Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
1/*
2 * Transparent proxy support for Linux/iptables
3 *
4 * Copyright (C) 2007-2008 BalaBit IT Ltd.
5 * Author: Krisztian Kovacs
6 *
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.
10 *
11 */
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>
17#include <net/tcp.h>
18#include <net/udp.h>
19#include <net/icmp.h>
20#include <net/sock.h>
21#include <net/inet_sock.h>
22#include <net/netfilter/ipv4/nf_defrag_ipv4.h>
23
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>
29#endif
30
31#include <linux/netfilter/xt_socket.h>
32
33#if IS_ENABLED(CONFIG_NF_CONNTRACK)
34#define XT_SOCKET_HAVE_CONNTRACK 1
35#include <net/netfilter/nf_conntrack.h>
36#endif
37
38static int
39extract_icmp4_fields(const struct sk_buff *skb,
40 u8 *protocol,
41 __be32 *raddr,
42 __be32 *laddr,
43 __be16 *rport,
44 __be16 *lport)
45{
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];
50
51 icmph = skb_header_pointer(skb, outside_hdrlen,
52 sizeof(_icmph), &_icmph);
53 if (icmph == NULL)
54 return 1;
55
56 switch (icmph->type) {
57 case ICMP_DEST_UNREACH:
58 case ICMP_SOURCE_QUENCH:
59 case ICMP_REDIRECT:
60 case ICMP_TIME_EXCEEDED:
61 case ICMP_PARAMETERPROB:
62 break;
63 default:
64 return 1;
65 }
66
67 inside_iph = skb_header_pointer(skb, outside_hdrlen +
68 sizeof(struct icmphdr),
69 sizeof(_inside_iph), &_inside_iph);
70 if (inside_iph == NULL)
71 return 1;
72
73 if (inside_iph->protocol != IPPROTO_TCP &&
74 inside_iph->protocol != IPPROTO_UDP)
75 return 1;
76
77 ports = skb_header_pointer(skb, outside_hdrlen +
78 sizeof(struct icmphdr) +
79 (inside_iph->ihl << 2),
80 sizeof(_ports), &_ports);
81 if (ports == NULL)
82 return 1;
83
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;
88 *lport = ports[0];
89 *raddr = inside_iph->daddr;
90 *rport = ports[1];
91
92 return 0;
93}
94
95/* "socket" match based redirection (no specific rule)
96 * ===================================================
97 *
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:
105 *
106 * - match: if there's a fully established connection matching the
107 * _packet_ tuple
108 *
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
112 * box.
113 */
114static struct sock *
115xt_socket_get_sock_v4(struct net *net, const u8 protocol,
116 const __be32 saddr, const __be32 daddr,
117 const __be16 sport, const __be16 dport,
118 const struct net_device *in)
119{
120 switch (protocol) {
121 case IPPROTO_TCP:
122 return __inet_lookup(net, &tcp_hashinfo,
123 saddr, sport, daddr, dport,
124 in->ifindex);
125 case IPPROTO_UDP:
126 return udp4_lib_lookup(net, saddr, sport, daddr, dport,
127 in->ifindex);
128 }
129 return NULL;
130}
131
132static bool
133socket_match(const struct sk_buff *skb, struct xt_action_param *par,
134 const struct xt_socket_mtinfo1 *info)
135{
136 const struct iphdr *iph = ip_hdr(skb);
137 struct udphdr _hdr, *hp = NULL;
138 struct sock *sk = skb->sk;
139 __be32 uninitialized_var(daddr), uninitialized_var(saddr);
140 __be16 uninitialized_var(dport), uninitialized_var(sport);
141 u8 uninitialized_var(protocol);
142#ifdef XT_SOCKET_HAVE_CONNTRACK
143 struct nf_conn const *ct;
144 enum ip_conntrack_info ctinfo;
145#endif
146
147 if (iph->protocol == IPPROTO_UDP || iph->protocol == IPPROTO_TCP) {
148 hp = skb_header_pointer(skb, ip_hdrlen(skb),
149 sizeof(_hdr), &_hdr);
150 if (hp == NULL)
151 return false;
152
153 protocol = iph->protocol;
154 saddr = iph->saddr;
155 sport = hp->source;
156 daddr = iph->daddr;
157 dport = hp->dest;
158
159 } else if (iph->protocol == IPPROTO_ICMP) {
160 if (extract_icmp4_fields(skb, &protocol, &saddr, &daddr,
161 &sport, &dport))
162 return false;
163 } else {
164 return false;
165 }
166
167#ifdef XT_SOCKET_HAVE_CONNTRACK
168 /* Do the lookup with the original socket address in case this is a
169 * reply packet of an established SNAT-ted connection. */
170
171 ct = nf_ct_get(skb, &ctinfo);
172 if (ct && !nf_ct_is_untracked(ct) &&
173 ((iph->protocol != IPPROTO_ICMP &&
174 ctinfo == IP_CT_ESTABLISHED_REPLY) ||
175 (iph->protocol == IPPROTO_ICMP &&
176 ctinfo == IP_CT_RELATED_REPLY)) &&
177 (ct->status & IPS_SRC_NAT_DONE)) {
178
179 daddr = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip;
180 dport = (iph->protocol == IPPROTO_TCP) ?
181 ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u.tcp.port :
182 ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u.udp.port;
183 }
184#endif
185
186 if (!sk)
187 sk = xt_socket_get_sock_v4(dev_net(skb->dev), protocol,
188 saddr, daddr, sport, dport,
189 par->in);
190 if (sk) {
191 bool wildcard;
192 bool transparent = true;
193
194 /* Ignore sockets listening on INADDR_ANY,
195 * unless XT_SOCKET_NOWILDCARD is set
196 */
197 wildcard = (!(info->flags & XT_SOCKET_NOWILDCARD) &&
198 sk->sk_state != TCP_TIME_WAIT &&
199 inet_sk(sk)->inet_rcv_saddr == 0);
200
201 /* Ignore non-transparent sockets,
202 if XT_SOCKET_TRANSPARENT is used */
203 if (info->flags & XT_SOCKET_TRANSPARENT)
204 transparent = ((sk->sk_state != TCP_TIME_WAIT &&
205 inet_sk(sk)->transparent) ||
206 (sk->sk_state == TCP_TIME_WAIT &&
207 inet_twsk(sk)->tw_transparent));
208
209 if (sk != skb->sk)
210 sock_gen_put(sk);
211
212 if (wildcard || !transparent)
213 sk = NULL;
214 }
215
216 pr_debug("proto %hhu %pI4:%hu -> %pI4:%hu (orig %pI4:%hu) sock %p\n",
217 protocol, &saddr, ntohs(sport),
218 &daddr, ntohs(dport),
219 &iph->daddr, hp ? ntohs(hp->dest) : 0, sk);
220
221 return (sk != NULL);
222}
223
224static bool
225socket_mt4_v0(const struct sk_buff *skb, struct xt_action_param *par)
226{
227 static struct xt_socket_mtinfo1 xt_info_v0 = {
228 .flags = 0,
229 };
230
231 return socket_match(skb, par, &xt_info_v0);
232}
233
234static bool
235socket_mt4_v1_v2(const struct sk_buff *skb, struct xt_action_param *par)
236{
237 return socket_match(skb, par, par->matchinfo);
238}
239
240#ifdef XT_SOCKET_HAVE_IPV6
241
242static int
243extract_icmp6_fields(const struct sk_buff *skb,
244 unsigned int outside_hdrlen,
245 int *protocol,
246 const struct in6_addr **raddr,
247 const struct in6_addr **laddr,
248 __be16 *rport,
249 __be16 *lport,
250 struct ipv6hdr *ipv6_var)
251{
252 const struct ipv6hdr *inside_iph;
253 struct icmp6hdr *icmph, _icmph;
254 __be16 *ports, _ports[2];
255 u8 inside_nexthdr;
256 __be16 inside_fragoff;
257 int inside_hdrlen;
258
259 icmph = skb_header_pointer(skb, outside_hdrlen,
260 sizeof(_icmph), &_icmph);
261 if (icmph == NULL)
262 return 1;
263
264 if (icmph->icmp6_type & ICMPV6_INFOMSG_MASK)
265 return 1;
266
267 inside_iph = skb_header_pointer(skb, outside_hdrlen + sizeof(_icmph),
268 sizeof(*ipv6_var), ipv6_var);
269 if (inside_iph == NULL)
270 return 1;
271 inside_nexthdr = inside_iph->nexthdr;
272
273 inside_hdrlen = ipv6_skip_exthdr(skb, outside_hdrlen + sizeof(_icmph) +
274 sizeof(*ipv6_var),
275 &inside_nexthdr, &inside_fragoff);
276 if (inside_hdrlen < 0)
277 return 1; /* hjm: Packet has no/incomplete transport layer headers. */
278
279 if (inside_nexthdr != IPPROTO_TCP &&
280 inside_nexthdr != IPPROTO_UDP)
281 return 1;
282
283 ports = skb_header_pointer(skb, inside_hdrlen,
284 sizeof(_ports), &_ports);
285 if (ports == NULL)
286 return 1;
287
288 /* the inside IP packet is the one quoted from our side, thus
289 * its saddr is the local address */
290 *protocol = inside_nexthdr;
291 *laddr = &inside_iph->saddr;
292 *lport = ports[0];
293 *raddr = &inside_iph->daddr;
294 *rport = ports[1];
295
296 return 0;
297}
298
299static struct sock *
300xt_socket_get_sock_v6(struct net *net, const u8 protocol,
301 const struct in6_addr *saddr, const struct in6_addr *daddr,
302 const __be16 sport, const __be16 dport,
303 const struct net_device *in)
304{
305 switch (protocol) {
306 case IPPROTO_TCP:
307 return inet6_lookup(net, &tcp_hashinfo,
308 saddr, sport, daddr, dport,
309 in->ifindex);
310 case IPPROTO_UDP:
311 return udp6_lib_lookup(net, saddr, sport, daddr, dport,
312 in->ifindex);
313 }
314
315 return NULL;
316}
317
318static bool
319socket_mt6_v1_v2(const struct sk_buff *skb, struct xt_action_param *par)
320{
321 struct ipv6hdr ipv6_var, *iph = ipv6_hdr(skb);
322 struct udphdr _hdr, *hp = NULL;
323 struct sock *sk = skb->sk;
324 const struct in6_addr *daddr = NULL, *saddr = NULL;
325 __be16 uninitialized_var(dport), uninitialized_var(sport);
326 int thoff = 0, uninitialized_var(tproto);
327 const struct xt_socket_mtinfo1 *info = (struct xt_socket_mtinfo1 *) par->matchinfo;
328
329 tproto = ipv6_find_hdr(skb, &thoff, -1, NULL, NULL);
330 if (tproto < 0) {
331 pr_debug("unable to find transport header in IPv6 packet, dropping\n");
332 return NF_DROP;
333 }
334
335 if (tproto == IPPROTO_UDP || tproto == IPPROTO_TCP) {
336 hp = skb_header_pointer(skb, thoff,
337 sizeof(_hdr), &_hdr);
338 if (hp == NULL)
339 return false;
340
341 saddr = &iph->saddr;
342 sport = hp->source;
343 daddr = &iph->daddr;
344 dport = hp->dest;
345
346 } else if (tproto == IPPROTO_ICMPV6) {
347 if (extract_icmp6_fields(skb, thoff, &tproto, &saddr, &daddr,
348 &sport, &dport, &ipv6_var))
349 return false;
350 } else {
351 return false;
352 }
353
354 if (!sk)
355 sk = xt_socket_get_sock_v6(dev_net(skb->dev), tproto,
356 saddr, daddr, sport, dport,
357 par->in);
358 if (sk) {
359 bool wildcard;
360 bool transparent = true;
361
362 /* Ignore sockets listening on INADDR_ANY
363 * unless XT_SOCKET_NOWILDCARD is set
364 */
365 wildcard = (!(info->flags & XT_SOCKET_NOWILDCARD) &&
366 sk->sk_state != TCP_TIME_WAIT &&
367 ipv6_addr_any(&sk->sk_v6_rcv_saddr));
368
369 /* Ignore non-transparent sockets,
370 if XT_SOCKET_TRANSPARENT is used */
371 if (info->flags & XT_SOCKET_TRANSPARENT)
372 transparent = ((sk->sk_state != TCP_TIME_WAIT &&
373 inet_sk(sk)->transparent) ||
374 (sk->sk_state == TCP_TIME_WAIT &&
375 inet_twsk(sk)->tw_transparent));
376
377 if (sk != skb->sk)
378 sock_gen_put(sk);
379
380 if (wildcard || !transparent)
381 sk = NULL;
382 }
383
384 pr_debug("proto %hhd %pI6:%hu -> %pI6:%hu "
385 "(orig %pI6:%hu) sock %p\n",
386 tproto, saddr, ntohs(sport),
387 daddr, ntohs(dport),
388 &iph->daddr, hp ? ntohs(hp->dest) : 0, sk);
389
390 return (sk != NULL);
391}
392#endif
393
394static int socket_mt_v1_check(const struct xt_mtchk_param *par)
395{
396 const struct xt_socket_mtinfo1 *info = (struct xt_socket_mtinfo1 *) par->matchinfo;
397
398 if (info->flags & ~XT_SOCKET_FLAGS_V1) {
399 pr_info("unknown flags 0x%x\n", info->flags & ~XT_SOCKET_FLAGS_V1);
400 return -EINVAL;
401 }
402 return 0;
403}
404
405static int socket_mt_v2_check(const struct xt_mtchk_param *par)
406{
407 const struct xt_socket_mtinfo2 *info = (struct xt_socket_mtinfo2 *) par->matchinfo;
408
409 if (info->flags & ~XT_SOCKET_FLAGS_V2) {
410 pr_info("unknown flags 0x%x\n", info->flags & ~XT_SOCKET_FLAGS_V2);
411 return -EINVAL;
412 }
413 return 0;
414}
415
416static struct xt_match socket_mt_reg[] __read_mostly = {
417 {
418 .name = "socket",
419 .revision = 0,
420 .family = NFPROTO_IPV4,
421 .match = socket_mt4_v0,
422 .hooks = (1 << NF_INET_PRE_ROUTING) |
423 (1 << NF_INET_LOCAL_IN),
424 .me = THIS_MODULE,
425 },
426 {
427 .name = "socket",
428 .revision = 1,
429 .family = NFPROTO_IPV4,
430 .match = socket_mt4_v1_v2,
431 .checkentry = socket_mt_v1_check,
432 .matchsize = sizeof(struct xt_socket_mtinfo1),
433 .hooks = (1 << NF_INET_PRE_ROUTING) |
434 (1 << NF_INET_LOCAL_IN),
435 .me = THIS_MODULE,
436 },
437#ifdef XT_SOCKET_HAVE_IPV6
438 {
439 .name = "socket",
440 .revision = 1,
441 .family = NFPROTO_IPV6,
442 .match = socket_mt6_v1_v2,
443 .checkentry = socket_mt_v1_check,
444 .matchsize = sizeof(struct xt_socket_mtinfo1),
445 .hooks = (1 << NF_INET_PRE_ROUTING) |
446 (1 << NF_INET_LOCAL_IN),
447 .me = THIS_MODULE,
448 },
449#endif
450 {
451 .name = "socket",
452 .revision = 2,
453 .family = NFPROTO_IPV4,
454 .match = socket_mt4_v1_v2,
455 .checkentry = socket_mt_v2_check,
456 .matchsize = sizeof(struct xt_socket_mtinfo1),
457 .hooks = (1 << NF_INET_PRE_ROUTING) |
458 (1 << NF_INET_LOCAL_IN),
459 .me = THIS_MODULE,
460 },
461#ifdef XT_SOCKET_HAVE_IPV6
462 {
463 .name = "socket",
464 .revision = 2,
465 .family = NFPROTO_IPV6,
466 .match = socket_mt6_v1_v2,
467 .checkentry = socket_mt_v2_check,
468 .matchsize = sizeof(struct xt_socket_mtinfo1),
469 .hooks = (1 << NF_INET_PRE_ROUTING) |
470 (1 << NF_INET_LOCAL_IN),
471 .me = THIS_MODULE,
472 },
473#endif
474};
475
476static int __init socket_mt_init(void)
477{
478 nf_defrag_ipv4_enable();
479#ifdef XT_SOCKET_HAVE_IPV6
480 nf_defrag_ipv6_enable();
481#endif
482
483 return xt_register_matches(socket_mt_reg, ARRAY_SIZE(socket_mt_reg));
484}
485
486static void __exit socket_mt_exit(void)
487{
488 xt_unregister_matches(socket_mt_reg, ARRAY_SIZE(socket_mt_reg));
489}
490
491module_init(socket_mt_init);
492module_exit(socket_mt_exit);
493
494MODULE_LICENSE("GPL");
495MODULE_AUTHOR("Krisztian Kovacs, Balazs Scheidler");
496MODULE_DESCRIPTION("x_tables socket match module");
497MODULE_ALIAS("ipt_socket");
498MODULE_ALIAS("ip6t_socket");