[NETFILTER] nf_conntrack: clean up to reduce size of 'struct nf_conn'
[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/config.h>
24 #include <linux/types.h>
25 #include <linux/ipv6.h>
26 #include <linux/in6.h>
27 #include <linux/netfilter.h>
28 #include <linux/module.h>
29 #include <linux/skbuff.h>
30 #include <linux/icmp.h>
31 #include <linux/sysctl.h>
32 #include <net/ipv6.h>
33
34 #include <linux/netfilter_ipv6.h>
35 #include <net/netfilter/nf_conntrack.h>
36 #include <net/netfilter/nf_conntrack_helper.h>
37 #include <net/netfilter/nf_conntrack_protocol.h>
38 #include <net/netfilter/nf_conntrack_l3proto.h>
39 #include <net/netfilter/nf_conntrack_core.h>
40
41 #if 0
42 #define DEBUGP printk
43 #else
44 #define DEBUGP(format, args...)
45 #endif
46
47 DECLARE_PER_CPU(struct ip_conntrack_stat, nf_conntrack_stat);
48
49 static int ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
50                              struct nf_conntrack_tuple *tuple)
51 {
52         u_int32_t _addrs[8], *ap;
53
54         ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
55                                 sizeof(_addrs), _addrs);
56         if (ap == NULL)
57                 return 0;
58
59         memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
60         memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
61
62         return 1;
63 }
64
65 static int ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
66                              const struct nf_conntrack_tuple *orig)
67 {
68         memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
69         memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
70
71         return 1;
72 }
73
74 static int ipv6_print_tuple(struct seq_file *s,
75                             const struct nf_conntrack_tuple *tuple)
76 {
77         return seq_printf(s, "src=" NIP6_FMT " dst=" NIP6_FMT " ",
78                           NIP6(*((struct in6_addr *)tuple->src.u3.ip6)),
79                           NIP6(*((struct in6_addr *)tuple->dst.u3.ip6)));
80 }
81
82 static int ipv6_print_conntrack(struct seq_file *s,
83                                 const struct nf_conn *conntrack)
84 {
85         return 0;
86 }
87
88 /*
89  * Based on ipv6_skip_exthdr() in net/ipv6/exthdr.c
90  *
91  * This function parses (probably truncated) exthdr set "hdr"
92  * of length "len". "nexthdrp" initially points to some place,
93  * where type of the first header can be found.
94  *
95  * It skips all well-known exthdrs, and returns pointer to the start
96  * of unparsable area i.e. the first header with unknown type.
97  * if success, *nexthdr is updated by type/protocol of this header.
98  *
99  * NOTES: - it may return pointer pointing beyond end of packet,
100  *          if the last recognized header is truncated in the middle.
101  *        - if packet is truncated, so that all parsed headers are skipped,
102  *          it returns -1.
103  *        - if packet is fragmented, return pointer of the fragment header.
104  *        - ESP is unparsable for now and considered like
105  *          normal payload protocol.
106  *        - Note also special handling of AUTH header. Thanks to IPsec wizards.
107  */
108
109 int nf_ct_ipv6_skip_exthdr(struct sk_buff *skb, int start, u8 *nexthdrp,
110                            int len)
111 {
112         u8 nexthdr = *nexthdrp;
113
114         while (ipv6_ext_hdr(nexthdr)) {
115                 struct ipv6_opt_hdr hdr;
116                 int hdrlen;
117
118                 if (len < (int)sizeof(struct ipv6_opt_hdr))
119                         return -1;
120                 if (nexthdr == NEXTHDR_NONE)
121                         break;
122                 if (nexthdr == NEXTHDR_FRAGMENT)
123                         break;
124                 if (skb_copy_bits(skb, start, &hdr, sizeof(hdr)))
125                         BUG();
126                 if (nexthdr == NEXTHDR_AUTH)
127                         hdrlen = (hdr.hdrlen+2)<<2;
128                 else
129                         hdrlen = ipv6_optlen(&hdr);
130
131                 nexthdr = hdr.nexthdr;
132                 len -= hdrlen;
133                 start += hdrlen;
134         }
135
136         *nexthdrp = nexthdr;
137         return start;
138 }
139
140 static int
141 ipv6_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
142              u_int8_t *protonum)
143 {
144         unsigned int extoff;
145         unsigned char pnum;
146         int protoff;
147
148         extoff = (u8*)((*pskb)->nh.ipv6h + 1) - (*pskb)->data;
149         pnum = (*pskb)->nh.ipv6h->nexthdr;
150
151         protoff = nf_ct_ipv6_skip_exthdr(*pskb, extoff, &pnum,
152                                          (*pskb)->len - extoff);
153
154         /*
155          * (protoff == (*pskb)->len) mean that the packet doesn't have no data
156          * except of IPv6 & ext headers. but it's tracked anyway. - YK
157          */
158         if ((protoff < 0) || (protoff > (*pskb)->len)) {
159                 DEBUGP("ip6_conntrack_core: can't find proto in pkt\n");
160                 NF_CT_STAT_INC(error);
161                 NF_CT_STAT_INC(invalid);
162                 return -NF_ACCEPT;
163         }
164
165         *dataoff = protoff;
166         *protonum = pnum;
167         return NF_ACCEPT;
168 }
169
170 static u_int32_t ipv6_get_features(const struct nf_conntrack_tuple *tuple)
171 {
172         return NF_CT_F_BASIC;
173 }
174
175 static unsigned int ipv6_confirm(unsigned int hooknum,
176                                  struct sk_buff **pskb,
177                                  const struct net_device *in,
178                                  const struct net_device *out,
179                                  int (*okfn)(struct sk_buff *))
180 {
181         struct nf_conn *ct;
182         struct nf_conn_help *help;
183         enum ip_conntrack_info ctinfo;
184         unsigned int ret, protoff;
185         unsigned int extoff = (u8*)((*pskb)->nh.ipv6h + 1)
186                               - (*pskb)->data;
187         unsigned char pnum = (*pskb)->nh.ipv6h->nexthdr;
188
189
190         /* This is where we call the helper: as the packet goes out. */
191         ct = nf_ct_get(*pskb, &ctinfo);
192         if (!ct)
193                 goto out;
194
195         help = nfct_help(ct);
196         if (!help || !help->helper)
197                 goto out;
198
199         protoff = nf_ct_ipv6_skip_exthdr(*pskb, extoff, &pnum,
200                                          (*pskb)->len - extoff);
201         if (protoff < 0 || protoff > (*pskb)->len ||
202             pnum == NEXTHDR_FRAGMENT) {
203                 DEBUGP("proto header not found\n");
204                 return NF_ACCEPT;
205         }
206
207         ret = help->helper->help(pskb, protoff, ct, ctinfo);
208         if (ret != NF_ACCEPT)
209                 return ret;
210 out:
211         /* We've seen it coming out the other side: confirm it */
212         return nf_conntrack_confirm(pskb);
213 }
214
215 extern struct sk_buff *nf_ct_frag6_gather(struct sk_buff *skb);
216 extern void nf_ct_frag6_output(unsigned int hooknum, struct sk_buff *skb,
217                                struct net_device *in,
218                                struct net_device *out,
219                                int (*okfn)(struct sk_buff *));
220 static unsigned int ipv6_defrag(unsigned int hooknum,
221                                 struct sk_buff **pskb,
222                                 const struct net_device *in,
223                                 const struct net_device *out,
224                                 int (*okfn)(struct sk_buff *))
225 {
226         struct sk_buff *reasm;
227
228         /* Previously seen (loopback)?  */
229         if ((*pskb)->nfct)
230                 return NF_ACCEPT;
231
232         reasm = nf_ct_frag6_gather(*pskb);
233
234         /* queued */
235         if (reasm == NULL)
236                 return NF_STOLEN;
237
238         /* error occured or not fragmented */
239         if (reasm == *pskb)
240                 return NF_ACCEPT;
241
242         nf_ct_frag6_output(hooknum, reasm, (struct net_device *)in,
243                            (struct net_device *)out, okfn);
244
245         return NF_STOLEN;
246 }
247
248 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
249                                       struct sk_buff **pskb,
250                                       const struct net_device *in,
251                                       const struct net_device *out,
252                                       int (*okfn)(struct sk_buff *))
253 {
254         struct sk_buff *reasm = (*pskb)->nfct_reasm;
255
256         /* This packet is fragmented and has reassembled packet. */
257         if (reasm) {
258                 /* Reassembled packet isn't parsed yet ? */
259                 if (!reasm->nfct) {
260                         unsigned int ret;
261
262                         ret = nf_conntrack_in(PF_INET6, hooknum, &reasm);
263                         if (ret != NF_ACCEPT)
264                                 return ret;
265                 }
266                 nf_conntrack_get(reasm->nfct);
267                 (*pskb)->nfct = reasm->nfct;
268                 return NF_ACCEPT;
269         }
270
271         return nf_conntrack_in(PF_INET6, hooknum, pskb);
272 }
273
274 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
275                                          struct sk_buff **pskb,
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 ((*pskb)->len < sizeof(struct ipv6hdr)) {
282                 if (net_ratelimit())
283                         printk("ipv6_conntrack_local: packet too short\n");
284                 return NF_ACCEPT;
285         }
286         return ipv6_conntrack_in(hooknum, pskb, in, out, okfn);
287 }
288
289 /* Connection tracking may drop packets, but never alters them, so
290    make it the first hook. */
291 static struct nf_hook_ops ipv6_conntrack_defrag_ops = {
292         .hook           = ipv6_defrag,
293         .owner          = THIS_MODULE,
294         .pf             = PF_INET6,
295         .hooknum        = NF_IP6_PRE_ROUTING,
296         .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
297 };
298
299 static struct nf_hook_ops ipv6_conntrack_in_ops = {
300         .hook           = ipv6_conntrack_in,
301         .owner          = THIS_MODULE,
302         .pf             = PF_INET6,
303         .hooknum        = NF_IP6_PRE_ROUTING,
304         .priority       = NF_IP6_PRI_CONNTRACK,
305 };
306
307 static struct nf_hook_ops ipv6_conntrack_local_out_ops = {
308         .hook           = ipv6_conntrack_local,
309         .owner          = THIS_MODULE,
310         .pf             = PF_INET6,
311         .hooknum        = NF_IP6_LOCAL_OUT,
312         .priority       = NF_IP6_PRI_CONNTRACK,
313 };
314
315 static struct nf_hook_ops ipv6_conntrack_defrag_local_out_ops = {
316         .hook           = ipv6_defrag,
317         .owner          = THIS_MODULE,
318         .pf             = PF_INET6,
319         .hooknum        = NF_IP6_LOCAL_OUT,
320         .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
321 };
322
323 /* Refragmenter; last chance. */
324 static struct nf_hook_ops ipv6_conntrack_out_ops = {
325         .hook           = ipv6_confirm,
326         .owner          = THIS_MODULE,
327         .pf             = PF_INET6,
328         .hooknum        = NF_IP6_POST_ROUTING,
329         .priority       = NF_IP6_PRI_LAST,
330 };
331
332 static struct nf_hook_ops ipv6_conntrack_local_in_ops = {
333         .hook           = ipv6_confirm,
334         .owner          = THIS_MODULE,
335         .pf             = PF_INET6,
336         .hooknum        = NF_IP6_LOCAL_IN,
337         .priority       = NF_IP6_PRI_LAST-1,
338 };
339
340 #ifdef CONFIG_SYSCTL
341
342 /* From nf_conntrack_proto_icmpv6.c */
343 extern unsigned int nf_ct_icmpv6_timeout;
344
345 /* From nf_conntrack_frag6.c */
346 extern unsigned int nf_ct_frag6_timeout;
347 extern unsigned int nf_ct_frag6_low_thresh;
348 extern unsigned int nf_ct_frag6_high_thresh;
349
350 static struct ctl_table_header *nf_ct_ipv6_sysctl_header;
351
352 static ctl_table nf_ct_sysctl_table[] = {
353         {
354                 .ctl_name       = NET_NF_CONNTRACK_ICMPV6_TIMEOUT,
355                 .procname       = "nf_conntrack_icmpv6_timeout",
356                 .data           = &nf_ct_icmpv6_timeout,
357                 .maxlen         = sizeof(unsigned int),
358                 .mode           = 0644,
359                 .proc_handler   = &proc_dointvec_jiffies,
360         },
361         {
362                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_TIMEOUT,
363                 .procname       = "nf_conntrack_frag6_timeout",
364                 .data           = &nf_ct_frag6_timeout,
365                 .maxlen         = sizeof(unsigned int),
366                 .mode           = 0644,
367                 .proc_handler   = &proc_dointvec_jiffies,
368         },
369         {
370                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_LOW_THRESH,
371                 .procname       = "nf_conntrack_frag6_low_thresh",
372                 .data           = &nf_ct_frag6_low_thresh,
373                 .maxlen         = sizeof(unsigned int),
374                 .mode           = 0644,
375                 .proc_handler   = &proc_dointvec,
376         },
377         {
378                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_HIGH_THRESH,
379                 .procname       = "nf_conntrack_frag6_high_thresh",
380                 .data           = &nf_ct_frag6_high_thresh,
381                 .maxlen         = sizeof(unsigned int),
382                 .mode           = 0644,
383                 .proc_handler   = &proc_dointvec,
384         },
385         { .ctl_name = 0 }
386 };
387
388 static ctl_table nf_ct_netfilter_table[] = {
389         {
390                 .ctl_name       = NET_NETFILTER,
391                 .procname       = "netfilter",
392                 .mode           = 0555,
393                 .child          = nf_ct_sysctl_table,
394         },
395         { .ctl_name = 0 }
396 };
397
398 static ctl_table nf_ct_net_table[] = {
399         {
400                 .ctl_name       = CTL_NET,
401                 .procname       = "net",
402                 .mode           = 0555,
403                 .child          = nf_ct_netfilter_table,
404         },
405         { .ctl_name = 0 }
406 };
407 #endif
408
409 #if defined(CONFIG_NF_CT_NETLINK) || \
410     defined(CONFIG_NF_CT_NETLINK_MODULE)
411
412 #include <linux/netfilter/nfnetlink.h>
413 #include <linux/netfilter/nfnetlink_conntrack.h>
414
415 static int ipv6_tuple_to_nfattr(struct sk_buff *skb,
416                                 const struct nf_conntrack_tuple *tuple)
417 {
418         NFA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
419                 &tuple->src.u3.ip6);
420         NFA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
421                 &tuple->dst.u3.ip6);
422         return 0;
423
424 nfattr_failure:
425         return -1;
426 }
427
428 static const size_t cta_min_ip[CTA_IP_MAX] = {
429         [CTA_IP_V6_SRC-1]       = sizeof(u_int32_t)*4,
430         [CTA_IP_V6_DST-1]       = sizeof(u_int32_t)*4,
431 };
432
433 static int ipv6_nfattr_to_tuple(struct nfattr *tb[],
434                                 struct nf_conntrack_tuple *t)
435 {
436         if (!tb[CTA_IP_V6_SRC-1] || !tb[CTA_IP_V6_DST-1])
437                 return -EINVAL;
438
439         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
440                 return -EINVAL;
441
442         memcpy(&t->src.u3.ip6, NFA_DATA(tb[CTA_IP_V6_SRC-1]), 
443                sizeof(u_int32_t) * 4);
444         memcpy(&t->dst.u3.ip6, NFA_DATA(tb[CTA_IP_V6_DST-1]),
445                sizeof(u_int32_t) * 4);
446
447         return 0;
448 }
449 #endif
450
451 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 = {
452         .l3proto                = PF_INET6,
453         .name                   = "ipv6",
454         .pkt_to_tuple           = ipv6_pkt_to_tuple,
455         .invert_tuple           = ipv6_invert_tuple,
456         .print_tuple            = ipv6_print_tuple,
457         .print_conntrack        = ipv6_print_conntrack,
458         .prepare                = ipv6_prepare,
459 #if defined(CONFIG_NF_CT_NETLINK) || \
460     defined(CONFIG_NF_CT_NETLINK_MODULE)
461         .tuple_to_nfattr        = ipv6_tuple_to_nfattr,
462         .nfattr_to_tuple        = ipv6_nfattr_to_tuple,
463 #endif
464         .get_features           = ipv6_get_features,
465         .me                     = THIS_MODULE,
466 };
467
468 extern struct nf_conntrack_protocol nf_conntrack_protocol_tcp6;
469 extern struct nf_conntrack_protocol nf_conntrack_protocol_udp6;
470 extern struct nf_conntrack_protocol nf_conntrack_protocol_icmpv6;
471 extern int nf_ct_frag6_init(void);
472 extern void nf_ct_frag6_cleanup(void);
473 static int init_or_cleanup(int init)
474 {
475         int ret = 0;
476
477         if (!init) goto cleanup;
478
479         ret = nf_ct_frag6_init();
480         if (ret < 0) {
481                 printk("nf_conntrack_ipv6: can't initialize frag6.\n");
482                 goto cleanup_nothing;
483         }
484         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_tcp6);
485         if (ret < 0) {
486                 printk("nf_conntrack_ipv6: can't register tcp.\n");
487                 goto cleanup_frag6;
488         }
489
490         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_udp6);
491         if (ret < 0) {
492                 printk("nf_conntrack_ipv6: can't register udp.\n");
493                 goto cleanup_tcp;
494         }
495
496         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_icmpv6);
497         if (ret < 0) {
498                 printk("nf_conntrack_ipv6: can't register icmpv6.\n");
499                 goto cleanup_udp;
500         }
501
502         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6);
503         if (ret < 0) {
504                 printk("nf_conntrack_ipv6: can't register ipv6\n");
505                 goto cleanup_icmpv6;
506         }
507
508         ret = nf_register_hook(&ipv6_conntrack_defrag_ops);
509         if (ret < 0) {
510                 printk("nf_conntrack_ipv6: can't register pre-routing defrag "
511                        "hook.\n");
512                 goto cleanup_ipv6;
513         }
514
515         ret = nf_register_hook(&ipv6_conntrack_defrag_local_out_ops);
516         if (ret < 0) {
517                 printk("nf_conntrack_ipv6: can't register local_out defrag "
518                        "hook.\n");
519                 goto cleanup_defragops;
520         }
521
522         ret = nf_register_hook(&ipv6_conntrack_in_ops);
523         if (ret < 0) {
524                 printk("nf_conntrack_ipv6: can't register pre-routing hook.\n");
525                 goto cleanup_defraglocalops;
526         }
527
528         ret = nf_register_hook(&ipv6_conntrack_local_out_ops);
529         if (ret < 0) {
530                 printk("nf_conntrack_ipv6: can't register local out hook.\n");
531                 goto cleanup_inops;
532         }
533
534         ret = nf_register_hook(&ipv6_conntrack_out_ops);
535         if (ret < 0) {
536                 printk("nf_conntrack_ipv6: can't register post-routing hook.\n");
537                 goto cleanup_inandlocalops;
538         }
539
540         ret = nf_register_hook(&ipv6_conntrack_local_in_ops);
541         if (ret < 0) {
542                 printk("nf_conntrack_ipv6: can't register local in hook.\n");
543                 goto cleanup_inoutandlocalops;
544         }
545
546 #ifdef CONFIG_SYSCTL
547         nf_ct_ipv6_sysctl_header = register_sysctl_table(nf_ct_net_table, 0);
548         if (nf_ct_ipv6_sysctl_header == NULL) {
549                 printk("nf_conntrack: can't register to sysctl.\n");
550                 ret = -ENOMEM;
551                 goto cleanup_localinops;
552         }
553 #endif
554         return ret;
555
556  cleanup:
557         synchronize_net();
558 #ifdef CONFIG_SYSCTL
559         unregister_sysctl_table(nf_ct_ipv6_sysctl_header);
560  cleanup_localinops:
561 #endif
562         nf_unregister_hook(&ipv6_conntrack_local_in_ops);
563  cleanup_inoutandlocalops:
564         nf_unregister_hook(&ipv6_conntrack_out_ops);
565  cleanup_inandlocalops:
566         nf_unregister_hook(&ipv6_conntrack_local_out_ops);
567  cleanup_inops:
568         nf_unregister_hook(&ipv6_conntrack_in_ops);
569  cleanup_defraglocalops:
570         nf_unregister_hook(&ipv6_conntrack_defrag_local_out_ops);
571  cleanup_defragops:
572         nf_unregister_hook(&ipv6_conntrack_defrag_ops);
573  cleanup_ipv6:
574         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
575  cleanup_icmpv6:
576         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_icmpv6);
577  cleanup_udp:
578         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp6);
579  cleanup_tcp:
580         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_tcp6);
581  cleanup_frag6:
582         nf_ct_frag6_cleanup();
583  cleanup_nothing:
584         return ret;
585 }
586
587 MODULE_LICENSE("GPL");
588 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
589
590 static int __init init(void)
591 {
592         need_conntrack();
593         return init_or_cleanup(1);
594 }
595
596 static void __exit fini(void)
597 {
598         init_or_cleanup(0);
599 }
600
601 module_init(init);
602 module_exit(fini);