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_zones.h>
26 #include <net/netfilter/nf_conntrack_core.h>
27 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
28 #include <net/netfilter/nf_nat_helper.h>
29 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
30 #include <net/netfilter/nf_log.h>
32 int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb,
34 enum ip_conntrack_info ctinfo);
35 EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook);
37 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
38 struct nf_conntrack_tuple *tuple)
42 ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
43 sizeof(u_int32_t) * 2, _addrs);
47 tuple->src.u3.ip = ap[0];
48 tuple->dst.u3.ip = ap[1];
53 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
54 const struct nf_conntrack_tuple *orig)
56 tuple->src.u3.ip = orig->dst.u3.ip;
57 tuple->dst.u3.ip = orig->src.u3.ip;
62 static int ipv4_print_tuple(struct seq_file *s,
63 const struct nf_conntrack_tuple *tuple)
65 return seq_printf(s, "src=%pI4 dst=%pI4 ",
66 &tuple->src.u3.ip, &tuple->dst.u3.ip);
69 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
70 unsigned int *dataoff, u_int8_t *protonum)
72 const struct iphdr *iph;
75 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
79 /* Conntrack defragments packets, we might still see fragments
80 * inside ICMP packets though. */
81 if (iph->frag_off & htons(IP_OFFSET))
84 *dataoff = nhoff + (iph->ihl << 2);
85 *protonum = iph->protocol;
87 /* Check bogus IP headers */
88 if (*dataoff > skb->len) {
89 pr_debug("nf_conntrack_ipv4: bogus IPv4 packet: "
90 "nhoff %u, ihl %u, skblen %u\n",
91 nhoff, iph->ihl << 2, skb->len);
98 static unsigned int ipv4_helper(unsigned int hooknum,
100 const struct net_device *in,
101 const struct net_device *out,
102 int (*okfn)(struct sk_buff *))
105 enum ip_conntrack_info ctinfo;
106 const struct nf_conn_help *help;
107 const struct nf_conntrack_helper *helper;
110 /* This is where we call the helper: as the packet goes out. */
111 ct = nf_ct_get(skb, &ctinfo);
112 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
115 help = nfct_help(ct);
119 /* rcu_read_lock()ed by nf_hook_slow */
120 helper = rcu_dereference(help->helper);
124 ret = helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
126 if (ret != NF_ACCEPT && (ret & NF_VERDICT_MASK) != NF_QUEUE) {
127 nf_log_packet(NFPROTO_IPV4, hooknum, skb, in, out, NULL,
128 "nf_ct_%s: dropping packet", helper->name);
133 static unsigned int ipv4_confirm(unsigned int hooknum,
135 const struct net_device *in,
136 const struct net_device *out,
137 int (*okfn)(struct sk_buff *))
140 enum ip_conntrack_info ctinfo;
142 ct = nf_ct_get(skb, &ctinfo);
143 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
146 /* adjust seqs for loopback traffic only in outgoing direction */
147 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
148 !nf_is_loopback_packet(skb)) {
149 typeof(nf_nat_seq_adjust_hook) seq_adjust;
151 seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
152 if (!seq_adjust || !seq_adjust(skb, ct, ctinfo)) {
153 NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
158 /* We've seen it coming out the other side: confirm it */
159 return nf_conntrack_confirm(skb);
162 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
164 const struct net_device *in,
165 const struct net_device *out,
166 int (*okfn)(struct sk_buff *))
168 return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
171 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
173 const struct net_device *in,
174 const struct net_device *out,
175 int (*okfn)(struct sk_buff *))
177 /* root is playing with raw sockets. */
178 if (skb->len < sizeof(struct iphdr) ||
179 ip_hdrlen(skb) < sizeof(struct iphdr))
181 return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
184 /* Connection tracking may drop packets, but never alters them, so
185 make it the first hook. */
186 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
188 .hook = ipv4_conntrack_in,
189 .owner = THIS_MODULE,
191 .hooknum = NF_INET_PRE_ROUTING,
192 .priority = NF_IP_PRI_CONNTRACK,
195 .hook = ipv4_conntrack_local,
196 .owner = THIS_MODULE,
198 .hooknum = NF_INET_LOCAL_OUT,
199 .priority = NF_IP_PRI_CONNTRACK,
203 .owner = THIS_MODULE,
205 .hooknum = NF_INET_POST_ROUTING,
206 .priority = NF_IP_PRI_CONNTRACK_HELPER,
209 .hook = ipv4_confirm,
210 .owner = THIS_MODULE,
212 .hooknum = NF_INET_POST_ROUTING,
213 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
217 .owner = THIS_MODULE,
219 .hooknum = NF_INET_LOCAL_IN,
220 .priority = NF_IP_PRI_CONNTRACK_HELPER,
223 .hook = ipv4_confirm,
224 .owner = THIS_MODULE,
226 .hooknum = NF_INET_LOCAL_IN,
227 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
231 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
232 static int log_invalid_proto_min = 0;
233 static int log_invalid_proto_max = 255;
235 static ctl_table ip_ct_sysctl_table[] = {
237 .procname = "ip_conntrack_max",
238 .maxlen = sizeof(int),
240 .proc_handler = proc_dointvec,
243 .procname = "ip_conntrack_count",
244 .maxlen = sizeof(int),
246 .proc_handler = proc_dointvec,
249 .procname = "ip_conntrack_buckets",
250 .maxlen = sizeof(unsigned int),
252 .proc_handler = proc_dointvec,
255 .procname = "ip_conntrack_checksum",
256 .maxlen = sizeof(int),
258 .proc_handler = proc_dointvec,
261 .procname = "ip_conntrack_log_invalid",
262 .maxlen = sizeof(unsigned int),
264 .proc_handler = proc_dointvec_minmax,
265 .extra1 = &log_invalid_proto_min,
266 .extra2 = &log_invalid_proto_max,
270 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
272 /* Fast function for those who don't want to parse /proc (and I don't
274 /* Reversing the socket's dst/src point of view gives us the reply
277 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
279 const struct inet_sock *inet = inet_sk(sk);
280 const struct nf_conntrack_tuple_hash *h;
281 struct nf_conntrack_tuple tuple;
283 memset(&tuple, 0, sizeof(tuple));
284 tuple.src.u3.ip = inet->inet_rcv_saddr;
285 tuple.src.u.tcp.port = inet->inet_sport;
286 tuple.dst.u3.ip = inet->inet_daddr;
287 tuple.dst.u.tcp.port = inet->inet_dport;
288 tuple.src.l3num = PF_INET;
289 tuple.dst.protonum = sk->sk_protocol;
291 /* We only do TCP and SCTP at the moment: is there a better way? */
292 if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP) {
293 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
297 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
298 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
299 *len, sizeof(struct sockaddr_in));
303 h = nf_conntrack_find_get(sock_net(sk), NF_CT_DEFAULT_ZONE, &tuple);
305 struct sockaddr_in sin;
306 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
308 sin.sin_family = AF_INET;
309 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
310 .tuple.dst.u.tcp.port;
311 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
313 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
315 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
316 &sin.sin_addr.s_addr, ntohs(sin.sin_port));
318 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
323 pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
324 &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
325 &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
329 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
331 #include <linux/netfilter/nfnetlink.h>
332 #include <linux/netfilter/nfnetlink_conntrack.h>
334 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
335 const struct nf_conntrack_tuple *tuple)
337 if (nla_put_be32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) ||
338 nla_put_be32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip))
339 goto nla_put_failure;
346 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
347 [CTA_IP_V4_SRC] = { .type = NLA_U32 },
348 [CTA_IP_V4_DST] = { .type = NLA_U32 },
351 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
352 struct nf_conntrack_tuple *t)
354 if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
357 t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
358 t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
363 static int ipv4_nlattr_tuple_size(void)
365 return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
369 static struct nf_sockopt_ops so_getorigdst = {
371 .get_optmin = SO_ORIGINAL_DST,
372 .get_optmax = SO_ORIGINAL_DST+1,
374 .owner = THIS_MODULE,
377 static int ipv4_init_net(struct net *net)
379 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
380 struct nf_ip_net *in = &net->ct.nf_ct_proto;
381 in->ctl_table = kmemdup(ip_ct_sysctl_table,
382 sizeof(ip_ct_sysctl_table),
387 in->ctl_table[0].data = &nf_conntrack_max;
388 in->ctl_table[1].data = &net->ct.count;
389 in->ctl_table[2].data = &net->ct.htable_size;
390 in->ctl_table[3].data = &net->ct.sysctl_checksum;
391 in->ctl_table[4].data = &net->ct.sysctl_log_invalid;
396 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
399 .pkt_to_tuple = ipv4_pkt_to_tuple,
400 .invert_tuple = ipv4_invert_tuple,
401 .print_tuple = ipv4_print_tuple,
402 .get_l4proto = ipv4_get_l4proto,
403 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
404 .tuple_to_nlattr = ipv4_tuple_to_nlattr,
405 .nlattr_tuple_size = ipv4_nlattr_tuple_size,
406 .nlattr_to_tuple = ipv4_nlattr_to_tuple,
407 .nla_policy = ipv4_nla_policy,
409 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
410 .ctl_table_path = "net/ipv4/netfilter",
412 .init_net = ipv4_init_net,
416 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
417 &nf_conntrack_htable_size, 0600);
419 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
420 MODULE_ALIAS("ip_conntrack");
421 MODULE_LICENSE("GPL");
423 static int ipv4_net_init(struct net *net)
427 ret = nf_conntrack_l4proto_register(net,
428 &nf_conntrack_l4proto_tcp4);
430 pr_err("nf_conntrack_l4proto_tcp4 :protocol register failed\n");
433 ret = nf_conntrack_l4proto_register(net,
434 &nf_conntrack_l4proto_udp4);
436 pr_err("nf_conntrack_l4proto_udp4 :protocol register failed\n");
439 ret = nf_conntrack_l4proto_register(net,
440 &nf_conntrack_l4proto_icmp);
442 pr_err("nf_conntrack_l4proto_icmp4 :protocol register failed\n");
445 ret = nf_conntrack_l3proto_register(net,
446 &nf_conntrack_l3proto_ipv4);
448 pr_err("nf_conntrack_l3proto_ipv4 :protocol register failed\n");
453 nf_conntrack_l4proto_unregister(net,
454 &nf_conntrack_l4proto_icmp);
456 nf_conntrack_l4proto_unregister(net,
457 &nf_conntrack_l4proto_udp4);
459 nf_conntrack_l4proto_unregister(net,
460 &nf_conntrack_l4proto_tcp4);
465 static void ipv4_net_exit(struct net *net)
467 nf_conntrack_l3proto_unregister(net,
468 &nf_conntrack_l3proto_ipv4);
469 nf_conntrack_l4proto_unregister(net,
470 &nf_conntrack_l4proto_icmp);
471 nf_conntrack_l4proto_unregister(net,
472 &nf_conntrack_l4proto_udp4);
473 nf_conntrack_l4proto_unregister(net,
474 &nf_conntrack_l4proto_tcp4);
477 static struct pernet_operations ipv4_net_ops = {
478 .init = ipv4_net_init,
479 .exit = ipv4_net_exit,
482 static int __init nf_conntrack_l3proto_ipv4_init(void)
487 nf_defrag_ipv4_enable();
489 ret = nf_register_sockopt(&so_getorigdst);
491 printk(KERN_ERR "Unable to register netfilter socket option\n");
495 ret = register_pernet_subsys(&ipv4_net_ops);
497 pr_err("nf_conntrack_ipv4: can't register pernet ops\n");
498 goto cleanup_sockopt;
501 ret = nf_register_hooks(ipv4_conntrack_ops,
502 ARRAY_SIZE(ipv4_conntrack_ops));
504 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
507 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
508 ret = nf_conntrack_ipv4_compat_init();
513 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
515 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
518 unregister_pernet_subsys(&ipv4_net_ops);
520 nf_unregister_sockopt(&so_getorigdst);
524 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
527 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
528 nf_conntrack_ipv4_compat_fini();
530 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
531 unregister_pernet_subsys(&ipv4_net_ops);
532 nf_unregister_sockopt(&so_getorigdst);
535 module_init(nf_conntrack_l3proto_ipv4_init);
536 module_exit(nf_conntrack_l3proto_ipv4_fini);
538 void need_ipv4_conntrack(void)
542 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);