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
;
90 static unsigned int ipv4_confirm(unsigned int hooknum
,
92 const struct net_device
*in
,
93 const struct net_device
*out
,
94 int (*okfn
)(struct sk_buff
*))
97 enum ip_conntrack_info ctinfo
;
98 const struct nf_conn_help
*help
;
99 const struct nf_conntrack_helper
*helper
;
102 /* This is where we call the helper: as the packet goes out. */
103 ct
= nf_ct_get(skb
, &ctinfo
);
104 if (!ct
|| ctinfo
== IP_CT_RELATED_REPLY
)
107 help
= nfct_help(ct
);
111 /* rcu_read_lock()ed by nf_hook_slow */
112 helper
= rcu_dereference(help
->helper
);
116 ret
= helper
->help(skb
, skb_network_offset(skb
) + ip_hdrlen(skb
),
118 if (ret
!= NF_ACCEPT
) {
119 nf_log_packet(NFPROTO_IPV4
, hooknum
, skb
, in
, out
, NULL
,
120 "nf_ct_%s: dropping packet", helper
->name
);
124 /* adjust seqs for loopback traffic only in outgoing direction */
125 if (test_bit(IPS_SEQ_ADJUST_BIT
, &ct
->status
) &&
126 !nf_is_loopback_packet(skb
)) {
127 typeof(nf_nat_seq_adjust_hook
) seq_adjust
;
129 seq_adjust
= rcu_dereference(nf_nat_seq_adjust_hook
);
130 if (!seq_adjust
|| !seq_adjust(skb
, ct
, ctinfo
)) {
131 NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct
), drop
);
136 /* We've seen it coming out the other side: confirm it */
137 return nf_conntrack_confirm(skb
);
140 static unsigned int ipv4_conntrack_in(unsigned int hooknum
,
142 const struct net_device
*in
,
143 const struct net_device
*out
,
144 int (*okfn
)(struct sk_buff
*))
146 return nf_conntrack_in(dev_net(in
), PF_INET
, hooknum
, skb
);
149 static unsigned int ipv4_conntrack_local(unsigned int hooknum
,
151 const struct net_device
*in
,
152 const struct net_device
*out
,
153 int (*okfn
)(struct sk_buff
*))
155 /* root is playing with raw sockets. */
156 if (skb
->len
< sizeof(struct iphdr
) ||
157 ip_hdrlen(skb
) < sizeof(struct iphdr
))
159 return nf_conntrack_in(dev_net(out
), PF_INET
, hooknum
, skb
);
162 /* Connection tracking may drop packets, but never alters them, so
163 make it the first hook. */
164 static struct nf_hook_ops ipv4_conntrack_ops
[] __read_mostly
= {
166 .hook
= ipv4_conntrack_in
,
167 .owner
= THIS_MODULE
,
169 .hooknum
= NF_INET_PRE_ROUTING
,
170 .priority
= NF_IP_PRI_CONNTRACK
,
173 .hook
= ipv4_conntrack_local
,
174 .owner
= THIS_MODULE
,
176 .hooknum
= NF_INET_LOCAL_OUT
,
177 .priority
= NF_IP_PRI_CONNTRACK
,
180 .hook
= ipv4_confirm
,
181 .owner
= THIS_MODULE
,
183 .hooknum
= NF_INET_POST_ROUTING
,
184 .priority
= NF_IP_PRI_CONNTRACK_CONFIRM
,
187 .hook
= ipv4_confirm
,
188 .owner
= THIS_MODULE
,
190 .hooknum
= NF_INET_LOCAL_IN
,
191 .priority
= NF_IP_PRI_CONNTRACK_CONFIRM
,
195 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
196 static int log_invalid_proto_min
= 0;
197 static int log_invalid_proto_max
= 255;
199 static ctl_table ip_ct_sysctl_table
[] = {
201 .procname
= "ip_conntrack_max",
202 .data
= &nf_conntrack_max
,
203 .maxlen
= sizeof(int),
205 .proc_handler
= proc_dointvec
,
208 .procname
= "ip_conntrack_count",
209 .data
= &init_net
.ct
.count
,
210 .maxlen
= sizeof(int),
212 .proc_handler
= proc_dointvec
,
215 .procname
= "ip_conntrack_buckets",
216 .data
= &init_net
.ct
.htable_size
,
217 .maxlen
= sizeof(unsigned int),
219 .proc_handler
= proc_dointvec
,
222 .procname
= "ip_conntrack_checksum",
223 .data
= &init_net
.ct
.sysctl_checksum
,
224 .maxlen
= sizeof(int),
226 .proc_handler
= proc_dointvec
,
229 .procname
= "ip_conntrack_log_invalid",
230 .data
= &init_net
.ct
.sysctl_log_invalid
,
231 .maxlen
= sizeof(unsigned int),
233 .proc_handler
= proc_dointvec_minmax
,
234 .extra1
= &log_invalid_proto_min
,
235 .extra2
= &log_invalid_proto_max
,
239 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
241 /* Fast function for those who don't want to parse /proc (and I don't
243 /* Reversing the socket's dst/src point of view gives us the reply
246 getorigdst(struct sock
*sk
, int optval
, void __user
*user
, int *len
)
248 const struct inet_sock
*inet
= inet_sk(sk
);
249 const struct nf_conntrack_tuple_hash
*h
;
250 struct nf_conntrack_tuple tuple
;
252 memset(&tuple
, 0, sizeof(tuple
));
253 tuple
.src
.u3
.ip
= inet
->inet_rcv_saddr
;
254 tuple
.src
.u
.tcp
.port
= inet
->inet_sport
;
255 tuple
.dst
.u3
.ip
= inet
->inet_daddr
;
256 tuple
.dst
.u
.tcp
.port
= inet
->inet_dport
;
257 tuple
.src
.l3num
= PF_INET
;
258 tuple
.dst
.protonum
= sk
->sk_protocol
;
260 /* We only do TCP and SCTP at the moment: is there a better way? */
261 if (sk
->sk_protocol
!= IPPROTO_TCP
&& sk
->sk_protocol
!= IPPROTO_SCTP
) {
262 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
266 if ((unsigned int) *len
< sizeof(struct sockaddr_in
)) {
267 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
268 *len
, sizeof(struct sockaddr_in
));
272 h
= nf_conntrack_find_get(sock_net(sk
), NF_CT_DEFAULT_ZONE
, &tuple
);
274 struct sockaddr_in sin
;
275 struct nf_conn
*ct
= nf_ct_tuplehash_to_ctrack(h
);
277 sin
.sin_family
= AF_INET
;
278 sin
.sin_port
= ct
->tuplehash
[IP_CT_DIR_ORIGINAL
]
279 .tuple
.dst
.u
.tcp
.port
;
280 sin
.sin_addr
.s_addr
= ct
->tuplehash
[IP_CT_DIR_ORIGINAL
]
282 memset(sin
.sin_zero
, 0, sizeof(sin
.sin_zero
));
284 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
285 &sin
.sin_addr
.s_addr
, ntohs(sin
.sin_port
));
287 if (copy_to_user(user
, &sin
, sizeof(sin
)) != 0)
292 pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
293 &tuple
.src
.u3
.ip
, ntohs(tuple
.src
.u
.tcp
.port
),
294 &tuple
.dst
.u3
.ip
, ntohs(tuple
.dst
.u
.tcp
.port
));
298 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
300 #include <linux/netfilter/nfnetlink.h>
301 #include <linux/netfilter/nfnetlink_conntrack.h>
303 static int ipv4_tuple_to_nlattr(struct sk_buff
*skb
,
304 const struct nf_conntrack_tuple
*tuple
)
306 NLA_PUT_BE32(skb
, CTA_IP_V4_SRC
, tuple
->src
.u3
.ip
);
307 NLA_PUT_BE32(skb
, CTA_IP_V4_DST
, tuple
->dst
.u3
.ip
);
314 static const struct nla_policy ipv4_nla_policy
[CTA_IP_MAX
+1] = {
315 [CTA_IP_V4_SRC
] = { .type
= NLA_U32
},
316 [CTA_IP_V4_DST
] = { .type
= NLA_U32
},
319 static int ipv4_nlattr_to_tuple(struct nlattr
*tb
[],
320 struct nf_conntrack_tuple
*t
)
322 if (!tb
[CTA_IP_V4_SRC
] || !tb
[CTA_IP_V4_DST
])
325 t
->src
.u3
.ip
= nla_get_be32(tb
[CTA_IP_V4_SRC
]);
326 t
->dst
.u3
.ip
= nla_get_be32(tb
[CTA_IP_V4_DST
]);
331 static int ipv4_nlattr_tuple_size(void)
333 return nla_policy_len(ipv4_nla_policy
, CTA_IP_MAX
+ 1);
337 static struct nf_sockopt_ops so_getorigdst
= {
339 .get_optmin
= SO_ORIGINAL_DST
,
340 .get_optmax
= SO_ORIGINAL_DST
+1,
342 .owner
= THIS_MODULE
,
345 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly
= {
348 .pkt_to_tuple
= ipv4_pkt_to_tuple
,
349 .invert_tuple
= ipv4_invert_tuple
,
350 .print_tuple
= ipv4_print_tuple
,
351 .get_l4proto
= ipv4_get_l4proto
,
352 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
353 .tuple_to_nlattr
= ipv4_tuple_to_nlattr
,
354 .nlattr_tuple_size
= ipv4_nlattr_tuple_size
,
355 .nlattr_to_tuple
= ipv4_nlattr_to_tuple
,
356 .nla_policy
= ipv4_nla_policy
,
358 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
359 .ctl_table_path
= nf_net_ipv4_netfilter_sysctl_path
,
360 .ctl_table
= ip_ct_sysctl_table
,
365 module_param_call(hashsize
, nf_conntrack_set_hashsize
, param_get_uint
,
366 &nf_conntrack_htable_size
, 0600);
368 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET
));
369 MODULE_ALIAS("ip_conntrack");
370 MODULE_LICENSE("GPL");
372 static int __init
nf_conntrack_l3proto_ipv4_init(void)
377 nf_defrag_ipv4_enable();
379 ret
= nf_register_sockopt(&so_getorigdst
);
381 printk(KERN_ERR
"Unable to register netfilter socket option\n");
385 ret
= nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4
);
387 pr_err("nf_conntrack_ipv4: can't register tcp.\n");
388 goto cleanup_sockopt
;
391 ret
= nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4
);
393 pr_err("nf_conntrack_ipv4: can't register udp.\n");
397 ret
= nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp
);
399 pr_err("nf_conntrack_ipv4: can't register icmp.\n");
403 ret
= nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4
);
405 pr_err("nf_conntrack_ipv4: can't register ipv4\n");
409 ret
= nf_register_hooks(ipv4_conntrack_ops
,
410 ARRAY_SIZE(ipv4_conntrack_ops
));
412 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
415 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
416 ret
= nf_conntrack_ipv4_compat_init();
421 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
423 nf_unregister_hooks(ipv4_conntrack_ops
, ARRAY_SIZE(ipv4_conntrack_ops
));
426 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4
);
428 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp
);
430 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4
);
432 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4
);
434 nf_unregister_sockopt(&so_getorigdst
);
438 static void __exit
nf_conntrack_l3proto_ipv4_fini(void)
441 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
442 nf_conntrack_ipv4_compat_fini();
444 nf_unregister_hooks(ipv4_conntrack_ops
, ARRAY_SIZE(ipv4_conntrack_ops
));
445 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4
);
446 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp
);
447 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4
);
448 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4
);
449 nf_unregister_sockopt(&so_getorigdst
);
452 module_init(nf_conntrack_l3proto_ipv4_init
);
453 module_exit(nf_conntrack_l3proto_ipv4_fini
);
455 void need_ipv4_conntrack(void)
459 EXPORT_SYMBOL_GPL(need_ipv4_conntrack
);