linux 2.6.16.38 w/ vs2.0.3-rc1
[linux-2.6.git] / net / ipv4 / netfilter / nf_conntrack_l3proto_ipv4.c
1 /* (C) 1999-2001 Paul `Rusty' Russell
2  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
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  * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
9  *      - move L3 protocol dependent part to this file.
10  * 23 Mar 2004: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
11  *      - add get_features() to support various size of conntrack
12  *        structures.
13  *
14  * Derived from net/ipv4/netfilter/ip_conntrack_standalone.c
15  */
16
17 #include <linux/config.h>
18 #include <linux/types.h>
19 #include <linux/ip.h>
20 #include <linux/netfilter.h>
21 #include <linux/module.h>
22 #include <linux/skbuff.h>
23 #include <linux/icmp.h>
24 #include <linux/sysctl.h>
25 #include <net/route.h>
26 #include <net/ip.h>
27
28 #include <linux/netfilter_ipv4.h>
29 #include <net/netfilter/nf_conntrack.h>
30 #include <net/netfilter/nf_conntrack_helper.h>
31 #include <net/netfilter/nf_conntrack_protocol.h>
32 #include <net/netfilter/nf_conntrack_l3proto.h>
33 #include <net/netfilter/nf_conntrack_core.h>
34 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
35
36 #if 0
37 #define DEBUGP printk
38 #else
39 #define DEBUGP(format, args...)
40 #endif
41
42 DECLARE_PER_CPU(struct nf_conntrack_stat, nf_conntrack_stat);
43
44 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
45                              struct nf_conntrack_tuple *tuple)
46 {
47         u_int32_t _addrs[2], *ap;
48         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
49                                 sizeof(u_int32_t) * 2, _addrs);
50         if (ap == NULL)
51                 return 0;
52
53         tuple->src.u3.ip = ap[0];
54         tuple->dst.u3.ip = ap[1];
55
56         return 1;
57 }
58
59 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
60                            const struct nf_conntrack_tuple *orig)
61 {
62         tuple->src.u3.ip = orig->dst.u3.ip;
63         tuple->dst.u3.ip = orig->src.u3.ip;
64
65         return 1;
66 }
67
68 static int ipv4_print_tuple(struct seq_file *s,
69                             const struct nf_conntrack_tuple *tuple)
70 {
71         return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
72                           NIPQUAD(tuple->src.u3.ip),
73                           NIPQUAD(tuple->dst.u3.ip));
74 }
75
76 static int ipv4_print_conntrack(struct seq_file *s,
77                                 const struct nf_conn *conntrack)
78 {
79         return 0;
80 }
81
82 /* Returns new sk_buff, or NULL */
83 static struct sk_buff *
84 nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
85 {
86         skb_orphan(skb);
87
88         local_bh_disable();
89         skb = ip_defrag(skb, user);
90         local_bh_enable();
91
92         if (skb)
93                 ip_send_check(skb->nh.iph);
94
95         return skb;
96 }
97
98 static int
99 ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
100              u_int8_t *protonum)
101 {
102         /* Never happen */
103         if ((*pskb)->nh.iph->frag_off & htons(IP_OFFSET)) {
104                 if (net_ratelimit()) {
105                         printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
106                         (*pskb)->nh.iph->protocol, hooknum);
107                 }
108                 return -NF_DROP;
109         }
110
111         *dataoff = (*pskb)->nh.raw - (*pskb)->data + (*pskb)->nh.iph->ihl*4;
112         *protonum = (*pskb)->nh.iph->protocol;
113
114         return NF_ACCEPT;
115 }
116
117 int nat_module_is_loaded = 0;
118 static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
119 {
120         if (nat_module_is_loaded)
121                 return NF_CT_F_NAT;
122
123         return NF_CT_F_BASIC;
124 }
125
126 static unsigned int ipv4_confirm(unsigned int hooknum,
127                                  struct sk_buff **pskb,
128                                  const struct net_device *in,
129                                  const struct net_device *out,
130                                  int (*okfn)(struct sk_buff *))
131 {
132         /* We've seen it coming out the other side: confirm it */
133         return nf_conntrack_confirm(pskb);
134 }
135
136 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
137                                       struct sk_buff **pskb,
138                                       const struct net_device *in,
139                                       const struct net_device *out,
140                                       int (*okfn)(struct sk_buff *))
141 {
142         struct nf_conn *ct;
143         enum ip_conntrack_info ctinfo;
144
145         /* This is where we call the helper: as the packet goes out. */
146         ct = nf_ct_get(*pskb, &ctinfo);
147         if (ct && ct->helper) {
148                 unsigned int ret;
149                 ret = ct->helper->help(pskb,
150                                        (*pskb)->nh.raw - (*pskb)->data
151                                                        + (*pskb)->nh.iph->ihl*4,
152                                        ct, ctinfo);
153                 if (ret != NF_ACCEPT)
154                         return ret;
155         }
156         return NF_ACCEPT;
157 }
158
159 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
160                                           struct sk_buff **pskb,
161                                           const struct net_device *in,
162                                           const struct net_device *out,
163                                           int (*okfn)(struct sk_buff *))
164 {
165 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
166         /* Previously seen (loopback)?  Ignore.  Do this before
167            fragment check. */
168         if ((*pskb)->nfct)
169                 return NF_ACCEPT;
170 #endif
171
172         /* Gather fragments. */
173         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
174                 *pskb = nf_ct_ipv4_gather_frags(*pskb,
175                                                 hooknum == NF_IP_PRE_ROUTING ?
176                                                 IP_DEFRAG_CONNTRACK_IN :
177                                                 IP_DEFRAG_CONNTRACK_OUT);
178                 if (!*pskb)
179                         return NF_STOLEN;
180         }
181         return NF_ACCEPT;
182 }
183
184 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
185                                       struct sk_buff **pskb,
186                                       const struct net_device *in,
187                                       const struct net_device *out,
188                                       int (*okfn)(struct sk_buff *))
189 {
190         return nf_conntrack_in(PF_INET, hooknum, pskb);
191 }
192
193 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
194                                          struct sk_buff **pskb,
195                                          const struct net_device *in,
196                                          const struct net_device *out,
197                                          int (*okfn)(struct sk_buff *))
198 {
199         /* root is playing with raw sockets. */
200         if ((*pskb)->len < sizeof(struct iphdr)
201             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
202                 if (net_ratelimit())
203                         printk("ipt_hook: happy cracking.\n");
204                 return NF_ACCEPT;
205         }
206         return nf_conntrack_in(PF_INET, hooknum, pskb);
207 }
208
209 /* Connection tracking may drop packets, but never alters them, so
210    make it the first hook. */
211 static struct nf_hook_ops ipv4_conntrack_defrag_ops = {
212         .hook           = ipv4_conntrack_defrag,
213         .owner          = THIS_MODULE,
214         .pf             = PF_INET,
215         .hooknum        = NF_IP_PRE_ROUTING,
216         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
217 };
218
219 static struct nf_hook_ops ipv4_conntrack_in_ops = {
220         .hook           = ipv4_conntrack_in,
221         .owner          = THIS_MODULE,
222         .pf             = PF_INET,
223         .hooknum        = NF_IP_PRE_ROUTING,
224         .priority       = NF_IP_PRI_CONNTRACK,
225 };
226
227 static struct nf_hook_ops ipv4_conntrack_defrag_local_out_ops = {
228         .hook           = ipv4_conntrack_defrag,
229         .owner          = THIS_MODULE,
230         .pf             = PF_INET,
231         .hooknum        = NF_IP_LOCAL_OUT,
232         .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
233 };
234
235 static struct nf_hook_ops ipv4_conntrack_local_out_ops = {
236         .hook           = ipv4_conntrack_local,
237         .owner          = THIS_MODULE,
238         .pf             = PF_INET,
239         .hooknum        = NF_IP_LOCAL_OUT,
240         .priority       = NF_IP_PRI_CONNTRACK,
241 };
242
243 /* helpers */
244 static struct nf_hook_ops ipv4_conntrack_helper_out_ops = {
245         .hook           = ipv4_conntrack_help,
246         .owner          = THIS_MODULE,
247         .pf             = PF_INET,
248         .hooknum        = NF_IP_POST_ROUTING,
249         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
250 };
251
252 static struct nf_hook_ops ipv4_conntrack_helper_in_ops = {
253         .hook           = ipv4_conntrack_help,
254         .owner          = THIS_MODULE,
255         .pf             = PF_INET,
256         .hooknum        = NF_IP_LOCAL_IN,
257         .priority       = NF_IP_PRI_CONNTRACK_HELPER,
258 };
259
260
261 /* Refragmenter; last chance. */
262 static struct nf_hook_ops ipv4_conntrack_out_ops = {
263         .hook           = ipv4_confirm,
264         .owner          = THIS_MODULE,
265         .pf             = PF_INET,
266         .hooknum        = NF_IP_POST_ROUTING,
267         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
268 };
269
270 static struct nf_hook_ops ipv4_conntrack_local_in_ops = {
271         .hook           = ipv4_confirm,
272         .owner          = THIS_MODULE,
273         .pf             = PF_INET,
274         .hooknum        = NF_IP_LOCAL_IN,
275         .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
276 };
277
278 #ifdef CONFIG_SYSCTL
279 /* From nf_conntrack_proto_icmp.c */
280 extern unsigned int nf_ct_icmp_timeout;
281 static struct ctl_table_header *nf_ct_ipv4_sysctl_header;
282
283 static ctl_table nf_ct_sysctl_table[] = {
284         {
285                 .ctl_name       = NET_NF_CONNTRACK_ICMP_TIMEOUT,
286                 .procname       = "nf_conntrack_icmp_timeout",
287                 .data           = &nf_ct_icmp_timeout,
288                 .maxlen         = sizeof(unsigned int),
289                 .mode           = 0644,
290                 .proc_handler   = &proc_dointvec_jiffies,
291         },
292         { .ctl_name = 0 }
293 };
294
295 static ctl_table nf_ct_netfilter_table[] = {
296         {
297                 .ctl_name       = NET_NETFILTER,
298                 .procname       = "netfilter",
299                 .mode           = 0555,
300                 .child          = nf_ct_sysctl_table,
301         },
302         { .ctl_name = 0 }
303 };
304
305 static ctl_table nf_ct_net_table[] = {
306         {
307                 .ctl_name       = CTL_NET,
308                 .procname       = "net",
309                 .mode           = 0555,
310                 .child          = nf_ct_netfilter_table,
311         },
312         { .ctl_name = 0 }
313 };
314 #endif
315
316 /* Fast function for those who don't want to parse /proc (and I don't
317    blame them). */
318 /* Reversing the socket's dst/src point of view gives us the reply
319    mapping. */
320 static int
321 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
322 {
323         struct inet_sock *inet = inet_sk(sk);
324         struct nf_conntrack_tuple_hash *h;
325         struct nf_conntrack_tuple tuple;
326         
327         NF_CT_TUPLE_U_BLANK(&tuple);
328         tuple.src.u3.ip = inet->rcv_saddr;
329         tuple.src.u.tcp.port = inet->sport;
330         tuple.dst.u3.ip = inet->daddr;
331         tuple.dst.u.tcp.port = inet->dport;
332         tuple.src.l3num = PF_INET;
333         tuple.dst.protonum = IPPROTO_TCP;
334
335         /* We only do TCP at the moment: is there a better way? */
336         if (strcmp(sk->sk_prot->name, "TCP")) {
337                 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
338                 return -ENOPROTOOPT;
339         }
340
341         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
342                 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
343                        *len, sizeof(struct sockaddr_in));
344                 return -EINVAL;
345         }
346
347         h = nf_conntrack_find_get(&tuple, NULL);
348         if (h) {
349                 struct sockaddr_in sin;
350                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
351
352                 sin.sin_family = AF_INET;
353                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
354                         .tuple.dst.u.tcp.port;
355                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
356                         .tuple.dst.u3.ip;
357                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
358
359                 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
360                        NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
361                 nf_ct_put(ct);
362                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
363                         return -EFAULT;
364                 else
365                         return 0;
366         }
367         DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
368                NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
369                NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
370         return -ENOENT;
371 }
372
373 #if defined(CONFIG_NF_CT_NETLINK) || \
374     defined(CONFIG_NF_CT_NETLINK_MODULE)
375
376 #include <linux/netfilter/nfnetlink.h>
377 #include <linux/netfilter/nfnetlink_conntrack.h>
378
379 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
380                                 const struct nf_conntrack_tuple *tuple)
381 {
382         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
383                 &tuple->src.u3.ip);
384         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
385                 &tuple->dst.u3.ip);
386         return 0;
387
388 nfattr_failure:
389         return -1;
390 }
391
392 static const size_t cta_min_ip[CTA_IP_MAX] = {
393         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
394         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
395 };
396
397 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
398                                 struct nf_conntrack_tuple *t)
399 {
400         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
401                 return -EINVAL;
402
403         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
404                 return -EINVAL;
405
406         t->src.u3.ip =
407                 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
408         t->dst.u3.ip =
409                 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
410
411         return 0;
412 }
413 #endif
414
415 static struct nf_sockopt_ops so_getorigdst = {
416         .pf             = PF_INET,
417         .get_optmin     = SO_ORIGINAL_DST,
418         .get_optmax     = SO_ORIGINAL_DST+1,
419         .get            = &getorigdst,
420 };
421
422 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
423         .l3proto         = PF_INET,
424         .name            = "ipv4",
425         .pkt_to_tuple    = ipv4_pkt_to_tuple,
426         .invert_tuple    = ipv4_invert_tuple,
427         .print_tuple     = ipv4_print_tuple,
428         .print_conntrack = ipv4_print_conntrack,
429         .prepare         = ipv4_prepare,
430         .get_features    = ipv4_get_features,
431 #if defined(CONFIG_NF_CT_NETLINK) || \
432     defined(CONFIG_NF_CT_NETLINK_MODULE)
433         .tuple_to_nfattr = ipv4_tuple_to_nfattr,
434         .nfattr_to_tuple = ipv4_nfattr_to_tuple,
435 #endif
436         .me              = THIS_MODULE,
437 };
438
439 extern struct nf_conntrack_protocol nf_conntrack_protocol_tcp4;
440 extern struct nf_conntrack_protocol nf_conntrack_protocol_udp4;
441 extern struct nf_conntrack_protocol nf_conntrack_protocol_icmp;
442 static int init_or_cleanup(int init)
443 {
444         int ret = 0;
445
446         if (!init) goto cleanup;
447
448         ret = nf_register_sockopt(&so_getorigdst);
449         if (ret < 0) {
450                 printk(KERN_ERR "Unable to register netfilter socket option\n");
451                 goto cleanup_nothing;
452         }
453
454         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_tcp4);
455         if (ret < 0) {
456                 printk("nf_conntrack_ipv4: can't register tcp.\n");
457                 goto cleanup_sockopt;
458         }
459
460         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_udp4);
461         if (ret < 0) {
462                 printk("nf_conntrack_ipv4: can't register udp.\n");
463                 goto cleanup_tcp;
464         }
465
466         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_icmp);
467         if (ret < 0) {
468                 printk("nf_conntrack_ipv4: can't register icmp.\n");
469                 goto cleanup_udp;
470         }
471
472         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
473         if (ret < 0) {
474                 printk("nf_conntrack_ipv4: can't register ipv4\n");
475                 goto cleanup_icmp;
476         }
477
478         ret = nf_register_hook(&ipv4_conntrack_defrag_ops);
479         if (ret < 0) {
480                 printk("nf_conntrack_ipv4: can't register pre-routing defrag hook.\n");
481                 goto cleanup_ipv4;
482         }
483         ret = nf_register_hook(&ipv4_conntrack_defrag_local_out_ops);
484         if (ret < 0) {
485                 printk("nf_conntrack_ipv4: can't register local_out defrag hook.\n");
486                 goto cleanup_defragops;
487         }
488
489         ret = nf_register_hook(&ipv4_conntrack_in_ops);
490         if (ret < 0) {
491                 printk("nf_conntrack_ipv4: can't register pre-routing hook.\n");
492                 goto cleanup_defraglocalops;
493         }
494
495         ret = nf_register_hook(&ipv4_conntrack_local_out_ops);
496         if (ret < 0) {
497                 printk("nf_conntrack_ipv4: can't register local out hook.\n");
498                 goto cleanup_inops;
499         }
500
501         ret = nf_register_hook(&ipv4_conntrack_helper_in_ops);
502         if (ret < 0) {
503                 printk("nf_conntrack_ipv4: can't register local helper hook.\n");
504                 goto cleanup_inandlocalops;
505         }
506
507         ret = nf_register_hook(&ipv4_conntrack_helper_out_ops);
508         if (ret < 0) {
509                 printk("nf_conntrack_ipv4: can't register postrouting helper hook.\n");
510                 goto cleanup_helperinops;
511         }
512
513         ret = nf_register_hook(&ipv4_conntrack_out_ops);
514         if (ret < 0) {
515                 printk("nf_conntrack_ipv4: can't register post-routing hook.\n");
516                 goto cleanup_helperoutops;
517         }
518
519         ret = nf_register_hook(&ipv4_conntrack_local_in_ops);
520         if (ret < 0) {
521                 printk("nf_conntrack_ipv4: can't register local in hook.\n");
522                 goto cleanup_inoutandlocalops;
523         }
524
525 #ifdef CONFIG_SYSCTL
526         nf_ct_ipv4_sysctl_header = register_sysctl_table(nf_ct_net_table, 0);
527         if (nf_ct_ipv4_sysctl_header == NULL) {
528                 printk("nf_conntrack: can't register to sysctl.\n");
529                 ret = -ENOMEM;
530                 goto cleanup_localinops;
531         }
532 #endif
533         return ret;
534
535  cleanup:
536         synchronize_net();
537 #ifdef CONFIG_SYSCTL
538         unregister_sysctl_table(nf_ct_ipv4_sysctl_header);
539  cleanup_localinops:
540 #endif
541         nf_unregister_hook(&ipv4_conntrack_local_in_ops);
542  cleanup_inoutandlocalops:
543         nf_unregister_hook(&ipv4_conntrack_out_ops);
544  cleanup_helperoutops:
545         nf_unregister_hook(&ipv4_conntrack_helper_out_ops);
546  cleanup_helperinops:
547         nf_unregister_hook(&ipv4_conntrack_helper_in_ops);
548  cleanup_inandlocalops:
549         nf_unregister_hook(&ipv4_conntrack_local_out_ops);
550  cleanup_inops:
551         nf_unregister_hook(&ipv4_conntrack_in_ops);
552  cleanup_defraglocalops:
553         nf_unregister_hook(&ipv4_conntrack_defrag_local_out_ops);
554  cleanup_defragops:
555         nf_unregister_hook(&ipv4_conntrack_defrag_ops);
556  cleanup_ipv4:
557         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
558  cleanup_icmp:
559         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_icmp);
560  cleanup_udp:
561         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp4);
562  cleanup_tcp:
563         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_tcp4);
564  cleanup_sockopt:
565         nf_unregister_sockopt(&so_getorigdst);
566  cleanup_nothing:
567         return ret;
568 }
569
570 MODULE_LICENSE("GPL");
571
572 static int __init init(void)
573 {
574         need_conntrack();
575         return init_or_cleanup(1);
576 }
577
578 static void __exit fini(void)
579 {
580         init_or_cleanup(0);
581 }
582
583 module_init(init);
584 module_exit(fini);
585
586 EXPORT_SYMBOL(nf_ct_ipv4_gather_frags);