8f3e92d20df82ce0c402f947189025978259a29e
[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/types.h>
18 #include <linux/ip.h>
19 #include <linux/netfilter.h>
20 #include <linux/module.h>
21 #include <linux/skbuff.h>
22 #include <linux/icmp.h>
23 #include <linux/sysctl.h>
24 #include <net/route.h>
25 #include <net/ip.h>
26
27 #include <linux/netfilter_ipv4.h>
28 #include <net/netfilter/nf_conntrack.h>
29 #include <net/netfilter/nf_conntrack_helper.h>
30 #include <net/netfilter/nf_conntrack_l4proto.h>
31 #include <net/netfilter/nf_conntrack_l3proto.h>
32 #include <net/netfilter/nf_conntrack_core.h>
33 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
34
35 #if 0
36 #define DEBUGP printk
37 #else
38 #define DEBUGP(format, args...)
39 #endif
40
41 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
42                              struct nf_conntrack_tuple *tuple)
43 {
44         __be32 _addrs[2], *ap;
45         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
46                                 sizeof(u_int32_t) * 2, _addrs);
47         if (ap == NULL)
48                 return 0;
49
50         tuple->src.u3.ip = ap[0];
51         tuple->dst.u3.ip = ap[1];
52
53         return 1;
54 }
55
56 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
57                            const struct nf_conntrack_tuple *orig)
58 {
59         tuple->src.u3.ip = orig->dst.u3.ip;
60         tuple->dst.u3.ip = orig->src.u3.ip;
61
62         return 1;
63 }
64
65 static int ipv4_print_tuple(struct seq_file *s,
66                             const struct nf_conntrack_tuple *tuple)
67 {
68         return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
69                           NIPQUAD(tuple->src.u3.ip),
70                           NIPQUAD(tuple->dst.u3.ip));
71 }
72
73 static int ipv4_print_conntrack(struct seq_file *s,
74                                 const struct nf_conn *conntrack)
75 {
76         return 0;
77 }
78
79 /* Returns new sk_buff, or NULL */
80 static struct sk_buff *
81 nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
82 {
83         skb_orphan(skb);
84
85         local_bh_disable();
86         skb = ip_defrag(skb, user);
87         local_bh_enable();
88
89         if (skb)
90                 ip_send_check(skb->nh.iph);
91
92         return skb;
93 }
94
95 static int
96 ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
97              u_int8_t *protonum)
98 {
99         /* Never happen */
100         if ((*pskb)->nh.iph->frag_off & htons(IP_OFFSET)) {
101                 if (net_ratelimit()) {
102                         printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
103                         (*pskb)->nh.iph->protocol, hooknum);
104                 }
105                 return -NF_DROP;
106         }
107
108         *dataoff = (*pskb)->nh.raw - (*pskb)->data + (*pskb)->nh.iph->ihl*4;
109         *protonum = (*pskb)->nh.iph->protocol;
110
111         return NF_ACCEPT;
112 }
113
114 int nf_nat_module_is_loaded = 0;
115 EXPORT_SYMBOL_GPL(nf_nat_module_is_loaded);
116
117 static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
118 {
119         if (nf_nat_module_is_loaded)
120                 return NF_CT_F_NAT;
121
122         return NF_CT_F_BASIC;
123 }
124
125 static unsigned int ipv4_confirm(unsigned int hooknum,
126                                  struct sk_buff **pskb,
127                                  const struct net_device *in,
128                                  const struct net_device *out,
129                                  int (*okfn)(struct sk_buff *))
130 {
131         /* We've seen it coming out the other side: confirm it */
132         return nf_conntrack_confirm(pskb);
133 }
134
135 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
136                                       struct sk_buff **pskb,
137                                       const struct net_device *in,
138                                       const struct net_device *out,
139                                       int (*okfn)(struct sk_buff *))
140 {
141         struct nf_conn *ct;
142         enum ip_conntrack_info ctinfo;
143         struct nf_conn_help *help;
144
145         /* This is where we call the helper: as the packet goes out. */
146         ct = nf_ct_get(*pskb, &ctinfo);
147         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
148                 return NF_ACCEPT;
149
150         help = nfct_help(ct);
151         if (!help || !help->helper)
152                 return NF_ACCEPT;
153
154         return help->helper->help(pskb,
155                                (*pskb)->nh.raw - (*pskb)->data
156                                                + (*pskb)->nh.iph->ihl*4,
157                                ct, ctinfo);
158 }
159
160 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
161                                           struct sk_buff **pskb,
162                                           const struct net_device *in,
163                                           const struct net_device *out,
164                                           int (*okfn)(struct sk_buff *))
165 {
166 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
167         /* Previously seen (loopback)?  Ignore.  Do this before
168            fragment check. */
169         if ((*pskb)->nfct)
170                 return NF_ACCEPT;
171 #endif
172
173         /* Gather fragments. */
174         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
175                 *pskb = nf_ct_ipv4_gather_frags(*pskb,
176                                                 hooknum == NF_IP_PRE_ROUTING ?
177                                                 IP_DEFRAG_CONNTRACK_IN :
178                                                 IP_DEFRAG_CONNTRACK_OUT);
179                 if (!*pskb)
180                         return NF_STOLEN;
181         }
182         return NF_ACCEPT;
183 }
184
185 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
186                                       struct sk_buff **pskb,
187                                       const struct net_device *in,
188                                       const struct net_device *out,
189                                       int (*okfn)(struct sk_buff *))
190 {
191         return nf_conntrack_in(PF_INET, hooknum, pskb);
192 }
193
194 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
195                                          struct sk_buff **pskb,
196                                          const struct net_device *in,
197                                          const struct net_device *out,
198                                          int (*okfn)(struct sk_buff *))
199 {
200         /* root is playing with raw sockets. */
201         if ((*pskb)->len < sizeof(struct iphdr)
202             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
203                 if (net_ratelimit())
204                         printk("ipt_hook: happy cracking.\n");
205                 return NF_ACCEPT;
206         }
207         return nf_conntrack_in(PF_INET, hooknum, pskb);
208 }
209
210 /* Connection tracking may drop packets, but never alters them, so
211    make it the first hook. */
212 static struct nf_hook_ops ipv4_conntrack_ops[] = {
213         {
214                 .hook           = ipv4_conntrack_defrag,
215                 .owner          = THIS_MODULE,
216                 .pf             = PF_INET,
217                 .hooknum        = NF_IP_PRE_ROUTING,
218                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
219         },
220         {
221                 .hook           = ipv4_conntrack_in,
222                 .owner          = THIS_MODULE,
223                 .pf             = PF_INET,
224                 .hooknum        = NF_IP_PRE_ROUTING,
225                 .priority       = NF_IP_PRI_CONNTRACK,
226         },
227         {
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                 .hook           = ipv4_conntrack_local,
236                 .owner          = THIS_MODULE,
237                 .pf             = PF_INET,
238                 .hooknum        = NF_IP_LOCAL_OUT,
239                 .priority       = NF_IP_PRI_CONNTRACK,
240         },
241         {
242                 .hook           = ipv4_conntrack_help,
243                 .owner          = THIS_MODULE,
244                 .pf             = PF_INET,
245                 .hooknum        = NF_IP_POST_ROUTING,
246                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
247         },
248         {
249                 .hook           = ipv4_conntrack_help,
250                 .owner          = THIS_MODULE,
251                 .pf             = PF_INET,
252                 .hooknum        = NF_IP_LOCAL_IN,
253                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
254         },
255         {
256                 .hook           = ipv4_confirm,
257                 .owner          = THIS_MODULE,
258                 .pf             = PF_INET,
259                 .hooknum        = NF_IP_POST_ROUTING,
260                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
261         },
262         {
263                 .hook           = ipv4_confirm,
264                 .owner          = THIS_MODULE,
265                 .pf             = PF_INET,
266                 .hooknum        = NF_IP_LOCAL_IN,
267                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
268         },
269 };
270
271 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
272 static int log_invalid_proto_min = 0;
273 static int log_invalid_proto_max = 255;
274
275 static ctl_table ip_ct_sysctl_table[] = {
276         {
277                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
278                 .procname       = "ip_conntrack_max",
279                 .data           = &nf_conntrack_max,
280                 .maxlen         = sizeof(int),
281                 .mode           = 0644,
282                 .proc_handler   = &proc_dointvec,
283         },
284         {
285                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
286                 .procname       = "ip_conntrack_count",
287                 .data           = &nf_conntrack_count,
288                 .maxlen         = sizeof(int),
289                 .mode           = 0444,
290                 .proc_handler   = &proc_dointvec,
291         },
292         {
293                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
294                 .procname       = "ip_conntrack_buckets",
295                 .data           = &nf_conntrack_htable_size,
296                 .maxlen         = sizeof(unsigned int),
297                 .mode           = 0444,
298                 .proc_handler   = &proc_dointvec,
299         },
300         {
301                 .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
302                 .procname       = "ip_conntrack_checksum",
303                 .data           = &nf_conntrack_checksum,
304                 .maxlen         = sizeof(int),
305                 .mode           = 0644,
306                 .proc_handler   = &proc_dointvec,
307         },
308         {
309                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
310                 .procname       = "ip_conntrack_log_invalid",
311                 .data           = &nf_ct_log_invalid,
312                 .maxlen         = sizeof(unsigned int),
313                 .mode           = 0644,
314                 .proc_handler   = &proc_dointvec_minmax,
315                 .strategy       = &sysctl_intvec,
316                 .extra1         = &log_invalid_proto_min,
317                 .extra2         = &log_invalid_proto_max,
318         },
319         {
320                 .ctl_name       = 0
321         }
322 };
323 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
324
325 /* Fast function for those who don't want to parse /proc (and I don't
326    blame them). */
327 /* Reversing the socket's dst/src point of view gives us the reply
328    mapping. */
329 static int
330 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
331 {
332         struct inet_sock *inet = inet_sk(sk);
333         struct nf_conntrack_tuple_hash *h;
334         struct nf_conntrack_tuple tuple;
335
336         NF_CT_TUPLE_U_BLANK(&tuple);
337         tuple.src.u3.ip = inet->rcv_saddr;
338         tuple.src.u.tcp.port = inet->sport;
339         tuple.dst.u3.ip = inet->daddr;
340         tuple.dst.u.tcp.port = inet->dport;
341         tuple.src.l3num = PF_INET;
342         tuple.dst.protonum = IPPROTO_TCP;
343
344         /* We only do TCP at the moment: is there a better way? */
345         if (strcmp(sk->sk_prot->name, "TCP")) {
346                 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
347                 return -ENOPROTOOPT;
348         }
349
350         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
351                 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
352                        *len, sizeof(struct sockaddr_in));
353                 return -EINVAL;
354         }
355
356         h = nf_conntrack_find_get(&tuple, NULL);
357         if (h) {
358                 struct sockaddr_in sin;
359                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
360
361                 sin.sin_family = AF_INET;
362                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
363                         .tuple.dst.u.tcp.port;
364                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
365                         .tuple.dst.u3.ip;
366                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
367
368                 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
369                        NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
370                 nf_ct_put(ct);
371                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
372                         return -EFAULT;
373                 else
374                         return 0;
375         }
376         DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
377                NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
378                NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
379         return -ENOENT;
380 }
381
382 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
383
384 #include <linux/netfilter/nfnetlink.h>
385 #include <linux/netfilter/nfnetlink_conntrack.h>
386
387 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
388                                 const struct nf_conntrack_tuple *tuple)
389 {
390         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
391                 &tuple->src.u3.ip);
392         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
393                 &tuple->dst.u3.ip);
394         return 0;
395
396 nfattr_failure:
397         return -1;
398 }
399
400 static const size_t cta_min_ip[CTA_IP_MAX] = {
401         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
402         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
403 };
404
405 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
406                                 struct nf_conntrack_tuple *t)
407 {
408         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
409                 return -EINVAL;
410
411         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
412                 return -EINVAL;
413
414         t->src.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
415         t->dst.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
416
417         return 0;
418 }
419 #endif
420
421 static struct nf_sockopt_ops so_getorigdst = {
422         .pf             = PF_INET,
423         .get_optmin     = SO_ORIGINAL_DST,
424         .get_optmax     = SO_ORIGINAL_DST+1,
425         .get            = &getorigdst,
426 };
427
428 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
429         .l3proto         = PF_INET,
430         .name            = "ipv4",
431         .pkt_to_tuple    = ipv4_pkt_to_tuple,
432         .invert_tuple    = ipv4_invert_tuple,
433         .print_tuple     = ipv4_print_tuple,
434         .print_conntrack = ipv4_print_conntrack,
435         .prepare         = ipv4_prepare,
436         .get_features    = ipv4_get_features,
437 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
438         .tuple_to_nfattr = ipv4_tuple_to_nfattr,
439         .nfattr_to_tuple = ipv4_nfattr_to_tuple,
440 #endif
441 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
442         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
443         .ctl_table       = ip_ct_sysctl_table,
444 #endif
445         .me              = THIS_MODULE,
446 };
447
448 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
449 MODULE_ALIAS("ip_conntrack");
450 MODULE_LICENSE("GPL");
451
452 static int __init nf_conntrack_l3proto_ipv4_init(void)
453 {
454         int ret = 0;
455
456         need_conntrack();
457
458         ret = nf_register_sockopt(&so_getorigdst);
459         if (ret < 0) {
460                 printk(KERN_ERR "Unable to register netfilter socket option\n");
461                 return ret;
462         }
463
464         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
465         if (ret < 0) {
466                 printk("nf_conntrack_ipv4: can't register tcp.\n");
467                 goto cleanup_sockopt;
468         }
469
470         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
471         if (ret < 0) {
472                 printk("nf_conntrack_ipv4: can't register udp.\n");
473                 goto cleanup_tcp;
474         }
475
476         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
477         if (ret < 0) {
478                 printk("nf_conntrack_ipv4: can't register icmp.\n");
479                 goto cleanup_udp;
480         }
481
482         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
483         if (ret < 0) {
484                 printk("nf_conntrack_ipv4: can't register ipv4\n");
485                 goto cleanup_icmp;
486         }
487
488         ret = nf_register_hooks(ipv4_conntrack_ops,
489                                 ARRAY_SIZE(ipv4_conntrack_ops));
490         if (ret < 0) {
491                 printk("nf_conntrack_ipv4: can't register hooks.\n");
492                 goto cleanup_ipv4;
493         }
494 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
495         ret = nf_conntrack_ipv4_compat_init();
496         if (ret < 0)
497                 goto cleanup_hooks;
498 #endif
499         return ret;
500 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
501  cleanup_hooks:
502         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
503 #endif
504  cleanup_ipv4:
505         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
506  cleanup_icmp:
507         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
508  cleanup_udp:
509         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
510  cleanup_tcp:
511         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
512  cleanup_sockopt:
513         nf_unregister_sockopt(&so_getorigdst);
514         return ret;
515 }
516
517 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
518 {
519         synchronize_net();
520 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
521         nf_conntrack_ipv4_compat_fini();
522 #endif
523         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
524         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
525         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
526         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
527         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
528         nf_unregister_sockopt(&so_getorigdst);
529 }
530
531 module_init(nf_conntrack_l3proto_ipv4_init);
532 module_exit(nf_conntrack_l3proto_ipv4_fini);