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 extract_icmp4_fields(const struct sk_buff *skb,
46 unsigned int outside_hdrlen = ip_hdrlen(skb);
47 struct iphdr *inside_iph, _inside_iph;
48 struct icmphdr *icmph, _icmph;
49 __be16 *ports, _ports[2];
51 icmph = skb_header_pointer(skb, outside_hdrlen,
52 sizeof(_icmph), &_icmph);
56 switch (icmph->type) {
57 case ICMP_DEST_UNREACH:
58 case ICMP_SOURCE_QUENCH:
60 case ICMP_TIME_EXCEEDED:
61 case ICMP_PARAMETERPROB:
67 inside_iph = skb_header_pointer(skb, outside_hdrlen +
68 sizeof(struct icmphdr),
69 sizeof(_inside_iph), &_inside_iph);
70 if (inside_iph == NULL)
73 if (inside_iph->protocol != IPPROTO_TCP &&
74 inside_iph->protocol != IPPROTO_UDP)
77 ports = skb_header_pointer(skb, outside_hdrlen +
78 sizeof(struct icmphdr) +
79 (inside_iph->ihl << 2),
80 sizeof(_ports), &_ports);
84 /* the inside IP packet is the one quoted from our side, thus
85 * its saddr is the local address */
86 *protocol = inside_iph->protocol;
87 *laddr = inside_iph->saddr;
89 *raddr = inside_iph->daddr;
95 /* "socket" match based redirection (no specific rule)
96 * ===================================================
98 * There are connections with dynamic endpoints (e.g. FTP data
99 * connection) that the user is unable to add explicit rules
100 * for. These are taken care of by a generic "socket" rule. It is
101 * assumed that the proxy application is trusted to open such
102 * connections without explicit iptables rule (except of course the
103 * generic 'socket' rule). In this case the following sockets are
104 * matched in preference order:
106 * - match: if there's a fully established connection matching the
109 * - match: if there's a non-zero bound listener (possibly with a
110 * non-local address) We don't accept zero-bound listeners, since
111 * then local services could intercept traffic going through the
115 xt_socket_get_sock_v4(struct net *net, struct sk_buff *skb, const int doff,
117 const __be32 saddr, const __be32 daddr,
118 const __be16 sport, const __be16 dport,
119 const struct net_device *in)
123 return inet_lookup(net, &tcp_hashinfo, skb, doff,
124 saddr, sport, daddr, dport,
127 return udp4_lib_lookup(net, saddr, sport, daddr, dport,
133 static bool xt_socket_sk_is_transparent(struct sock *sk)
135 switch (sk->sk_state) {
137 return inet_twsk(sk)->tw_transparent;
139 case TCP_NEW_SYN_RECV:
140 return inet_rsk(inet_reqsk(sk))->no_srccheck;
143 return inet_sk(sk)->transparent;
147 static struct sock *xt_socket_lookup_slow_v4(struct net *net,
148 const struct sk_buff *skb,
149 const struct net_device *indev)
151 const struct iphdr *iph = ip_hdr(skb);
152 struct sk_buff *data_skb = NULL;
154 __be32 uninitialized_var(daddr), uninitialized_var(saddr);
155 __be16 uninitialized_var(dport), uninitialized_var(sport);
156 u8 uninitialized_var(protocol);
157 #ifdef XT_SOCKET_HAVE_CONNTRACK
158 struct nf_conn const *ct;
159 enum ip_conntrack_info ctinfo;
162 if (iph->protocol == IPPROTO_UDP || iph->protocol == IPPROTO_TCP) {
163 struct udphdr _hdr, *hp;
165 hp = skb_header_pointer(skb, ip_hdrlen(skb),
166 sizeof(_hdr), &_hdr);
170 protocol = iph->protocol;
175 data_skb = (struct sk_buff *)skb;
176 doff = iph->protocol == IPPROTO_TCP ?
177 ip_hdrlen(skb) + __tcp_hdrlen((struct tcphdr *)hp) :
178 ip_hdrlen(skb) + sizeof(*hp);
180 } else if (iph->protocol == IPPROTO_ICMP) {
181 if (extract_icmp4_fields(skb, &protocol, &saddr, &daddr,
188 #ifdef XT_SOCKET_HAVE_CONNTRACK
189 /* Do the lookup with the original socket address in
190 * case this is a reply packet of an established
191 * SNAT-ted connection.
193 ct = nf_ct_get(skb, &ctinfo);
194 if (ct && !nf_ct_is_untracked(ct) &&
195 ((iph->protocol != IPPROTO_ICMP &&
196 ctinfo == IP_CT_ESTABLISHED_REPLY) ||
197 (iph->protocol == IPPROTO_ICMP &&
198 ctinfo == IP_CT_RELATED_REPLY)) &&
199 (ct->status & IPS_SRC_NAT_DONE)) {
201 daddr = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip;
202 dport = (iph->protocol == IPPROTO_TCP) ?
203 ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u.tcp.port :
204 ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u.udp.port;
208 return xt_socket_get_sock_v4(net, data_skb, doff, protocol, saddr,
209 daddr, sport, dport, indev);
213 socket_match(const struct sk_buff *skb, struct xt_action_param *par,
214 const struct xt_socket_mtinfo1 *info)
216 struct sk_buff *pskb = (struct sk_buff *)skb;
217 struct sock *sk = skb->sk;
220 sk = xt_socket_lookup_slow_v4(par->net, skb, par->in);
223 bool transparent = true;
225 /* Ignore sockets listening on INADDR_ANY,
226 * unless XT_SOCKET_NOWILDCARD is set
228 wildcard = (!(info->flags & XT_SOCKET_NOWILDCARD) &&
230 inet_sk(sk)->inet_rcv_saddr == 0);
232 /* Ignore non-transparent sockets,
233 * if XT_SOCKET_TRANSPARENT is used
235 if (info->flags & XT_SOCKET_TRANSPARENT)
236 transparent = xt_socket_sk_is_transparent(sk);
238 if (info->flags & XT_SOCKET_RESTORESKMARK && !wildcard &&
240 pskb->mark = sk->sk_mark;
245 if (wildcard || !transparent)
253 socket_mt4_v0(const struct sk_buff *skb, struct xt_action_param *par)
255 static struct xt_socket_mtinfo1 xt_info_v0 = {
259 return socket_match(skb, par, &xt_info_v0);
263 socket_mt4_v1_v2_v3(const struct sk_buff *skb, struct xt_action_param *par)
265 return socket_match(skb, par, par->matchinfo);
268 #ifdef XT_SOCKET_HAVE_IPV6
271 extract_icmp6_fields(const struct sk_buff *skb,
272 unsigned int outside_hdrlen,
274 const struct in6_addr **raddr,
275 const struct in6_addr **laddr,
278 struct ipv6hdr *ipv6_var)
280 const struct ipv6hdr *inside_iph;
281 struct icmp6hdr *icmph, _icmph;
282 __be16 *ports, _ports[2];
284 __be16 inside_fragoff;
287 icmph = skb_header_pointer(skb, outside_hdrlen,
288 sizeof(_icmph), &_icmph);
292 if (icmph->icmp6_type & ICMPV6_INFOMSG_MASK)
295 inside_iph = skb_header_pointer(skb, outside_hdrlen + sizeof(_icmph),
296 sizeof(*ipv6_var), ipv6_var);
297 if (inside_iph == NULL)
299 inside_nexthdr = inside_iph->nexthdr;
301 inside_hdrlen = ipv6_skip_exthdr(skb, outside_hdrlen + sizeof(_icmph) +
303 &inside_nexthdr, &inside_fragoff);
304 if (inside_hdrlen < 0)
305 return 1; /* hjm: Packet has no/incomplete transport layer headers. */
307 if (inside_nexthdr != IPPROTO_TCP &&
308 inside_nexthdr != IPPROTO_UDP)
311 ports = skb_header_pointer(skb, inside_hdrlen,
312 sizeof(_ports), &_ports);
316 /* the inside IP packet is the one quoted from our side, thus
317 * its saddr is the local address */
318 *protocol = inside_nexthdr;
319 *laddr = &inside_iph->saddr;
321 *raddr = &inside_iph->daddr;
328 xt_socket_get_sock_v6(struct net *net, struct sk_buff *skb, int doff,
330 const struct in6_addr *saddr, const struct in6_addr *daddr,
331 const __be16 sport, const __be16 dport,
332 const struct net_device *in)
336 return inet6_lookup(net, &tcp_hashinfo, skb, doff,
337 saddr, sport, daddr, dport,
340 return udp6_lib_lookup(net, saddr, sport, daddr, dport,
347 static struct sock *xt_socket_lookup_slow_v6(struct net *net,
348 const struct sk_buff *skb,
349 const struct net_device *indev)
351 __be16 uninitialized_var(dport), uninitialized_var(sport);
352 const struct in6_addr *daddr = NULL, *saddr = NULL;
353 struct ipv6hdr *iph = ipv6_hdr(skb);
354 struct sk_buff *data_skb = NULL;
356 int thoff = 0, tproto;
358 tproto = ipv6_find_hdr(skb, &thoff, -1, NULL, NULL);
360 pr_debug("unable to find transport header in IPv6 packet, dropping\n");
364 if (tproto == IPPROTO_UDP || tproto == IPPROTO_TCP) {
365 struct udphdr _hdr, *hp;
367 hp = skb_header_pointer(skb, thoff, sizeof(_hdr), &_hdr);
375 data_skb = (struct sk_buff *)skb;
376 doff = tproto == IPPROTO_TCP ?
377 thoff + __tcp_hdrlen((struct tcphdr *)hp) :
380 } else if (tproto == IPPROTO_ICMPV6) {
381 struct ipv6hdr ipv6_var;
383 if (extract_icmp6_fields(skb, thoff, &tproto, &saddr, &daddr,
384 &sport, &dport, &ipv6_var))
390 return xt_socket_get_sock_v6(net, data_skb, doff, tproto, saddr, daddr,
391 sport, dport, indev);
395 socket_mt6_v1_v2_v3(const struct sk_buff *skb, struct xt_action_param *par)
397 const struct xt_socket_mtinfo1 *info = (struct xt_socket_mtinfo1 *) par->matchinfo;
398 struct sk_buff *pskb = (struct sk_buff *)skb;
399 struct sock *sk = skb->sk;
402 sk = xt_socket_lookup_slow_v6(par->net, skb, par->in);
405 bool transparent = true;
407 /* Ignore sockets listening on INADDR_ANY
408 * unless XT_SOCKET_NOWILDCARD is set
410 wildcard = (!(info->flags & XT_SOCKET_NOWILDCARD) &&
412 ipv6_addr_any(&sk->sk_v6_rcv_saddr));
414 /* Ignore non-transparent sockets,
415 * if XT_SOCKET_TRANSPARENT is used
417 if (info->flags & XT_SOCKET_TRANSPARENT)
418 transparent = xt_socket_sk_is_transparent(sk);
420 if (info->flags & XT_SOCKET_RESTORESKMARK && !wildcard &&
422 pskb->mark = sk->sk_mark;
427 if (wildcard || !transparent)
435 static int socket_mt_v1_check(const struct xt_mtchk_param *par)
437 const struct xt_socket_mtinfo1 *info = (struct xt_socket_mtinfo1 *) par->matchinfo;
439 if (info->flags & ~XT_SOCKET_FLAGS_V1) {
440 pr_info("unknown flags 0x%x\n", info->flags & ~XT_SOCKET_FLAGS_V1);
446 static int socket_mt_v2_check(const struct xt_mtchk_param *par)
448 const struct xt_socket_mtinfo2 *info = (struct xt_socket_mtinfo2 *) par->matchinfo;
450 if (info->flags & ~XT_SOCKET_FLAGS_V2) {
451 pr_info("unknown flags 0x%x\n", info->flags & ~XT_SOCKET_FLAGS_V2);
457 static int socket_mt_v3_check(const struct xt_mtchk_param *par)
459 const struct xt_socket_mtinfo3 *info =
460 (struct xt_socket_mtinfo3 *)par->matchinfo;
462 if (info->flags & ~XT_SOCKET_FLAGS_V3) {
463 pr_info("unknown flags 0x%x\n",
464 info->flags & ~XT_SOCKET_FLAGS_V3);
470 static struct xt_match socket_mt_reg[] __read_mostly = {
474 .family = NFPROTO_IPV4,
475 .match = socket_mt4_v0,
476 .hooks = (1 << NF_INET_PRE_ROUTING) |
477 (1 << NF_INET_LOCAL_IN),
483 .family = NFPROTO_IPV4,
484 .match = socket_mt4_v1_v2_v3,
485 .checkentry = socket_mt_v1_check,
486 .matchsize = sizeof(struct xt_socket_mtinfo1),
487 .hooks = (1 << NF_INET_PRE_ROUTING) |
488 (1 << NF_INET_LOCAL_IN),
491 #ifdef XT_SOCKET_HAVE_IPV6
495 .family = NFPROTO_IPV6,
496 .match = socket_mt6_v1_v2_v3,
497 .checkentry = socket_mt_v1_check,
498 .matchsize = sizeof(struct xt_socket_mtinfo1),
499 .hooks = (1 << NF_INET_PRE_ROUTING) |
500 (1 << NF_INET_LOCAL_IN),
507 .family = NFPROTO_IPV4,
508 .match = socket_mt4_v1_v2_v3,
509 .checkentry = socket_mt_v2_check,
510 .matchsize = sizeof(struct xt_socket_mtinfo1),
511 .hooks = (1 << NF_INET_PRE_ROUTING) |
512 (1 << NF_INET_LOCAL_IN),
515 #ifdef XT_SOCKET_HAVE_IPV6
519 .family = NFPROTO_IPV6,
520 .match = socket_mt6_v1_v2_v3,
521 .checkentry = socket_mt_v2_check,
522 .matchsize = sizeof(struct xt_socket_mtinfo1),
523 .hooks = (1 << NF_INET_PRE_ROUTING) |
524 (1 << NF_INET_LOCAL_IN),
531 .family = NFPROTO_IPV4,
532 .match = socket_mt4_v1_v2_v3,
533 .checkentry = socket_mt_v3_check,
534 .matchsize = sizeof(struct xt_socket_mtinfo1),
535 .hooks = (1 << NF_INET_PRE_ROUTING) |
536 (1 << NF_INET_LOCAL_IN),
539 #ifdef XT_SOCKET_HAVE_IPV6
543 .family = NFPROTO_IPV6,
544 .match = socket_mt6_v1_v2_v3,
545 .checkentry = socket_mt_v3_check,
546 .matchsize = sizeof(struct xt_socket_mtinfo1),
547 .hooks = (1 << NF_INET_PRE_ROUTING) |
548 (1 << NF_INET_LOCAL_IN),
554 static int __init socket_mt_init(void)
556 nf_defrag_ipv4_enable();
557 #ifdef XT_SOCKET_HAVE_IPV6
558 nf_defrag_ipv6_enable();
561 return xt_register_matches(socket_mt_reg, ARRAY_SIZE(socket_mt_reg));
564 static void __exit socket_mt_exit(void)
566 xt_unregister_matches(socket_mt_reg, ARRAY_SIZE(socket_mt_reg));
569 module_init(socket_mt_init);
570 module_exit(socket_mt_exit);
572 MODULE_LICENSE("GPL");
573 MODULE_AUTHOR("Krisztian Kovacs, Balazs Scheidler");
574 MODULE_DESCRIPTION("x_tables socket match module");
575 MODULE_ALIAS("ipt_socket");
576 MODULE_ALIAS("ip6t_socket");