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 void ipv4_print_tuple(struct seq_file
*s
,
60 const struct nf_conntrack_tuple
*tuple
)
62 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 nf_hook_state
*state
)
100 enum ip_conntrack_info ctinfo
;
101 const struct nf_conn_help
*help
;
102 const struct nf_conntrack_helper
*helper
;
104 /* This is where we call the helper: as the packet goes out. */
105 ct
= nf_ct_get(skb
, &ctinfo
);
106 if (!ct
|| ctinfo
== IP_CT_RELATED_REPLY
)
109 help
= nfct_help(ct
);
113 /* rcu_read_lock()ed by nf_hook_slow */
114 helper
= rcu_dereference(help
->helper
);
118 return helper
->help(skb
, skb_network_offset(skb
) + ip_hdrlen(skb
),
122 static unsigned int ipv4_confirm(const struct nf_hook_ops
*ops
,
124 const struct nf_hook_state
*state
)
127 enum ip_conntrack_info ctinfo
;
129 ct
= nf_ct_get(skb
, &ctinfo
);
130 if (!ct
|| ctinfo
== IP_CT_RELATED_REPLY
)
133 /* adjust seqs for loopback traffic only in outgoing direction */
134 if (test_bit(IPS_SEQ_ADJUST_BIT
, &ct
->status
) &&
135 !nf_is_loopback_packet(skb
)) {
136 if (!nf_ct_seq_adjust(skb
, ct
, ctinfo
, ip_hdrlen(skb
))) {
137 NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct
), drop
);
142 /* We've seen it coming out the other side: confirm it */
143 return nf_conntrack_confirm(skb
);
146 static unsigned int ipv4_conntrack_in(const struct nf_hook_ops
*ops
,
148 const struct nf_hook_state
*state
)
150 return nf_conntrack_in(dev_net(state
->in
), PF_INET
, ops
->hooknum
, skb
);
153 static unsigned int ipv4_conntrack_local(const struct nf_hook_ops
*ops
,
155 const struct nf_hook_state
*state
)
157 /* root is playing with raw sockets. */
158 if (skb
->len
< sizeof(struct iphdr
) ||
159 ip_hdrlen(skb
) < sizeof(struct iphdr
))
161 return nf_conntrack_in(dev_net(state
->out
), PF_INET
, ops
->hooknum
, skb
);
164 /* Connection tracking may drop packets, but never alters them, so
165 make it the first hook. */
166 static struct nf_hook_ops ipv4_conntrack_ops
[] __read_mostly
= {
168 .hook
= ipv4_conntrack_in
,
169 .owner
= THIS_MODULE
,
171 .hooknum
= NF_INET_PRE_ROUTING
,
172 .priority
= NF_IP_PRI_CONNTRACK
,
175 .hook
= ipv4_conntrack_local
,
176 .owner
= THIS_MODULE
,
178 .hooknum
= NF_INET_LOCAL_OUT
,
179 .priority
= NF_IP_PRI_CONNTRACK
,
183 .owner
= THIS_MODULE
,
185 .hooknum
= NF_INET_POST_ROUTING
,
186 .priority
= NF_IP_PRI_CONNTRACK_HELPER
,
189 .hook
= ipv4_confirm
,
190 .owner
= THIS_MODULE
,
192 .hooknum
= NF_INET_POST_ROUTING
,
193 .priority
= NF_IP_PRI_CONNTRACK_CONFIRM
,
197 .owner
= THIS_MODULE
,
199 .hooknum
= NF_INET_LOCAL_IN
,
200 .priority
= NF_IP_PRI_CONNTRACK_HELPER
,
203 .hook
= ipv4_confirm
,
204 .owner
= THIS_MODULE
,
206 .hooknum
= NF_INET_LOCAL_IN
,
207 .priority
= NF_IP_PRI_CONNTRACK_CONFIRM
,
211 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
212 static int log_invalid_proto_min
= 0;
213 static int log_invalid_proto_max
= 255;
215 static struct ctl_table ip_ct_sysctl_table
[] = {
217 .procname
= "ip_conntrack_max",
218 .maxlen
= sizeof(int),
220 .proc_handler
= proc_dointvec
,
223 .procname
= "ip_conntrack_count",
224 .maxlen
= sizeof(int),
226 .proc_handler
= proc_dointvec
,
229 .procname
= "ip_conntrack_buckets",
230 .maxlen
= sizeof(unsigned int),
232 .proc_handler
= proc_dointvec
,
235 .procname
= "ip_conntrack_checksum",
236 .maxlen
= sizeof(int),
238 .proc_handler
= proc_dointvec
,
241 .procname
= "ip_conntrack_log_invalid",
242 .maxlen
= sizeof(unsigned int),
244 .proc_handler
= proc_dointvec_minmax
,
245 .extra1
= &log_invalid_proto_min
,
246 .extra2
= &log_invalid_proto_max
,
250 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
252 /* Fast function for those who don't want to parse /proc (and I don't
254 /* Reversing the socket's dst/src point of view gives us the reply
257 getorigdst(struct sock
*sk
, int optval
, void __user
*user
, int *len
)
259 const struct inet_sock
*inet
= inet_sk(sk
);
260 const struct nf_conntrack_tuple_hash
*h
;
261 struct nf_conntrack_tuple tuple
;
263 memset(&tuple
, 0, sizeof(tuple
));
264 tuple
.src
.u3
.ip
= inet
->inet_rcv_saddr
;
265 tuple
.src
.u
.tcp
.port
= inet
->inet_sport
;
266 tuple
.dst
.u3
.ip
= inet
->inet_daddr
;
267 tuple
.dst
.u
.tcp
.port
= inet
->inet_dport
;
268 tuple
.src
.l3num
= PF_INET
;
269 tuple
.dst
.protonum
= sk
->sk_protocol
;
271 /* We only do TCP and SCTP at the moment: is there a better way? */
272 if (sk
->sk_protocol
!= IPPROTO_TCP
&& sk
->sk_protocol
!= IPPROTO_SCTP
) {
273 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
277 if ((unsigned int) *len
< sizeof(struct sockaddr_in
)) {
278 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
279 *len
, sizeof(struct sockaddr_in
));
283 h
= nf_conntrack_find_get(sock_net(sk
), NF_CT_DEFAULT_ZONE
, &tuple
);
285 struct sockaddr_in sin
;
286 struct nf_conn
*ct
= nf_ct_tuplehash_to_ctrack(h
);
288 sin
.sin_family
= AF_INET
;
289 sin
.sin_port
= ct
->tuplehash
[IP_CT_DIR_ORIGINAL
]
290 .tuple
.dst
.u
.tcp
.port
;
291 sin
.sin_addr
.s_addr
= ct
->tuplehash
[IP_CT_DIR_ORIGINAL
]
293 memset(sin
.sin_zero
, 0, sizeof(sin
.sin_zero
));
295 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
296 &sin
.sin_addr
.s_addr
, ntohs(sin
.sin_port
));
298 if (copy_to_user(user
, &sin
, sizeof(sin
)) != 0)
303 pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
304 &tuple
.src
.u3
.ip
, ntohs(tuple
.src
.u
.tcp
.port
),
305 &tuple
.dst
.u3
.ip
, ntohs(tuple
.dst
.u
.tcp
.port
));
309 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
311 #include <linux/netfilter/nfnetlink.h>
312 #include <linux/netfilter/nfnetlink_conntrack.h>
314 static int ipv4_tuple_to_nlattr(struct sk_buff
*skb
,
315 const struct nf_conntrack_tuple
*tuple
)
317 if (nla_put_in_addr(skb
, CTA_IP_V4_SRC
, tuple
->src
.u3
.ip
) ||
318 nla_put_in_addr(skb
, CTA_IP_V4_DST
, tuple
->dst
.u3
.ip
))
319 goto nla_put_failure
;
326 static const struct nla_policy ipv4_nla_policy
[CTA_IP_MAX
+1] = {
327 [CTA_IP_V4_SRC
] = { .type
= NLA_U32
},
328 [CTA_IP_V4_DST
] = { .type
= NLA_U32
},
331 static int ipv4_nlattr_to_tuple(struct nlattr
*tb
[],
332 struct nf_conntrack_tuple
*t
)
334 if (!tb
[CTA_IP_V4_SRC
] || !tb
[CTA_IP_V4_DST
])
337 t
->src
.u3
.ip
= nla_get_in_addr(tb
[CTA_IP_V4_SRC
]);
338 t
->dst
.u3
.ip
= nla_get_in_addr(tb
[CTA_IP_V4_DST
]);
343 static int ipv4_nlattr_tuple_size(void)
345 return nla_policy_len(ipv4_nla_policy
, CTA_IP_MAX
+ 1);
349 static struct nf_sockopt_ops so_getorigdst
= {
351 .get_optmin
= SO_ORIGINAL_DST
,
352 .get_optmax
= SO_ORIGINAL_DST
+1,
354 .owner
= THIS_MODULE
,
357 static int ipv4_init_net(struct net
*net
)
359 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
360 struct nf_ip_net
*in
= &net
->ct
.nf_ct_proto
;
361 in
->ctl_table
= kmemdup(ip_ct_sysctl_table
,
362 sizeof(ip_ct_sysctl_table
),
367 in
->ctl_table
[0].data
= &nf_conntrack_max
;
368 in
->ctl_table
[1].data
= &net
->ct
.count
;
369 in
->ctl_table
[2].data
= &net
->ct
.htable_size
;
370 in
->ctl_table
[3].data
= &net
->ct
.sysctl_checksum
;
371 in
->ctl_table
[4].data
= &net
->ct
.sysctl_log_invalid
;
376 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly
= {
379 .pkt_to_tuple
= ipv4_pkt_to_tuple
,
380 .invert_tuple
= ipv4_invert_tuple
,
381 .print_tuple
= ipv4_print_tuple
,
382 .get_l4proto
= ipv4_get_l4proto
,
383 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
384 .tuple_to_nlattr
= ipv4_tuple_to_nlattr
,
385 .nlattr_tuple_size
= ipv4_nlattr_tuple_size
,
386 .nlattr_to_tuple
= ipv4_nlattr_to_tuple
,
387 .nla_policy
= ipv4_nla_policy
,
389 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
390 .ctl_table_path
= "net/ipv4/netfilter",
392 .init_net
= ipv4_init_net
,
396 module_param_call(hashsize
, nf_conntrack_set_hashsize
, param_get_uint
,
397 &nf_conntrack_htable_size
, 0600);
399 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET
));
400 MODULE_ALIAS("ip_conntrack");
401 MODULE_LICENSE("GPL");
403 static int ipv4_net_init(struct net
*net
)
407 ret
= nf_ct_l4proto_pernet_register(net
, &nf_conntrack_l4proto_tcp4
);
409 pr_err("nf_conntrack_tcp4: pernet registration failed\n");
412 ret
= nf_ct_l4proto_pernet_register(net
, &nf_conntrack_l4proto_udp4
);
414 pr_err("nf_conntrack_udp4: pernet registration failed\n");
417 ret
= nf_ct_l4proto_pernet_register(net
, &nf_conntrack_l4proto_icmp
);
419 pr_err("nf_conntrack_icmp4: pernet registration failed\n");
422 ret
= nf_ct_l3proto_pernet_register(net
, &nf_conntrack_l3proto_ipv4
);
424 pr_err("nf_conntrack_ipv4: pernet registration failed\n");
429 nf_ct_l4proto_pernet_unregister(net
, &nf_conntrack_l4proto_icmp
);
431 nf_ct_l4proto_pernet_unregister(net
, &nf_conntrack_l4proto_udp4
);
433 nf_ct_l4proto_pernet_unregister(net
, &nf_conntrack_l4proto_tcp4
);
438 static void ipv4_net_exit(struct net
*net
)
440 nf_ct_l3proto_pernet_unregister(net
, &nf_conntrack_l3proto_ipv4
);
441 nf_ct_l4proto_pernet_unregister(net
, &nf_conntrack_l4proto_icmp
);
442 nf_ct_l4proto_pernet_unregister(net
, &nf_conntrack_l4proto_udp4
);
443 nf_ct_l4proto_pernet_unregister(net
, &nf_conntrack_l4proto_tcp4
);
446 static struct pernet_operations ipv4_net_ops
= {
447 .init
= ipv4_net_init
,
448 .exit
= ipv4_net_exit
,
451 static int __init
nf_conntrack_l3proto_ipv4_init(void)
456 nf_defrag_ipv4_enable();
458 ret
= nf_register_sockopt(&so_getorigdst
);
460 printk(KERN_ERR
"Unable to register netfilter socket option\n");
464 ret
= register_pernet_subsys(&ipv4_net_ops
);
466 pr_err("nf_conntrack_ipv4: can't register pernet ops\n");
467 goto cleanup_sockopt
;
470 ret
= nf_register_hooks(ipv4_conntrack_ops
,
471 ARRAY_SIZE(ipv4_conntrack_ops
));
473 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
477 ret
= nf_ct_l4proto_register(&nf_conntrack_l4proto_tcp4
);
479 pr_err("nf_conntrack_ipv4: can't register tcp4 proto.\n");
483 ret
= nf_ct_l4proto_register(&nf_conntrack_l4proto_udp4
);
485 pr_err("nf_conntrack_ipv4: can't register udp4 proto.\n");
489 ret
= nf_ct_l4proto_register(&nf_conntrack_l4proto_icmp
);
491 pr_err("nf_conntrack_ipv4: can't register icmpv4 proto.\n");
495 ret
= nf_ct_l3proto_register(&nf_conntrack_l3proto_ipv4
);
497 pr_err("nf_conntrack_ipv4: can't register ipv4 proto.\n");
501 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
502 ret
= nf_conntrack_ipv4_compat_init();
507 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
509 nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4
);
512 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmp
);
514 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp4
);
516 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp4
);
518 nf_unregister_hooks(ipv4_conntrack_ops
, ARRAY_SIZE(ipv4_conntrack_ops
));
520 unregister_pernet_subsys(&ipv4_net_ops
);
522 nf_unregister_sockopt(&so_getorigdst
);
526 static void __exit
nf_conntrack_l3proto_ipv4_fini(void)
529 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
530 nf_conntrack_ipv4_compat_fini();
532 nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4
);
533 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmp
);
534 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp4
);
535 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp4
);
536 nf_unregister_hooks(ipv4_conntrack_ops
, ARRAY_SIZE(ipv4_conntrack_ops
));
537 unregister_pernet_subsys(&ipv4_net_ops
);
538 nf_unregister_sockopt(&so_getorigdst
);
541 module_init(nf_conntrack_l3proto_ipv4_init
);
542 module_exit(nf_conntrack_l3proto_ipv4_fini
);