1 /* (C) 1999-2001 Paul `Rusty' Russell
2 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
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.
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
14 * Derived from net/ipv4/netfilter/ip_conntrack_standalone.c
17 #include <linux/types.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>
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_protocol.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>
38 #define DEBUGP(format, args...)
41 DECLARE_PER_CPU(struct nf_conntrack_stat, nf_conntrack_stat);
43 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
44 struct nf_conntrack_tuple *tuple)
46 u_int32_t _addrs[2], *ap;
47 ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
48 sizeof(u_int32_t) * 2, _addrs);
52 tuple->src.u3.ip = ap[0];
53 tuple->dst.u3.ip = ap[1];
58 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
59 const struct nf_conntrack_tuple *orig)
61 tuple->src.u3.ip = orig->dst.u3.ip;
62 tuple->dst.u3.ip = orig->src.u3.ip;
67 static int ipv4_print_tuple(struct seq_file *s,
68 const struct nf_conntrack_tuple *tuple)
70 return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
71 NIPQUAD(tuple->src.u3.ip),
72 NIPQUAD(tuple->dst.u3.ip));
75 static int ipv4_print_conntrack(struct seq_file *s,
76 const struct nf_conn *conntrack)
81 /* Returns new sk_buff, or NULL */
82 static struct sk_buff *
83 nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
88 skb = ip_defrag(skb, user);
92 ip_send_check(skb->nh.iph);
98 ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
102 if ((*pskb)->nh.iph->frag_off & htons(IP_OFFSET)) {
103 if (net_ratelimit()) {
104 printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
105 (*pskb)->nh.iph->protocol, hooknum);
110 *dataoff = (*pskb)->nh.raw - (*pskb)->data + (*pskb)->nh.iph->ihl*4;
111 *protonum = (*pskb)->nh.iph->protocol;
116 int nat_module_is_loaded = 0;
117 static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
119 if (nat_module_is_loaded)
122 return NF_CT_F_BASIC;
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 *))
131 /* We've seen it coming out the other side: confirm it */
132 return nf_conntrack_confirm(pskb);
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 *))
142 enum ip_conntrack_info ctinfo;
143 struct nf_conn_help *help;
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)
150 help = nfct_help(ct);
151 if (!help || !help->helper)
154 return help->helper->help(pskb,
155 (*pskb)->nh.raw - (*pskb)->data
156 + (*pskb)->nh.iph->ihl*4,
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 *))
166 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
167 /* Previously seen (loopback)? Ignore. Do this before
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);
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 *))
191 return nf_conntrack_in(PF_INET, hooknum, pskb);
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 *))
200 /* root is playing with raw sockets. */
201 if ((*pskb)->len < sizeof(struct iphdr)
202 || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
204 printk("ipt_hook: happy cracking.\n");
207 return nf_conntrack_in(PF_INET, hooknum, pskb);
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[] = {
214 .hook = ipv4_conntrack_defrag,
215 .owner = THIS_MODULE,
217 .hooknum = NF_IP_PRE_ROUTING,
218 .priority = NF_IP_PRI_CONNTRACK_DEFRAG,
221 .hook = ipv4_conntrack_in,
222 .owner = THIS_MODULE,
224 .hooknum = NF_IP_PRE_ROUTING,
225 .priority = NF_IP_PRI_CONNTRACK,
228 .hook = ipv4_conntrack_defrag,
229 .owner = THIS_MODULE,
231 .hooknum = NF_IP_LOCAL_OUT,
232 .priority = NF_IP_PRI_CONNTRACK_DEFRAG,
235 .hook = ipv4_conntrack_local,
236 .owner = THIS_MODULE,
238 .hooknum = NF_IP_LOCAL_OUT,
239 .priority = NF_IP_PRI_CONNTRACK,
242 .hook = ipv4_conntrack_help,
243 .owner = THIS_MODULE,
245 .hooknum = NF_IP_POST_ROUTING,
246 .priority = NF_IP_PRI_CONNTRACK_HELPER,
249 .hook = ipv4_conntrack_help,
250 .owner = THIS_MODULE,
252 .hooknum = NF_IP_LOCAL_IN,
253 .priority = NF_IP_PRI_CONNTRACK_HELPER,
256 .hook = ipv4_confirm,
257 .owner = THIS_MODULE,
259 .hooknum = NF_IP_POST_ROUTING,
260 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
263 .hook = ipv4_confirm,
264 .owner = THIS_MODULE,
266 .hooknum = NF_IP_LOCAL_IN,
267 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
272 /* From nf_conntrack_proto_icmp.c */
273 extern unsigned int nf_ct_icmp_timeout;
274 static struct ctl_table_header *nf_ct_ipv4_sysctl_header;
276 static ctl_table nf_ct_sysctl_table[] = {
278 .ctl_name = NET_NF_CONNTRACK_ICMP_TIMEOUT,
279 .procname = "nf_conntrack_icmp_timeout",
280 .data = &nf_ct_icmp_timeout,
281 .maxlen = sizeof(unsigned int),
283 .proc_handler = &proc_dointvec_jiffies,
288 static ctl_table nf_ct_netfilter_table[] = {
290 .ctl_name = NET_NETFILTER,
291 .procname = "netfilter",
293 .child = nf_ct_sysctl_table,
298 static ctl_table nf_ct_net_table[] = {
303 .child = nf_ct_netfilter_table,
309 /* Fast function for those who don't want to parse /proc (and I don't
311 /* Reversing the socket's dst/src point of view gives us the reply
314 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
316 struct inet_sock *inet = inet_sk(sk);
317 struct nf_conntrack_tuple_hash *h;
318 struct nf_conntrack_tuple tuple;
320 NF_CT_TUPLE_U_BLANK(&tuple);
321 tuple.src.u3.ip = inet->rcv_saddr;
322 tuple.src.u.tcp.port = inet->sport;
323 tuple.dst.u3.ip = inet->daddr;
324 tuple.dst.u.tcp.port = inet->dport;
325 tuple.src.l3num = PF_INET;
326 tuple.dst.protonum = IPPROTO_TCP;
328 /* We only do TCP at the moment: is there a better way? */
329 if (strcmp(sk->sk_prot->name, "TCP")) {
330 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
334 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
335 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
336 *len, sizeof(struct sockaddr_in));
340 h = nf_conntrack_find_get(&tuple, NULL);
342 struct sockaddr_in sin;
343 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
345 sin.sin_family = AF_INET;
346 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
347 .tuple.dst.u.tcp.port;
348 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
350 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
352 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
353 NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
355 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
360 DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
361 NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
362 NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
366 #if defined(CONFIG_NF_CT_NETLINK) || \
367 defined(CONFIG_NF_CT_NETLINK_MODULE)
369 #include <linux/netfilter/nfnetlink.h>
370 #include <linux/netfilter/nfnetlink_conntrack.h>
372 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
373 const struct nf_conntrack_tuple *tuple)
375 NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
377 NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
385 static const size_t cta_min_ip[CTA_IP_MAX] = {
386 [CTA_IP_V4_SRC-1] = sizeof(u_int32_t),
387 [CTA_IP_V4_DST-1] = sizeof(u_int32_t),
390 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
391 struct nf_conntrack_tuple *t)
393 if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
396 if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
400 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
402 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
408 static struct nf_sockopt_ops so_getorigdst = {
410 .get_optmin = SO_ORIGINAL_DST,
411 .get_optmax = SO_ORIGINAL_DST+1,
415 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
418 .pkt_to_tuple = ipv4_pkt_to_tuple,
419 .invert_tuple = ipv4_invert_tuple,
420 .print_tuple = ipv4_print_tuple,
421 .print_conntrack = ipv4_print_conntrack,
422 .prepare = ipv4_prepare,
423 .get_features = ipv4_get_features,
424 #if defined(CONFIG_NF_CT_NETLINK) || \
425 defined(CONFIG_NF_CT_NETLINK_MODULE)
426 .tuple_to_nfattr = ipv4_tuple_to_nfattr,
427 .nfattr_to_tuple = ipv4_nfattr_to_tuple,
432 extern struct nf_conntrack_protocol nf_conntrack_protocol_tcp4;
433 extern struct nf_conntrack_protocol nf_conntrack_protocol_udp4;
434 extern struct nf_conntrack_protocol nf_conntrack_protocol_icmp;
436 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
437 MODULE_LICENSE("GPL");
439 static int __init nf_conntrack_l3proto_ipv4_init(void)
445 ret = nf_register_sockopt(&so_getorigdst);
447 printk(KERN_ERR "Unable to register netfilter socket option\n");
451 ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_tcp4);
453 printk("nf_conntrack_ipv4: can't register tcp.\n");
454 goto cleanup_sockopt;
457 ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_udp4);
459 printk("nf_conntrack_ipv4: can't register udp.\n");
463 ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_icmp);
465 printk("nf_conntrack_ipv4: can't register icmp.\n");
469 ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
471 printk("nf_conntrack_ipv4: can't register ipv4\n");
475 ret = nf_register_hooks(ipv4_conntrack_ops,
476 ARRAY_SIZE(ipv4_conntrack_ops));
478 printk("nf_conntrack_ipv4: can't register hooks.\n");
482 nf_ct_ipv4_sysctl_header = register_sysctl_table(nf_ct_net_table, 0);
483 if (nf_ct_ipv4_sysctl_header == NULL) {
484 printk("nf_conntrack: can't register to sysctl.\n");
493 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
496 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
498 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_icmp);
500 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp4);
502 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_tcp4);
504 nf_unregister_sockopt(&so_getorigdst);
508 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
512 unregister_sysctl_table(nf_ct_ipv4_sysctl_header);
514 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
515 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
516 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_icmp);
517 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp4);
518 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_tcp4);
519 nf_unregister_sockopt(&so_getorigdst);
522 module_init(nf_conntrack_l3proto_ipv4_init);
523 module_exit(nf_conntrack_l3proto_ipv4_fini);
525 EXPORT_SYMBOL(nf_ct_ipv4_gather_frags);