git.rot13.org Git - linux/blob - net/netfilter/nf_conntrack_proto.c

   1 // SPDX-License-Identifier: GPL-2.0
   2
   3 #include <linux/types.h>
   4 #include <linux/netfilter.h>
   5 #include <linux/module.h>
   6 #include <linux/slab.h>
   7 #include <linux/mutex.h>
   8 #include <linux/vmalloc.h>
   9 #include <linux/stddef.h>
  10 #include <linux/err.h>
  11 #include <linux/percpu.h>
  12 #include <linux/notifier.h>
  13 #include <linux/kernel.h>
  14 #include <linux/netdevice.h>
  15
  16 #include <net/netfilter/nf_conntrack.h>
  17 #include <net/netfilter/nf_conntrack_l4proto.h>
  18 #include <net/netfilter/nf_conntrack_core.h>
  19 #include <net/netfilter/nf_log.h>
  20
  21 #include <linux/ip.h>
  22 #include <linux/icmp.h>
  23 #include <linux/sysctl.h>
  24 #include <net/route.h>
  25 #include <net/ip.h>
  26
  27 #include <linux/netfilter_ipv4.h>
  28 #include <linux/netfilter_ipv6.h>
  29 #include <linux/netfilter_ipv6/ip6_tables.h>
  30 #include <net/netfilter/nf_conntrack_helper.h>
  31 #include <net/netfilter/nf_conntrack_zones.h>
  32 #include <net/netfilter/nf_conntrack_seqadj.h>
  33 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
  34 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
  35 #include <net/netfilter/nf_nat_helper.h>
  36 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
  37 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
  38
  39 #include <linux/ipv6.h>
  40 #include <linux/in6.h>
  41 #include <net/ipv6.h>
  42 #include <net/inet_frag.h>
  43
  44 extern unsigned int nf_conntrack_net_id;
  45
  46 static struct nf_conntrack_l4proto __rcu *nf_ct_protos[MAX_NF_CT_PROTO + 1] __read_mostly;
  47
  48 static DEFINE_MUTEX(nf_ct_proto_mutex);
  49
  50 #ifdef CONFIG_SYSCTL
  51 static int
  52 nf_ct_register_sysctl(struct net *net,
  53                       struct ctl_table_header **header,
  54                       const char *path,
  55                       struct ctl_table *table)
  56 {
  57         if (*header == NULL) {
  58                 *header = register_net_sysctl(net, path, table);
  59                 if (*header == NULL)
  60                         return -ENOMEM;
  61         }
  62
  63         return 0;
  64 }
  65
  66 static void
  67 nf_ct_unregister_sysctl(struct ctl_table_header **header,
  68                         struct ctl_table **table,
  69                         unsigned int users)
  70 {
  71         if (users > 0)
  72                 return;
  73
  74         unregister_net_sysctl_table(*header);
  75         kfree(*table);
  76         *header = NULL;
  77         *table = NULL;
  78 }
  79
  80 __printf(5, 6)
  81 void nf_l4proto_log_invalid(const struct sk_buff *skb,
  82                             struct net *net,
  83                             u16 pf, u8 protonum,
  84                             const char *fmt, ...)
  85 {
  86         struct va_format vaf;
  87         va_list args;
  88
  89         if (net->ct.sysctl_log_invalid != protonum ||
  90             net->ct.sysctl_log_invalid != IPPROTO_RAW)
  91                 return;
  92
  93         va_start(args, fmt);
  94         vaf.fmt = fmt;
  95         vaf.va = &args;
  96
  97         nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
  98                       "nf_ct_proto_%d: %pV ", protonum, &vaf);
  99         va_end(args);
 100 }
 101 EXPORT_SYMBOL_GPL(nf_l4proto_log_invalid);
 102
 103 __printf(3, 4)
 104 void nf_ct_l4proto_log_invalid(const struct sk_buff *skb,
 105                                const struct nf_conn *ct,
 106                                const char *fmt, ...)
 107 {
 108         struct va_format vaf;
 109         struct net *net;
 110         va_list args;
 111
 112         net = nf_ct_net(ct);
 113         if (likely(net->ct.sysctl_log_invalid == 0))
 114                 return;
 115
 116         va_start(args, fmt);
 117         vaf.fmt = fmt;
 118         vaf.va = &args;
 119
 120         nf_l4proto_log_invalid(skb, net, nf_ct_l3num(ct),
 121                                nf_ct_protonum(ct), "%pV", &vaf);
 122         va_end(args);
 123 }
 124 EXPORT_SYMBOL_GPL(nf_ct_l4proto_log_invalid);
 125 #endif
 126
 127 const struct nf_conntrack_l4proto *__nf_ct_l4proto_find(u8 l4proto)
 128 {
 129         if (unlikely(l4proto >= ARRAY_SIZE(nf_ct_protos)))
 130                 return &nf_conntrack_l4proto_generic;
 131
 132         return rcu_dereference(nf_ct_protos[l4proto]);
 133 }
 134 EXPORT_SYMBOL_GPL(__nf_ct_l4proto_find);
 135
 136 const struct nf_conntrack_l4proto *nf_ct_l4proto_find_get(u8 l4num)
 137 {
 138         const struct nf_conntrack_l4proto *p;
 139
 140         rcu_read_lock();
 141         p = __nf_ct_l4proto_find(l4num);
 142         if (!try_module_get(p->me))
 143                 p = &nf_conntrack_l4proto_generic;
 144         rcu_read_unlock();
 145
 146         return p;
 147 }
 148 EXPORT_SYMBOL_GPL(nf_ct_l4proto_find_get);
 149
 150 void nf_ct_l4proto_put(const struct nf_conntrack_l4proto *p)
 151 {
 152         module_put(p->me);
 153 }
 154 EXPORT_SYMBOL_GPL(nf_ct_l4proto_put);
 155
 156 static int kill_l4proto(struct nf_conn *i, void *data)
 157 {
 158         const struct nf_conntrack_l4proto *l4proto;
 159         l4proto = data;
 160         return nf_ct_protonum(i) == l4proto->l4proto;
 161 }
 162
 163 static struct nf_proto_net *nf_ct_l4proto_net(struct net *net,
 164                                 const struct nf_conntrack_l4proto *l4proto)
 165 {
 166         if (l4proto->get_net_proto) {
 167                 /* statically built-in protocols use static per-net */
 168                 return l4proto->get_net_proto(net);
 169         } else if (l4proto->net_id) {
 170                 /* ... and loadable protocols use dynamic per-net */
 171                 return net_generic(net, *l4proto->net_id);
 172         }
 173         return NULL;
 174 }
 175
 176 static
 177 int nf_ct_l4proto_register_sysctl(struct net *net,
 178                                   struct nf_proto_net *pn)
 179 {
 180         int err = 0;
 181
 182 #ifdef CONFIG_SYSCTL
 183         if (pn->ctl_table != NULL) {
 184                 err = nf_ct_register_sysctl(net,
 185                                             &pn->ctl_table_header,
 186                                             "net/netfilter",
 187                                             pn->ctl_table);
 188                 if (err < 0) {
 189                         if (!pn->users) {
 190                                 kfree(pn->ctl_table);
 191                                 pn->ctl_table = NULL;
 192                         }
 193                 }
 194         }
 195 #endif /* CONFIG_SYSCTL */
 196         return err;
 197 }
 198
 199 static
 200 void nf_ct_l4proto_unregister_sysctl(struct nf_proto_net *pn)
 201 {
 202 #ifdef CONFIG_SYSCTL
 203         if (pn->ctl_table_header != NULL)
 204                 nf_ct_unregister_sysctl(&pn->ctl_table_header,
 205                                         &pn->ctl_table,
 206                                         pn->users);
 207 #endif /* CONFIG_SYSCTL */
 208 }
 209
 210 /* FIXME: Allow NULL functions and sub in pointers to generic for
 211    them. --RR */
 212 int nf_ct_l4proto_register_one(const struct nf_conntrack_l4proto *l4proto)
 213 {
 214         int ret = 0;
 215
 216         if ((l4proto->to_nlattr && l4proto->nlattr_size == 0) ||
 217             (l4proto->tuple_to_nlattr && !l4proto->nlattr_tuple_size))
 218                 return -EINVAL;
 219
 220         mutex_lock(&nf_ct_proto_mutex);
 221         if (rcu_dereference_protected(
 222                         nf_ct_protos[l4proto->l4proto],
 223                         lockdep_is_held(&nf_ct_proto_mutex)
 224                         ) != &nf_conntrack_l4proto_generic) {
 225                 ret = -EBUSY;
 226                 goto out_unlock;
 227         }
 228
 229         rcu_assign_pointer(nf_ct_protos[l4proto->l4proto], l4proto);
 230 out_unlock:
 231         mutex_unlock(&nf_ct_proto_mutex);
 232         return ret;
 233 }
 234 EXPORT_SYMBOL_GPL(nf_ct_l4proto_register_one);
 235
 236 int nf_ct_l4proto_pernet_register_one(struct net *net,
 237                                 const struct nf_conntrack_l4proto *l4proto)
 238 {
 239         int ret = 0;
 240         struct nf_proto_net *pn = NULL;
 241
 242         if (l4proto->init_net) {
 243                 ret = l4proto->init_net(net);
 244                 if (ret < 0)
 245                         goto out;
 246         }
 247
 248         pn = nf_ct_l4proto_net(net, l4proto);
 249         if (pn == NULL)
 250                 goto out;
 251
 252         ret = nf_ct_l4proto_register_sysctl(net, pn);
 253         if (ret < 0)
 254                 goto out;
 255
 256         pn->users++;
 257 out:
 258         return ret;
 259 }
 260 EXPORT_SYMBOL_GPL(nf_ct_l4proto_pernet_register_one);
 261
 262 static void __nf_ct_l4proto_unregister_one(const struct nf_conntrack_l4proto *l4proto)
 263
 264 {
 265         BUG_ON(l4proto->l4proto >= ARRAY_SIZE(nf_ct_protos));
 266
 267         BUG_ON(rcu_dereference_protected(
 268                         nf_ct_protos[l4proto->l4proto],
 269                         lockdep_is_held(&nf_ct_proto_mutex)
 270                         ) != l4proto);
 271         rcu_assign_pointer(nf_ct_protos[l4proto->l4proto],
 272                            &nf_conntrack_l4proto_generic);
 273 }
 274
 275 void nf_ct_l4proto_unregister_one(const struct nf_conntrack_l4proto *l4proto)
 276 {
 277         mutex_lock(&nf_ct_proto_mutex);
 278         __nf_ct_l4proto_unregister_one(l4proto);
 279         mutex_unlock(&nf_ct_proto_mutex);
 280
 281         synchronize_net();
 282         /* Remove all contrack entries for this protocol */
 283         nf_ct_iterate_destroy(kill_l4proto, (void *)l4proto);
 284 }
 285 EXPORT_SYMBOL_GPL(nf_ct_l4proto_unregister_one);
 286
 287 void nf_ct_l4proto_pernet_unregister_one(struct net *net,
 288                                 const struct nf_conntrack_l4proto *l4proto)
 289 {
 290         struct nf_proto_net *pn = nf_ct_l4proto_net(net, l4proto);
 291
 292         if (pn == NULL)
 293                 return;
 294
 295         pn->users--;
 296         nf_ct_l4proto_unregister_sysctl(pn);
 297 }
 298 EXPORT_SYMBOL_GPL(nf_ct_l4proto_pernet_unregister_one);
 299
 300 static void
 301 nf_ct_l4proto_unregister(const struct nf_conntrack_l4proto * const l4proto[],
 302                          unsigned int num_proto)
 303 {
 304         int i;
 305
 306         mutex_lock(&nf_ct_proto_mutex);
 307         for (i = 0; i < num_proto; i++)
 308                 __nf_ct_l4proto_unregister_one(l4proto[i]);
 309         mutex_unlock(&nf_ct_proto_mutex);
 310
 311         synchronize_net();
 312
 313         for (i = 0; i < num_proto; i++)
 314                 nf_ct_iterate_destroy(kill_l4proto, (void *)l4proto[i]);
 315 }
 316
 317 static int
 318 nf_ct_l4proto_register(const struct nf_conntrack_l4proto * const l4proto[],
 319                        unsigned int num_proto)
 320 {
 321         int ret = -EINVAL;
 322         unsigned int i;
 323
 324         for (i = 0; i < num_proto; i++) {
 325                 ret = nf_ct_l4proto_register_one(l4proto[i]);
 326                 if (ret < 0)
 327                         break;
 328         }
 329         if (i != num_proto) {
 330                 pr_err("nf_conntrack: can't register l4 %d proto.\n",
 331                        l4proto[i]->l4proto);
 332                 nf_ct_l4proto_unregister(l4proto, i);
 333         }
 334         return ret;
 335 }
 336
 337 int nf_ct_l4proto_pernet_register(struct net *net,
 338                                   const struct nf_conntrack_l4proto *const l4proto[],
 339                                   unsigned int num_proto)
 340 {
 341         int ret = -EINVAL;
 342         unsigned int i;
 343
 344         for (i = 0; i < num_proto; i++) {
 345                 ret = nf_ct_l4proto_pernet_register_one(net, l4proto[i]);
 346                 if (ret < 0)
 347                         break;
 348         }
 349         if (i != num_proto) {
 350                 pr_err("nf_conntrack %d: pernet registration failed\n",
 351                        l4proto[i]->l4proto);
 352                 nf_ct_l4proto_pernet_unregister(net, l4proto, i);
 353         }
 354         return ret;
 355 }
 356 EXPORT_SYMBOL_GPL(nf_ct_l4proto_pernet_register);
 357
 358 void nf_ct_l4proto_pernet_unregister(struct net *net,
 359                                 const struct nf_conntrack_l4proto *const l4proto[],
 360                                 unsigned int num_proto)
 361 {
 362         while (num_proto-- != 0)
 363                 nf_ct_l4proto_pernet_unregister_one(net, l4proto[num_proto]);
 364 }
 365 EXPORT_SYMBOL_GPL(nf_ct_l4proto_pernet_unregister);
 366
 367 static unsigned int ipv4_helper(void *priv,
 368                                 struct sk_buff *skb,
 369                                 const struct nf_hook_state *state)
 370 {
 371         struct nf_conn *ct;
 372         enum ip_conntrack_info ctinfo;
 373         const struct nf_conn_help *help;
 374         const struct nf_conntrack_helper *helper;
 375
 376         /* This is where we call the helper: as the packet goes out. */
 377         ct = nf_ct_get(skb, &ctinfo);
 378         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 379                 return NF_ACCEPT;
 380
 381         help = nfct_help(ct);
 382         if (!help)
 383                 return NF_ACCEPT;
 384
 385         /* rcu_read_lock()ed by nf_hook_thresh */
 386         helper = rcu_dereference(help->helper);
 387         if (!helper)
 388                 return NF_ACCEPT;
 389
 390         return helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
 391                             ct, ctinfo);
 392 }
 393
 394 static unsigned int ipv4_confirm(void *priv,
 395                                  struct sk_buff *skb,
 396                                  const struct nf_hook_state *state)
 397 {
 398         struct nf_conn *ct;
 399         enum ip_conntrack_info ctinfo;
 400
 401         ct = nf_ct_get(skb, &ctinfo);
 402         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 403                 goto out;
 404
 405         /* adjust seqs for loopback traffic only in outgoing direction */
 406         if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
 407             !nf_is_loopback_packet(skb)) {
 408                 if (!nf_ct_seq_adjust(skb, ct, ctinfo, ip_hdrlen(skb))) {
 409                         NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
 410                         return NF_DROP;
 411                 }
 412         }
 413 out:
 414         /* We've seen it coming out the other side: confirm it */
 415         return nf_conntrack_confirm(skb);
 416 }
 417
 418 static unsigned int ipv4_conntrack_in(void *priv,
 419                                       struct sk_buff *skb,
 420                                       const struct nf_hook_state *state)
 421 {
 422         return nf_conntrack_in(skb, state);
 423 }
 424
 425 static unsigned int ipv4_conntrack_local(void *priv,
 426                                          struct sk_buff *skb,
 427                                          const struct nf_hook_state *state)
 428 {
 429         if (ip_is_fragment(ip_hdr(skb))) { /* IP_NODEFRAG setsockopt set */
 430                 enum ip_conntrack_info ctinfo;
 431                 struct nf_conn *tmpl;
 432
 433                 tmpl = nf_ct_get(skb, &ctinfo);
 434                 if (tmpl && nf_ct_is_template(tmpl)) {
 435                         /* when skipping ct, clear templates to avoid fooling
 436                          * later targets/matches
 437                          */
 438                         skb->_nfct = 0;
 439                         nf_ct_put(tmpl);
 440                 }
 441                 return NF_ACCEPT;
 442         }
 443
 444         return nf_conntrack_in(skb, state);
 445 }
 446
 447 /* Connection tracking may drop packets, but never alters them, so
 448  * make it the first hook.
 449  */
 450 static const struct nf_hook_ops ipv4_conntrack_ops[] = {
 451         {
 452                 .hook           = ipv4_conntrack_in,
 453                 .pf             = NFPROTO_IPV4,
 454                 .hooknum        = NF_INET_PRE_ROUTING,
 455                 .priority       = NF_IP_PRI_CONNTRACK,
 456         },
 457         {
 458                 .hook           = ipv4_conntrack_local,
 459                 .pf             = NFPROTO_IPV4,
 460                 .hooknum        = NF_INET_LOCAL_OUT,
 461                 .priority       = NF_IP_PRI_CONNTRACK,
 462         },
 463         {
 464                 .hook           = ipv4_helper,
 465                 .pf             = NFPROTO_IPV4,
 466                 .hooknum        = NF_INET_POST_ROUTING,
 467                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 468         },
 469         {
 470                 .hook           = ipv4_confirm,
 471                 .pf             = NFPROTO_IPV4,
 472                 .hooknum        = NF_INET_POST_ROUTING,
 473                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 474         },
 475         {
 476                 .hook           = ipv4_helper,
 477                 .pf             = NFPROTO_IPV4,
 478                 .hooknum        = NF_INET_LOCAL_IN,
 479                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 480         },
 481         {
 482                 .hook           = ipv4_confirm,
 483                 .pf             = NFPROTO_IPV4,
 484                 .hooknum        = NF_INET_LOCAL_IN,
 485                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 486         },
 487 };
 488
 489 /* Fast function for those who don't want to parse /proc (and I don't
 490  * blame them).
 491  * Reversing the socket's dst/src point of view gives us the reply
 492  * mapping.
 493  */
 494 static int
 495 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 496 {
 497         const struct inet_sock *inet = inet_sk(sk);
 498         const struct nf_conntrack_tuple_hash *h;
 499         struct nf_conntrack_tuple tuple;
 500
 501         memset(&tuple, 0, sizeof(tuple));
 502
 503         lock_sock(sk);
 504         tuple.src.u3.ip = inet->inet_rcv_saddr;
 505         tuple.src.u.tcp.port = inet->inet_sport;
 506         tuple.dst.u3.ip = inet->inet_daddr;
 507         tuple.dst.u.tcp.port = inet->inet_dport;
 508         tuple.src.l3num = PF_INET;
 509         tuple.dst.protonum = sk->sk_protocol;
 510         release_sock(sk);
 511
 512         /* We only do TCP and SCTP at the moment: is there a better way? */
 513         if (tuple.dst.protonum != IPPROTO_TCP &&
 514             tuple.dst.protonum != IPPROTO_SCTP) {
 515                 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
 516                 return -ENOPROTOOPT;
 517         }
 518
 519         if ((unsigned int)*len < sizeof(struct sockaddr_in)) {
 520                 pr_debug("SO_ORIGINAL_DST: len %d not %zu\n",
 521                          *len, sizeof(struct sockaddr_in));
 522                 return -EINVAL;
 523         }
 524
 525         h = nf_conntrack_find_get(sock_net(sk), &nf_ct_zone_dflt, &tuple);
 526         if (h) {
 527                 struct sockaddr_in sin;
 528                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 529
 530                 sin.sin_family = AF_INET;
 531                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 532                         .tuple.dst.u.tcp.port;
 533                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 534                         .tuple.dst.u3.ip;
 535                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 536
 537                 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
 538                          &sin.sin_addr.s_addr, ntohs(sin.sin_port));
 539                 nf_ct_put(ct);
 540                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 541                         return -EFAULT;
 542                 else
 543                         return 0;
 544         }
 545         pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
 546                  &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
 547                  &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
 548         return -ENOENT;
 549 }
 550
 551 static struct nf_sockopt_ops so_getorigdst = {
 552         .pf             = PF_INET,
 553         .get_optmin     = SO_ORIGINAL_DST,
 554         .get_optmax     = SO_ORIGINAL_DST + 1,
 555         .get            = getorigdst,
 556         .owner          = THIS_MODULE,
 557 };
 558
 559 #if IS_ENABLED(CONFIG_IPV6)
 560 static int
 561 ipv6_getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 562 {
 563         struct nf_conntrack_tuple tuple = { .src.l3num = NFPROTO_IPV6 };
 564         const struct ipv6_pinfo *inet6 = inet6_sk(sk);
 565         const struct inet_sock *inet = inet_sk(sk);
 566         const struct nf_conntrack_tuple_hash *h;
 567         struct sockaddr_in6 sin6;
 568         struct nf_conn *ct;
 569         __be32 flow_label;
 570         int bound_dev_if;
 571
 572         lock_sock(sk);
 573         tuple.src.u3.in6 = sk->sk_v6_rcv_saddr;
 574         tuple.src.u.tcp.port = inet->inet_sport;
 575         tuple.dst.u3.in6 = sk->sk_v6_daddr;
 576         tuple.dst.u.tcp.port = inet->inet_dport;
 577         tuple.dst.protonum = sk->sk_protocol;
 578         bound_dev_if = sk->sk_bound_dev_if;
 579         flow_label = inet6->flow_label;
 580         release_sock(sk);
 581
 582         if (tuple.dst.protonum != IPPROTO_TCP &&
 583             tuple.dst.protonum != IPPROTO_SCTP)
 584                 return -ENOPROTOOPT;
 585
 586         if (*len < 0 || (unsigned int)*len < sizeof(sin6))
 587                 return -EINVAL;
 588
 589         h = nf_conntrack_find_get(sock_net(sk), &nf_ct_zone_dflt, &tuple);
 590         if (!h) {
 591                 pr_debug("IP6T_SO_ORIGINAL_DST: Can't find %pI6c/%u-%pI6c/%u.\n",
 592                          &tuple.src.u3.ip6, ntohs(tuple.src.u.tcp.port),
 593                          &tuple.dst.u3.ip6, ntohs(tuple.dst.u.tcp.port));
 594                 return -ENOENT;
 595         }
 596
 597         ct = nf_ct_tuplehash_to_ctrack(h);
 598
 599         sin6.sin6_family = AF_INET6;
 600         sin6.sin6_port = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.u.tcp.port;
 601         sin6.sin6_flowinfo = flow_label & IPV6_FLOWINFO_MASK;
 602         memcpy(&sin6.sin6_addr,
 603                &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.u3.in6,
 604                sizeof(sin6.sin6_addr));
 605
 606         nf_ct_put(ct);
 607         sin6.sin6_scope_id = ipv6_iface_scope_id(&sin6.sin6_addr, bound_dev_if);
 608         return copy_to_user(user, &sin6, sizeof(sin6)) ? -EFAULT : 0;
 609 }
 610
 611 static struct nf_sockopt_ops so_getorigdst6 = {
 612         .pf             = NFPROTO_IPV6,
 613         .get_optmin     = IP6T_SO_ORIGINAL_DST,
 614         .get_optmax     = IP6T_SO_ORIGINAL_DST + 1,
 615         .get            = ipv6_getorigdst,
 616         .owner          = THIS_MODULE,
 617 };
 618
 619 static unsigned int ipv6_confirm(void *priv,
 620                                  struct sk_buff *skb,
 621                                  const struct nf_hook_state *state)
 622 {
 623         struct nf_conn *ct;
 624         enum ip_conntrack_info ctinfo;
 625         unsigned char pnum = ipv6_hdr(skb)->nexthdr;
 626         int protoff;
 627         __be16 frag_off;
 628
 629         ct = nf_ct_get(skb, &ctinfo);
 630         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 631                 goto out;
 632
 633         protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &pnum,
 634                                    &frag_off);
 635         if (protoff < 0 || (frag_off & htons(~0x7)) != 0) {
 636                 pr_debug("proto header not found\n");
 637                 goto out;
 638         }
 639
 640         /* adjust seqs for loopback traffic only in outgoing direction */
 641         if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
 642             !nf_is_loopback_packet(skb)) {
 643                 if (!nf_ct_seq_adjust(skb, ct, ctinfo, protoff)) {
 644                         NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
 645                         return NF_DROP;
 646                 }
 647         }
 648 out:
 649         /* We've seen it coming out the other side: confirm it */
 650         return nf_conntrack_confirm(skb);
 651 }
 652
 653 static unsigned int ipv6_conntrack_in(void *priv,
 654                                       struct sk_buff *skb,
 655                                       const struct nf_hook_state *state)
 656 {
 657         return nf_conntrack_in(skb, state);
 658 }
 659
 660 static unsigned int ipv6_conntrack_local(void *priv,
 661                                          struct sk_buff *skb,
 662                                          const struct nf_hook_state *state)
 663 {
 664         return nf_conntrack_in(skb, state);
 665 }
 666
 667 static unsigned int ipv6_helper(void *priv,
 668                                 struct sk_buff *skb,
 669                                 const struct nf_hook_state *state)
 670 {
 671         struct nf_conn *ct;
 672         const struct nf_conn_help *help;
 673         const struct nf_conntrack_helper *helper;
 674         enum ip_conntrack_info ctinfo;
 675         __be16 frag_off;
 676         int protoff;
 677         u8 nexthdr;
 678
 679         /* This is where we call the helper: as the packet goes out. */
 680         ct = nf_ct_get(skb, &ctinfo);
 681         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 682                 return NF_ACCEPT;
 683
 684         help = nfct_help(ct);
 685         if (!help)
 686                 return NF_ACCEPT;
 687         /* rcu_read_lock()ed by nf_hook_thresh */
 688         helper = rcu_dereference(help->helper);
 689         if (!helper)
 690                 return NF_ACCEPT;
 691
 692         nexthdr = ipv6_hdr(skb)->nexthdr;
 693         protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
 694                                    &frag_off);
 695         if (protoff < 0 || (frag_off & htons(~0x7)) != 0) {
 696                 pr_debug("proto header not found\n");
 697                 return NF_ACCEPT;
 698         }
 699
 700         return helper->help(skb, protoff, ct, ctinfo);
 701 }
 702
 703 static const struct nf_hook_ops ipv6_conntrack_ops[] = {
 704         {
 705                 .hook           = ipv6_conntrack_in,
 706                 .pf             = NFPROTO_IPV6,
 707                 .hooknum        = NF_INET_PRE_ROUTING,
 708                 .priority       = NF_IP6_PRI_CONNTRACK,
 709         },
 710         {
 711                 .hook           = ipv6_conntrack_local,
 712                 .pf             = NFPROTO_IPV6,
 713                 .hooknum        = NF_INET_LOCAL_OUT,
 714                 .priority       = NF_IP6_PRI_CONNTRACK,
 715         },
 716         {
 717                 .hook           = ipv6_helper,
 718                 .pf             = NFPROTO_IPV6,
 719                 .hooknum        = NF_INET_POST_ROUTING,
 720                 .priority       = NF_IP6_PRI_CONNTRACK_HELPER,
 721         },
 722         {
 723                 .hook           = ipv6_confirm,
 724                 .pf             = NFPROTO_IPV6,
 725                 .hooknum        = NF_INET_POST_ROUTING,
 726                 .priority       = NF_IP6_PRI_LAST,
 727         },
 728         {
 729                 .hook           = ipv6_helper,
 730                 .pf             = NFPROTO_IPV6,
 731                 .hooknum        = NF_INET_LOCAL_IN,
 732                 .priority       = NF_IP6_PRI_CONNTRACK_HELPER,
 733         },
 734         {
 735                 .hook           = ipv6_confirm,
 736                 .pf             = NFPROTO_IPV6,
 737                 .hooknum        = NF_INET_LOCAL_IN,
 738                 .priority       = NF_IP6_PRI_LAST - 1,
 739         },
 740 };
 741 #endif
 742
 743 static int nf_ct_tcp_fixup(struct nf_conn *ct, void *_nfproto)
 744 {
 745         u8 nfproto = (unsigned long)_nfproto;
 746
 747         if (nf_ct_l3num(ct) != nfproto)
 748                 return 0;
 749
 750         if (nf_ct_protonum(ct) == IPPROTO_TCP &&
 751             ct->proto.tcp.state == TCP_CONNTRACK_ESTABLISHED) {
 752                 ct->proto.tcp.seen[0].td_maxwin = 0;
 753                 ct->proto.tcp.seen[1].td_maxwin = 0;
 754         }
 755
 756         return 0;
 757 }
 758
 759 static int nf_ct_netns_do_get(struct net *net, u8 nfproto)
 760 {
 761         struct nf_conntrack_net *cnet = net_generic(net, nf_conntrack_net_id);
 762         bool fixup_needed = false;
 763         int err = 0;
 764
 765         mutex_lock(&nf_ct_proto_mutex);
 766
 767         switch (nfproto) {
 768         case NFPROTO_IPV4:
 769                 cnet->users4++;
 770                 if (cnet->users4 > 1)
 771                         goto out_unlock;
 772                 err = nf_defrag_ipv4_enable(net);
 773                 if (err) {
 774                         cnet->users4 = 0;
 775                         goto out_unlock;
 776                 }
 777
 778                 err = nf_register_net_hooks(net, ipv4_conntrack_ops,
 779                                             ARRAY_SIZE(ipv4_conntrack_ops));
 780                 if (err)
 781                         cnet->users4 = 0;
 782                 else
 783                         fixup_needed = true;
 784                 break;
 785 #if IS_ENABLED(CONFIG_IPV6)
 786         case NFPROTO_IPV6:
 787                 cnet->users6++;
 788                 if (cnet->users6 > 1)
 789                         goto out_unlock;
 790                 err = nf_defrag_ipv6_enable(net);
 791                 if (err < 0) {
 792                         cnet->users6 = 0;
 793                         goto out_unlock;
 794                 }
 795
 796                 err = nf_register_net_hooks(net, ipv6_conntrack_ops,
 797                                             ARRAY_SIZE(ipv6_conntrack_ops));
 798                 if (err)
 799                         cnet->users6 = 0;
 800                 else
 801                         fixup_needed = true;
 802                 break;
 803 #endif
 804         default:
 805                 err = -EPROTO;
 806                 break;
 807         }
 808  out_unlock:
 809         mutex_unlock(&nf_ct_proto_mutex);
 810
 811         if (fixup_needed)
 812                 nf_ct_iterate_cleanup_net(net, nf_ct_tcp_fixup,
 813                                           (void *)(unsigned long)nfproto, 0, 0);
 814
 815         return err;
 816 }
 817
 818 static void nf_ct_netns_do_put(struct net *net, u8 nfproto)
 819 {
 820         struct nf_conntrack_net *cnet = net_generic(net, nf_conntrack_net_id);
 821
 822         mutex_lock(&nf_ct_proto_mutex);
 823         switch (nfproto) {
 824         case NFPROTO_IPV4:
 825                 if (cnet->users4 && (--cnet->users4 == 0))
 826                         nf_unregister_net_hooks(net, ipv4_conntrack_ops,
 827                                                 ARRAY_SIZE(ipv4_conntrack_ops));
 828                 break;
 829 #if IS_ENABLED(CONFIG_IPV6)
 830         case NFPROTO_IPV6:
 831                 if (cnet->users6 && (--cnet->users6 == 0))
 832                         nf_unregister_net_hooks(net, ipv6_conntrack_ops,
 833                                                 ARRAY_SIZE(ipv6_conntrack_ops));
 834                 break;
 835 #endif
 836         }
 837
 838         mutex_unlock(&nf_ct_proto_mutex);
 839 }
 840
 841 int nf_ct_netns_get(struct net *net, u8 nfproto)
 842 {
 843         int err;
 844
 845         if (nfproto == NFPROTO_INET) {
 846                 err = nf_ct_netns_do_get(net, NFPROTO_IPV4);
 847                 if (err < 0)
 848                         goto err1;
 849                 err = nf_ct_netns_do_get(net, NFPROTO_IPV6);
 850                 if (err < 0)
 851                         goto err2;
 852         } else {
 853                 err = nf_ct_netns_do_get(net, nfproto);
 854                 if (err < 0)
 855                         goto err1;
 856         }
 857         return 0;
 858
 859 err2:
 860         nf_ct_netns_put(net, NFPROTO_IPV4);
 861 err1:
 862         return err;
 863 }
 864 EXPORT_SYMBOL_GPL(nf_ct_netns_get);
 865
 866 void nf_ct_netns_put(struct net *net, uint8_t nfproto)
 867 {
 868         if (nfproto == NFPROTO_INET) {
 869                 nf_ct_netns_do_put(net, NFPROTO_IPV4);
 870                 nf_ct_netns_do_put(net, NFPROTO_IPV6);
 871         } else {
 872                 nf_ct_netns_do_put(net, nfproto);
 873         }
 874 }
 875 EXPORT_SYMBOL_GPL(nf_ct_netns_put);
 876
 877 static const struct nf_conntrack_l4proto * const builtin_l4proto[] = {
 878         &nf_conntrack_l4proto_tcp,
 879         &nf_conntrack_l4proto_udp,
 880         &nf_conntrack_l4proto_icmp,
 881 #ifdef CONFIG_NF_CT_PROTO_DCCP
 882         &nf_conntrack_l4proto_dccp,
 883 #endif
 884 #ifdef CONFIG_NF_CT_PROTO_SCTP
 885         &nf_conntrack_l4proto_sctp,
 886 #endif
 887 #ifdef CONFIG_NF_CT_PROTO_UDPLITE
 888         &nf_conntrack_l4proto_udplite,
 889 #endif
 890 #if IS_ENABLED(CONFIG_IPV6)
 891         &nf_conntrack_l4proto_icmpv6,
 892 #endif /* CONFIG_IPV6 */
 893 };
 894
 895 int nf_conntrack_proto_init(void)
 896 {
 897         int ret = 0, i;
 898
 899         ret = nf_register_sockopt(&so_getorigdst);
 900         if (ret < 0)
 901                 return ret;
 902
 903 #if IS_ENABLED(CONFIG_IPV6)
 904         ret = nf_register_sockopt(&so_getorigdst6);
 905         if (ret < 0)
 906                 goto cleanup_sockopt;
 907 #endif
 908
 909         for (i = 0; i < ARRAY_SIZE(nf_ct_protos); i++)
 910                 RCU_INIT_POINTER(nf_ct_protos[i],
 911                                  &nf_conntrack_l4proto_generic);
 912
 913         ret = nf_ct_l4proto_register(builtin_l4proto,
 914                                      ARRAY_SIZE(builtin_l4proto));
 915         if (ret < 0)
 916                 goto cleanup_sockopt2;
 917
 918         return ret;
 919 cleanup_sockopt2:
 920         nf_unregister_sockopt(&so_getorigdst);
 921 #if IS_ENABLED(CONFIG_IPV6)
 922 cleanup_sockopt:
 923         nf_unregister_sockopt(&so_getorigdst6);
 924 #endif
 925         return ret;
 926 }
 927
 928 void nf_conntrack_proto_fini(void)
 929 {
 930         nf_unregister_sockopt(&so_getorigdst);
 931 #if IS_ENABLED(CONFIG_IPV6)
 932         nf_unregister_sockopt(&so_getorigdst6);
 933 #endif
 934 }
 935
 936 int nf_conntrack_proto_pernet_init(struct net *net)
 937 {
 938         int err;
 939         struct nf_proto_net *pn = nf_ct_l4proto_net(net,
 940                                         &nf_conntrack_l4proto_generic);
 941
 942         err = nf_conntrack_l4proto_generic.init_net(net);
 943         if (err < 0)
 944                 return err;
 945         err = nf_ct_l4proto_register_sysctl(net,
 946                                             pn);
 947         if (err < 0)
 948                 return err;
 949
 950         err = nf_ct_l4proto_pernet_register(net, builtin_l4proto,
 951                                             ARRAY_SIZE(builtin_l4proto));
 952         if (err < 0) {
 953                 nf_ct_l4proto_unregister_sysctl(pn);
 954                 return err;
 955         }
 956
 957         pn->users++;
 958         return 0;
 959 }
 960
 961 void nf_conntrack_proto_pernet_fini(struct net *net)
 962 {
 963         struct nf_proto_net *pn = nf_ct_l4proto_net(net,
 964                                         &nf_conntrack_l4proto_generic);
 965
 966         nf_ct_l4proto_pernet_unregister(net, builtin_l4proto,
 967                                         ARRAY_SIZE(builtin_l4proto));
 968         pn->users--;
 969         nf_ct_l4proto_unregister_sysctl(pn);
 970 }
 971
 972
 973 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
 974                   &nf_conntrack_htable_size, 0600);
 975
 976 MODULE_ALIAS("ip_conntrack");
 977 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 978 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
 979 MODULE_LICENSE("GPL");