2 /* (C) 1999-2001 Paul `Rusty' Russell
3 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
4 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
11 #include <linux/types.h>
13 #include <linux/netfilter.h>
14 #include <linux/module.h>
15 #include <linux/skbuff.h>
16 #include <linux/icmp.h>
17 #include <linux/sysctl.h>
18 #include <net/route.h>
21 #include <linux/netfilter_ipv4.h>
22 #include <net/netfilter/nf_conntrack.h>
23 #include <net/netfilter/nf_conntrack_helper.h>
24 #include <net/netfilter/nf_conntrack_l4proto.h>
25 #include <net/netfilter/nf_conntrack_l3proto.h>
26 #include <net/netfilter/nf_conntrack_zones.h>
27 #include <net/netfilter/nf_conntrack_core.h>
28 #include <net/netfilter/nf_conntrack_seqadj.h>
29 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
30 #include <net/netfilter/nf_nat_helper.h>
31 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
32 #include <net/netfilter/nf_log.h>
34 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
35 struct nf_conntrack_tuple *tuple)
39 ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
40 sizeof(u_int32_t) * 2, _addrs);
44 tuple->src.u3.ip = ap[0];
45 tuple->dst.u3.ip = ap[1];
50 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
51 const struct nf_conntrack_tuple *orig)
53 tuple->src.u3.ip = orig->dst.u3.ip;
54 tuple->dst.u3.ip = orig->src.u3.ip;
59 static int ipv4_print_tuple(struct seq_file *s,
60 const struct nf_conntrack_tuple *tuple)
62 return seq_printf(s, "src=%pI4 dst=%pI4 ",
63 &tuple->src.u3.ip, &tuple->dst.u3.ip);
66 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
67 unsigned int *dataoff, u_int8_t *protonum)
69 const struct iphdr *iph;
72 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
76 /* Conntrack defragments packets, we might still see fragments
77 * inside ICMP packets though. */
78 if (iph->frag_off & htons(IP_OFFSET))
81 *dataoff = nhoff + (iph->ihl << 2);
82 *protonum = iph->protocol;
84 /* Check bogus IP headers */
85 if (*dataoff > skb->len) {
86 pr_debug("nf_conntrack_ipv4: bogus IPv4 packet: "
87 "nhoff %u, ihl %u, skblen %u\n",
88 nhoff, iph->ihl << 2, skb->len);
95 static unsigned int ipv4_helper(const struct nf_hook_ops *ops,
97 const struct net_device *in,
98 const struct net_device *out,
99 int (*okfn)(struct sk_buff *))
102 enum ip_conntrack_info ctinfo;
103 const struct nf_conn_help *help;
104 const struct nf_conntrack_helper *helper;
106 /* This is where we call the helper: as the packet goes out. */
107 ct = nf_ct_get(skb, &ctinfo);
108 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
111 help = nfct_help(ct);
115 /* rcu_read_lock()ed by nf_hook_slow */
116 helper = rcu_dereference(help->helper);
120 return helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
124 static unsigned int ipv4_confirm(const struct nf_hook_ops *ops,
126 const struct net_device *in,
127 const struct net_device *out,
128 int (*okfn)(struct sk_buff *))
131 enum ip_conntrack_info ctinfo;
133 ct = nf_ct_get(skb, &ctinfo);
134 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
137 /* adjust seqs for loopback traffic only in outgoing direction */
138 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
139 !nf_is_loopback_packet(skb)) {
140 if (!nf_ct_seq_adjust(skb, ct, ctinfo, ip_hdrlen(skb))) {
141 NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
146 /* We've seen it coming out the other side: confirm it */
147 return nf_conntrack_confirm(skb);
150 static unsigned int ipv4_conntrack_in(const struct nf_hook_ops *ops,
152 const struct net_device *in,
153 const struct net_device *out,
154 int (*okfn)(struct sk_buff *))
156 return nf_conntrack_in(dev_net(in), PF_INET, ops->hooknum, skb);
159 static unsigned int ipv4_conntrack_local(const struct nf_hook_ops *ops,
161 const struct net_device *in,
162 const struct net_device *out,
163 int (*okfn)(struct sk_buff *))
165 /* root is playing with raw sockets. */
166 if (skb->len < sizeof(struct iphdr) ||
167 ip_hdrlen(skb) < sizeof(struct iphdr))
169 return nf_conntrack_in(dev_net(out), PF_INET, ops->hooknum, skb);
172 /* Connection tracking may drop packets, but never alters them, so
173 make it the first hook. */
174 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
176 .hook = ipv4_conntrack_in,
177 .owner = THIS_MODULE,
179 .hooknum = NF_INET_PRE_ROUTING,
180 .priority = NF_IP_PRI_CONNTRACK,
183 .hook = ipv4_conntrack_local,
184 .owner = THIS_MODULE,
186 .hooknum = NF_INET_LOCAL_OUT,
187 .priority = NF_IP_PRI_CONNTRACK,
191 .owner = THIS_MODULE,
193 .hooknum = NF_INET_POST_ROUTING,
194 .priority = NF_IP_PRI_CONNTRACK_HELPER,
197 .hook = ipv4_confirm,
198 .owner = THIS_MODULE,
200 .hooknum = NF_INET_POST_ROUTING,
201 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
205 .owner = THIS_MODULE,
207 .hooknum = NF_INET_LOCAL_IN,
208 .priority = NF_IP_PRI_CONNTRACK_HELPER,
211 .hook = ipv4_confirm,
212 .owner = THIS_MODULE,
214 .hooknum = NF_INET_LOCAL_IN,
215 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
219 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
220 static int log_invalid_proto_min = 0;
221 static int log_invalid_proto_max = 255;
223 static struct ctl_table ip_ct_sysctl_table[] = {
225 .procname = "ip_conntrack_max",
226 .maxlen = sizeof(int),
228 .proc_handler = proc_dointvec,
231 .procname = "ip_conntrack_count",
232 .maxlen = sizeof(int),
234 .proc_handler = proc_dointvec,
237 .procname = "ip_conntrack_buckets",
238 .maxlen = sizeof(unsigned int),
240 .proc_handler = proc_dointvec,
243 .procname = "ip_conntrack_checksum",
244 .maxlen = sizeof(int),
246 .proc_handler = proc_dointvec,
249 .procname = "ip_conntrack_log_invalid",
250 .maxlen = sizeof(unsigned int),
252 .proc_handler = proc_dointvec_minmax,
253 .extra1 = &log_invalid_proto_min,
254 .extra2 = &log_invalid_proto_max,
258 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
260 /* Fast function for those who don't want to parse /proc (and I don't
262 /* Reversing the socket's dst/src point of view gives us the reply
265 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
267 const struct inet_sock *inet = inet_sk(sk);
268 const struct nf_conntrack_tuple_hash *h;
269 struct nf_conntrack_tuple tuple;
271 memset(&tuple, 0, sizeof(tuple));
272 tuple.src.u3.ip = inet->inet_rcv_saddr;
273 tuple.src.u.tcp.port = inet->inet_sport;
274 tuple.dst.u3.ip = inet->inet_daddr;
275 tuple.dst.u.tcp.port = inet->inet_dport;
276 tuple.src.l3num = PF_INET;
277 tuple.dst.protonum = sk->sk_protocol;
279 /* We only do TCP and SCTP at the moment: is there a better way? */
280 if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP) {
281 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
285 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
286 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
287 *len, sizeof(struct sockaddr_in));
291 h = nf_conntrack_find_get(sock_net(sk), NF_CT_DEFAULT_ZONE, &tuple);
293 struct sockaddr_in sin;
294 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
296 sin.sin_family = AF_INET;
297 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
298 .tuple.dst.u.tcp.port;
299 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
301 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
303 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
304 &sin.sin_addr.s_addr, ntohs(sin.sin_port));
306 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
311 pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
312 &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
313 &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
317 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
319 #include <linux/netfilter/nfnetlink.h>
320 #include <linux/netfilter/nfnetlink_conntrack.h>
322 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
323 const struct nf_conntrack_tuple *tuple)
325 if (nla_put_be32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) ||
326 nla_put_be32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip))
327 goto nla_put_failure;
334 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
335 [CTA_IP_V4_SRC] = { .type = NLA_U32 },
336 [CTA_IP_V4_DST] = { .type = NLA_U32 },
339 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
340 struct nf_conntrack_tuple *t)
342 if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
345 t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
346 t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
351 static int ipv4_nlattr_tuple_size(void)
353 return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
357 static struct nf_sockopt_ops so_getorigdst = {
359 .get_optmin = SO_ORIGINAL_DST,
360 .get_optmax = SO_ORIGINAL_DST+1,
362 .owner = THIS_MODULE,
365 static int ipv4_init_net(struct net *net)
367 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
368 struct nf_ip_net *in = &net->ct.nf_ct_proto;
369 in->ctl_table = kmemdup(ip_ct_sysctl_table,
370 sizeof(ip_ct_sysctl_table),
375 in->ctl_table[0].data = &nf_conntrack_max;
376 in->ctl_table[1].data = &net->ct.count;
377 in->ctl_table[2].data = &net->ct.htable_size;
378 in->ctl_table[3].data = &net->ct.sysctl_checksum;
379 in->ctl_table[4].data = &net->ct.sysctl_log_invalid;
384 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
387 .pkt_to_tuple = ipv4_pkt_to_tuple,
388 .invert_tuple = ipv4_invert_tuple,
389 .print_tuple = ipv4_print_tuple,
390 .get_l4proto = ipv4_get_l4proto,
391 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
392 .tuple_to_nlattr = ipv4_tuple_to_nlattr,
393 .nlattr_tuple_size = ipv4_nlattr_tuple_size,
394 .nlattr_to_tuple = ipv4_nlattr_to_tuple,
395 .nla_policy = ipv4_nla_policy,
397 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
398 .ctl_table_path = "net/ipv4/netfilter",
400 .init_net = ipv4_init_net,
404 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
405 &nf_conntrack_htable_size, 0600);
407 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
408 MODULE_ALIAS("ip_conntrack");
409 MODULE_LICENSE("GPL");
411 static int ipv4_net_init(struct net *net)
415 ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_tcp4);
417 pr_err("nf_conntrack_tcp4: pernet registration failed\n");
420 ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_udp4);
422 pr_err("nf_conntrack_udp4: pernet registration failed\n");
425 ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_icmp);
427 pr_err("nf_conntrack_icmp4: pernet registration failed\n");
430 ret = nf_ct_l3proto_pernet_register(net, &nf_conntrack_l3proto_ipv4);
432 pr_err("nf_conntrack_ipv4: pernet registration failed\n");
437 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_icmp);
439 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_udp4);
441 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_tcp4);
446 static void ipv4_net_exit(struct net *net)
448 nf_ct_l3proto_pernet_unregister(net, &nf_conntrack_l3proto_ipv4);
449 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_icmp);
450 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_udp4);
451 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_tcp4);
454 static struct pernet_operations ipv4_net_ops = {
455 .init = ipv4_net_init,
456 .exit = ipv4_net_exit,
459 static int __init nf_conntrack_l3proto_ipv4_init(void)
464 nf_defrag_ipv4_enable();
466 ret = nf_register_sockopt(&so_getorigdst);
468 printk(KERN_ERR "Unable to register netfilter socket option\n");
472 ret = register_pernet_subsys(&ipv4_net_ops);
474 pr_err("nf_conntrack_ipv4: can't register pernet ops\n");
475 goto cleanup_sockopt;
478 ret = nf_register_hooks(ipv4_conntrack_ops,
479 ARRAY_SIZE(ipv4_conntrack_ops));
481 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
485 ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_tcp4);
487 pr_err("nf_conntrack_ipv4: can't register tcp4 proto.\n");
491 ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_udp4);
493 pr_err("nf_conntrack_ipv4: can't register udp4 proto.\n");
497 ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_icmp);
499 pr_err("nf_conntrack_ipv4: can't register icmpv4 proto.\n");
503 ret = nf_ct_l3proto_register(&nf_conntrack_l3proto_ipv4);
505 pr_err("nf_conntrack_ipv4: can't register ipv4 proto.\n");
509 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
510 ret = nf_conntrack_ipv4_compat_init();
515 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
517 nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
520 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmp);
522 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp4);
524 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
526 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
528 unregister_pernet_subsys(&ipv4_net_ops);
530 nf_unregister_sockopt(&so_getorigdst);
534 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
537 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
538 nf_conntrack_ipv4_compat_fini();
540 nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
541 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmp);
542 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp4);
543 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
544 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
545 unregister_pernet_subsys(&ipv4_net_ops);
546 nf_unregister_sockopt(&so_getorigdst);
549 module_init(nf_conntrack_l3proto_ipv4_init);
550 module_exit(nf_conntrack_l3proto_ipv4_fini);