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/config.h>
18 #include <linux/types.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>
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>
39 #define DEBUGP(format, args...)
42 DECLARE_PER_CPU(struct nf_conntrack_stat, nf_conntrack_stat);
44 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
45 struct nf_conntrack_tuple *tuple)
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);
53 tuple->src.u3.ip = ap[0];
54 tuple->dst.u3.ip = ap[1];
59 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
60 const struct nf_conntrack_tuple *orig)
62 tuple->src.u3.ip = orig->dst.u3.ip;
63 tuple->dst.u3.ip = orig->src.u3.ip;
68 static int ipv4_print_tuple(struct seq_file *s,
69 const struct nf_conntrack_tuple *tuple)
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));
76 static int ipv4_print_conntrack(struct seq_file *s,
77 const struct nf_conn *conntrack)
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)
89 skb = ip_defrag(skb, user);
93 ip_send_check(skb->nh.iph);
99 ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
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);
111 *dataoff = (*pskb)->nh.raw - (*pskb)->data + (*pskb)->nh.iph->ihl*4;
112 *protonum = (*pskb)->nh.iph->protocol;
117 int nat_module_is_loaded = 0;
118 static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
120 if (nat_module_is_loaded)
123 return NF_CT_F_BASIC;
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 *))
132 /* We've seen it coming out the other side: confirm it */
133 return nf_conntrack_confirm(pskb);
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 *))
143 enum ip_conntrack_info ctinfo;
144 struct nf_conn_help *help;
146 /* This is where we call the helper: as the packet goes out. */
147 ct = nf_ct_get(*pskb, &ctinfo);
151 help = nfct_help(ct);
152 if (!help || !help->helper)
155 return help->helper->help(pskb,
156 (*pskb)->nh.raw - (*pskb)->data
157 + (*pskb)->nh.iph->ihl*4,
161 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
162 struct sk_buff **pskb,
163 const struct net_device *in,
164 const struct net_device *out,
165 int (*okfn)(struct sk_buff *))
167 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
168 /* Previously seen (loopback)? Ignore. Do this before
174 /* Gather fragments. */
175 if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
176 *pskb = nf_ct_ipv4_gather_frags(*pskb,
177 hooknum == NF_IP_PRE_ROUTING ?
178 IP_DEFRAG_CONNTRACK_IN :
179 IP_DEFRAG_CONNTRACK_OUT);
186 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
187 struct sk_buff **pskb,
188 const struct net_device *in,
189 const struct net_device *out,
190 int (*okfn)(struct sk_buff *))
192 return nf_conntrack_in(PF_INET, hooknum, pskb);
195 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
196 struct sk_buff **pskb,
197 const struct net_device *in,
198 const struct net_device *out,
199 int (*okfn)(struct sk_buff *))
201 /* root is playing with raw sockets. */
202 if ((*pskb)->len < sizeof(struct iphdr)
203 || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
205 printk("ipt_hook: happy cracking.\n");
208 return nf_conntrack_in(PF_INET, hooknum, pskb);
211 /* Connection tracking may drop packets, but never alters them, so
212 make it the first hook. */
213 static struct nf_hook_ops ipv4_conntrack_ops[] = {
215 .hook = ipv4_conntrack_defrag,
216 .owner = THIS_MODULE,
218 .hooknum = NF_IP_PRE_ROUTING,
219 .priority = NF_IP_PRI_CONNTRACK_DEFRAG,
222 .hook = ipv4_conntrack_in,
223 .owner = THIS_MODULE,
225 .hooknum = NF_IP_PRE_ROUTING,
226 .priority = NF_IP_PRI_CONNTRACK,
229 .hook = ipv4_conntrack_defrag,
230 .owner = THIS_MODULE,
232 .hooknum = NF_IP_LOCAL_OUT,
233 .priority = NF_IP_PRI_CONNTRACK_DEFRAG,
236 .hook = ipv4_conntrack_local,
237 .owner = THIS_MODULE,
239 .hooknum = NF_IP_LOCAL_OUT,
240 .priority = NF_IP_PRI_CONNTRACK,
243 .hook = ipv4_conntrack_help,
244 .owner = THIS_MODULE,
246 .hooknum = NF_IP_POST_ROUTING,
247 .priority = NF_IP_PRI_CONNTRACK_HELPER,
250 .hook = ipv4_conntrack_help,
251 .owner = THIS_MODULE,
253 .hooknum = NF_IP_LOCAL_IN,
254 .priority = NF_IP_PRI_CONNTRACK_HELPER,
257 .hook = ipv4_confirm,
258 .owner = THIS_MODULE,
260 .hooknum = NF_IP_POST_ROUTING,
261 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
264 .hook = ipv4_confirm,
265 .owner = THIS_MODULE,
267 .hooknum = NF_IP_LOCAL_IN,
268 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
273 /* From nf_conntrack_proto_icmp.c */
274 extern unsigned int nf_ct_icmp_timeout;
275 static struct ctl_table_header *nf_ct_ipv4_sysctl_header;
277 static ctl_table nf_ct_sysctl_table[] = {
279 .ctl_name = NET_NF_CONNTRACK_ICMP_TIMEOUT,
280 .procname = "nf_conntrack_icmp_timeout",
281 .data = &nf_ct_icmp_timeout,
282 .maxlen = sizeof(unsigned int),
284 .proc_handler = &proc_dointvec_jiffies,
289 static ctl_table nf_ct_netfilter_table[] = {
291 .ctl_name = NET_NETFILTER,
292 .procname = "netfilter",
294 .child = nf_ct_sysctl_table,
299 static ctl_table nf_ct_net_table[] = {
304 .child = nf_ct_netfilter_table,
310 /* Fast function for those who don't want to parse /proc (and I don't
312 /* Reversing the socket's dst/src point of view gives us the reply
315 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
317 struct inet_sock *inet = inet_sk(sk);
318 struct nf_conntrack_tuple_hash *h;
319 struct nf_conntrack_tuple tuple;
321 NF_CT_TUPLE_U_BLANK(&tuple);
322 tuple.src.u3.ip = inet->rcv_saddr;
323 tuple.src.u.tcp.port = inet->sport;
324 tuple.dst.u3.ip = inet->daddr;
325 tuple.dst.u.tcp.port = inet->dport;
326 tuple.src.l3num = PF_INET;
327 tuple.dst.protonum = IPPROTO_TCP;
329 /* We only do TCP at the moment: is there a better way? */
330 if (strcmp(sk->sk_prot->name, "TCP")) {
331 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
335 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
336 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
337 *len, sizeof(struct sockaddr_in));
341 h = nf_conntrack_find_get(&tuple, NULL);
343 struct sockaddr_in sin;
344 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
346 sin.sin_family = AF_INET;
347 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
348 .tuple.dst.u.tcp.port;
349 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
351 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
353 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
354 NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
356 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
361 DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
362 NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
363 NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
367 #if defined(CONFIG_NF_CT_NETLINK) || \
368 defined(CONFIG_NF_CT_NETLINK_MODULE)
370 #include <linux/netfilter/nfnetlink.h>
371 #include <linux/netfilter/nfnetlink_conntrack.h>
373 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
374 const struct nf_conntrack_tuple *tuple)
376 NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
378 NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
386 static const size_t cta_min_ip[CTA_IP_MAX] = {
387 [CTA_IP_V4_SRC-1] = sizeof(u_int32_t),
388 [CTA_IP_V4_DST-1] = sizeof(u_int32_t),
391 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
392 struct nf_conntrack_tuple *t)
394 if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
397 if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
401 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
403 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
409 static struct nf_sockopt_ops so_getorigdst = {
411 .get_optmin = SO_ORIGINAL_DST,
412 .get_optmax = SO_ORIGINAL_DST+1,
416 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
419 .pkt_to_tuple = ipv4_pkt_to_tuple,
420 .invert_tuple = ipv4_invert_tuple,
421 .print_tuple = ipv4_print_tuple,
422 .print_conntrack = ipv4_print_conntrack,
423 .prepare = ipv4_prepare,
424 .get_features = ipv4_get_features,
425 #if defined(CONFIG_NF_CT_NETLINK) || \
426 defined(CONFIG_NF_CT_NETLINK_MODULE)
427 .tuple_to_nfattr = ipv4_tuple_to_nfattr,
428 .nfattr_to_tuple = ipv4_nfattr_to_tuple,
433 extern struct nf_conntrack_protocol nf_conntrack_protocol_tcp4;
434 extern struct nf_conntrack_protocol nf_conntrack_protocol_udp4;
435 extern struct nf_conntrack_protocol nf_conntrack_protocol_icmp;
437 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
438 MODULE_LICENSE("GPL");
440 static int __init nf_conntrack_l3proto_ipv4_init(void)
446 ret = nf_register_sockopt(&so_getorigdst);
448 printk(KERN_ERR "Unable to register netfilter socket option\n");
452 ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_tcp4);
454 printk("nf_conntrack_ipv4: can't register tcp.\n");
455 goto cleanup_sockopt;
458 ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_udp4);
460 printk("nf_conntrack_ipv4: can't register udp.\n");
464 ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_icmp);
466 printk("nf_conntrack_ipv4: can't register icmp.\n");
470 ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
472 printk("nf_conntrack_ipv4: can't register ipv4\n");
476 ret = nf_register_hooks(ipv4_conntrack_ops,
477 ARRAY_SIZE(ipv4_conntrack_ops));
479 printk("nf_conntrack_ipv4: can't register hooks.\n");
483 nf_ct_ipv4_sysctl_header = register_sysctl_table(nf_ct_net_table, 0);
484 if (nf_ct_ipv4_sysctl_header == NULL) {
485 printk("nf_conntrack: can't register to sysctl.\n");
494 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
497 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
499 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_icmp);
501 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp4);
503 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_tcp4);
505 nf_unregister_sockopt(&so_getorigdst);
509 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
513 unregister_sysctl_table(nf_ct_ipv4_sysctl_header);
515 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
516 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
517 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_icmp);
518 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp4);
519 nf_conntrack_protocol_unregister(&nf_conntrack_protocol_tcp4);
520 nf_unregister_sockopt(&so_getorigdst);
523 module_init(nf_conntrack_l3proto_ipv4_init);
524 module_exit(nf_conntrack_l3proto_ipv4_fini);
526 EXPORT_SYMBOL(nf_ct_ipv4_gather_frags);