1
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>
5 *
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.
9 */
10
11 #include <linux/types.h>
12 #include <linux/ip.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>
19 #include <net/ip.h>
20
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>
33
34 static int conntrack4_net_id __read_mostly;
35 static DEFINE_MUTEX(register_ipv4_hooks);
36
37 struct conntrack4_net {
38 unsigned int users;
39 };
40
41 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
42 struct nf_conntrack_tuple *tuple)
43 {
44 const __be32 *ap;
45 __be32 _addrs[2];
46 ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
47 sizeof(u_int32_t) * 2, _addrs);
48 if (ap == NULL)
49 return false;
50
51 tuple->src.u3.ip = ap[0];
52 tuple->dst.u3.ip = ap[1];
53
54 return true;
55 }
56
57 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
58 const struct nf_conntrack_tuple *orig)
59 {
60 tuple->src.u3.ip = orig->dst.u3.ip;
61 tuple->dst.u3.ip = orig->src.u3.ip;
62
63 return true;
64 }
65
66 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
67 unsigned int *dataoff, u_int8_t *protonum)
68 {
69 const struct iphdr *iph;
70 struct iphdr _iph;
71
72 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
73 if (iph == NULL)
74 return -NF_ACCEPT;
75
76 /* Conntrack defragments packets, we might still see fragments
77 * inside ICMP packets though. */
78 if (iph->frag_off & htons(IP_OFFSET))
79 return -NF_ACCEPT;
80
81 *dataoff = nhoff + (iph->ihl << 2);
82 *protonum = iph->protocol;
83
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);
89 return -NF_ACCEPT;
90 }
91
92 return NF_ACCEPT;
93 }
94
95 static unsigned int ipv4_helper(void *priv,
96 struct sk_buff *skb,
97 const struct nf_hook_state *state)
98 {
99 struct nf_conn *ct;
100 enum ip_conntrack_info ctinfo;
101 const struct nf_conn_help *help;
102 const struct nf_conntrack_helper *helper;
103
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)
107 return NF_ACCEPT;
108
109 help = nfct_help(ct);
110 if (!help)
111 return NF_ACCEPT;
112
113 /* rcu_read_lock()ed by nf_hook_thresh */
114 helper = rcu_dereference(help->helper);
115 if (!helper)
116 return NF_ACCEPT;
117
118 return helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
119 ct, ctinfo);
120 }
121
122 static unsigned int ipv4_confirm(void *priv,
123 struct sk_buff *skb,
124 const struct nf_hook_state *state)
125 {
126 struct nf_conn *ct;
127 enum ip_conntrack_info ctinfo;
128
129 ct = nf_ct_get(skb, &ctinfo);
130 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
131 goto out;
132
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);
138 return NF_DROP;
139 }
140 }
141 out:
142 /* We've seen it coming out the other side: confirm it */
143 return nf_conntrack_confirm(skb);
144 }
145
146 static unsigned int ipv4_conntrack_in(void *priv,
147 struct sk_buff *skb,
148 const struct nf_hook_state *state)
149 {
150 return nf_conntrack_in(state->net, PF_INET, state->hook, skb);
151 }
152
153 static unsigned int ipv4_conntrack_local(void *priv,
154 struct sk_buff *skb,
155 const struct nf_hook_state *state)
156 {
157 if (ip_is_fragment(ip_hdr(skb))) { /* IP_NODEFRAG setsockopt set */
158 enum ip_conntrack_info ctinfo;
159 struct nf_conn *tmpl;
160
161 tmpl = nf_ct_get(skb, &ctinfo);
162 if (tmpl && nf_ct_is_template(tmpl)) {
163 /* when skipping ct, clear templates to avoid fooling
164 * later targets/matches
165 */
166 skb->_nfct = 0;
167 nf_ct_put(tmpl);
168 }
169 return NF_ACCEPT;
170 }
171
172 return nf_conntrack_in(state->net, PF_INET, state->hook, skb);
173 }
174
175 /* Connection tracking may drop packets, but never alters them, so
176 make it the first hook. */
177 static const struct nf_hook_ops ipv4_conntrack_ops[] = {
178 {
179 .hook = ipv4_conntrack_in,
180 .pf = NFPROTO_IPV4,
181 .hooknum = NF_INET_PRE_ROUTING,
182 .priority = NF_IP_PRI_CONNTRACK,
183 },
184 {
185 .hook = ipv4_conntrack_local,
186 .pf = NFPROTO_IPV4,
187 .hooknum = NF_INET_LOCAL_OUT,
188 .priority = NF_IP_PRI_CONNTRACK,
189 },
190 {
191 .hook = ipv4_helper,
192 .pf = NFPROTO_IPV4,
193 .hooknum = NF_INET_POST_ROUTING,
194 .priority = NF_IP_PRI_CONNTRACK_HELPER,
195 },
196 {
197 .hook = ipv4_confirm,
198 .pf = NFPROTO_IPV4,
199 .hooknum = NF_INET_POST_ROUTING,
200 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
201 },
202 {
203 .hook = ipv4_helper,
204 .pf = NFPROTO_IPV4,
205 .hooknum = NF_INET_LOCAL_IN,
206 .priority = NF_IP_PRI_CONNTRACK_HELPER,
207 },
208 {
209 .hook = ipv4_confirm,
210 .pf = NFPROTO_IPV4,
211 .hooknum = NF_INET_LOCAL_IN,
212 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
213 },
214 };
215
216 /* Fast function for those who don't want to parse /proc (and I don't
217 blame them). */
218 /* Reversing the socket's dst/src point of view gives us the reply
219 mapping. */
220 static int
221 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
222 {
223 const struct inet_sock *inet = inet_sk(sk);
224 const struct nf_conntrack_tuple_hash *h;
225 struct nf_conntrack_tuple tuple;
226
227 memset(&tuple, 0, sizeof(tuple));
228
229 lock_sock(sk);
230 tuple.src.u3.ip = inet->inet_rcv_saddr;
231 tuple.src.u.tcp.port = inet->inet_sport;
232 tuple.dst.u3.ip = inet->inet_daddr;
233 tuple.dst.u.tcp.port = inet->inet_dport;
234 tuple.src.l3num = PF_INET;
235 tuple.dst.protonum = sk->sk_protocol;
236 release_sock(sk);
237
238 /* We only do TCP and SCTP at the moment: is there a better way? */
239 if (tuple.dst.protonum != IPPROTO_TCP &&
240 tuple.dst.protonum != IPPROTO_SCTP) {
241 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
242 return -ENOPROTOOPT;
243 }
244
245 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
246 pr_debug("SO_ORIGINAL_DST: len %d not %zu\n",
247 *len, sizeof(struct sockaddr_in));
248 return -EINVAL;
249 }
250
251 h = nf_conntrack_find_get(sock_net(sk), &nf_ct_zone_dflt, &tuple);
252 if (h) {
253 struct sockaddr_in sin;
254 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
255
256 sin.sin_family = AF_INET;
257 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
258 .tuple.dst.u.tcp.port;
259 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
260 .tuple.dst.u3.ip;
261 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
262
263 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
264 &sin.sin_addr.s_addr, ntohs(sin.sin_port));
265 nf_ct_put(ct);
266 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
267 return -EFAULT;
268 else
269 return 0;
270 }
271 pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
272 &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
273 &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
274 return -ENOENT;
275 }
276
277 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
278
279 #include <linux/netfilter/nfnetlink.h>
280 #include <linux/netfilter/nfnetlink_conntrack.h>
281
282 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
283 const struct nf_conntrack_tuple *tuple)
284 {
285 if (nla_put_in_addr(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) ||
286 nla_put_in_addr(skb, CTA_IP_V4_DST, tuple->dst.u3.ip))
287 goto nla_put_failure;
288 return 0;
289
290 nla_put_failure:
291 return -1;
292 }
293
294 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
295 [CTA_IP_V4_SRC] = { .type = NLA_U32 },
296 [CTA_IP_V4_DST] = { .type = NLA_U32 },
297 };
298
299 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
300 struct nf_conntrack_tuple *t)
301 {
302 if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
303 return -EINVAL;
304
305 t->src.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_SRC]);
306 t->dst.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_DST]);
307
308 return 0;
309 }
310 #endif
311
312 static struct nf_sockopt_ops so_getorigdst = {
313 .pf = PF_INET,
314 .get_optmin = SO_ORIGINAL_DST,
315 .get_optmax = SO_ORIGINAL_DST+1,
316 .get = getorigdst,
317 .owner = THIS_MODULE,
318 };
319
320 static int ipv4_hooks_register(struct net *net)
321 {
322 struct conntrack4_net *cnet = net_generic(net, conntrack4_net_id);
323 int err = 0;
324
325 mutex_lock(®ister_ipv4_hooks);
326
327 cnet->users++;
328 if (cnet->users > 1)
329 goto out_unlock;
330
331 err = nf_defrag_ipv4_enable(net);
332 if (err) {
333 cnet->users = 0;
334 goto out_unlock;
335 }
336
337 err = nf_register_net_hooks(net, ipv4_conntrack_ops,
338 ARRAY_SIZE(ipv4_conntrack_ops));
339
340 if (err)
341 cnet->users = 0;
342 out_unlock:
343 mutex_unlock(®ister_ipv4_hooks);
344 return err;
345 }
346
347 static void ipv4_hooks_unregister(struct net *net)
348 {
349 struct conntrack4_net *cnet = net_generic(net, conntrack4_net_id);
350
351 mutex_lock(®ister_ipv4_hooks);
352 if (cnet->users && (--cnet->users == 0))
353 nf_unregister_net_hooks(net, ipv4_conntrack_ops,
354 ARRAY_SIZE(ipv4_conntrack_ops));
355 mutex_unlock(®ister_ipv4_hooks);
356 }
357
358 const struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
359 .l3proto = PF_INET,
360 .pkt_to_tuple = ipv4_pkt_to_tuple,
361 .invert_tuple = ipv4_invert_tuple,
362 .get_l4proto = ipv4_get_l4proto,
363 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
364 .tuple_to_nlattr = ipv4_tuple_to_nlattr,
365 .nlattr_to_tuple = ipv4_nlattr_to_tuple,
366 .nla_policy = ipv4_nla_policy,
367 .nla_size = NLA_ALIGN(NLA_HDRLEN + sizeof(u32)) + /* CTA_IP_V4_SRC */
368 NLA_ALIGN(NLA_HDRLEN + sizeof(u32)), /* CTA_IP_V4_DST */
369 #endif
370 .net_ns_get = ipv4_hooks_register,
371 .net_ns_put = ipv4_hooks_unregister,
372 .me = THIS_MODULE,
373 };
374
375 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
376 &nf_conntrack_htable_size, 0600);
377
378 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
379 MODULE_ALIAS("ip_conntrack");
380 MODULE_LICENSE("GPL");
381
382 static const struct nf_conntrack_l4proto * const builtin_l4proto4[] = {
383 &nf_conntrack_l4proto_tcp4,
384 &nf_conntrack_l4proto_udp4,
385 &nf_conntrack_l4proto_icmp,
386 #ifdef CONFIG_NF_CT_PROTO_DCCP
387 &nf_conntrack_l4proto_dccp4,
388 #endif
389 #ifdef CONFIG_NF_CT_PROTO_SCTP
390 &nf_conntrack_l4proto_sctp4,
391 #endif
392 #ifdef CONFIG_NF_CT_PROTO_UDPLITE
393 &nf_conntrack_l4proto_udplite4,
394 #endif
395 };
396
397 static int ipv4_net_init(struct net *net)
398 {
399 return nf_ct_l4proto_pernet_register(net, builtin_l4proto4,
400 ARRAY_SIZE(builtin_l4proto4));
401 }
402
403 static void ipv4_net_exit(struct net *net)
404 {
405 nf_ct_l4proto_pernet_unregister(net, builtin_l4proto4,
406 ARRAY_SIZE(builtin_l4proto4));
407 }
408
409 static struct pernet_operations ipv4_net_ops = {
410 .init = ipv4_net_init,
411 .exit = ipv4_net_exit,
412 .id = &conntrack4_net_id,
413 .size = sizeof(struct conntrack4_net),
414 };
415
416 static int __init nf_conntrack_l3proto_ipv4_init(void)
417 {
418 int ret = 0;
419
420 need_conntrack();
421
422 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
423 if (WARN_ON(nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1) !=
424 nf_conntrack_l3proto_ipv4.nla_size))
425 return -EINVAL;
426 #endif
427 ret = nf_register_sockopt(&so_getorigdst);
428 if (ret < 0) {
429 pr_err("Unable to register netfilter socket option\n");
430 return ret;
431 }
432
433 ret = register_pernet_subsys(&ipv4_net_ops);
434 if (ret < 0) {
435 pr_err("nf_conntrack_ipv4: can't register pernet ops\n");
436 goto cleanup_sockopt;
437 }
438
439 ret = nf_ct_l4proto_register(builtin_l4proto4,
440 ARRAY_SIZE(builtin_l4proto4));
441 if (ret < 0)
442 goto cleanup_pernet;
443
444 ret = nf_ct_l3proto_register(&nf_conntrack_l3proto_ipv4);
445 if (ret < 0) {
446 pr_err("nf_conntrack_ipv4: can't register ipv4 proto.\n");
447 goto cleanup_l4proto;
448 }
449
450 return ret;
451 cleanup_l4proto:
452 nf_ct_l4proto_unregister(builtin_l4proto4,
453 ARRAY_SIZE(builtin_l4proto4));
454 cleanup_pernet:
455 unregister_pernet_subsys(&ipv4_net_ops);
456 cleanup_sockopt:
457 nf_unregister_sockopt(&so_getorigdst);
458 return ret;
459 }
460
461 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
462 {
463 synchronize_net();
464 nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
465 nf_ct_l4proto_unregister(builtin_l4proto4,
466 ARRAY_SIZE(builtin_l4proto4));
467 unregister_pernet_subsys(&ipv4_net_ops);
468 nf_unregister_sockopt(&so_getorigdst);
469 }
470
471 module_init(nf_conntrack_l3proto_ipv4_init);
472 module_exit(nf_conntrack_l3proto_ipv4_fini);
473
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.