242         if (skb->len <= mtu)
 243                 return false;
 244
 245         if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
 246                 return false;
 247
 248         return true;
 249 }
 250
 251 static inline bool nf_flow_dst_check(struct flow_offload_tuple *tuple)
 252 {
 253         if (tuple->xmit_type != FLOW_OFFLOAD_XMIT_NEIGH &&
 254             tuple->xmit_type != FLOW_OFFLOAD_XMIT_XFRM)
 255                 return true;
 256
 257         return dst_check(tuple->dst_cache, tuple->dst_cookie);
 258 }
 259
 260 static unsigned int nf_flow_xmit_xfrm(struct sk_buff *skb,
 261                                       const struct nf_hook_state *state,
 262                                       struct dst_entry *dst)
 263 {
 264         skb_orphan(skb);
 265         skb_dst_set_noref(skb, dst);
 266         dst_output(state->net, state->sk, skb);
 267         return NF_STOLEN;
 268 }
 269
 270 static bool nf_flow_skb_encap_protocol(struct sk_buff *skb, __be16 proto,
 271                                        u32 *offset)
 272 {
 273         struct vlan_ethhdr *veth;
 274         __be16 inner_proto;
 275
 276         switch (skb->protocol) {
 277         case htons(ETH_P_8021Q):
 278                 veth = (struct vlan_ethhdr *)skb_mac_header(skb);
 279                 if (veth->h_vlan_encapsulated_proto == proto) {
 280                         *offset += VLAN_HLEN;
 281                         return true;
 282                 }
 283                 break;
 284         case htons(ETH_P_PPP_SES):
 285                 if (nf_flow_pppoe_proto(skb, &inner_proto) &&
 286                     inner_proto == proto) {
 287                         *offset += PPPOE_SES_HLEN;
 288                         return true;
 289                 }
 290                 break;
 291         }
 292
 293         return false;
 294 }
 295
 296 static void nf_flow_encap_pop(struct sk_buff *skb,
 297                               struct flow_offload_tuple_rhash *tuplehash)
 298 {
 299         struct vlan_hdr *vlan_hdr;
 300         int i;
 301
 302         for (i = 0; i < tuplehash->tuple.encap_num; i++) {
 303                 if (skb_vlan_tag_present(skb)) {
 304                         __vlan_hwaccel_clear_tag(skb);
 305                         continue;
 306                 }
 307                 switch (skb->protocol) {
 308                 case htons(ETH_P_8021Q):
 309                         vlan_hdr = (struct vlan_hdr *)skb->data;
 310                         __skb_pull(skb, VLAN_HLEN);
 311                         vlan_set_encap_proto(skb, vlan_hdr);
 312                         skb_reset_network_header(skb);
 313                         break;
 314                 case htons(ETH_P_PPP_SES):
 315                         skb->protocol = __nf_flow_pppoe_proto(skb);
 316                         skb_pull(skb, PPPOE_SES_HLEN);
 317                         skb_reset_network_header(skb);
 318                         break;
 319                 }
 320         }
 321 }
 322
 323 static unsigned int nf_flow_queue_xmit(struct net *net, struct sk_buff *skb,
 324                                        const struct flow_offload_tuple_rhash *tuplehash,
 325                                        unsigned short type)
 326 {
 327         struct net_device *outdev;
 328
 329         outdev = dev_get_by_index_rcu(net, tuplehash->tuple.out.ifidx);
 330         if (!outdev)
 331                 return NF_DROP;
 332
 333         skb->dev = outdev;
 334         dev_hard_header(skb, skb->dev, type, tuplehash->tuple.out.h_dest,
 335                         tuplehash->tuple.out.h_source, skb->len);
 336         dev_queue_xmit(skb);
 337
 338         return NF_STOLEN;
 339 }
 340
 341 unsigned int
 342 nf_flow_offload_ip_hook(void *priv, struct sk_buff *skb,
 343                         const struct nf_hook_state *state)
 344 {
 345         struct flow_offload_tuple_rhash *tuplehash;
 346         struct nf_flowtable *flow_table = priv;
 347         struct flow_offload_tuple tuple = {};
 348         enum flow_offload_tuple_dir dir;
 349         struct flow_offload *flow;
 350         struct net_device *outdev;
 351         u32 hdrsize, offset = 0;
 352         unsigned int thoff, mtu;
 353         struct rtable *rt;
 354         struct iphdr *iph;
 355         __be32 nexthop;
 356         int ret;
 357
 358         if (skb->protocol != htons(ETH_P_IP) &&
 359             !nf_flow_skb_encap_protocol(skb, htons(ETH_P_IP), &offset))
 360                 return NF_ACCEPT;
 361
 362         if (nf_flow_tuple_ip(skb, state->in, &tuple, &hdrsize, offset) < 0)
 363                 return NF_ACCEPT;
 364
 365         tuplehash = flow_offload_lookup(flow_table, &tuple);
 366         if (tuplehash == NULL)
 367                 return NF_ACCEPT;
 368
 369         dir = tuplehash->tuple.dir;
 370         flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
 371
 372         mtu = flow->tuplehash[dir].tuple.mtu + offset;
 373         if (unlikely(nf_flow_exceeds_mtu(skb, mtu)))
 374                 return NF_ACCEPT;
 375
 376         iph = (struct iphdr *)(skb_network_header(skb) + offset);
 377         thoff = (iph->ihl * 4) + offset;
 378         if (nf_flow_state_check(flow, iph->protocol, skb, thoff))
 379                 return NF_ACCEPT;
 380
 381         if (!nf_flow_dst_check(&tuplehash->tuple)) {
 382                 flow_offload_teardown(flow);
 383                 return NF_ACCEPT;
 384         }
 385
 386         if (skb_try_make_writable(skb, thoff + hdrsize))
 387                 return NF_DROP;
 388
 389         flow_offload_refresh(flow_table, flow, false);
 390
 391         nf_flow_encap_pop(skb, tuplehash);
 392         thoff -= offset;
 393
 394         iph = ip_hdr(skb);
 395         nf_flow_nat_ip(flow, skb, thoff, dir, iph);
 396
 397         ip_decrease_ttl(iph);
 398         skb_clear_tstamp(skb);
 399
 400         if (flow_table->flags & NF_FLOWTABLE_COUNTER)
 401                 nf_ct_acct_update(flow->ct, tuplehash->tuple.dir, skb->len);
 402
 403         if (unlikely(tuplehash->tuple.xmit_type == FLOW_OFFLOAD_XMIT_XFRM)) {
 404                 rt = (struct rtable *)tuplehash->tuple.dst_cache;
 405                 memset(skb->cb, 0, sizeof(struct inet_skb_parm));
 406                 IPCB(skb)->iif = skb->dev->ifindex;
 407                 IPCB(skb)->flags = IPSKB_FORWARDED;
 408                 return nf_flow_xmit_xfrm(skb, state, &rt->dst);
 409         }
 410
 411         switch (tuplehash->tuple.xmit_type) {
 412         case FLOW_OFFLOAD_XMIT_NEIGH:
 413                 rt = (struct rtable *)tuplehash->tuple.dst_cache;
 414                 outdev = rt->dst.dev;
 415                 skb->dev = outdev;
 416                 nexthop = rt_nexthop(rt, flow->tuplehash[!dir].tuple.src_v4.s_addr);
 417                 skb_dst_set_noref(skb, &rt->dst);
 418                 neigh_xmit(NEIGH_ARP_TABLE, outdev, &nexthop, skb);
 419                 ret = NF_STOLEN;
 420                 break;
 421         case FLOW_OFFLOAD_XMIT_DIRECT:
 422                 ret = nf_flow_queue_xmit(state->net, skb, tuplehash, ETH_P_IP);
 423                 if (ret == NF_DROP)
 424                         flow_offload_teardown(flow);
 425                 break;
 426         }
 427
 428         return ret;
 429 }
 430 EXPORT_SYMBOL_GPL(nf_flow_offload_ip_hook);
 431
 432 static void nf_flow_nat_ipv6_tcp(struct sk_buff *skb, unsigned int thoff,
 433                                  struct in6_addr *addr,
 434                                  struct in6_addr *new_addr,
 435                                  struct ipv6hdr *ip6h)
 436 {
 437         struct tcphdr *tcph;
 438
 439         tcph = (void *)(skb_network_header(skb) + thoff);
 440         inet_proto_csum_replace16(&tcph->check, skb, addr->s6_addr32,
 441                                   new_addr->s6_addr32, true);
 442 }
 443
 444 static void nf_flow_nat_ipv6_udp(struct sk_buff *skb, unsigned int thoff,
 445                                  struct in6_addr *addr,
 446                                  struct in6_addr *new_addr)
 447 {
 448         struct udphdr *udph;
 449
 450         udph = (void *)(skb_network_header(skb) + thoff);
 451         if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
 452                 inet_proto_csum_replace16(&udph->check, skb, addr->s6_addr32,
 453                                           new_addr->s6_addr32, true);
 454                 if (!udph->check)
 455                         udph->check = CSUM_MANGLED_0;
 456         }
 457 }
 458
 459 static void nf_flow_nat_ipv6_l4proto(struct sk_buff *skb, struct ipv6hdr *ip6h,
 460                                      unsigned int thoff, struct in6_addr *addr,
 461                                      struct in6_addr *new_addr)
 462 {
 463         switch (ip6h->nexthdr) {
 464         case IPPROTO_TCP:
 465                 nf_flow_nat_ipv6_tcp(skb, thoff, addr, new_addr, ip6h);
 466                 break;
 467         case IPPROTO_UDP:
 468                 nf_flow_nat_ipv6_udp(skb, thoff, addr, new_addr);
 469                 break;
 470         }
 471 }
 472
 473 static void nf_flow_snat_ipv6(const struct flow_offload *flow,
 474                               struct sk_buff *skb, struct ipv6hdr *ip6h,
 475                               unsigned int thoff,
 476                               enum flow_offload_tuple_dir dir)
 477 {
 478         struct in6_addr addr, new_addr;
 479
 480         switch (dir) {
 481         case FLOW_OFFLOAD_DIR_ORIGINAL:
 482                 addr = ip6h->saddr;
 483                 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_v6;
 484                 ip6h->saddr = new_addr;
 485                 break;
 486         case FLOW_OFFLOAD_DIR_REPLY:
 487                 addr = ip6h->daddr;
 488                 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_v6;
 489                 ip6h->daddr = new_addr;
 490                 break;
 491         }
 492
 493         nf_flow_nat_ipv6_l4proto(skb, ip6h, thoff, &addr, &new_addr);
 494 }
 495
 496 static void nf_flow_dnat_ipv6(const struct flow_offload *flow,
 497                               struct sk_buff *skb, struct ipv6hdr *ip6h,
 498                               unsigned int thoff,
 499                               enum flow_offload_tuple_dir dir)
 500 {
 501         struct in6_addr addr, new_addr;
 502
 503         switch (dir) {
 504         case FLOW_OFFLOAD_DIR_ORIGINAL:
 505                 addr = ip6h->daddr;
 506                 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_v6;
 507                 ip6h->daddr = new_addr;
 508                 break;
 509         case FLOW_OFFLOAD_DIR_REPLY:
 510                 addr = ip6h->saddr;
 511                 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_v6;
 512                 ip6h->saddr = new_addr;
 513                 break;
 514         }
 515
 516         nf_flow_nat_ipv6_l4proto(skb, ip6h, thoff, &addr, &new_addr);
 517 }
 518
 519 static void nf_flow_nat_ipv6(const struct flow_offload *flow,
 520                              struct sk_buff *skb,
 521                              enum flow_offload_tuple_dir dir,
 522                              struct ipv6hdr *ip6h)
 523 {
 524         unsigned int thoff = sizeof(*ip6h);
 525
 526         if (test_bit(NF_FLOW_SNAT, &flow->flags)) {
 527                 nf_flow_snat_port(flow, skb, thoff, ip6h->nexthdr, dir);
 528                 nf_flow_snat_ipv6(flow, skb, ip6h, thoff, dir);
 529         }
 530         if (test_bit(NF_FLOW_DNAT, &flow->flags)) {
 531                 nf_flow_dnat_port(flow, skb, thoff, ip6h->nexthdr, dir);
 532                 nf_flow_dnat_ipv6(flow, skb, ip6h, thoff, dir);
 533         }
 534 }
 535
 536 static int nf_flow_tuple_ipv6(struct sk_buff *skb, const struct net_device *dev,
 537                               struct flow_offload_tuple *tuple, u32 *hdrsize,
 538                               u32 offset)
 539 {
 540         struct flow_ports *ports;
 541         struct ipv6hdr *ip6h;
 542         unsigned int thoff;
 543         u8 nexthdr;
 544
 545         thoff = sizeof(*ip6h) + offset;
 546         if (!pskb_may_pull(skb, thoff))
 547                 return -1;
 548
 549         ip6h = (struct ipv6hdr *)(skb_network_header(skb) + offset);
 550
 551         nexthdr = ip6h->nexthdr;
 552         switch (nexthdr) {
 553         case IPPROTO_TCP:
 554                 *hdrsize = sizeof(struct tcphdr);
 555                 break;
 556         case IPPROTO_UDP:
 557                 *hdrsize = sizeof(struct udphdr);
 558                 break;
 559 #ifdef CONFIG_NF_CT_PROTO_GRE
 560         case IPPROTO_GRE:
 561                 *hdrsize = sizeof(struct gre_base_hdr);
 562                 break;
 563 #endif
 564         default:
 565                 return -1;
 566         }
 567
 568         if (ip6h->hop_limit <= 1)
 569                 return -1;
 570
 571         if (!pskb_may_pull(skb, thoff + *hdrsize))
 572                 return -1;
 573
 574         switch (nexthdr) {
 575         case IPPROTO_TCP:
 576         case IPPROTO_UDP:
 577                 ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
 578                 tuple->src_port         = ports->source;
 579                 tuple->dst_port         = ports->dest;
 580                 break;
 581         case IPPROTO_GRE: {
 582                 struct gre_base_hdr *greh;
 583
 584                 greh = (struct gre_base_hdr *)(skb_network_header(skb) + thoff);
 585                 if ((greh->flags & GRE_VERSION) != GRE_VERSION_0)
 586                         return -1;
 587                 break;
 588         }
 589         }
 590
 591         ip6h = (struct ipv6hdr *)(skb_network_header(skb) + offset);
 592
 593         tuple->src_v6           = ip6h->saddr;
 594         tuple->dst_v6           = ip6h->daddr;
 595         tuple->l3proto          = AF_INET6;
 596         tuple->l4proto          = nexthdr;
 597         tuple->iifidx           = dev->ifindex;
 598         nf_flow_tuple_encap(skb, tuple);
 599
 600         return 0;
 601 }
 602
 603 unsigned int
 604 nf_flow_offload_ipv6_hook(void *priv, struct sk_buff *skb,
 605                           const struct nf_hook_state *state)
 606 {
 607         struct flow_offload_tuple_rhash *tuplehash;
 608         struct nf_flowtable *flow_table = priv;
 609         struct flow_offload_tuple tuple = {};
 610         enum flow_offload_tuple_dir dir;
 611         const struct in6_addr *nexthop;
 612         struct flow_offload *flow;
 613         struct net_device *outdev;
 614         unsigned int thoff, mtu;
 615         u32 hdrsize, offset = 0;
 616         struct ipv6hdr *ip6h;
 617         struct rt6_info *rt;
 618         int ret;
 619
 620         if (skb->protocol != htons(ETH_P_IPV6) &&
 621             !nf_flow_skb_encap_protocol(skb, htons(ETH_P_IPV6), &offset))
 622                 return NF_ACCEPT;
 623
 624         if (nf_flow_tuple_ipv6(skb, state->in, &tuple, &hdrsize, offset) < 0)
 625                 return NF_ACCEPT;
 626
 627         tuplehash = flow_offload_lookup(flow_table, &tuple);
 628         if (tuplehash == NULL)
 629                 return NF_ACCEPT;
 630
 631         dir = tuplehash->tuple.dir;
 632         flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
 633
 634         mtu = flow->tuplehash[dir].tuple.mtu + offset;
 635         if (unlikely(nf_flow_exceeds_mtu(skb, mtu)))
 636                 return NF_ACCEPT;
 637
 638         ip6h = (struct ipv6hdr *)(skb_network_header(skb) + offset);
 639         thoff = sizeof(*ip6h) + offset;
 640         if (nf_flow_state_check(flow, ip6h->nexthdr, skb, thoff))
 641                 return NF_ACCEPT;
 642
 643         if (!nf_flow_dst_check(&tuplehash->tuple)) {
 644                 flow_offload_teardown(flow);
 645                 return NF_ACCEPT;
 646         }
 647
 648         if (skb_try_make_writable(skb, thoff + hdrsize))
 649                 return NF_DROP;
 650
 651         flow_offload_refresh(flow_table, flow, false);
 652
 653         nf_flow_encap_pop(skb, tuplehash);
 654
 655         ip6h = ipv6_hdr(skb);
 656         nf_flow_nat_ipv6(flow, skb, dir, ip6h);
 657
 658         ip6h->hop_limit--;
 659         skb_clear_tstamp(skb);
 660
 661         if (flow_table->flags & NF_FLOWTABLE_COUNTER)
 662                 nf_ct_acct_update(flow->ct, tuplehash->tuple.dir, skb->len);
 663
 664         if (unlikely(tuplehash->tuple.xmit_type == FLOW_OFFLOAD_XMIT_XFRM)) {
 665                 rt = (struct rt6_info *)tuplehash->tuple.dst_cache;
 666                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 667                 IP6CB(skb)->iif = skb->dev->ifindex;
 668                 IP6CB(skb)->flags = IP6SKB_FORWARDED;
 669                 return nf_flow_xmit_xfrm(skb, state, &rt->dst);
 670         }
 671
 672         switch (tuplehash->tuple.xmit_type) {
 673         case FLOW_OFFLOAD_XMIT_NEIGH:
 674                 rt = (struct rt6_info *)tuplehash->tuple.dst_cache;
 675                 outdev = rt->dst.dev;
 676                 skb->dev = outdev;
 677                 nexthop = rt6_nexthop(rt, &flow->tuplehash[!dir].tuple.src_v6);
 678                 skb_dst_set_noref(skb, &rt->dst);
 679                 neigh_xmit(NEIGH_ND_TABLE, outdev, nexthop, skb);
 680                 ret = NF_STOLEN;
 681                 break;
 682         case FLOW_OFFLOAD_XMIT_DIRECT:
 683                 ret = nf_flow_queue_xmit(state->net, skb, tuplehash, ETH_P_IPV6);
 684                 if (ret == NF_DROP)
 685                         flow_offload_teardown(flow);
 686                 break;
 687         }
 688
 689         return ret;
 690 }
 691 EXPORT_SYMBOL_GPL(nf_flow_offload_ipv6_hook);