net/ipv6/netfilter/nf_conntrack_l3proto_ipv6.c

   1 /*
   2  * Copyright (C)2004 USAGI/WIDE Project
   3  *
   4  * This program is free software; you can redistribute it and/or modify
   5  * it under the terms of the GNU General Public License version 2 as
   6  * published by the Free Software Foundation.
   7  *
   8  * Author:
   9  *      Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
  10  */
  11
  12 #include <linux/types.h>
  13 #include <linux/ipv6.h>
  14 #include <linux/in6.h>
  15 #include <linux/netfilter.h>
  16 #include <linux/module.h>
  17 #include <linux/skbuff.h>
  18 #include <linux/icmp.h>
  19 #include <linux/sysctl.h>
  20 #include <net/ipv6.h>
  21 #include <net/inet_frag.h>
  22
  23 #include <linux/netfilter_ipv6.h>
  24 #include <net/netfilter/nf_conntrack.h>
  25 #include <net/netfilter/nf_conntrack_helper.h>
  26 #include <net/netfilter/nf_conntrack_l4proto.h>
  27 #include <net/netfilter/nf_conntrack_l3proto.h>
  28 #include <net/netfilter/nf_conntrack_core.h>
  29
  30 static bool ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  31                               struct nf_conntrack_tuple *tuple)
  32 {
  33         const u_int32_t *ap;
  34         u_int32_t _addrs[8];
  35
  36         ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
  37                                 sizeof(_addrs), _addrs);
  38         if (ap == NULL)
  39                 return false;
  40
  41         memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
  42         memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
  43
  44         return true;
  45 }
  46
  47 static bool ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
  48                               const struct nf_conntrack_tuple *orig)
  49 {
  50         memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
  51         memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
  52
  53         return true;
  54 }
  55
  56 static int ipv6_print_tuple(struct seq_file *s,
  57                             const struct nf_conntrack_tuple *tuple)
  58 {
  59         return seq_printf(s, "src=%pI6 dst=%pI6 ",
  60                           tuple->src.u3.ip6, tuple->dst.u3.ip6);
  61 }
  62
  63 /*
  64  * Based on ipv6_skip_exthdr() in net/ipv6/exthdr.c
  65  *
  66  * This function parses (probably truncated) exthdr set "hdr"
  67  * of length "len". "nexthdrp" initially points to some place,
  68  * where type of the first header can be found.
  69  *
  70  * It skips all well-known exthdrs, and returns pointer to the start
  71  * of unparsable area i.e. the first header with unknown type.
  72  * if success, *nexthdr is updated by type/protocol of this header.
  73  *
  74  * NOTES: - it may return pointer pointing beyond end of packet,
  75  *          if the last recognized header is truncated in the middle.
  76  *        - if packet is truncated, so that all parsed headers are skipped,
  77  *          it returns -1.
  78  *        - if packet is fragmented, return pointer of the fragment header.
  79  *        - ESP is unparsable for now and considered like
  80  *          normal payload protocol.
  81  *        - Note also special handling of AUTH header. Thanks to IPsec wizards.
  82  */
  83
  84 static int nf_ct_ipv6_skip_exthdr(const struct sk_buff *skb, int start,
  85                                   u8 *nexthdrp, int len)
  86 {
  87         u8 nexthdr = *nexthdrp;
  88
  89         while (ipv6_ext_hdr(nexthdr)) {
  90                 struct ipv6_opt_hdr hdr;
  91                 int hdrlen;
  92
  93                 if (len < (int)sizeof(struct ipv6_opt_hdr))
  94                         return -1;
  95                 if (nexthdr == NEXTHDR_NONE)
  96                         break;
  97                 if (nexthdr == NEXTHDR_FRAGMENT)
  98                         break;
  99                 if (skb_copy_bits(skb, start, &hdr, sizeof(hdr)))
 100                         BUG();
 101                 if (nexthdr == NEXTHDR_AUTH)
 102                         hdrlen = (hdr.hdrlen+2)<<2;
 103                 else
 104                         hdrlen = ipv6_optlen(&hdr);
 105
 106                 nexthdr = hdr.nexthdr;
 107                 len -= hdrlen;
 108                 start += hdrlen;
 109         }
 110
 111         *nexthdrp = nexthdr;
 112         return start;
 113 }
 114
 115 static int ipv6_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
 116                             unsigned int *dataoff, u_int8_t *protonum)
 117 {
 118         unsigned int extoff = nhoff + sizeof(struct ipv6hdr);
 119         unsigned char pnum;
 120         int protoff;
 121
 122         if (skb_copy_bits(skb, nhoff + offsetof(struct ipv6hdr, nexthdr),
 123                           &pnum, sizeof(pnum)) != 0) {
 124                 pr_debug("ip6_conntrack_core: can't get nexthdr\n");
 125                 return -NF_ACCEPT;
 126         }
 127         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum, skb->len - extoff);
 128         /*
 129          * (protoff == skb->len) mean that the packet doesn't have no data
 130          * except of IPv6 & ext headers. but it's tracked anyway. - YK
 131          */
 132         if ((protoff < 0) || (protoff > skb->len)) {
 133                 pr_debug("ip6_conntrack_core: can't find proto in pkt\n");
 134                 return -NF_ACCEPT;
 135         }
 136
 137         *dataoff = protoff;
 138         *protonum = pnum;
 139         return NF_ACCEPT;
 140 }
 141
 142 static unsigned int ipv6_confirm(unsigned int hooknum,
 143                                  struct sk_buff *skb,
 144                                  const struct net_device *in,
 145                                  const struct net_device *out,
 146                                  int (*okfn)(struct sk_buff *))
 147 {
 148         struct nf_conn *ct;
 149         const struct nf_conn_help *help;
 150         const struct nf_conntrack_helper *helper;
 151         enum ip_conntrack_info ctinfo;
 152         unsigned int ret, protoff;
 153         unsigned int extoff = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
 154         unsigned char pnum = ipv6_hdr(skb)->nexthdr;
 155
 156
 157         /* This is where we call the helper: as the packet goes out. */
 158         ct = nf_ct_get(skb, &ctinfo);
 159         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 160                 goto out;
 161
 162         help = nfct_help(ct);
 163         if (!help)
 164                 goto out;
 165         /* rcu_read_lock()ed by nf_hook_slow */
 166         helper = rcu_dereference(help->helper);
 167         if (!helper)
 168                 goto out;
 169
 170         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum,
 171                                          skb->len - extoff);
 172         if (protoff > skb->len || pnum == NEXTHDR_FRAGMENT) {
 173                 pr_debug("proto header not found\n");
 174                 return NF_ACCEPT;
 175         }
 176
 177         ret = helper->help(skb, protoff, ct, ctinfo);
 178         if (ret != NF_ACCEPT)
 179                 return ret;
 180 out:
 181         /* We've seen it coming out the other side: confirm it */
 182         return nf_conntrack_confirm(skb);
 183 }
 184
 185 static unsigned int ipv6_defrag(unsigned int hooknum,
 186                                 struct sk_buff *skb,
 187                                 const struct net_device *in,
 188                                 const struct net_device *out,
 189                                 int (*okfn)(struct sk_buff *))
 190 {
 191         struct sk_buff *reasm;
 192
 193         /* Previously seen (loopback)?  */
 194         if (skb->nfct)
 195                 return NF_ACCEPT;
 196
 197         reasm = nf_ct_frag6_gather(skb);
 198
 199         /* queued */
 200         if (reasm == NULL)
 201                 return NF_STOLEN;
 202
 203         /* error occured or not fragmented */
 204         if (reasm == skb)
 205                 return NF_ACCEPT;
 206
 207         nf_ct_frag6_output(hooknum, reasm, (struct net_device *)in,
 208                            (struct net_device *)out, okfn);
 209
 210         return NF_STOLEN;
 211 }
 212
 213 static unsigned int __ipv6_conntrack_in(struct net *net,
 214                                         unsigned int hooknum,
 215                                         struct sk_buff *skb,
 216                                         int (*okfn)(struct sk_buff *))
 217 {
 218         struct sk_buff *reasm = skb->nfct_reasm;
 219
 220         /* This packet is fragmented and has reassembled packet. */
 221         if (reasm) {
 222                 /* Reassembled packet isn't parsed yet ? */
 223                 if (!reasm->nfct) {
 224                         unsigned int ret;
 225
 226                         ret = nf_conntrack_in(net, PF_INET6, hooknum, reasm);
 227                         if (ret != NF_ACCEPT)
 228                                 return ret;
 229                 }
 230                 nf_conntrack_get(reasm->nfct);
 231                 skb->nfct = reasm->nfct;
 232                 skb->nfctinfo = reasm->nfctinfo;
 233                 return NF_ACCEPT;
 234         }
 235
 236         return nf_conntrack_in(net, PF_INET6, hooknum, skb);
 237 }
 238
 239 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
 240                                       struct sk_buff *skb,
 241                                       const struct net_device *in,
 242                                       const struct net_device *out,
 243                                       int (*okfn)(struct sk_buff *))
 244 {
 245         return __ipv6_conntrack_in(dev_net(in), hooknum, skb, okfn);
 246 }
 247
 248 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
 249                                          struct sk_buff *skb,
 250                                          const struct net_device *in,
 251                                          const struct net_device *out,
 252                                          int (*okfn)(struct sk_buff *))
 253 {
 254         /* root is playing with raw sockets. */
 255         if (skb->len < sizeof(struct ipv6hdr)) {
 256                 if (net_ratelimit())
 257                         printk("ipv6_conntrack_local: packet too short\n");
 258                 return NF_ACCEPT;
 259         }
 260         return __ipv6_conntrack_in(dev_net(out), hooknum, skb, okfn);
 261 }
 262
 263 static struct nf_hook_ops ipv6_conntrack_ops[] __read_mostly = {
 264         {
 265                 .hook           = ipv6_defrag,
 266                 .owner          = THIS_MODULE,
 267                 .pf             = PF_INET6,
 268                 .hooknum        = NF_INET_PRE_ROUTING,
 269                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
 270         },
 271         {
 272                 .hook           = ipv6_conntrack_in,
 273                 .owner          = THIS_MODULE,
 274                 .pf             = PF_INET6,
 275                 .hooknum        = NF_INET_PRE_ROUTING,
 276                 .priority       = NF_IP6_PRI_CONNTRACK,
 277         },
 278         {
 279                 .hook           = ipv6_conntrack_local,
 280                 .owner          = THIS_MODULE,
 281                 .pf             = PF_INET6,
 282                 .hooknum        = NF_INET_LOCAL_OUT,
 283                 .priority       = NF_IP6_PRI_CONNTRACK,
 284         },
 285         {
 286                 .hook           = ipv6_defrag,
 287                 .owner          = THIS_MODULE,
 288                 .pf             = PF_INET6,
 289                 .hooknum        = NF_INET_LOCAL_OUT,
 290                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
 291         },
 292         {
 293                 .hook           = ipv6_confirm,
 294                 .owner          = THIS_MODULE,
 295                 .pf             = PF_INET6,
 296                 .hooknum        = NF_INET_POST_ROUTING,
 297                 .priority       = NF_IP6_PRI_LAST,
 298         },
 299         {
 300                 .hook           = ipv6_confirm,
 301                 .owner          = THIS_MODULE,
 302                 .pf             = PF_INET6,
 303                 .hooknum        = NF_INET_LOCAL_IN,
 304                 .priority       = NF_IP6_PRI_LAST-1,
 305         },
 306 };
 307
 308 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 309
 310 #include <linux/netfilter/nfnetlink.h>
 311 #include <linux/netfilter/nfnetlink_conntrack.h>
 312
 313 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
 314                                 const struct nf_conntrack_tuple *tuple)
 315 {
 316         NLA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
 317                 &tuple->src.u3.ip6);
 318         NLA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
 319                 &tuple->dst.u3.ip6);
 320         return 0;
 321
 322 nla_put_failure:
 323         return -1;
 324 }
 325
 326 static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = {
 327         [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 },
 328         [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 },
 329 };
 330
 331 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
 332                                 struct nf_conntrack_tuple *t)
 333 {
 334         if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
 335                 return -EINVAL;
 336
 337         memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]),
 338                sizeof(u_int32_t) * 4);
 339         memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]),
 340                sizeof(u_int32_t) * 4);
 341
 342         return 0;
 343 }
 344 #endif
 345
 346 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = {
 347         .l3proto                = PF_INET6,
 348         .name                   = "ipv6",
 349         .pkt_to_tuple           = ipv6_pkt_to_tuple,
 350         .invert_tuple           = ipv6_invert_tuple,
 351         .print_tuple            = ipv6_print_tuple,
 352         .get_l4proto            = ipv6_get_l4proto,
 353 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 354         .tuple_to_nlattr        = ipv6_tuple_to_nlattr,
 355         .nlattr_to_tuple        = ipv6_nlattr_to_tuple,
 356         .nla_policy             = ipv6_nla_policy,
 357 #endif
 358 #ifdef CONFIG_SYSCTL
 359         .ctl_table_path         = nf_net_netfilter_sysctl_path,
 360         .ctl_table              = nf_ct_ipv6_sysctl_table,
 361 #endif
 362         .me                     = THIS_MODULE,
 363 };
 364
 365 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
 366 MODULE_LICENSE("GPL");
 367 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
 368
 369 static int __init nf_conntrack_l3proto_ipv6_init(void)
 370 {
 371         int ret = 0;
 372
 373         need_conntrack();
 374
 375         ret = nf_ct_frag6_init();
 376         if (ret < 0) {
 377                 printk("nf_conntrack_ipv6: can't initialize frag6.\n");
 378                 return ret;
 379         }
 380         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp6);
 381         if (ret < 0) {
 382                 printk("nf_conntrack_ipv6: can't register tcp.\n");
 383                 goto cleanup_frag6;
 384         }
 385
 386         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp6);
 387         if (ret < 0) {
 388                 printk("nf_conntrack_ipv6: can't register udp.\n");
 389                 goto cleanup_tcp;
 390         }
 391
 392         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmpv6);
 393         if (ret < 0) {
 394                 printk("nf_conntrack_ipv6: can't register icmpv6.\n");
 395                 goto cleanup_udp;
 396         }
 397
 398         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6);
 399         if (ret < 0) {
 400                 printk("nf_conntrack_ipv6: can't register ipv6\n");
 401                 goto cleanup_icmpv6;
 402         }
 403
 404         ret = nf_register_hooks(ipv6_conntrack_ops,
 405                                 ARRAY_SIZE(ipv6_conntrack_ops));
 406         if (ret < 0) {
 407                 printk("nf_conntrack_ipv6: can't register pre-routing defrag "
 408                        "hook.\n");
 409                 goto cleanup_ipv6;
 410         }
 411         return ret;
 412
 413  cleanup_ipv6:
 414         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
 415  cleanup_icmpv6:
 416         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
 417  cleanup_udp:
 418         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
 419  cleanup_tcp:
 420         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
 421  cleanup_frag6:
 422         nf_ct_frag6_cleanup();
 423         return ret;
 424 }
 425
 426 static void __exit nf_conntrack_l3proto_ipv6_fini(void)
 427 {
 428         synchronize_net();
 429         nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
 430         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
 431         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
 432         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
 433         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
 434         nf_ct_frag6_cleanup();
 435 }
 436
 437 module_init(nf_conntrack_l3proto_ipv6_init);
 438 module_exit(nf_conntrack_l3proto_ipv6_fini);