netfilter: nf_conntrack: calculate per-protocol nlattr size
[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_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 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
30
31 static bool ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
32                               struct nf_conntrack_tuple *tuple)
33 {
34         const u_int32_t *ap;
35         u_int32_t _addrs[8];
36
37         ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
38                                 sizeof(_addrs), _addrs);
39         if (ap == NULL)
40                 return false;
41
42         memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
43         memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
44
45         return true;
46 }
47
48 static bool ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
49                               const struct nf_conntrack_tuple *orig)
50 {
51         memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
52         memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
53
54         return true;
55 }
56
57 static int ipv6_print_tuple(struct seq_file *s,
58                             const struct nf_conntrack_tuple *tuple)
59 {
60         return seq_printf(s, "src=%pI6 dst=%pI6 ",
61                           tuple->src.u3.ip6, 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(struct net *net,
215                                         unsigned int hooknum,
216                                         struct sk_buff *skb,
217                                         int (*okfn)(struct sk_buff *))
218 {
219         struct sk_buff *reasm = skb->nfct_reasm;
220
221         /* This packet is fragmented and has reassembled packet. */
222         if (reasm) {
223                 /* Reassembled packet isn't parsed yet ? */
224                 if (!reasm->nfct) {
225                         unsigned int ret;
226
227                         ret = nf_conntrack_in(net, PF_INET6, hooknum, reasm);
228                         if (ret != NF_ACCEPT)
229                                 return ret;
230                 }
231                 nf_conntrack_get(reasm->nfct);
232                 skb->nfct = reasm->nfct;
233                 skb->nfctinfo = reasm->nfctinfo;
234                 return NF_ACCEPT;
235         }
236
237         return nf_conntrack_in(net, PF_INET6, hooknum, skb);
238 }
239
240 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
241                                       struct sk_buff *skb,
242                                       const struct net_device *in,
243                                       const struct net_device *out,
244                                       int (*okfn)(struct sk_buff *))
245 {
246         return __ipv6_conntrack_in(dev_net(in), hooknum, skb, okfn);
247 }
248
249 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
250                                          struct sk_buff *skb,
251                                          const struct net_device *in,
252                                          const struct net_device *out,
253                                          int (*okfn)(struct sk_buff *))
254 {
255         /* root is playing with raw sockets. */
256         if (skb->len < sizeof(struct ipv6hdr)) {
257                 if (net_ratelimit())
258                         printk("ipv6_conntrack_local: packet too short\n");
259                 return NF_ACCEPT;
260         }
261         return __ipv6_conntrack_in(dev_net(out), hooknum, skb, okfn);
262 }
263
264 static struct nf_hook_ops ipv6_conntrack_ops[] __read_mostly = {
265         {
266                 .hook           = ipv6_defrag,
267                 .owner          = THIS_MODULE,
268                 .pf             = PF_INET6,
269                 .hooknum        = NF_INET_PRE_ROUTING,
270                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
271         },
272         {
273                 .hook           = ipv6_conntrack_in,
274                 .owner          = THIS_MODULE,
275                 .pf             = PF_INET6,
276                 .hooknum        = NF_INET_PRE_ROUTING,
277                 .priority       = NF_IP6_PRI_CONNTRACK,
278         },
279         {
280                 .hook           = ipv6_conntrack_local,
281                 .owner          = THIS_MODULE,
282                 .pf             = PF_INET6,
283                 .hooknum        = NF_INET_LOCAL_OUT,
284                 .priority       = NF_IP6_PRI_CONNTRACK,
285         },
286         {
287                 .hook           = ipv6_defrag,
288                 .owner          = THIS_MODULE,
289                 .pf             = PF_INET6,
290                 .hooknum        = NF_INET_LOCAL_OUT,
291                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
292         },
293         {
294                 .hook           = ipv6_confirm,
295                 .owner          = THIS_MODULE,
296                 .pf             = PF_INET6,
297                 .hooknum        = NF_INET_POST_ROUTING,
298                 .priority       = NF_IP6_PRI_LAST,
299         },
300         {
301                 .hook           = ipv6_confirm,
302                 .owner          = THIS_MODULE,
303                 .pf             = PF_INET6,
304                 .hooknum        = NF_INET_LOCAL_IN,
305                 .priority       = NF_IP6_PRI_LAST-1,
306         },
307 };
308
309 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
310
311 #include <linux/netfilter/nfnetlink.h>
312 #include <linux/netfilter/nfnetlink_conntrack.h>
313
314 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
315                                 const struct nf_conntrack_tuple *tuple)
316 {
317         NLA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
318                 &tuple->src.u3.ip6);
319         NLA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
320                 &tuple->dst.u3.ip6);
321         return 0;
322
323 nla_put_failure:
324         return -1;
325 }
326
327 static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = {
328         [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 },
329         [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 },
330 };
331
332 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
333                                 struct nf_conntrack_tuple *t)
334 {
335         if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
336                 return -EINVAL;
337
338         memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]),
339                sizeof(u_int32_t) * 4);
340         memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]),
341                sizeof(u_int32_t) * 4);
342
343         return 0;
344 }
345
346 static int ipv6_nlattr_tuple_size(void)
347 {
348         return nla_policy_len(ipv6_nla_policy, CTA_IP_MAX + 1);
349 }
350 #endif
351
352 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = {
353         .l3proto                = PF_INET6,
354         .name                   = "ipv6",
355         .pkt_to_tuple           = ipv6_pkt_to_tuple,
356         .invert_tuple           = ipv6_invert_tuple,
357         .print_tuple            = ipv6_print_tuple,
358         .get_l4proto            = ipv6_get_l4proto,
359 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
360         .tuple_to_nlattr        = ipv6_tuple_to_nlattr,
361         .nlattr_tuple_size      = ipv6_nlattr_tuple_size,
362         .nlattr_to_tuple        = ipv6_nlattr_to_tuple,
363         .nla_policy             = ipv6_nla_policy,
364 #endif
365 #ifdef CONFIG_SYSCTL
366         .ctl_table_path         = nf_net_netfilter_sysctl_path,
367         .ctl_table              = nf_ct_ipv6_sysctl_table,
368 #endif
369         .me                     = THIS_MODULE,
370 };
371
372 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
373 MODULE_LICENSE("GPL");
374 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
375
376 static int __init nf_conntrack_l3proto_ipv6_init(void)
377 {
378         int ret = 0;
379
380         need_conntrack();
381
382         ret = nf_ct_frag6_init();
383         if (ret < 0) {
384                 printk("nf_conntrack_ipv6: can't initialize frag6.\n");
385                 return ret;
386         }
387         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp6);
388         if (ret < 0) {
389                 printk("nf_conntrack_ipv6: can't register tcp.\n");
390                 goto cleanup_frag6;
391         }
392
393         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp6);
394         if (ret < 0) {
395                 printk("nf_conntrack_ipv6: can't register udp.\n");
396                 goto cleanup_tcp;
397         }
398
399         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmpv6);
400         if (ret < 0) {
401                 printk("nf_conntrack_ipv6: can't register icmpv6.\n");
402                 goto cleanup_udp;
403         }
404
405         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6);
406         if (ret < 0) {
407                 printk("nf_conntrack_ipv6: can't register ipv6\n");
408                 goto cleanup_icmpv6;
409         }
410
411         ret = nf_register_hooks(ipv6_conntrack_ops,
412                                 ARRAY_SIZE(ipv6_conntrack_ops));
413         if (ret < 0) {
414                 printk("nf_conntrack_ipv6: can't register pre-routing defrag "
415                        "hook.\n");
416                 goto cleanup_ipv6;
417         }
418         return ret;
419
420  cleanup_ipv6:
421         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
422  cleanup_icmpv6:
423         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
424  cleanup_udp:
425         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
426  cleanup_tcp:
427         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
428  cleanup_frag6:
429         nf_ct_frag6_cleanup();
430         return ret;
431 }
432
433 static void __exit nf_conntrack_l3proto_ipv6_fini(void)
434 {
435         synchronize_net();
436         nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
437         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
438         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
439         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
440         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
441         nf_ct_frag6_cleanup();
442 }
443
444 module_init(nf_conntrack_l3proto_ipv6_init);
445 module_exit(nf_conntrack_l3proto_ipv6_fini);