- Update to 2.6.25-rc3.
[linux-flexiantxendom0-3.2.10.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_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 int 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 0;
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 1;
45 }
46
47 static int 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 1;
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=" NIP6_FMT " dst=" NIP6_FMT " ",
60                           NIP6(*((struct in6_addr *)tuple->src.u3.ip6)),
61                           NIP6(*((struct in6_addr *)tuple->dst.u3.ip6)));
62 }
63
64 /*
65  * Based on ipv6_skip_exthdr() in net/ipv6/exthdr.c
66  *
67  * This function parses (probably truncated) exthdr set "hdr"
68  * of length "len". "nexthdrp" initially points to some place,
69  * where type of the first header can be found.
70  *
71  * It skips all well-known exthdrs, and returns pointer to the start
72  * of unparsable area i.e. the first header with unknown type.
73  * if success, *nexthdr is updated by type/protocol of this header.
74  *
75  * NOTES: - it may return pointer pointing beyond end of packet,
76  *          if the last recognized header is truncated in the middle.
77  *        - if packet is truncated, so that all parsed headers are skipped,
78  *          it returns -1.
79  *        - if packet is fragmented, return pointer of the fragment header.
80  *        - ESP is unparsable for now and considered like
81  *          normal payload protocol.
82  *        - Note also special handling of AUTH header. Thanks to IPsec wizards.
83  */
84
85 static int nf_ct_ipv6_skip_exthdr(const struct sk_buff *skb, int start,
86                                   u8 *nexthdrp, int len)
87 {
88         u8 nexthdr = *nexthdrp;
89
90         while (ipv6_ext_hdr(nexthdr)) {
91                 struct ipv6_opt_hdr hdr;
92                 int hdrlen;
93
94                 if (len < (int)sizeof(struct ipv6_opt_hdr))
95                         return -1;
96                 if (nexthdr == NEXTHDR_NONE)
97                         break;
98                 if (nexthdr == NEXTHDR_FRAGMENT)
99                         break;
100                 if (skb_copy_bits(skb, start, &hdr, sizeof(hdr)))
101                         BUG();
102                 if (nexthdr == NEXTHDR_AUTH)
103                         hdrlen = (hdr.hdrlen+2)<<2;
104                 else
105                         hdrlen = ipv6_optlen(&hdr);
106
107                 nexthdr = hdr.nexthdr;
108                 len -= hdrlen;
109                 start += hdrlen;
110         }
111
112         *nexthdrp = nexthdr;
113         return start;
114 }
115
116 static int ipv6_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
117                             unsigned int *dataoff, u_int8_t *protonum)
118 {
119         unsigned int extoff = nhoff + sizeof(struct ipv6hdr);
120         unsigned char pnum;
121         int protoff;
122
123         if (skb_copy_bits(skb, nhoff + offsetof(struct ipv6hdr, nexthdr),
124                           &pnum, sizeof(pnum)) != 0) {
125                 pr_debug("ip6_conntrack_core: can't get nexthdr\n");
126                 return -NF_ACCEPT;
127         }
128         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum, skb->len - extoff);
129         /*
130          * (protoff == skb->len) mean that the packet doesn't have no data
131          * except of IPv6 & ext headers. but it's tracked anyway. - YK
132          */
133         if ((protoff < 0) || (protoff > skb->len)) {
134                 pr_debug("ip6_conntrack_core: can't find proto in pkt\n");
135                 return -NF_ACCEPT;
136         }
137
138         *dataoff = protoff;
139         *protonum = pnum;
140         return NF_ACCEPT;
141 }
142
143 static unsigned int ipv6_confirm(unsigned int hooknum,
144                                  struct sk_buff *skb,
145                                  const struct net_device *in,
146                                  const struct net_device *out,
147                                  int (*okfn)(struct sk_buff *))
148 {
149         struct nf_conn *ct;
150         const struct nf_conn_help *help;
151         const struct nf_conntrack_helper *helper;
152         enum ip_conntrack_info ctinfo;
153         unsigned int ret, protoff;
154         unsigned int extoff = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
155         unsigned char pnum = ipv6_hdr(skb)->nexthdr;
156
157
158         /* This is where we call the helper: as the packet goes out. */
159         ct = nf_ct_get(skb, &ctinfo);
160         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
161                 goto out;
162
163         help = nfct_help(ct);
164         if (!help)
165                 goto out;
166         /* rcu_read_lock()ed by nf_hook_slow */
167         helper = rcu_dereference(help->helper);
168         if (!helper)
169                 goto out;
170
171         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum,
172                                          skb->len - extoff);
173         if (protoff > skb->len || pnum == NEXTHDR_FRAGMENT) {
174                 pr_debug("proto header not found\n");
175                 return NF_ACCEPT;
176         }
177
178         ret = helper->help(skb, protoff, ct, ctinfo);
179         if (ret != NF_ACCEPT)
180                 return ret;
181 out:
182         /* We've seen it coming out the other side: confirm it */
183         return nf_conntrack_confirm(skb);
184 }
185
186 static unsigned int ipv6_defrag(unsigned int hooknum,
187                                 struct sk_buff *skb,
188                                 const struct net_device *in,
189                                 const struct net_device *out,
190                                 int (*okfn)(struct sk_buff *))
191 {
192         struct sk_buff *reasm;
193
194         /* Previously seen (loopback)?  */
195         if (skb->nfct)
196                 return NF_ACCEPT;
197
198         reasm = nf_ct_frag6_gather(skb);
199
200         /* queued */
201         if (reasm == NULL)
202                 return NF_STOLEN;
203
204         /* error occured or not fragmented */
205         if (reasm == skb)
206                 return NF_ACCEPT;
207
208         nf_ct_frag6_output(hooknum, reasm, (struct net_device *)in,
209                            (struct net_device *)out, okfn);
210
211         return NF_STOLEN;
212 }
213
214 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
215                                       struct sk_buff *skb,
216                                       const struct net_device *in,
217                                       const struct net_device *out,
218                                       int (*okfn)(struct sk_buff *))
219 {
220         struct sk_buff *reasm = skb->nfct_reasm;
221
222         /* This packet is fragmented and has reassembled packet. */
223         if (reasm) {
224                 /* Reassembled packet isn't parsed yet ? */
225                 if (!reasm->nfct) {
226                         unsigned int ret;
227
228                         ret = nf_conntrack_in(PF_INET6, hooknum, reasm);
229                         if (ret != NF_ACCEPT)
230                                 return ret;
231                 }
232                 nf_conntrack_get(reasm->nfct);
233                 skb->nfct = reasm->nfct;
234                 skb->nfctinfo = reasm->nfctinfo;
235                 return NF_ACCEPT;
236         }
237
238         return nf_conntrack_in(PF_INET6, hooknum, skb);
239 }
240
241 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
242                                          struct sk_buff *skb,
243                                          const struct net_device *in,
244                                          const struct net_device *out,
245                                          int (*okfn)(struct sk_buff *))
246 {
247         /* root is playing with raw sockets. */
248         if (skb->len < sizeof(struct ipv6hdr)) {
249                 if (net_ratelimit())
250                         printk("ipv6_conntrack_local: packet too short\n");
251                 return NF_ACCEPT;
252         }
253         return ipv6_conntrack_in(hooknum, skb, in, out, okfn);
254 }
255
256 static struct nf_hook_ops ipv6_conntrack_ops[] __read_mostly = {
257         {
258                 .hook           = ipv6_defrag,
259                 .owner          = THIS_MODULE,
260                 .pf             = PF_INET6,
261                 .hooknum        = NF_INET_PRE_ROUTING,
262                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
263         },
264         {
265                 .hook           = ipv6_conntrack_in,
266                 .owner          = THIS_MODULE,
267                 .pf             = PF_INET6,
268                 .hooknum        = NF_INET_PRE_ROUTING,
269                 .priority       = NF_IP6_PRI_CONNTRACK,
270         },
271         {
272                 .hook           = ipv6_conntrack_local,
273                 .owner          = THIS_MODULE,
274                 .pf             = PF_INET6,
275                 .hooknum        = NF_INET_LOCAL_OUT,
276                 .priority       = NF_IP6_PRI_CONNTRACK,
277         },
278         {
279                 .hook           = ipv6_defrag,
280                 .owner          = THIS_MODULE,
281                 .pf             = PF_INET6,
282                 .hooknum        = NF_INET_LOCAL_OUT,
283                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
284         },
285         {
286                 .hook           = ipv6_confirm,
287                 .owner          = THIS_MODULE,
288                 .pf             = PF_INET6,
289                 .hooknum        = NF_INET_POST_ROUTING,
290                 .priority       = NF_IP6_PRI_LAST,
291         },
292         {
293                 .hook           = ipv6_confirm,
294                 .owner          = THIS_MODULE,
295                 .pf             = PF_INET6,
296                 .hooknum        = NF_INET_LOCAL_IN,
297                 .priority       = NF_IP6_PRI_LAST-1,
298         },
299 };
300
301 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
302
303 #include <linux/netfilter/nfnetlink.h>
304 #include <linux/netfilter/nfnetlink_conntrack.h>
305
306 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
307                                 const struct nf_conntrack_tuple *tuple)
308 {
309         NLA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
310                 &tuple->src.u3.ip6);
311         NLA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
312                 &tuple->dst.u3.ip6);
313         return 0;
314
315 nla_put_failure:
316         return -1;
317 }
318
319 static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = {
320         [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 },
321         [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 },
322 };
323
324 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
325                                 struct nf_conntrack_tuple *t)
326 {
327         if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
328                 return -EINVAL;
329
330         memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]),
331                sizeof(u_int32_t) * 4);
332         memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]),
333                sizeof(u_int32_t) * 4);
334
335         return 0;
336 }
337 #endif
338
339 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = {
340         .l3proto                = PF_INET6,
341         .name                   = "ipv6",
342         .pkt_to_tuple           = ipv6_pkt_to_tuple,
343         .invert_tuple           = ipv6_invert_tuple,
344         .print_tuple            = ipv6_print_tuple,
345         .get_l4proto            = ipv6_get_l4proto,
346 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
347         .tuple_to_nlattr        = ipv6_tuple_to_nlattr,
348         .nlattr_to_tuple        = ipv6_nlattr_to_tuple,
349         .nla_policy             = ipv6_nla_policy,
350 #endif
351 #ifdef CONFIG_SYSCTL
352         .ctl_table_path         = nf_net_netfilter_sysctl_path,
353         .ctl_table              = nf_ct_ipv6_sysctl_table,
354 #endif
355         .me                     = THIS_MODULE,
356 };
357
358 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
359 MODULE_LICENSE("GPL");
360 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
361
362 static int __init nf_conntrack_l3proto_ipv6_init(void)
363 {
364         int ret = 0;
365
366         need_conntrack();
367
368         ret = nf_ct_frag6_init();
369         if (ret < 0) {
370                 printk("nf_conntrack_ipv6: can't initialize frag6.\n");
371                 return ret;
372         }
373         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp6);
374         if (ret < 0) {
375                 printk("nf_conntrack_ipv6: can't register tcp.\n");
376                 goto cleanup_frag6;
377         }
378
379         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp6);
380         if (ret < 0) {
381                 printk("nf_conntrack_ipv6: can't register udp.\n");
382                 goto cleanup_tcp;
383         }
384
385         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmpv6);
386         if (ret < 0) {
387                 printk("nf_conntrack_ipv6: can't register icmpv6.\n");
388                 goto cleanup_udp;
389         }
390
391         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6);
392         if (ret < 0) {
393                 printk("nf_conntrack_ipv6: can't register ipv6\n");
394                 goto cleanup_icmpv6;
395         }
396
397         ret = nf_register_hooks(ipv6_conntrack_ops,
398                                 ARRAY_SIZE(ipv6_conntrack_ops));
399         if (ret < 0) {
400                 printk("nf_conntrack_ipv6: can't register pre-routing defrag "
401                        "hook.\n");
402                 goto cleanup_ipv6;
403         }
404         return ret;
405
406  cleanup_ipv6:
407         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
408  cleanup_icmpv6:
409         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
410  cleanup_udp:
411         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
412  cleanup_tcp:
413         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
414  cleanup_frag6:
415         nf_ct_frag6_cleanup();
416         return ret;
417 }
418
419 static void __exit nf_conntrack_l3proto_ipv6_fini(void)
420 {
421         synchronize_net();
422         nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
423         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
424         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
425         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
426         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
427         nf_ct_frag6_cleanup();
428 }
429
430 module_init(nf_conntrack_l3proto_ipv6_init);
431 module_exit(nf_conntrack_l3proto_ipv6_fini);