9108ecc22beafc64e61e6a42954e7984fd537640
[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  * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
12  *      - support Layer 3 protocol independent connection tracking.
13  *        Based on the original ip_conntrack code which had the following
14  *        copyright information:
15  *              (C) 1999-2001 Paul `Rusty' Russell
16  *              (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
17  *
18  * 23 Mar 2004: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
19  *      - add get_features() to support various size of conntrack
20  *        structures.
21  */
22
23 #include <linux/types.h>
24 #include <linux/ipv6.h>
25 #include <linux/in6.h>
26 #include <linux/netfilter.h>
27 #include <linux/module.h>
28 #include <linux/skbuff.h>
29 #include <linux/icmp.h>
30 #include <linux/sysctl.h>
31 #include <net/ipv6.h>
32
33 #include <linux/netfilter_ipv6.h>
34 #include <net/netfilter/nf_conntrack.h>
35 #include <net/netfilter/nf_conntrack_helper.h>
36 #include <net/netfilter/nf_conntrack_l4proto.h>
37 #include <net/netfilter/nf_conntrack_l3proto.h>
38 #include <net/netfilter/nf_conntrack_core.h>
39
40 #if 0
41 #define DEBUGP printk
42 #else
43 #define DEBUGP(format, args...)
44 #endif
45
46 static int ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
47                              struct nf_conntrack_tuple *tuple)
48 {
49         u_int32_t _addrs[8], *ap;
50
51         ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
52                                 sizeof(_addrs), _addrs);
53         if (ap == NULL)
54                 return 0;
55
56         memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
57         memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
58
59         return 1;
60 }
61
62 static int ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
63                              const struct nf_conntrack_tuple *orig)
64 {
65         memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
66         memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
67
68         return 1;
69 }
70
71 static int ipv6_print_tuple(struct seq_file *s,
72                             const struct nf_conntrack_tuple *tuple)
73 {
74         return seq_printf(s, "src=" NIP6_FMT " dst=" NIP6_FMT " ",
75                           NIP6(*((struct in6_addr *)tuple->src.u3.ip6)),
76                           NIP6(*((struct in6_addr *)tuple->dst.u3.ip6)));
77 }
78
79 static int ipv6_print_conntrack(struct seq_file *s,
80                                 const struct nf_conn *conntrack)
81 {
82         return 0;
83 }
84
85 /*
86  * Based on ipv6_skip_exthdr() in net/ipv6/exthdr.c
87  *
88  * This function parses (probably truncated) exthdr set "hdr"
89  * of length "len". "nexthdrp" initially points to some place,
90  * where type of the first header can be found.
91  *
92  * It skips all well-known exthdrs, and returns pointer to the start
93  * of unparsable area i.e. the first header with unknown type.
94  * if success, *nexthdr is updated by type/protocol of this header.
95  *
96  * NOTES: - it may return pointer pointing beyond end of packet,
97  *          if the last recognized header is truncated in the middle.
98  *        - if packet is truncated, so that all parsed headers are skipped,
99  *          it returns -1.
100  *        - if packet is fragmented, return pointer of the fragment header.
101  *        - ESP is unparsable for now and considered like
102  *          normal payload protocol.
103  *        - Note also special handling of AUTH header. Thanks to IPsec wizards.
104  */
105
106 int nf_ct_ipv6_skip_exthdr(struct sk_buff *skb, int start, u8 *nexthdrp,
107                            int len)
108 {
109         u8 nexthdr = *nexthdrp;
110
111         while (ipv6_ext_hdr(nexthdr)) {
112                 struct ipv6_opt_hdr hdr;
113                 int hdrlen;
114
115                 if (len < (int)sizeof(struct ipv6_opt_hdr))
116                         return -1;
117                 if (nexthdr == NEXTHDR_NONE)
118                         break;
119                 if (nexthdr == NEXTHDR_FRAGMENT)
120                         break;
121                 if (skb_copy_bits(skb, start, &hdr, sizeof(hdr)))
122                         BUG();
123                 if (nexthdr == NEXTHDR_AUTH)
124                         hdrlen = (hdr.hdrlen+2)<<2;
125                 else
126                         hdrlen = ipv6_optlen(&hdr);
127
128                 nexthdr = hdr.nexthdr;
129                 len -= hdrlen;
130                 start += hdrlen;
131         }
132
133         *nexthdrp = nexthdr;
134         return start;
135 }
136
137 static int
138 ipv6_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
139              u_int8_t *protonum)
140 {
141         unsigned int extoff;
142         unsigned char pnum;
143         int protoff;
144
145         extoff = (u8*)((*pskb)->nh.ipv6h + 1) - (*pskb)->data;
146         pnum = (*pskb)->nh.ipv6h->nexthdr;
147
148         protoff = nf_ct_ipv6_skip_exthdr(*pskb, extoff, &pnum,
149                                          (*pskb)->len - extoff);
150
151         /*
152          * (protoff == (*pskb)->len) mean that the packet doesn't have no data
153          * except of IPv6 & ext headers. but it's tracked anyway. - YK
154          */
155         if ((protoff < 0) || (protoff > (*pskb)->len)) {
156                 DEBUGP("ip6_conntrack_core: can't find proto in pkt\n");
157                 NF_CT_STAT_INC(error);
158                 NF_CT_STAT_INC(invalid);
159                 return -NF_ACCEPT;
160         }
161
162         *dataoff = protoff;
163         *protonum = pnum;
164         return NF_ACCEPT;
165 }
166
167 static u_int32_t ipv6_get_features(const struct nf_conntrack_tuple *tuple)
168 {
169         return NF_CT_F_BASIC;
170 }
171
172 static unsigned int ipv6_confirm(unsigned int hooknum,
173                                  struct sk_buff **pskb,
174                                  const struct net_device *in,
175                                  const struct net_device *out,
176                                  int (*okfn)(struct sk_buff *))
177 {
178         struct nf_conn *ct;
179         struct nf_conn_help *help;
180         enum ip_conntrack_info ctinfo;
181         unsigned int ret, protoff;
182         unsigned int extoff = (u8*)((*pskb)->nh.ipv6h + 1)
183                               - (*pskb)->data;
184         unsigned char pnum = (*pskb)->nh.ipv6h->nexthdr;
185
186
187         /* This is where we call the helper: as the packet goes out. */
188         ct = nf_ct_get(*pskb, &ctinfo);
189         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
190                 goto out;
191
192         help = nfct_help(ct);
193         if (!help || !help->helper)
194                 goto out;
195
196         protoff = nf_ct_ipv6_skip_exthdr(*pskb, extoff, &pnum,
197                                          (*pskb)->len - extoff);
198         if (protoff < 0 || protoff > (*pskb)->len ||
199             pnum == NEXTHDR_FRAGMENT) {
200                 DEBUGP("proto header not found\n");
201                 return NF_ACCEPT;
202         }
203
204         ret = help->helper->help(pskb, protoff, ct, ctinfo);
205         if (ret != NF_ACCEPT)
206                 return ret;
207 out:
208         /* We've seen it coming out the other side: confirm it */
209         return nf_conntrack_confirm(pskb);
210 }
211
212 static unsigned int ipv6_defrag(unsigned int hooknum,
213                                 struct sk_buff **pskb,
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 ((*pskb)->nfct)
222                 return NF_ACCEPT;
223
224         reasm = nf_ct_frag6_gather(*pskb);
225
226         /* queued */
227         if (reasm == NULL)
228                 return NF_STOLEN;
229
230         /* error occured or not fragmented */
231         if (reasm == *pskb)
232                 return NF_ACCEPT;
233
234         nf_ct_frag6_output(hooknum, reasm, (struct net_device *)in,
235                            (struct net_device *)out, okfn);
236
237         return NF_STOLEN;
238 }
239
240 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
241                                       struct sk_buff **pskb,
242                                       const struct net_device *in,
243                                       const struct net_device *out,
244                                       int (*okfn)(struct sk_buff *))
245 {
246         struct sk_buff *reasm = (*pskb)->nfct_reasm;
247
248         /* This packet is fragmented and has reassembled packet. */
249         if (reasm) {
250                 /* Reassembled packet isn't parsed yet ? */
251                 if (!reasm->nfct) {
252                         unsigned int ret;
253
254                         ret = nf_conntrack_in(PF_INET6, hooknum, &reasm);
255                         if (ret != NF_ACCEPT)
256                                 return ret;
257                 }
258                 nf_conntrack_get(reasm->nfct);
259                 (*pskb)->nfct = reasm->nfct;
260                 return NF_ACCEPT;
261         }
262
263         return nf_conntrack_in(PF_INET6, hooknum, pskb);
264 }
265
266 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
267                                          struct sk_buff **pskb,
268                                          const struct net_device *in,
269                                          const struct net_device *out,
270                                          int (*okfn)(struct sk_buff *))
271 {
272         /* root is playing with raw sockets. */
273         if ((*pskb)->len < sizeof(struct ipv6hdr)) {
274                 if (net_ratelimit())
275                         printk("ipv6_conntrack_local: packet too short\n");
276                 return NF_ACCEPT;
277         }
278         return ipv6_conntrack_in(hooknum, pskb, in, out, okfn);
279 }
280
281 static struct nf_hook_ops ipv6_conntrack_ops[] = {
282         {
283                 .hook           = ipv6_defrag,
284                 .owner          = THIS_MODULE,
285                 .pf             = PF_INET6,
286                 .hooknum        = NF_IP6_PRE_ROUTING,
287                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
288         },
289         {
290                 .hook           = ipv6_conntrack_in,
291                 .owner          = THIS_MODULE,
292                 .pf             = PF_INET6,
293                 .hooknum        = NF_IP6_PRE_ROUTING,
294                 .priority       = NF_IP6_PRI_CONNTRACK,
295         },
296         {
297                 .hook           = ipv6_conntrack_local,
298                 .owner          = THIS_MODULE,
299                 .pf             = PF_INET6,
300                 .hooknum        = NF_IP6_LOCAL_OUT,
301                 .priority       = NF_IP6_PRI_CONNTRACK,
302         },
303         {
304                 .hook           = ipv6_defrag,
305                 .owner          = THIS_MODULE,
306                 .pf             = PF_INET6,
307                 .hooknum        = NF_IP6_LOCAL_OUT,
308                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
309         },
310         {
311                 .hook           = ipv6_confirm,
312                 .owner          = THIS_MODULE,
313                 .pf             = PF_INET6,
314                 .hooknum        = NF_IP6_POST_ROUTING,
315                 .priority       = NF_IP6_PRI_LAST,
316         },
317         {
318                 .hook           = ipv6_confirm,
319                 .owner          = THIS_MODULE,
320                 .pf             = PF_INET6,
321                 .hooknum        = NF_IP6_LOCAL_IN,
322                 .priority       = NF_IP6_PRI_LAST-1,
323         },
324 };
325
326 #ifdef CONFIG_SYSCTL
327
328 /* From nf_conntrack_proto_icmpv6.c */
329 extern unsigned int nf_ct_icmpv6_timeout;
330
331 static struct ctl_table_header *nf_ct_ipv6_sysctl_header;
332
333 static ctl_table nf_ct_sysctl_table[] = {
334         {
335                 .ctl_name       = NET_NF_CONNTRACK_ICMPV6_TIMEOUT,
336                 .procname       = "nf_conntrack_icmpv6_timeout",
337                 .data           = &nf_ct_icmpv6_timeout,
338                 .maxlen         = sizeof(unsigned int),
339                 .mode           = 0644,
340                 .proc_handler   = &proc_dointvec_jiffies,
341         },
342         {
343                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_TIMEOUT,
344                 .procname       = "nf_conntrack_frag6_timeout",
345                 .data           = &nf_ct_frag6_timeout,
346                 .maxlen         = sizeof(unsigned int),
347                 .mode           = 0644,
348                 .proc_handler   = &proc_dointvec_jiffies,
349         },
350         {
351                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_LOW_THRESH,
352                 .procname       = "nf_conntrack_frag6_low_thresh",
353                 .data           = &nf_ct_frag6_low_thresh,
354                 .maxlen         = sizeof(unsigned int),
355                 .mode           = 0644,
356                 .proc_handler   = &proc_dointvec,
357         },
358         {
359                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_HIGH_THRESH,
360                 .procname       = "nf_conntrack_frag6_high_thresh",
361                 .data           = &nf_ct_frag6_high_thresh,
362                 .maxlen         = sizeof(unsigned int),
363                 .mode           = 0644,
364                 .proc_handler   = &proc_dointvec,
365         },
366         { .ctl_name = 0 }
367 };
368
369 static ctl_table nf_ct_netfilter_table[] = {
370         {
371                 .ctl_name       = NET_NETFILTER,
372                 .procname       = "netfilter",
373                 .mode           = 0555,
374                 .child          = nf_ct_sysctl_table,
375         },
376         { .ctl_name = 0 }
377 };
378
379 static ctl_table nf_ct_net_table[] = {
380         {
381                 .ctl_name       = CTL_NET,
382                 .procname       = "net",
383                 .mode           = 0555,
384                 .child          = nf_ct_netfilter_table,
385         },
386         { .ctl_name = 0 }
387 };
388 #endif
389
390 #if defined(CONFIG_NF_CT_NETLINK) || \
391     defined(CONFIG_NF_CT_NETLINK_MODULE)
392
393 #include <linux/netfilter/nfnetlink.h>
394 #include <linux/netfilter/nfnetlink_conntrack.h>
395
396 static int ipv6_tuple_to_nfattr(struct sk_buff *skb,
397                                 const struct nf_conntrack_tuple *tuple)
398 {
399         NFA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
400                 &tuple->src.u3.ip6);
401         NFA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
402                 &tuple->dst.u3.ip6);
403         return 0;
404
405 nfattr_failure:
406         return -1;
407 }
408
409 static const size_t cta_min_ip[CTA_IP_MAX] = {
410         [CTA_IP_V6_SRC-1]       = sizeof(u_int32_t)*4,
411         [CTA_IP_V6_DST-1]       = sizeof(u_int32_t)*4,
412 };
413
414 static int ipv6_nfattr_to_tuple(struct nfattr *tb[],
415                                 struct nf_conntrack_tuple *t)
416 {
417         if (!tb[CTA_IP_V6_SRC-1] || !tb[CTA_IP_V6_DST-1])
418                 return -EINVAL;
419
420         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
421                 return -EINVAL;
422
423         memcpy(&t->src.u3.ip6, NFA_DATA(tb[CTA_IP_V6_SRC-1]), 
424                sizeof(u_int32_t) * 4);
425         memcpy(&t->dst.u3.ip6, NFA_DATA(tb[CTA_IP_V6_DST-1]),
426                sizeof(u_int32_t) * 4);
427
428         return 0;
429 }
430 #endif
431
432 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 = {
433         .l3proto                = PF_INET6,
434         .name                   = "ipv6",
435         .pkt_to_tuple           = ipv6_pkt_to_tuple,
436         .invert_tuple           = ipv6_invert_tuple,
437         .print_tuple            = ipv6_print_tuple,
438         .print_conntrack        = ipv6_print_conntrack,
439         .prepare                = ipv6_prepare,
440 #if defined(CONFIG_NF_CT_NETLINK) || \
441     defined(CONFIG_NF_CT_NETLINK_MODULE)
442         .tuple_to_nfattr        = ipv6_tuple_to_nfattr,
443         .nfattr_to_tuple        = ipv6_nfattr_to_tuple,
444 #endif
445         .get_features           = ipv6_get_features,
446         .me                     = THIS_MODULE,
447 };
448
449 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
450 MODULE_LICENSE("GPL");
451 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
452
453 static int __init nf_conntrack_l3proto_ipv6_init(void)
454 {
455         int ret = 0;
456
457         need_conntrack();
458
459         ret = nf_ct_frag6_init();
460         if (ret < 0) {
461                 printk("nf_conntrack_ipv6: can't initialize frag6.\n");
462                 return ret;
463         }
464         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp6);
465         if (ret < 0) {
466                 printk("nf_conntrack_ipv6: can't register tcp.\n");
467                 goto cleanup_frag6;
468         }
469
470         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp6);
471         if (ret < 0) {
472                 printk("nf_conntrack_ipv6: can't register udp.\n");
473                 goto cleanup_tcp;
474         }
475
476         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmpv6);
477         if (ret < 0) {
478                 printk("nf_conntrack_ipv6: can't register icmpv6.\n");
479                 goto cleanup_udp;
480         }
481
482         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6);
483         if (ret < 0) {
484                 printk("nf_conntrack_ipv6: can't register ipv6\n");
485                 goto cleanup_icmpv6;
486         }
487
488         ret = nf_register_hooks(ipv6_conntrack_ops,
489                                 ARRAY_SIZE(ipv6_conntrack_ops));
490         if (ret < 0) {
491                 printk("nf_conntrack_ipv6: can't register pre-routing defrag "
492                        "hook.\n");
493                 goto cleanup_ipv6;
494         }
495 #ifdef CONFIG_SYSCTL
496         nf_ct_ipv6_sysctl_header = register_sysctl_table(nf_ct_net_table, 0);
497         if (nf_ct_ipv6_sysctl_header == NULL) {
498                 printk("nf_conntrack: can't register to sysctl.\n");
499                 ret = -ENOMEM;
500                 goto cleanup_hooks;
501         }
502 #endif
503         return ret;
504
505 #ifdef CONFIG_SYSCTL
506  cleanup_hooks:
507         nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
508 #endif
509  cleanup_ipv6:
510         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
511  cleanup_icmpv6:
512         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
513  cleanup_udp:
514         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
515  cleanup_tcp:
516         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
517  cleanup_frag6:
518         nf_ct_frag6_cleanup();
519         return ret;
520 }
521
522 static void __exit nf_conntrack_l3proto_ipv6_fini(void)
523 {
524         synchronize_net();
525 #ifdef CONFIG_SYSCTL
526         unregister_sysctl_table(nf_ct_ipv6_sysctl_header);
527 #endif
528         nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
529         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
530         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
531         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
532         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
533         nf_ct_frag6_cleanup();
534 }
535
536 module_init(nf_conntrack_l3proto_ipv6_init);
537 module_exit(nf_conntrack_l3proto_ipv6_fini);