2 * Transparent proxy support for Linux/iptables
4 * Copyright (C) 2007-2008 BalaBit IT Ltd.
5 * Author: Krisztian Kovacs
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 #include <linux/module.h>
14 #include <linux/skbuff.h>
15 #include <linux/netfilter/x_tables.h>
16 #include <linux/netfilter_ipv4/ip_tables.h>
21 #include <net/inet_sock.h>
22 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
24 #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
25 #define XT_SOCKET_HAVE_IPV6 1
26 #include <linux/netfilter_ipv6/ip6_tables.h>
27 #include <net/inet6_hashtables.h>
28 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
31 #include <linux/netfilter/xt_socket.h>
33 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
34 #define XT_SOCKET_HAVE_CONNTRACK 1
35 #include <net/netfilter/nf_conntrack.h>
39 xt_socket_put_sk(struct sock
*sk
)
41 if (sk
->sk_state
== TCP_TIME_WAIT
)
42 inet_twsk_put(inet_twsk(sk
));
48 extract_icmp4_fields(const struct sk_buff
*skb
,
55 unsigned int outside_hdrlen
= ip_hdrlen(skb
);
56 struct iphdr
*inside_iph
, _inside_iph
;
57 struct icmphdr
*icmph
, _icmph
;
58 __be16
*ports
, _ports
[2];
60 icmph
= skb_header_pointer(skb
, outside_hdrlen
,
61 sizeof(_icmph
), &_icmph
);
65 switch (icmph
->type
) {
66 case ICMP_DEST_UNREACH
:
67 case ICMP_SOURCE_QUENCH
:
69 case ICMP_TIME_EXCEEDED
:
70 case ICMP_PARAMETERPROB
:
76 inside_iph
= skb_header_pointer(skb
, outside_hdrlen
+
77 sizeof(struct icmphdr
),
78 sizeof(_inside_iph
), &_inside_iph
);
79 if (inside_iph
== NULL
)
82 if (inside_iph
->protocol
!= IPPROTO_TCP
&&
83 inside_iph
->protocol
!= IPPROTO_UDP
)
86 ports
= skb_header_pointer(skb
, outside_hdrlen
+
87 sizeof(struct icmphdr
) +
88 (inside_iph
->ihl
<< 2),
89 sizeof(_ports
), &_ports
);
93 /* the inside IP packet is the one quoted from our side, thus
94 * its saddr is the local address */
95 *protocol
= inside_iph
->protocol
;
96 *laddr
= inside_iph
->saddr
;
98 *raddr
= inside_iph
->daddr
;
104 /* "socket" match based redirection (no specific rule)
105 * ===================================================
107 * There are connections with dynamic endpoints (e.g. FTP data
108 * connection) that the user is unable to add explicit rules
109 * for. These are taken care of by a generic "socket" rule. It is
110 * assumed that the proxy application is trusted to open such
111 * connections without explicit iptables rule (except of course the
112 * generic 'socket' rule). In this case the following sockets are
113 * matched in preference order:
115 * - match: if there's a fully established connection matching the
118 * - match: if there's a non-zero bound listener (possibly with a
119 * non-local address) We don't accept zero-bound listeners, since
120 * then local services could intercept traffic going through the
124 xt_socket_get_sock_v4(struct net
*net
, const u8 protocol
,
125 const __be32 saddr
, const __be32 daddr
,
126 const __be16 sport
, const __be16 dport
,
127 const struct net_device
*in
)
131 return __inet_lookup(net
, &tcp_hashinfo
,
132 saddr
, sport
, daddr
, dport
,
135 return udp4_lib_lookup(net
, saddr
, sport
, daddr
, dport
,
142 socket_match(const struct sk_buff
*skb
, struct xt_action_param
*par
,
143 const struct xt_socket_mtinfo1
*info
)
145 const struct iphdr
*iph
= ip_hdr(skb
);
146 struct udphdr _hdr
, *hp
= NULL
;
147 struct sock
*sk
= skb
->sk
;
148 __be32
uninitialized_var(daddr
), uninitialized_var(saddr
);
149 __be16
uninitialized_var(dport
), uninitialized_var(sport
);
150 u8
uninitialized_var(protocol
);
151 #ifdef XT_SOCKET_HAVE_CONNTRACK
152 struct nf_conn
const *ct
;
153 enum ip_conntrack_info ctinfo
;
156 if (iph
->protocol
== IPPROTO_UDP
|| iph
->protocol
== IPPROTO_TCP
) {
157 hp
= skb_header_pointer(skb
, ip_hdrlen(skb
),
158 sizeof(_hdr
), &_hdr
);
162 protocol
= iph
->protocol
;
168 } else if (iph
->protocol
== IPPROTO_ICMP
) {
169 if (extract_icmp4_fields(skb
, &protocol
, &saddr
, &daddr
,
176 #ifdef XT_SOCKET_HAVE_CONNTRACK
177 /* Do the lookup with the original socket address in case this is a
178 * reply packet of an established SNAT-ted connection. */
180 ct
= nf_ct_get(skb
, &ctinfo
);
181 if (ct
&& !nf_ct_is_untracked(ct
) &&
182 ((iph
->protocol
!= IPPROTO_ICMP
&&
183 ctinfo
== IP_CT_ESTABLISHED_REPLY
) ||
184 (iph
->protocol
== IPPROTO_ICMP
&&
185 ctinfo
== IP_CT_RELATED_REPLY
)) &&
186 (ct
->status
& IPS_SRC_NAT_DONE
)) {
188 daddr
= ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
.src
.u3
.ip
;
189 dport
= (iph
->protocol
== IPPROTO_TCP
) ?
190 ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
.src
.u
.tcp
.port
:
191 ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
.src
.u
.udp
.port
;
196 sk
= xt_socket_get_sock_v4(dev_net(skb
->dev
), protocol
,
197 saddr
, daddr
, sport
, dport
,
201 bool transparent
= true;
203 /* Ignore sockets listening on INADDR_ANY,
204 * unless XT_SOCKET_NOWILDCARD is set
206 wildcard
= (!(info
->flags
& XT_SOCKET_NOWILDCARD
) &&
207 sk
->sk_state
!= TCP_TIME_WAIT
&&
208 inet_sk(sk
)->inet_rcv_saddr
== 0);
210 /* Ignore non-transparent sockets,
211 if XT_SOCKET_TRANSPARENT is used */
212 if (info
->flags
& XT_SOCKET_TRANSPARENT
)
213 transparent
= ((sk
->sk_state
!= TCP_TIME_WAIT
&&
214 inet_sk(sk
)->transparent
) ||
215 (sk
->sk_state
== TCP_TIME_WAIT
&&
216 inet_twsk(sk
)->tw_transparent
));
219 xt_socket_put_sk(sk
);
221 if (wildcard
|| !transparent
)
225 pr_debug("proto %hhu %pI4:%hu -> %pI4:%hu (orig %pI4:%hu) sock %p\n",
226 protocol
, &saddr
, ntohs(sport
),
227 &daddr
, ntohs(dport
),
228 &iph
->daddr
, hp
? ntohs(hp
->dest
) : 0, sk
);
234 socket_mt4_v0(const struct sk_buff
*skb
, struct xt_action_param
*par
)
236 static struct xt_socket_mtinfo1 xt_info_v0
= {
240 return socket_match(skb
, par
, &xt_info_v0
);
244 socket_mt4_v1_v2(const struct sk_buff
*skb
, struct xt_action_param
*par
)
246 return socket_match(skb
, par
, par
->matchinfo
);
249 #ifdef XT_SOCKET_HAVE_IPV6
252 extract_icmp6_fields(const struct sk_buff
*skb
,
253 unsigned int outside_hdrlen
,
255 const struct in6_addr
**raddr
,
256 const struct in6_addr
**laddr
,
259 struct ipv6hdr
*ipv6_var
)
261 const struct ipv6hdr
*inside_iph
;
262 struct icmp6hdr
*icmph
, _icmph
;
263 __be16
*ports
, _ports
[2];
265 __be16 inside_fragoff
;
268 icmph
= skb_header_pointer(skb
, outside_hdrlen
,
269 sizeof(_icmph
), &_icmph
);
273 if (icmph
->icmp6_type
& ICMPV6_INFOMSG_MASK
)
276 inside_iph
= skb_header_pointer(skb
, outside_hdrlen
+ sizeof(_icmph
),
277 sizeof(*ipv6_var
), ipv6_var
);
278 if (inside_iph
== NULL
)
280 inside_nexthdr
= inside_iph
->nexthdr
;
282 inside_hdrlen
= ipv6_skip_exthdr(skb
, outside_hdrlen
+ sizeof(_icmph
) +
284 &inside_nexthdr
, &inside_fragoff
);
285 if (inside_hdrlen
< 0)
286 return 1; /* hjm: Packet has no/incomplete transport layer headers. */
288 if (inside_nexthdr
!= IPPROTO_TCP
&&
289 inside_nexthdr
!= IPPROTO_UDP
)
292 ports
= skb_header_pointer(skb
, inside_hdrlen
,
293 sizeof(_ports
), &_ports
);
297 /* the inside IP packet is the one quoted from our side, thus
298 * its saddr is the local address */
299 *protocol
= inside_nexthdr
;
300 *laddr
= &inside_iph
->saddr
;
302 *raddr
= &inside_iph
->daddr
;
309 xt_socket_get_sock_v6(struct net
*net
, const u8 protocol
,
310 const struct in6_addr
*saddr
, const struct in6_addr
*daddr
,
311 const __be16 sport
, const __be16 dport
,
312 const struct net_device
*in
)
316 return inet6_lookup(net
, &tcp_hashinfo
,
317 saddr
, sport
, daddr
, dport
,
320 return udp6_lib_lookup(net
, saddr
, sport
, daddr
, dport
,
328 socket_mt6_v1_v2(const struct sk_buff
*skb
, struct xt_action_param
*par
)
330 struct ipv6hdr ipv6_var
, *iph
= ipv6_hdr(skb
);
331 struct udphdr _hdr
, *hp
= NULL
;
332 struct sock
*sk
= skb
->sk
;
333 const struct in6_addr
*daddr
= NULL
, *saddr
= NULL
;
334 __be16
uninitialized_var(dport
), uninitialized_var(sport
);
335 int thoff
= 0, uninitialized_var(tproto
);
336 const struct xt_socket_mtinfo1
*info
= (struct xt_socket_mtinfo1
*) par
->matchinfo
;
338 tproto
= ipv6_find_hdr(skb
, &thoff
, -1, NULL
, NULL
);
340 pr_debug("unable to find transport header in IPv6 packet, dropping\n");
344 if (tproto
== IPPROTO_UDP
|| tproto
== IPPROTO_TCP
) {
345 hp
= skb_header_pointer(skb
, thoff
,
346 sizeof(_hdr
), &_hdr
);
355 } else if (tproto
== IPPROTO_ICMPV6
) {
356 if (extract_icmp6_fields(skb
, thoff
, &tproto
, &saddr
, &daddr
,
357 &sport
, &dport
, &ipv6_var
))
364 sk
= xt_socket_get_sock_v6(dev_net(skb
->dev
), tproto
,
365 saddr
, daddr
, sport
, dport
,
369 bool transparent
= true;
371 /* Ignore sockets listening on INADDR_ANY
372 * unless XT_SOCKET_NOWILDCARD is set
374 wildcard
= (!(info
->flags
& XT_SOCKET_NOWILDCARD
) &&
375 sk
->sk_state
!= TCP_TIME_WAIT
&&
376 ipv6_addr_any(&inet6_sk(sk
)->rcv_saddr
));
378 /* Ignore non-transparent sockets,
379 if XT_SOCKET_TRANSPARENT is used */
380 if (info
->flags
& XT_SOCKET_TRANSPARENT
)
381 transparent
= ((sk
->sk_state
!= TCP_TIME_WAIT
&&
382 inet_sk(sk
)->transparent
) ||
383 (sk
->sk_state
== TCP_TIME_WAIT
&&
384 inet_twsk(sk
)->tw_transparent
));
387 xt_socket_put_sk(sk
);
389 if (wildcard
|| !transparent
)
393 pr_debug("proto %hhd %pI6:%hu -> %pI6:%hu "
394 "(orig %pI6:%hu) sock %p\n",
395 tproto
, saddr
, ntohs(sport
),
397 &iph
->daddr
, hp
? ntohs(hp
->dest
) : 0, sk
);
403 static int socket_mt_v1_check(const struct xt_mtchk_param
*par
)
405 const struct xt_socket_mtinfo1
*info
= (struct xt_socket_mtinfo1
*) par
->matchinfo
;
407 if (info
->flags
& ~XT_SOCKET_FLAGS_V1
) {
408 pr_info("unknown flags 0x%x\n", info
->flags
& ~XT_SOCKET_FLAGS_V1
);
414 static int socket_mt_v2_check(const struct xt_mtchk_param
*par
)
416 const struct xt_socket_mtinfo2
*info
= (struct xt_socket_mtinfo2
*) par
->matchinfo
;
418 if (info
->flags
& ~XT_SOCKET_FLAGS_V2
) {
419 pr_info("unknown flags 0x%x\n", info
->flags
& ~XT_SOCKET_FLAGS_V2
);
425 static struct xt_match socket_mt_reg
[] __read_mostly
= {
429 .family
= NFPROTO_IPV4
,
430 .match
= socket_mt4_v0
,
431 .hooks
= (1 << NF_INET_PRE_ROUTING
) |
432 (1 << NF_INET_LOCAL_IN
),
438 .family
= NFPROTO_IPV4
,
439 .match
= socket_mt4_v1_v2
,
440 .checkentry
= socket_mt_v1_check
,
441 .matchsize
= sizeof(struct xt_socket_mtinfo1
),
442 .hooks
= (1 << NF_INET_PRE_ROUTING
) |
443 (1 << NF_INET_LOCAL_IN
),
446 #ifdef XT_SOCKET_HAVE_IPV6
450 .family
= NFPROTO_IPV6
,
451 .match
= socket_mt6_v1_v2
,
452 .checkentry
= socket_mt_v1_check
,
453 .matchsize
= sizeof(struct xt_socket_mtinfo1
),
454 .hooks
= (1 << NF_INET_PRE_ROUTING
) |
455 (1 << NF_INET_LOCAL_IN
),
462 .family
= NFPROTO_IPV4
,
463 .match
= socket_mt4_v1_v2
,
464 .checkentry
= socket_mt_v2_check
,
465 .matchsize
= sizeof(struct xt_socket_mtinfo1
),
466 .hooks
= (1 << NF_INET_PRE_ROUTING
) |
467 (1 << NF_INET_LOCAL_IN
),
470 #ifdef XT_SOCKET_HAVE_IPV6
474 .family
= NFPROTO_IPV6
,
475 .match
= socket_mt6_v1_v2
,
476 .checkentry
= socket_mt_v2_check
,
477 .matchsize
= sizeof(struct xt_socket_mtinfo1
),
478 .hooks
= (1 << NF_INET_PRE_ROUTING
) |
479 (1 << NF_INET_LOCAL_IN
),
485 static int __init
socket_mt_init(void)
487 nf_defrag_ipv4_enable();
488 #ifdef XT_SOCKET_HAVE_IPV6
489 nf_defrag_ipv6_enable();
492 return xt_register_matches(socket_mt_reg
, ARRAY_SIZE(socket_mt_reg
));
495 static void __exit
socket_mt_exit(void)
497 xt_unregister_matches(socket_mt_reg
, ARRAY_SIZE(socket_mt_reg
));
500 module_init(socket_mt_init
);
501 module_exit(socket_mt_exit
);
503 MODULE_LICENSE("GPL");
504 MODULE_AUTHOR("Krisztian Kovacs, Balazs Scheidler");
505 MODULE_DESCRIPTION("x_tables socket match module");
506 MODULE_ALIAS("ipt_socket");
507 MODULE_ALIAS("ip6t_socket");