2 /* (C) 1999-2001 Paul `Rusty' Russell
3 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
10 #include <linux/types.h>
12 #include <linux/netfilter.h>
13 #include <linux/module.h>
14 #include <linux/skbuff.h>
15 #include <linux/icmp.h>
16 #include <linux/sysctl.h>
17 #include <net/route.h>
20 #include <linux/netfilter_ipv4.h>
21 #include <net/netfilter/nf_conntrack.h>
22 #include <net/netfilter/nf_conntrack_helper.h>
23 #include <net/netfilter/nf_conntrack_l4proto.h>
24 #include <net/netfilter/nf_conntrack_l3proto.h>
25 #include <net/netfilter/nf_conntrack_core.h>
26 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
27 #include <net/netfilter/nf_nat_helper.h>
28 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
29 #include <net/netfilter/nf_log.h>
31 int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb,
33 enum ip_conntrack_info ctinfo);
34 EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook);
36 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
37 struct nf_conntrack_tuple *tuple)
41 ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
42 sizeof(u_int32_t) * 2, _addrs);
46 tuple->src.u3.ip = ap[0];
47 tuple->dst.u3.ip = ap[1];
52 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
53 const struct nf_conntrack_tuple *orig)
55 tuple->src.u3.ip = orig->dst.u3.ip;
56 tuple->dst.u3.ip = orig->src.u3.ip;
61 static int ipv4_print_tuple(struct seq_file *s,
62 const struct nf_conntrack_tuple *tuple)
64 return seq_printf(s, "src=%pI4 dst=%pI4 ",
65 &tuple->src.u3.ip, &tuple->dst.u3.ip);
68 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
69 unsigned int *dataoff, u_int8_t *protonum)
71 const struct iphdr *iph;
74 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
78 /* Conntrack defragments packets, we might still see fragments
79 * inside ICMP packets though. */
80 if (iph->frag_off & htons(IP_OFFSET))
83 *dataoff = nhoff + (iph->ihl << 2);
84 *protonum = iph->protocol;
89 static unsigned int ipv4_confirm(unsigned int hooknum,
91 const struct net_device *in,
92 const struct net_device *out,
93 int (*okfn)(struct sk_buff *))
96 enum ip_conntrack_info ctinfo;
97 const struct nf_conn_help *help;
98 const struct nf_conntrack_helper *helper;
101 /* This is where we call the helper: as the packet goes out. */
102 ct = nf_ct_get(skb, &ctinfo);
103 if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
106 help = nfct_help(ct);
110 /* rcu_read_lock()ed by nf_hook_slow */
111 helper = rcu_dereference(help->helper);
115 ret = helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
117 if (ret != NF_ACCEPT) {
118 nf_log_packet(NFPROTO_IPV4, hooknum, skb, in, out, NULL,
119 "nf_ct_%s: dropping packet", helper->name);
123 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
124 typeof(nf_nat_seq_adjust_hook) seq_adjust;
126 seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
127 if (!seq_adjust || !seq_adjust(skb, ct, ctinfo)) {
128 NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
133 /* We've seen it coming out the other side: confirm it */
134 return nf_conntrack_confirm(skb);
137 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
139 const struct net_device *in,
140 const struct net_device *out,
141 int (*okfn)(struct sk_buff *))
143 return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
146 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
148 const struct net_device *in,
149 const struct net_device *out,
150 int (*okfn)(struct sk_buff *))
152 /* root is playing with raw sockets. */
153 if (skb->len < sizeof(struct iphdr) ||
154 ip_hdrlen(skb) < sizeof(struct iphdr))
156 return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
159 /* Connection tracking may drop packets, but never alters them, so
160 make it the first hook. */
161 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
163 .hook = ipv4_conntrack_in,
164 .owner = THIS_MODULE,
166 .hooknum = NF_INET_PRE_ROUTING,
167 .priority = NF_IP_PRI_CONNTRACK,
170 .hook = ipv4_conntrack_local,
171 .owner = THIS_MODULE,
173 .hooknum = NF_INET_LOCAL_OUT,
174 .priority = NF_IP_PRI_CONNTRACK,
177 .hook = ipv4_confirm,
178 .owner = THIS_MODULE,
180 .hooknum = NF_INET_POST_ROUTING,
181 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
184 .hook = ipv4_confirm,
185 .owner = THIS_MODULE,
187 .hooknum = NF_INET_LOCAL_IN,
188 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
192 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
193 static int log_invalid_proto_min = 0;
194 static int log_invalid_proto_max = 255;
196 static ctl_table ip_ct_sysctl_table[] = {
198 .ctl_name = NET_IPV4_NF_CONNTRACK_MAX,
199 .procname = "ip_conntrack_max",
200 .data = &nf_conntrack_max,
201 .maxlen = sizeof(int),
203 .proc_handler = proc_dointvec,
206 .ctl_name = NET_IPV4_NF_CONNTRACK_COUNT,
207 .procname = "ip_conntrack_count",
208 .data = &init_net.ct.count,
209 .maxlen = sizeof(int),
211 .proc_handler = proc_dointvec,
214 .ctl_name = NET_IPV4_NF_CONNTRACK_BUCKETS,
215 .procname = "ip_conntrack_buckets",
216 .data = &nf_conntrack_htable_size,
217 .maxlen = sizeof(unsigned int),
219 .proc_handler = proc_dointvec,
222 .ctl_name = NET_IPV4_NF_CONNTRACK_CHECKSUM,
223 .procname = "ip_conntrack_checksum",
224 .data = &init_net.ct.sysctl_checksum,
225 .maxlen = sizeof(int),
227 .proc_handler = proc_dointvec,
230 .ctl_name = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
231 .procname = "ip_conntrack_log_invalid",
232 .data = &init_net.ct.sysctl_log_invalid,
233 .maxlen = sizeof(unsigned int),
235 .proc_handler = proc_dointvec_minmax,
236 .strategy = sysctl_intvec,
237 .extra1 = &log_invalid_proto_min,
238 .extra2 = &log_invalid_proto_max,
244 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
246 /* Fast function for those who don't want to parse /proc (and I don't
248 /* Reversing the socket's dst/src point of view gives us the reply
251 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
253 const struct inet_sock *inet = inet_sk(sk);
254 const struct nf_conntrack_tuple_hash *h;
255 struct nf_conntrack_tuple tuple;
257 memset(&tuple, 0, sizeof(tuple));
258 tuple.src.u3.ip = inet->rcv_saddr;
259 tuple.src.u.tcp.port = inet->sport;
260 tuple.dst.u3.ip = inet->daddr;
261 tuple.dst.u.tcp.port = inet->dport;
262 tuple.src.l3num = PF_INET;
263 tuple.dst.protonum = sk->sk_protocol;
265 /* We only do TCP and SCTP at the moment: is there a better way? */
266 if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP) {
267 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
271 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
272 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
273 *len, sizeof(struct sockaddr_in));
277 h = nf_conntrack_find_get(sock_net(sk), &tuple);
279 struct sockaddr_in sin;
280 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
282 sin.sin_family = AF_INET;
283 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
284 .tuple.dst.u.tcp.port;
285 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
287 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
289 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
290 &sin.sin_addr.s_addr, ntohs(sin.sin_port));
292 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
297 pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
298 &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
299 &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
303 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
305 #include <linux/netfilter/nfnetlink.h>
306 #include <linux/netfilter/nfnetlink_conntrack.h>
308 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
309 const struct nf_conntrack_tuple *tuple)
311 NLA_PUT_BE32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip);
312 NLA_PUT_BE32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip);
319 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
320 [CTA_IP_V4_SRC] = { .type = NLA_U32 },
321 [CTA_IP_V4_DST] = { .type = NLA_U32 },
324 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
325 struct nf_conntrack_tuple *t)
327 if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
330 t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
331 t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
336 static int ipv4_nlattr_tuple_size(void)
338 return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
342 static struct nf_sockopt_ops so_getorigdst = {
344 .get_optmin = SO_ORIGINAL_DST,
345 .get_optmax = SO_ORIGINAL_DST+1,
347 .owner = THIS_MODULE,
350 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
353 .pkt_to_tuple = ipv4_pkt_to_tuple,
354 .invert_tuple = ipv4_invert_tuple,
355 .print_tuple = ipv4_print_tuple,
356 .get_l4proto = ipv4_get_l4proto,
357 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
358 .tuple_to_nlattr = ipv4_tuple_to_nlattr,
359 .nlattr_tuple_size = ipv4_nlattr_tuple_size,
360 .nlattr_to_tuple = ipv4_nlattr_to_tuple,
361 .nla_policy = ipv4_nla_policy,
363 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
364 .ctl_table_path = nf_net_ipv4_netfilter_sysctl_path,
365 .ctl_table = ip_ct_sysctl_table,
370 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
371 &nf_conntrack_htable_size, 0600);
373 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
374 MODULE_ALIAS("ip_conntrack");
375 MODULE_LICENSE("GPL");
377 static int __init nf_conntrack_l3proto_ipv4_init(void)
382 nf_defrag_ipv4_enable();
384 ret = nf_register_sockopt(&so_getorigdst);
386 printk(KERN_ERR "Unable to register netfilter socket option\n");
390 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
392 printk("nf_conntrack_ipv4: can't register tcp.\n");
393 goto cleanup_sockopt;
396 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
398 printk("nf_conntrack_ipv4: can't register udp.\n");
402 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
404 printk("nf_conntrack_ipv4: can't register icmp.\n");
408 ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
410 printk("nf_conntrack_ipv4: can't register ipv4\n");
414 ret = nf_register_hooks(ipv4_conntrack_ops,
415 ARRAY_SIZE(ipv4_conntrack_ops));
417 printk("nf_conntrack_ipv4: can't register hooks.\n");
420 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
421 ret = nf_conntrack_ipv4_compat_init();
426 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
428 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
431 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
433 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
435 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
437 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
439 nf_unregister_sockopt(&so_getorigdst);
443 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
446 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
447 nf_conntrack_ipv4_compat_fini();
449 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
450 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
451 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
452 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
453 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
454 nf_unregister_sockopt(&so_getorigdst);
457 module_init(nf_conntrack_l3proto_ipv4_init);
458 module_exit(nf_conntrack_l3proto_ipv4_fini);
460 void need_ipv4_conntrack(void)
464 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);