ff43461704be5c9433d7468f1fe99c63aab7d613
[linux-2.6.git] / 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_bridge.h>
24 #include <linux/netfilter_ipv6.h>
25 #include <net/netfilter/nf_conntrack.h>
26 #include <net/netfilter/nf_conntrack_helper.h>
27 #include <net/netfilter/nf_conntrack_l4proto.h>
28 #include <net/netfilter/nf_conntrack_l3proto.h>
29 #include <net/netfilter/nf_conntrack_core.h>
30 #include <net/netfilter/nf_conntrack_zones.h>
31 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
32 #include <net/netfilter/nf_log.h>
33
34 static bool ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
35                               struct nf_conntrack_tuple *tuple)
36 {
37         const u_int32_t *ap;
38         u_int32_t _addrs[8];
39
40         ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
41                                 sizeof(_addrs), _addrs);
42         if (ap == NULL)
43                 return false;
44
45         memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
46         memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
47
48         return true;
49 }
50
51 static bool ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
52                               const struct nf_conntrack_tuple *orig)
53 {
54         memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
55         memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
56
57         return true;
58 }
59
60 static int ipv6_print_tuple(struct seq_file *s,
61                             const struct nf_conntrack_tuple *tuple)
62 {
63         return seq_printf(s, "src=%pI6 dst=%pI6 ",
64                           tuple->src.u3.ip6, tuple->dst.u3.ip6);
65 }
66
67 /*
68  * Based on ipv6_skip_exthdr() in net/ipv6/exthdr.c
69  *
70  * This function parses (probably truncated) exthdr set "hdr"
71  * of length "len". "nexthdrp" initially points to some place,
72  * where type of the first header can be found.
73  *
74  * It skips all well-known exthdrs, and returns pointer to the start
75  * of unparsable area i.e. the first header with unknown type.
76  * if success, *nexthdr is updated by type/protocol of this header.
77  *
78  * NOTES: - it may return pointer pointing beyond end of packet,
79  *          if the last recognized header is truncated in the middle.
80  *        - if packet is truncated, so that all parsed headers are skipped,
81  *          it returns -1.
82  *        - if packet is fragmented, return pointer of the fragment header.
83  *        - ESP is unparsable for now and considered like
84  *          normal payload protocol.
85  *        - Note also special handling of AUTH header. Thanks to IPsec wizards.
86  */
87
88 static int nf_ct_ipv6_skip_exthdr(const struct sk_buff *skb, int start,
89                                   u8 *nexthdrp, int len)
90 {
91         u8 nexthdr = *nexthdrp;
92
93         while (ipv6_ext_hdr(nexthdr)) {
94                 struct ipv6_opt_hdr hdr;
95                 int hdrlen;
96
97                 if (len < (int)sizeof(struct ipv6_opt_hdr))
98                         return -1;
99                 if (nexthdr == NEXTHDR_NONE)
100                         break;
101                 if (nexthdr == NEXTHDR_FRAGMENT)
102                         break;
103                 if (skb_copy_bits(skb, start, &hdr, sizeof(hdr)))
104                         BUG();
105                 if (nexthdr == NEXTHDR_AUTH)
106                         hdrlen = (hdr.hdrlen+2)<<2;
107                 else
108                         hdrlen = ipv6_optlen(&hdr);
109
110                 nexthdr = hdr.nexthdr;
111                 len -= hdrlen;
112                 start += hdrlen;
113         }
114
115         *nexthdrp = nexthdr;
116         return start;
117 }
118
119 static int ipv6_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
120                             unsigned int *dataoff, u_int8_t *protonum)
121 {
122         unsigned int extoff = nhoff + sizeof(struct ipv6hdr);
123         unsigned char pnum;
124         int protoff;
125
126         if (skb_copy_bits(skb, nhoff + offsetof(struct ipv6hdr, nexthdr),
127                           &pnum, sizeof(pnum)) != 0) {
128                 pr_debug("ip6_conntrack_core: can't get nexthdr\n");
129                 return -NF_ACCEPT;
130         }
131         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum, skb->len - extoff);
132         /*
133          * (protoff == skb->len) mean that the packet doesn't have no data
134          * except of IPv6 & ext headers. but it's tracked anyway. - YK
135          */
136         if ((protoff < 0) || (protoff > skb->len)) {
137                 pr_debug("ip6_conntrack_core: can't find proto in pkt\n");
138                 return -NF_ACCEPT;
139         }
140
141         *dataoff = protoff;
142         *protonum = pnum;
143         return NF_ACCEPT;
144 }
145
146 static unsigned int ipv6_confirm(unsigned int hooknum,
147                                  struct sk_buff *skb,
148                                  const struct net_device *in,
149                                  const struct net_device *out,
150                                  int (*okfn)(struct sk_buff *))
151 {
152         struct nf_conn *ct;
153         const struct nf_conn_help *help;
154         const struct nf_conntrack_helper *helper;
155         enum ip_conntrack_info ctinfo;
156         unsigned int ret, protoff;
157         unsigned int extoff = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
158         unsigned char pnum = ipv6_hdr(skb)->nexthdr;
159
160
161         /* This is where we call the helper: as the packet goes out. */
162         ct = nf_ct_get(skb, &ctinfo);
163         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
164                 goto out;
165
166         help = nfct_help(ct);
167         if (!help)
168                 goto out;
169         /* rcu_read_lock()ed by nf_hook_slow */
170         helper = rcu_dereference(help->helper);
171         if (!helper)
172                 goto out;
173
174         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum,
175                                          skb->len - extoff);
176         if (protoff > skb->len || pnum == NEXTHDR_FRAGMENT) {
177                 pr_debug("proto header not found\n");
178                 return NF_ACCEPT;
179         }
180
181         ret = helper->help(skb, protoff, ct, ctinfo);
182         if (ret != NF_ACCEPT) {
183                 nf_log_packet(NFPROTO_IPV6, hooknum, skb, in, out, NULL,
184                               "nf_ct_%s: dropping packet", helper->name);
185                 return ret;
186         }
187 out:
188         /* We've seen it coming out the other side: confirm it */
189         return nf_conntrack_confirm(skb);
190 }
191
192 static enum ip6_defrag_users nf_ct6_defrag_user(unsigned int hooknum,
193                                                 struct sk_buff *skb)
194 {
195         u16 zone = NF_CT_DEFAULT_ZONE;
196
197         if (skb->nfct)
198                 zone = nf_ct_zone((struct nf_conn *)skb->nfct);
199
200 #ifdef CONFIG_BRIDGE_NETFILTER
201         if (skb->nf_bridge &&
202             skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)
203                 return IP6_DEFRAG_CONNTRACK_BRIDGE_IN + zone;
204 #endif
205         if (hooknum == NF_INET_PRE_ROUTING)
206                 return IP6_DEFRAG_CONNTRACK_IN + zone;
207         else
208                 return IP6_DEFRAG_CONNTRACK_OUT + zone;
209
210 }
211
212 static unsigned int ipv6_defrag(unsigned int hooknum,
213                                 struct sk_buff *skb,
214                                 const struct net_device *in,
215                                 const struct net_device *out,
216                                 int (*okfn)(struct sk_buff *))
217 {
218         struct sk_buff *reasm;
219
220         /* Previously seen (loopback)?  */
221         if (skb->nfct && !nf_ct_is_template((struct nf_conn *)skb->nfct))
222                 return NF_ACCEPT;
223
224         reasm = nf_ct_frag6_gather(skb, nf_ct6_defrag_user(hooknum, skb));
225         /* queued */
226         if (reasm == NULL)
227                 return NF_STOLEN;
228
229         /* error occured or not fragmented */
230         if (reasm == skb)
231                 return NF_ACCEPT;
232
233         nf_ct_frag6_output(hooknum, reasm, (struct net_device *)in,
234                            (struct net_device *)out, okfn);
235
236         return NF_STOLEN;
237 }
238
239 static unsigned int __ipv6_conntrack_in(struct net *net,
240                                         unsigned int hooknum,
241                                         struct sk_buff *skb,
242                                         int (*okfn)(struct sk_buff *))
243 {
244         struct sk_buff *reasm = skb->nfct_reasm;
245
246         /* This packet is fragmented and has reassembled packet. */
247         if (reasm) {
248                 /* Reassembled packet isn't parsed yet ? */
249                 if (!reasm->nfct) {
250                         unsigned int ret;
251
252                         ret = nf_conntrack_in(net, PF_INET6, hooknum, reasm);
253                         if (ret != NF_ACCEPT)
254                                 return ret;
255                 }
256                 nf_conntrack_get(reasm->nfct);
257                 skb->nfct = reasm->nfct;
258                 skb->nfctinfo = reasm->nfctinfo;
259                 return NF_ACCEPT;
260         }
261
262         return nf_conntrack_in(net, PF_INET6, hooknum, skb);
263 }
264
265 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
266                                       struct sk_buff *skb,
267                                       const struct net_device *in,
268                                       const struct net_device *out,
269                                       int (*okfn)(struct sk_buff *))
270 {
271         return __ipv6_conntrack_in(dev_net(in), hooknum, skb, okfn);
272 }
273
274 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
275                                          struct sk_buff *skb,
276                                          const struct net_device *in,
277                                          const struct net_device *out,
278                                          int (*okfn)(struct sk_buff *))
279 {
280         /* root is playing with raw sockets. */
281         if (skb->len < sizeof(struct ipv6hdr)) {
282                 if (net_ratelimit())
283                         pr_notice("ipv6_conntrack_local: packet too short\n");
284                 return NF_ACCEPT;
285         }
286         return __ipv6_conntrack_in(dev_net(out), hooknum, skb, okfn);
287 }
288
289 static struct nf_hook_ops ipv6_conntrack_ops[] __read_mostly = {
290         {
291                 .hook           = ipv6_defrag,
292                 .owner          = THIS_MODULE,
293                 .pf             = NFPROTO_IPV6,
294                 .hooknum        = NF_INET_PRE_ROUTING,
295                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
296         },
297         {
298                 .hook           = ipv6_conntrack_in,
299                 .owner          = THIS_MODULE,
300                 .pf             = NFPROTO_IPV6,
301                 .hooknum        = NF_INET_PRE_ROUTING,
302                 .priority       = NF_IP6_PRI_CONNTRACK,
303         },
304         {
305                 .hook           = ipv6_conntrack_local,
306                 .owner          = THIS_MODULE,
307                 .pf             = NFPROTO_IPV6,
308                 .hooknum        = NF_INET_LOCAL_OUT,
309                 .priority       = NF_IP6_PRI_CONNTRACK,
310         },
311         {
312                 .hook           = ipv6_defrag,
313                 .owner          = THIS_MODULE,
314                 .pf             = NFPROTO_IPV6,
315                 .hooknum        = NF_INET_LOCAL_OUT,
316                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
317         },
318         {
319                 .hook           = ipv6_confirm,
320                 .owner          = THIS_MODULE,
321                 .pf             = NFPROTO_IPV6,
322                 .hooknum        = NF_INET_POST_ROUTING,
323                 .priority       = NF_IP6_PRI_LAST,
324         },
325         {
326                 .hook           = ipv6_confirm,
327                 .owner          = THIS_MODULE,
328                 .pf             = NFPROTO_IPV6,
329                 .hooknum        = NF_INET_LOCAL_IN,
330                 .priority       = NF_IP6_PRI_LAST-1,
331         },
332 };
333
334 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
335
336 #include <linux/netfilter/nfnetlink.h>
337 #include <linux/netfilter/nfnetlink_conntrack.h>
338
339 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
340                                 const struct nf_conntrack_tuple *tuple)
341 {
342         NLA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
343                 &tuple->src.u3.ip6);
344         NLA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
345                 &tuple->dst.u3.ip6);
346         return 0;
347
348 nla_put_failure:
349         return -1;
350 }
351
352 static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = {
353         [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 },
354         [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 },
355 };
356
357 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
358                                 struct nf_conntrack_tuple *t)
359 {
360         if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
361                 return -EINVAL;
362
363         memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]),
364                sizeof(u_int32_t) * 4);
365         memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]),
366                sizeof(u_int32_t) * 4);
367
368         return 0;
369 }
370
371 static int ipv6_nlattr_tuple_size(void)
372 {
373         return nla_policy_len(ipv6_nla_policy, CTA_IP_MAX + 1);
374 }
375 #endif
376
377 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = {
378         .l3proto                = PF_INET6,
379         .name                   = "ipv6",
380         .pkt_to_tuple           = ipv6_pkt_to_tuple,
381         .invert_tuple           = ipv6_invert_tuple,
382         .print_tuple            = ipv6_print_tuple,
383         .get_l4proto            = ipv6_get_l4proto,
384 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
385         .tuple_to_nlattr        = ipv6_tuple_to_nlattr,
386         .nlattr_tuple_size      = ipv6_nlattr_tuple_size,
387         .nlattr_to_tuple        = ipv6_nlattr_to_tuple,
388         .nla_policy             = ipv6_nla_policy,
389 #endif
390 #ifdef CONFIG_SYSCTL
391         .ctl_table_path         = nf_net_netfilter_sysctl_path,
392         .ctl_table              = nf_ct_ipv6_sysctl_table,
393 #endif
394         .me                     = THIS_MODULE,
395 };
396
397 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
398 MODULE_LICENSE("GPL");
399 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
400
401 static int __init nf_conntrack_l3proto_ipv6_init(void)
402 {
403         int ret = 0;
404
405         need_conntrack();
406
407         ret = nf_ct_frag6_init();
408         if (ret < 0) {
409                 pr_err("nf_conntrack_ipv6: can't initialize frag6.\n");
410                 return ret;
411         }
412         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp6);
413         if (ret < 0) {
414                 pr_err("nf_conntrack_ipv6: can't register tcp.\n");
415                 goto cleanup_frag6;
416         }
417
418         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp6);
419         if (ret < 0) {
420                 pr_err("nf_conntrack_ipv6: can't register udp.\n");
421                 goto cleanup_tcp;
422         }
423
424         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmpv6);
425         if (ret < 0) {
426                 pr_err("nf_conntrack_ipv6: can't register icmpv6.\n");
427                 goto cleanup_udp;
428         }
429
430         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6);
431         if (ret < 0) {
432                 pr_err("nf_conntrack_ipv6: can't register ipv6\n");
433                 goto cleanup_icmpv6;
434         }
435
436         ret = nf_register_hooks(ipv6_conntrack_ops,
437                                 ARRAY_SIZE(ipv6_conntrack_ops));
438         if (ret < 0) {
439                 pr_err("nf_conntrack_ipv6: can't register pre-routing defrag "
440                        "hook.\n");
441                 goto cleanup_ipv6;
442         }
443         return ret;
444
445  cleanup_ipv6:
446         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
447  cleanup_icmpv6:
448         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
449  cleanup_udp:
450         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
451  cleanup_tcp:
452         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
453  cleanup_frag6:
454         nf_ct_frag6_cleanup();
455         return ret;
456 }
457
458 static void __exit nf_conntrack_l3proto_ipv6_fini(void)
459 {
460         synchronize_net();
461         nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
462         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
463         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
464         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
465         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
466         nf_ct_frag6_cleanup();
467 }
468
469 module_init(nf_conntrack_l3proto_ipv6_init);
470 module_exit(nf_conntrack_l3proto_ipv6_fini);