net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c

   1
   2 /* (C) 1999-2001 Paul `Rusty' Russell
   3  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
   4  * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 as
   8  * published by the Free Software Foundation.
   9  */
  10
  11 #include <linux/types.h>
  12 #include <linux/ip.h>
  13 #include <linux/netfilter.h>
  14 #include <linux/module.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/icmp.h>
  17 #include <linux/sysctl.h>
  18 #include <net/route.h>
  19 #include <net/ip.h>
  20
  21 #include <linux/netfilter_ipv4.h>
  22 #include <net/netfilter/nf_conntrack.h>
  23 #include <net/netfilter/nf_conntrack_helper.h>
  24 #include <net/netfilter/nf_conntrack_l4proto.h>
  25 #include <net/netfilter/nf_conntrack_l3proto.h>
  26 #include <net/netfilter/nf_conntrack_zones.h>
  27 #include <net/netfilter/nf_conntrack_core.h>
  28 #include <net/netfilter/nf_conntrack_seqadj.h>
  29 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
  30 #include <net/netfilter/nf_nat_helper.h>
  31 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
  32 #include <net/netfilter/nf_log.h>
  33
  34 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  35                               struct nf_conntrack_tuple *tuple)
  36 {
  37         const __be32 *ap;
  38         __be32 _addrs[2];
  39         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  40                                 sizeof(u_int32_t) * 2, _addrs);
  41         if (ap == NULL)
  42                 return false;
  43
  44         tuple->src.u3.ip = ap[0];
  45         tuple->dst.u3.ip = ap[1];
  46
  47         return true;
  48 }
  49
  50 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
  51                               const struct nf_conntrack_tuple *orig)
  52 {
  53         tuple->src.u3.ip = orig->dst.u3.ip;
  54         tuple->dst.u3.ip = orig->src.u3.ip;
  55
  56         return true;
  57 }
  58
  59 static void ipv4_print_tuple(struct seq_file *s,
  60                             const struct nf_conntrack_tuple *tuple)
  61 {
  62         seq_printf(s, "src=%pI4 dst=%pI4 ",
  63                    &tuple->src.u3.ip, &tuple->dst.u3.ip);
  64 }
  65
  66 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
  67                             unsigned int *dataoff, u_int8_t *protonum)
  68 {
  69         const struct iphdr *iph;
  70         struct iphdr _iph;
  71
  72         iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
  73         if (iph == NULL)
  74                 return -NF_ACCEPT;
  75
  76         /* Conntrack defragments packets, we might still see fragments
  77          * inside ICMP packets though. */
  78         if (iph->frag_off & htons(IP_OFFSET))
  79                 return -NF_ACCEPT;
  80
  81         *dataoff = nhoff + (iph->ihl << 2);
  82         *protonum = iph->protocol;
  83
  84         /* Check bogus IP headers */
  85         if (*dataoff > skb->len) {
  86                 pr_debug("nf_conntrack_ipv4: bogus IPv4 packet: "
  87                          "nhoff %u, ihl %u, skblen %u\n",
  88                          nhoff, iph->ihl << 2, skb->len);
  89                 return -NF_ACCEPT;
  90         }
  91
  92         return NF_ACCEPT;
  93 }
  94
  95 static unsigned int ipv4_helper(void *priv,
  96                                 struct sk_buff *skb,
  97                                 const struct nf_hook_state *state)
  98 {
  99         struct nf_conn *ct;
 100         enum ip_conntrack_info ctinfo;
 101         const struct nf_conn_help *help;
 102         const struct nf_conntrack_helper *helper;
 103
 104         /* This is where we call the helper: as the packet goes out. */
 105         ct = nf_ct_get(skb, &ctinfo);
 106         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 107                 return NF_ACCEPT;
 108
 109         help = nfct_help(ct);
 110         if (!help)
 111                 return NF_ACCEPT;
 112
 113         /* rcu_read_lock()ed by nf_hook_slow */
 114         helper = rcu_dereference(help->helper);
 115         if (!helper)
 116                 return NF_ACCEPT;
 117
 118         return helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
 119                             ct, ctinfo);
 120 }
 121
 122 static unsigned int ipv4_confirm(void *priv,
 123                                  struct sk_buff *skb,
 124                                  const struct nf_hook_state *state)
 125 {
 126         struct nf_conn *ct;
 127         enum ip_conntrack_info ctinfo;
 128
 129         ct = nf_ct_get(skb, &ctinfo);
 130         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 131                 goto out;
 132
 133         /* adjust seqs for loopback traffic only in outgoing direction */
 134         if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
 135             !nf_is_loopback_packet(skb)) {
 136                 if (!nf_ct_seq_adjust(skb, ct, ctinfo, ip_hdrlen(skb))) {
 137                         NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
 138                         return NF_DROP;
 139                 }
 140         }
 141 out:
 142         /* We've seen it coming out the other side: confirm it */
 143         return nf_conntrack_confirm(skb);
 144 }
 145
 146 static unsigned int ipv4_conntrack_in(void *priv,
 147                                       struct sk_buff *skb,
 148                                       const struct nf_hook_state *state)
 149 {
 150         return nf_conntrack_in(state->net, PF_INET, state->hook, skb);
 151 }
 152
 153 static unsigned int ipv4_conntrack_local(void *priv,
 154                                          struct sk_buff *skb,
 155                                          const struct nf_hook_state *state)
 156 {
 157         /* root is playing with raw sockets. */
 158         if (skb->len < sizeof(struct iphdr) ||
 159             ip_hdrlen(skb) < sizeof(struct iphdr))
 160                 return NF_ACCEPT;
 161         return nf_conntrack_in(state->net, PF_INET, state->hook, skb);
 162 }
 163
 164 /* Connection tracking may drop packets, but never alters them, so
 165    make it the first hook. */
 166 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
 167         {
 168                 .hook           = ipv4_conntrack_in,
 169                 .pf             = NFPROTO_IPV4,
 170                 .hooknum        = NF_INET_PRE_ROUTING,
 171                 .priority       = NF_IP_PRI_CONNTRACK,
 172         },
 173         {
 174                 .hook           = ipv4_conntrack_local,
 175                 .pf             = NFPROTO_IPV4,
 176                 .hooknum        = NF_INET_LOCAL_OUT,
 177                 .priority       = NF_IP_PRI_CONNTRACK,
 178         },
 179         {
 180                 .hook           = ipv4_helper,
 181                 .pf             = NFPROTO_IPV4,
 182                 .hooknum        = NF_INET_POST_ROUTING,
 183                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 184         },
 185         {
 186                 .hook           = ipv4_confirm,
 187                 .pf             = NFPROTO_IPV4,
 188                 .hooknum        = NF_INET_POST_ROUTING,
 189                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 190         },
 191         {
 192                 .hook           = ipv4_helper,
 193                 .pf             = NFPROTO_IPV4,
 194                 .hooknum        = NF_INET_LOCAL_IN,
 195                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 196         },
 197         {
 198                 .hook           = ipv4_confirm,
 199                 .pf             = NFPROTO_IPV4,
 200                 .hooknum        = NF_INET_LOCAL_IN,
 201                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 202         },
 203 };
 204
 205 /* Fast function for those who don't want to parse /proc (and I don't
 206    blame them). */
 207 /* Reversing the socket's dst/src point of view gives us the reply
 208    mapping. */
 209 static int
 210 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 211 {
 212         const struct inet_sock *inet = inet_sk(sk);
 213         const struct nf_conntrack_tuple_hash *h;
 214         struct nf_conntrack_tuple tuple;
 215
 216         memset(&tuple, 0, sizeof(tuple));
 217         tuple.src.u3.ip = inet->inet_rcv_saddr;
 218         tuple.src.u.tcp.port = inet->inet_sport;
 219         tuple.dst.u3.ip = inet->inet_daddr;
 220         tuple.dst.u.tcp.port = inet->inet_dport;
 221         tuple.src.l3num = PF_INET;
 222         tuple.dst.protonum = sk->sk_protocol;
 223
 224         /* We only do TCP and SCTP at the moment: is there a better way? */
 225         if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP) {
 226                 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
 227                 return -ENOPROTOOPT;
 228         }
 229
 230         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 231                 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
 232                          *len, sizeof(struct sockaddr_in));
 233                 return -EINVAL;
 234         }
 235
 236         h = nf_conntrack_find_get(sock_net(sk), &nf_ct_zone_dflt, &tuple);
 237         if (h) {
 238                 struct sockaddr_in sin;
 239                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 240
 241                 sin.sin_family = AF_INET;
 242                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 243                         .tuple.dst.u.tcp.port;
 244                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 245                         .tuple.dst.u3.ip;
 246                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 247
 248                 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
 249                          &sin.sin_addr.s_addr, ntohs(sin.sin_port));
 250                 nf_ct_put(ct);
 251                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 252                         return -EFAULT;
 253                 else
 254                         return 0;
 255         }
 256         pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
 257                  &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
 258                  &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
 259         return -ENOENT;
 260 }
 261
 262 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 263
 264 #include <linux/netfilter/nfnetlink.h>
 265 #include <linux/netfilter/nfnetlink_conntrack.h>
 266
 267 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
 268                                 const struct nf_conntrack_tuple *tuple)
 269 {
 270         if (nla_put_in_addr(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) ||
 271             nla_put_in_addr(skb, CTA_IP_V4_DST, tuple->dst.u3.ip))
 272                 goto nla_put_failure;
 273         return 0;
 274
 275 nla_put_failure:
 276         return -1;
 277 }
 278
 279 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
 280         [CTA_IP_V4_SRC] = { .type = NLA_U32 },
 281         [CTA_IP_V4_DST] = { .type = NLA_U32 },
 282 };
 283
 284 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
 285                                 struct nf_conntrack_tuple *t)
 286 {
 287         if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
 288                 return -EINVAL;
 289
 290         t->src.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_SRC]);
 291         t->dst.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_DST]);
 292
 293         return 0;
 294 }
 295
 296 static int ipv4_nlattr_tuple_size(void)
 297 {
 298         return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
 299 }
 300 #endif
 301
 302 static struct nf_sockopt_ops so_getorigdst = {
 303         .pf             = PF_INET,
 304         .get_optmin     = SO_ORIGINAL_DST,
 305         .get_optmax     = SO_ORIGINAL_DST+1,
 306         .get            = getorigdst,
 307         .owner          = THIS_MODULE,
 308 };
 309
 310 static int ipv4_init_net(struct net *net)
 311 {
 312         return 0;
 313 }
 314
 315 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
 316         .l3proto         = PF_INET,
 317         .name            = "ipv4",
 318         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 319         .invert_tuple    = ipv4_invert_tuple,
 320         .print_tuple     = ipv4_print_tuple,
 321         .get_l4proto     = ipv4_get_l4proto,
 322 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 323         .tuple_to_nlattr = ipv4_tuple_to_nlattr,
 324         .nlattr_tuple_size = ipv4_nlattr_tuple_size,
 325         .nlattr_to_tuple = ipv4_nlattr_to_tuple,
 326         .nla_policy      = ipv4_nla_policy,
 327 #endif
 328         .init_net        = ipv4_init_net,
 329         .me              = THIS_MODULE,
 330 };
 331
 332 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
 333                   &nf_conntrack_htable_size, 0600);
 334
 335 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 336 MODULE_ALIAS("ip_conntrack");
 337 MODULE_LICENSE("GPL");
 338
 339 static int ipv4_net_init(struct net *net)
 340 {
 341         int ret = 0;
 342
 343         ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_tcp4);
 344         if (ret < 0) {
 345                 pr_err("nf_conntrack_tcp4: pernet registration failed\n");
 346                 goto out_tcp;
 347         }
 348         ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_udp4);
 349         if (ret < 0) {
 350                 pr_err("nf_conntrack_udp4: pernet registration failed\n");
 351                 goto out_udp;
 352         }
 353         ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_icmp);
 354         if (ret < 0) {
 355                 pr_err("nf_conntrack_icmp4: pernet registration failed\n");
 356                 goto out_icmp;
 357         }
 358         ret = nf_ct_l3proto_pernet_register(net, &nf_conntrack_l3proto_ipv4);
 359         if (ret < 0) {
 360                 pr_err("nf_conntrack_ipv4: pernet registration failed\n");
 361                 goto out_ipv4;
 362         }
 363         return 0;
 364 out_ipv4:
 365         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_icmp);
 366 out_icmp:
 367         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_udp4);
 368 out_udp:
 369         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_tcp4);
 370 out_tcp:
 371         return ret;
 372 }
 373
 374 static void ipv4_net_exit(struct net *net)
 375 {
 376         nf_ct_l3proto_pernet_unregister(net, &nf_conntrack_l3proto_ipv4);
 377         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_icmp);
 378         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_udp4);
 379         nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_tcp4);
 380 }
 381
 382 static struct pernet_operations ipv4_net_ops = {
 383         .init = ipv4_net_init,
 384         .exit = ipv4_net_exit,
 385 };
 386
 387 static int __init nf_conntrack_l3proto_ipv4_init(void)
 388 {
 389         int ret = 0;
 390
 391         need_conntrack();
 392         nf_defrag_ipv4_enable();
 393
 394         ret = nf_register_sockopt(&so_getorigdst);
 395         if (ret < 0) {
 396                 pr_err("Unable to register netfilter socket option\n");
 397                 return ret;
 398         }
 399
 400         ret = register_pernet_subsys(&ipv4_net_ops);
 401         if (ret < 0) {
 402                 pr_err("nf_conntrack_ipv4: can't register pernet ops\n");
 403                 goto cleanup_sockopt;
 404         }
 405
 406         ret = nf_register_hooks(ipv4_conntrack_ops,
 407                                 ARRAY_SIZE(ipv4_conntrack_ops));
 408         if (ret < 0) {
 409                 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
 410                 goto cleanup_pernet;
 411         }
 412
 413         ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_tcp4);
 414         if (ret < 0) {
 415                 pr_err("nf_conntrack_ipv4: can't register tcp4 proto.\n");
 416                 goto cleanup_hooks;
 417         }
 418
 419         ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_udp4);
 420         if (ret < 0) {
 421                 pr_err("nf_conntrack_ipv4: can't register udp4 proto.\n");
 422                 goto cleanup_tcp4;
 423         }
 424
 425         ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_icmp);
 426         if (ret < 0) {
 427                 pr_err("nf_conntrack_ipv4: can't register icmpv4 proto.\n");
 428                 goto cleanup_udp4;
 429         }
 430
 431         ret = nf_ct_l3proto_register(&nf_conntrack_l3proto_ipv4);
 432         if (ret < 0) {
 433                 pr_err("nf_conntrack_ipv4: can't register ipv4 proto.\n");
 434                 goto cleanup_icmpv4;
 435         }
 436
 437         return ret;
 438  cleanup_icmpv4:
 439         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 440  cleanup_udp4:
 441         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 442  cleanup_tcp4:
 443         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 444  cleanup_hooks:
 445         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 446  cleanup_pernet:
 447         unregister_pernet_subsys(&ipv4_net_ops);
 448  cleanup_sockopt:
 449         nf_unregister_sockopt(&so_getorigdst);
 450         return ret;
 451 }
 452
 453 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 454 {
 455         synchronize_net();
 456         nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 457         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 458         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 459         nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 460         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 461         unregister_pernet_subsys(&ipv4_net_ops);
 462         nf_unregister_sockopt(&so_getorigdst);
 463 }
 464
 465 module_init(nf_conntrack_l3proto_ipv4_init);
 466 module_exit(nf_conntrack_l3proto_ipv4_fini);