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.
9 #include <linux/types.h>
11 #include <linux/netfilter.h>
12 #include <linux/module.h>
13 #include <linux/skbuff.h>
14 #include <linux/icmp.h>
15 #include <linux/sysctl.h>
16 #include <net/route.h>
19 #include <linux/netfilter_ipv4.h>
20 #include <net/netfilter/nf_conntrack.h>
21 #include <net/netfilter/nf_conntrack_helper.h>
22 #include <net/netfilter/nf_conntrack_l4proto.h>
23 #include <net/netfilter/nf_conntrack_l3proto.h>
24 #include <net/netfilter/nf_conntrack_core.h>
25 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
27 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
28 struct nf_conntrack_tuple *tuple)
30 __be32 _addrs[2], *ap;
31 ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
32 sizeof(u_int32_t) * 2, _addrs);
36 tuple->src.u3.ip = ap[0];
37 tuple->dst.u3.ip = ap[1];
42 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
43 const struct nf_conntrack_tuple *orig)
45 tuple->src.u3.ip = orig->dst.u3.ip;
46 tuple->dst.u3.ip = orig->src.u3.ip;
51 static int ipv4_print_tuple(struct seq_file *s,
52 const struct nf_conntrack_tuple *tuple)
54 return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
55 NIPQUAD(tuple->src.u3.ip),
56 NIPQUAD(tuple->dst.u3.ip));
59 static int ipv4_print_conntrack(struct seq_file *s,
60 const struct nf_conn *conntrack)
65 /* Returns new sk_buff, or NULL */
66 static struct sk_buff *
67 nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
72 skb = ip_defrag(skb, user);
76 ip_send_check(ip_hdr(skb));
81 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
82 unsigned int *dataoff, u_int8_t *protonum)
84 struct iphdr _iph, *iph;
86 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
90 /* Conntrack defragments packets, we might still see fragments
91 * inside ICMP packets though. */
92 if (iph->frag_off & htons(IP_OFFSET))
95 *dataoff = nhoff + (iph->ihl << 2);
96 *protonum = iph->protocol;
101 static unsigned int ipv4_confirm(unsigned int hooknum,
102 struct sk_buff **pskb,
103 const struct net_device *in,
104 const struct net_device *out,
105 int (*okfn)(struct sk_buff *))
107 /* We've seen it coming out the other side: confirm it */
108 return nf_conntrack_confirm(pskb);
111 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
112 struct sk_buff **pskb,
113 const struct net_device *in,
114 const struct net_device *out,
115 int (*okfn)(struct sk_buff *))
118 enum ip_conntrack_info ctinfo;
119 struct nf_conn_help *help;
120 struct nf_conntrack_helper *helper;
122 /* This is where we call the helper: as the packet goes out. */
123 ct = nf_ct_get(*pskb, &ctinfo);
124 if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
127 help = nfct_help(ct);
130 /* rcu_read_lock()ed by nf_hook_slow */
131 helper = rcu_dereference(help->helper);
134 return helper->help(pskb, skb_network_offset(*pskb) + ip_hdrlen(*pskb),
138 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
139 struct sk_buff **pskb,
140 const struct net_device *in,
141 const struct net_device *out,
142 int (*okfn)(struct sk_buff *))
144 /* Previously seen (loopback)? Ignore. Do this before
149 /* Gather fragments. */
150 if (ip_hdr(*pskb)->frag_off & htons(IP_MF | IP_OFFSET)) {
151 *pskb = nf_ct_ipv4_gather_frags(*pskb,
152 hooknum == NF_IP_PRE_ROUTING ?
153 IP_DEFRAG_CONNTRACK_IN :
154 IP_DEFRAG_CONNTRACK_OUT);
161 static unsigned int ipv4_conntrack_in(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 return nf_conntrack_in(PF_INET, hooknum, pskb);
170 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
171 struct sk_buff **pskb,
172 const struct net_device *in,
173 const struct net_device *out,
174 int (*okfn)(struct sk_buff *))
176 /* root is playing with raw sockets. */
177 if ((*pskb)->len < sizeof(struct iphdr)
178 || ip_hdrlen(*pskb) < sizeof(struct iphdr)) {
180 printk("ipt_hook: happy cracking.\n");
183 return nf_conntrack_in(PF_INET, hooknum, pskb);
186 /* Connection tracking may drop packets, but never alters them, so
187 make it the first hook. */
188 static struct nf_hook_ops ipv4_conntrack_ops[] = {
190 .hook = ipv4_conntrack_defrag,
191 .owner = THIS_MODULE,
193 .hooknum = NF_IP_PRE_ROUTING,
194 .priority = NF_IP_PRI_CONNTRACK_DEFRAG,
197 .hook = ipv4_conntrack_in,
198 .owner = THIS_MODULE,
200 .hooknum = NF_IP_PRE_ROUTING,
201 .priority = NF_IP_PRI_CONNTRACK,
204 .hook = ipv4_conntrack_defrag,
205 .owner = THIS_MODULE,
207 .hooknum = NF_IP_LOCAL_OUT,
208 .priority = NF_IP_PRI_CONNTRACK_DEFRAG,
211 .hook = ipv4_conntrack_local,
212 .owner = THIS_MODULE,
214 .hooknum = NF_IP_LOCAL_OUT,
215 .priority = NF_IP_PRI_CONNTRACK,
218 .hook = ipv4_conntrack_help,
219 .owner = THIS_MODULE,
221 .hooknum = NF_IP_POST_ROUTING,
222 .priority = NF_IP_PRI_CONNTRACK_HELPER,
225 .hook = ipv4_conntrack_help,
226 .owner = THIS_MODULE,
228 .hooknum = NF_IP_LOCAL_IN,
229 .priority = NF_IP_PRI_CONNTRACK_HELPER,
232 .hook = ipv4_confirm,
233 .owner = THIS_MODULE,
235 .hooknum = NF_IP_POST_ROUTING,
236 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
239 .hook = ipv4_confirm,
240 .owner = THIS_MODULE,
242 .hooknum = NF_IP_LOCAL_IN,
243 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
247 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
248 static int log_invalid_proto_min = 0;
249 static int log_invalid_proto_max = 255;
251 static ctl_table ip_ct_sysctl_table[] = {
253 .ctl_name = NET_IPV4_NF_CONNTRACK_MAX,
254 .procname = "ip_conntrack_max",
255 .data = &nf_conntrack_max,
256 .maxlen = sizeof(int),
258 .proc_handler = &proc_dointvec,
261 .ctl_name = NET_IPV4_NF_CONNTRACK_COUNT,
262 .procname = "ip_conntrack_count",
263 .data = &nf_conntrack_count,
264 .maxlen = sizeof(int),
266 .proc_handler = &proc_dointvec,
269 .ctl_name = NET_IPV4_NF_CONNTRACK_BUCKETS,
270 .procname = "ip_conntrack_buckets",
271 .data = &nf_conntrack_htable_size,
272 .maxlen = sizeof(unsigned int),
274 .proc_handler = &proc_dointvec,
277 .ctl_name = NET_IPV4_NF_CONNTRACK_CHECKSUM,
278 .procname = "ip_conntrack_checksum",
279 .data = &nf_conntrack_checksum,
280 .maxlen = sizeof(int),
282 .proc_handler = &proc_dointvec,
285 .ctl_name = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
286 .procname = "ip_conntrack_log_invalid",
287 .data = &nf_ct_log_invalid,
288 .maxlen = sizeof(unsigned int),
290 .proc_handler = &proc_dointvec_minmax,
291 .strategy = &sysctl_intvec,
292 .extra1 = &log_invalid_proto_min,
293 .extra2 = &log_invalid_proto_max,
299 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
301 /* Fast function for those who don't want to parse /proc (and I don't
303 /* Reversing the socket's dst/src point of view gives us the reply
306 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
308 struct inet_sock *inet = inet_sk(sk);
309 struct nf_conntrack_tuple_hash *h;
310 struct nf_conntrack_tuple tuple;
312 NF_CT_TUPLE_U_BLANK(&tuple);
313 tuple.src.u3.ip = inet->rcv_saddr;
314 tuple.src.u.tcp.port = inet->sport;
315 tuple.dst.u3.ip = inet->daddr;
316 tuple.dst.u.tcp.port = inet->dport;
317 tuple.src.l3num = PF_INET;
318 tuple.dst.protonum = IPPROTO_TCP;
320 /* We only do TCP at the moment: is there a better way? */
321 if (strcmp(sk->sk_prot->name, "TCP")) {
322 pr_debug("SO_ORIGINAL_DST: Not a TCP socket\n");
326 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
327 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
328 *len, sizeof(struct sockaddr_in));
332 h = nf_conntrack_find_get(&tuple);
334 struct sockaddr_in sin;
335 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
337 sin.sin_family = AF_INET;
338 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
339 .tuple.dst.u.tcp.port;
340 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
342 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
344 pr_debug("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
345 NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
347 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
352 pr_debug("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
353 NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
354 NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
358 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
360 #include <linux/netfilter/nfnetlink.h>
361 #include <linux/netfilter/nfnetlink_conntrack.h>
363 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
364 const struct nf_conntrack_tuple *tuple)
366 NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
368 NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
376 static const size_t cta_min_ip[CTA_IP_MAX] = {
377 [CTA_IP_V4_SRC-1] = sizeof(u_int32_t),
378 [CTA_IP_V4_DST-1] = sizeof(u_int32_t),
381 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
382 struct nf_conntrack_tuple *t)
384 if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
387 if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
390 t->src.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
391 t->dst.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
397 static struct nf_sockopt_ops so_getorigdst = {
399 .get_optmin = SO_ORIGINAL_DST,
400 .get_optmax = SO_ORIGINAL_DST+1,
402 .owner = THIS_MODULE,
405 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
408 .pkt_to_tuple = ipv4_pkt_to_tuple,
409 .invert_tuple = ipv4_invert_tuple,
410 .print_tuple = ipv4_print_tuple,
411 .print_conntrack = ipv4_print_conntrack,
412 .get_l4proto = ipv4_get_l4proto,
413 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
414 .tuple_to_nfattr = ipv4_tuple_to_nfattr,
415 .nfattr_to_tuple = ipv4_nfattr_to_tuple,
417 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
418 .ctl_table_path = nf_net_ipv4_netfilter_sysctl_path,
419 .ctl_table = ip_ct_sysctl_table,
424 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
425 MODULE_ALIAS("ip_conntrack");
426 MODULE_LICENSE("GPL");
428 static int __init nf_conntrack_l3proto_ipv4_init(void)
434 ret = nf_register_sockopt(&so_getorigdst);
436 printk(KERN_ERR "Unable to register netfilter socket option\n");
440 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
442 printk("nf_conntrack_ipv4: can't register tcp.\n");
443 goto cleanup_sockopt;
446 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
448 printk("nf_conntrack_ipv4: can't register udp.\n");
452 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
454 printk("nf_conntrack_ipv4: can't register icmp.\n");
458 ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
460 printk("nf_conntrack_ipv4: can't register ipv4\n");
464 ret = nf_register_hooks(ipv4_conntrack_ops,
465 ARRAY_SIZE(ipv4_conntrack_ops));
467 printk("nf_conntrack_ipv4: can't register hooks.\n");
470 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
471 ret = nf_conntrack_ipv4_compat_init();
476 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
478 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
481 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
483 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
485 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
487 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
489 nf_unregister_sockopt(&so_getorigdst);
493 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
496 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
497 nf_conntrack_ipv4_compat_fini();
499 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
500 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
501 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
502 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
503 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
504 nf_unregister_sockopt(&so_getorigdst);
507 module_init(nf_conntrack_l3proto_ipv4_init);
508 module_exit(nf_conntrack_l3proto_ipv4_fini);
510 void need_ipv4_conntrack(void)
514 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);