ipv6: make lookups simpler and faster
[deliverable/linux.git] / net / ipv6 / netfilter / nf_conntrack_l3proto_ipv6.c
1 /*
2 * Copyright (C)2004 USAGI/WIDE Project
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
7 *
8 * Author:
9 * Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
10 */
11
12 #include <linux/types.h>
13 #include <linux/ipv6.h>
14 #include <linux/in6.h>
15 #include <linux/netfilter.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/icmp.h>
19 #include <net/ipv6.h>
20 #include <net/inet_frag.h>
21
22 #include <linux/netfilter_bridge.h>
23 #include <linux/netfilter_ipv6.h>
24 #include <linux/netfilter_ipv6/ip6_tables.h>
25 #include <net/netfilter/nf_conntrack.h>
26 #include <net/netfilter/nf_conntrack_helper.h>
27 #include <net/netfilter/nf_conntrack_l4proto.h>
28 #include <net/netfilter/nf_conntrack_l3proto.h>
29 #include <net/netfilter/nf_conntrack_core.h>
30 #include <net/netfilter/nf_conntrack_zones.h>
31 #include <net/netfilter/nf_conntrack_seqadj.h>
32 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
33 #include <net/netfilter/nf_nat_helper.h>
34 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
35 #include <net/netfilter/nf_log.h>
36
37 static bool ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
38 struct nf_conntrack_tuple *tuple)
39 {
40 const u_int32_t *ap;
41 u_int32_t _addrs[8];
42
43 ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
44 sizeof(_addrs), _addrs);
45 if (ap == NULL)
46 return false;
47
48 memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
49 memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
50
51 return true;
52 }
53
54 static bool ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
55 const struct nf_conntrack_tuple *orig)
56 {
57 memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
58 memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
59
60 return true;
61 }
62
63 static int ipv6_print_tuple(struct seq_file *s,
64 const struct nf_conntrack_tuple *tuple)
65 {
66 return seq_printf(s, "src=%pI6 dst=%pI6 ",
67 tuple->src.u3.ip6, tuple->dst.u3.ip6);
68 }
69
70 static int ipv6_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
71 unsigned int *dataoff, u_int8_t *protonum)
72 {
73 unsigned int extoff = nhoff + sizeof(struct ipv6hdr);
74 __be16 frag_off;
75 int protoff;
76 u8 nexthdr;
77
78 if (skb_copy_bits(skb, nhoff + offsetof(struct ipv6hdr, nexthdr),
79 &nexthdr, sizeof(nexthdr)) != 0) {
80 pr_debug("ip6_conntrack_core: can't get nexthdr\n");
81 return -NF_ACCEPT;
82 }
83 protoff = ipv6_skip_exthdr(skb, extoff, &nexthdr, &frag_off);
84 /*
85 * (protoff == skb->len) means the packet has not data, just
86 * IPv6 and possibly extensions headers, but it is tracked anyway
87 */
88 if (protoff < 0 || (frag_off & htons(~0x7)) != 0) {
89 pr_debug("ip6_conntrack_core: can't find proto in pkt\n");
90 return -NF_ACCEPT;
91 }
92
93 *dataoff = protoff;
94 *protonum = nexthdr;
95 return NF_ACCEPT;
96 }
97
98 static unsigned int ipv6_helper(unsigned int hooknum,
99 struct sk_buff *skb,
100 const struct net_device *in,
101 const struct net_device *out,
102 int (*okfn)(struct sk_buff *))
103 {
104 struct nf_conn *ct;
105 const struct nf_conn_help *help;
106 const struct nf_conntrack_helper *helper;
107 enum ip_conntrack_info ctinfo;
108 __be16 frag_off;
109 int protoff;
110 u8 nexthdr;
111
112 /* This is where we call the helper: as the packet goes out. */
113 ct = nf_ct_get(skb, &ctinfo);
114 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
115 return NF_ACCEPT;
116
117 help = nfct_help(ct);
118 if (!help)
119 return NF_ACCEPT;
120 /* rcu_read_lock()ed by nf_hook_slow */
121 helper = rcu_dereference(help->helper);
122 if (!helper)
123 return NF_ACCEPT;
124
125 nexthdr = ipv6_hdr(skb)->nexthdr;
126 protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
127 &frag_off);
128 if (protoff < 0 || (frag_off & htons(~0x7)) != 0) {
129 pr_debug("proto header not found\n");
130 return NF_ACCEPT;
131 }
132
133 return helper->help(skb, protoff, ct, ctinfo);
134 }
135
136 static unsigned int ipv6_confirm(unsigned int hooknum,
137 struct sk_buff *skb,
138 const struct net_device *in,
139 const struct net_device *out,
140 int (*okfn)(struct sk_buff *))
141 {
142 struct nf_conn *ct;
143 enum ip_conntrack_info ctinfo;
144 unsigned char pnum = ipv6_hdr(skb)->nexthdr;
145 int protoff;
146 __be16 frag_off;
147
148 ct = nf_ct_get(skb, &ctinfo);
149 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
150 goto out;
151
152 protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &pnum,
153 &frag_off);
154 if (protoff < 0 || (frag_off & htons(~0x7)) != 0) {
155 pr_debug("proto header not found\n");
156 goto out;
157 }
158
159 /* adjust seqs for loopback traffic only in outgoing direction */
160 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
161 !nf_is_loopback_packet(skb)) {
162 if (!nf_ct_seq_adjust(skb, ct, ctinfo, protoff)) {
163 NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
164 return NF_DROP;
165 }
166 }
167 out:
168 /* We've seen it coming out the other side: confirm it */
169 return nf_conntrack_confirm(skb);
170 }
171
172 static unsigned int __ipv6_conntrack_in(struct net *net,
173 unsigned int hooknum,
174 struct sk_buff *skb,
175 const struct net_device *in,
176 const struct net_device *out,
177 int (*okfn)(struct sk_buff *))
178 {
179 struct sk_buff *reasm = skb->nfct_reasm;
180 const struct nf_conn_help *help;
181 struct nf_conn *ct;
182 enum ip_conntrack_info ctinfo;
183
184 /* This packet is fragmented and has reassembled packet. */
185 if (reasm) {
186 /* Reassembled packet isn't parsed yet ? */
187 if (!reasm->nfct) {
188 unsigned int ret;
189
190 ret = nf_conntrack_in(net, PF_INET6, hooknum, reasm);
191 if (ret != NF_ACCEPT)
192 return ret;
193 }
194
195 /* Conntrack helpers need the entire reassembled packet in the
196 * POST_ROUTING hook. In case of unconfirmed connections NAT
197 * might reassign a helper, so the entire packet is also
198 * required.
199 */
200 ct = nf_ct_get(reasm, &ctinfo);
201 if (ct != NULL && !nf_ct_is_untracked(ct)) {
202 help = nfct_help(ct);
203 if ((help && help->helper) || !nf_ct_is_confirmed(ct)) {
204 nf_conntrack_get_reasm(reasm);
205 NF_HOOK_THRESH(NFPROTO_IPV6, hooknum, reasm,
206 (struct net_device *)in,
207 (struct net_device *)out,
208 okfn, NF_IP6_PRI_CONNTRACK + 1);
209 return NF_DROP_ERR(-ECANCELED);
210 }
211 }
212
213 nf_conntrack_get(reasm->nfct);
214 skb->nfct = reasm->nfct;
215 skb->nfctinfo = reasm->nfctinfo;
216 return NF_ACCEPT;
217 }
218
219 return nf_conntrack_in(net, PF_INET6, hooknum, skb);
220 }
221
222 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
223 struct sk_buff *skb,
224 const struct net_device *in,
225 const struct net_device *out,
226 int (*okfn)(struct sk_buff *))
227 {
228 return __ipv6_conntrack_in(dev_net(in), hooknum, skb, in, out, okfn);
229 }
230
231 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
232 struct sk_buff *skb,
233 const struct net_device *in,
234 const struct net_device *out,
235 int (*okfn)(struct sk_buff *))
236 {
237 /* root is playing with raw sockets. */
238 if (skb->len < sizeof(struct ipv6hdr)) {
239 net_notice_ratelimited("ipv6_conntrack_local: packet too short\n");
240 return NF_ACCEPT;
241 }
242 return __ipv6_conntrack_in(dev_net(out), hooknum, skb, in, out, okfn);
243 }
244
245 static struct nf_hook_ops ipv6_conntrack_ops[] __read_mostly = {
246 {
247 .hook = ipv6_conntrack_in,
248 .owner = THIS_MODULE,
249 .pf = NFPROTO_IPV6,
250 .hooknum = NF_INET_PRE_ROUTING,
251 .priority = NF_IP6_PRI_CONNTRACK,
252 },
253 {
254 .hook = ipv6_conntrack_local,
255 .owner = THIS_MODULE,
256 .pf = NFPROTO_IPV6,
257 .hooknum = NF_INET_LOCAL_OUT,
258 .priority = NF_IP6_PRI_CONNTRACK,
259 },
260 {
261 .hook = ipv6_helper,
262 .owner = THIS_MODULE,
263 .pf = NFPROTO_IPV6,
264 .hooknum = NF_INET_POST_ROUTING,
265 .priority = NF_IP6_PRI_CONNTRACK_HELPER,
266 },
267 {
268 .hook = ipv6_confirm,
269 .owner = THIS_MODULE,
270 .pf = NFPROTO_IPV6,
271 .hooknum = NF_INET_POST_ROUTING,
272 .priority = NF_IP6_PRI_LAST,
273 },
274 {
275 .hook = ipv6_helper,
276 .owner = THIS_MODULE,
277 .pf = NFPROTO_IPV6,
278 .hooknum = NF_INET_LOCAL_IN,
279 .priority = NF_IP6_PRI_CONNTRACK_HELPER,
280 },
281 {
282 .hook = ipv6_confirm,
283 .owner = THIS_MODULE,
284 .pf = NFPROTO_IPV6,
285 .hooknum = NF_INET_LOCAL_IN,
286 .priority = NF_IP6_PRI_LAST-1,
287 },
288 };
289
290 static int
291 ipv6_getorigdst(struct sock *sk, int optval, void __user *user, int *len)
292 {
293 const struct inet_sock *inet = inet_sk(sk);
294 const struct ipv6_pinfo *inet6 = inet6_sk(sk);
295 const struct nf_conntrack_tuple_hash *h;
296 struct sockaddr_in6 sin6;
297 struct nf_conntrack_tuple tuple = { .src.l3num = NFPROTO_IPV6 };
298 struct nf_conn *ct;
299
300 tuple.src.u3.in6 = sk->sk_v6_rcv_saddr;
301 tuple.src.u.tcp.port = inet->inet_sport;
302 tuple.dst.u3.in6 = sk->sk_v6_daddr;
303 tuple.dst.u.tcp.port = inet->inet_dport;
304 tuple.dst.protonum = sk->sk_protocol;
305
306 if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP)
307 return -ENOPROTOOPT;
308
309 if (*len < 0 || (unsigned int) *len < sizeof(sin6))
310 return -EINVAL;
311
312 h = nf_conntrack_find_get(sock_net(sk), NF_CT_DEFAULT_ZONE, &tuple);
313 if (!h) {
314 pr_debug("IP6T_SO_ORIGINAL_DST: Can't find %pI6c/%u-%pI6c/%u.\n",
315 &tuple.src.u3.ip6, ntohs(tuple.src.u.tcp.port),
316 &tuple.dst.u3.ip6, ntohs(tuple.dst.u.tcp.port));
317 return -ENOENT;
318 }
319
320 ct = nf_ct_tuplehash_to_ctrack(h);
321
322 sin6.sin6_family = AF_INET6;
323 sin6.sin6_port = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.u.tcp.port;
324 sin6.sin6_flowinfo = inet6->flow_label & IPV6_FLOWINFO_MASK;
325 memcpy(&sin6.sin6_addr,
326 &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.u3.in6,
327 sizeof(sin6.sin6_addr));
328
329 nf_ct_put(ct);
330 sin6.sin6_scope_id = ipv6_iface_scope_id(&sin6.sin6_addr,
331 sk->sk_bound_dev_if);
332 return copy_to_user(user, &sin6, sizeof(sin6)) ? -EFAULT : 0;
333 }
334
335 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
336
337 #include <linux/netfilter/nfnetlink.h>
338 #include <linux/netfilter/nfnetlink_conntrack.h>
339
340 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
341 const struct nf_conntrack_tuple *tuple)
342 {
343 if (nla_put(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
344 &tuple->src.u3.ip6) ||
345 nla_put(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
346 &tuple->dst.u3.ip6))
347 goto nla_put_failure;
348 return 0;
349
350 nla_put_failure:
351 return -1;
352 }
353
354 static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = {
355 [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 },
356 [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 },
357 };
358
359 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
360 struct nf_conntrack_tuple *t)
361 {
362 if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
363 return -EINVAL;
364
365 memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]),
366 sizeof(u_int32_t) * 4);
367 memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]),
368 sizeof(u_int32_t) * 4);
369
370 return 0;
371 }
372
373 static int ipv6_nlattr_tuple_size(void)
374 {
375 return nla_policy_len(ipv6_nla_policy, CTA_IP_MAX + 1);
376 }
377 #endif
378
379 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = {
380 .l3proto = PF_INET6,
381 .name = "ipv6",
382 .pkt_to_tuple = ipv6_pkt_to_tuple,
383 .invert_tuple = ipv6_invert_tuple,
384 .print_tuple = ipv6_print_tuple,
385 .get_l4proto = ipv6_get_l4proto,
386 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
387 .tuple_to_nlattr = ipv6_tuple_to_nlattr,
388 .nlattr_tuple_size = ipv6_nlattr_tuple_size,
389 .nlattr_to_tuple = ipv6_nlattr_to_tuple,
390 .nla_policy = ipv6_nla_policy,
391 #endif
392 .me = THIS_MODULE,
393 };
394
395 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
396 MODULE_LICENSE("GPL");
397 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
398
399 static struct nf_sockopt_ops so_getorigdst6 = {
400 .pf = NFPROTO_IPV6,
401 .get_optmin = IP6T_SO_ORIGINAL_DST,
402 .get_optmax = IP6T_SO_ORIGINAL_DST + 1,
403 .get = ipv6_getorigdst,
404 .owner = THIS_MODULE,
405 };
406
407 static int ipv6_net_init(struct net *net)
408 {
409 int ret = 0;
410
411 ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_tcp6);
412 if (ret < 0) {
413 pr_err("nf_conntrack_tcp6: pernet registration failed\n");
414 goto out;
415 }
416 ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_udp6);
417 if (ret < 0) {
418 pr_err("nf_conntrack_udp6: pernet registration failed\n");
419 goto cleanup_tcp6;
420 }
421 ret = nf_ct_l4proto_pernet_register(net, &nf_conntrack_l4proto_icmpv6);
422 if (ret < 0) {
423 pr_err("nf_conntrack_icmp6: pernet registration failed\n");
424 goto cleanup_udp6;
425 }
426 ret = nf_ct_l3proto_pernet_register(net, &nf_conntrack_l3proto_ipv6);
427 if (ret < 0) {
428 pr_err("nf_conntrack_ipv6: pernet registration failed.\n");
429 goto cleanup_icmpv6;
430 }
431 return 0;
432 cleanup_icmpv6:
433 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_icmpv6);
434 cleanup_udp6:
435 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_udp6);
436 cleanup_tcp6:
437 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_tcp6);
438 out:
439 return ret;
440 }
441
442 static void ipv6_net_exit(struct net *net)
443 {
444 nf_ct_l3proto_pernet_unregister(net, &nf_conntrack_l3proto_ipv6);
445 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_icmpv6);
446 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_udp6);
447 nf_ct_l4proto_pernet_unregister(net, &nf_conntrack_l4proto_tcp6);
448 }
449
450 static struct pernet_operations ipv6_net_ops = {
451 .init = ipv6_net_init,
452 .exit = ipv6_net_exit,
453 };
454
455 static int __init nf_conntrack_l3proto_ipv6_init(void)
456 {
457 int ret = 0;
458
459 need_conntrack();
460 nf_defrag_ipv6_enable();
461
462 ret = nf_register_sockopt(&so_getorigdst6);
463 if (ret < 0) {
464 pr_err("Unable to register netfilter socket option\n");
465 return ret;
466 }
467
468 ret = register_pernet_subsys(&ipv6_net_ops);
469 if (ret < 0)
470 goto cleanup_sockopt;
471
472 ret = nf_register_hooks(ipv6_conntrack_ops,
473 ARRAY_SIZE(ipv6_conntrack_ops));
474 if (ret < 0) {
475 pr_err("nf_conntrack_ipv6: can't register pre-routing defrag "
476 "hook.\n");
477 goto cleanup_pernet;
478 }
479
480 ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_tcp6);
481 if (ret < 0) {
482 pr_err("nf_conntrack_ipv6: can't register tcp6 proto.\n");
483 goto cleanup_hooks;
484 }
485
486 ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_udp6);
487 if (ret < 0) {
488 pr_err("nf_conntrack_ipv6: can't register udp6 proto.\n");
489 goto cleanup_tcp6;
490 }
491
492 ret = nf_ct_l4proto_register(&nf_conntrack_l4proto_icmpv6);
493 if (ret < 0) {
494 pr_err("nf_conntrack_ipv6: can't register icmpv6 proto.\n");
495 goto cleanup_udp6;
496 }
497
498 ret = nf_ct_l3proto_register(&nf_conntrack_l3proto_ipv6);
499 if (ret < 0) {
500 pr_err("nf_conntrack_ipv6: can't register ipv6 proto.\n");
501 goto cleanup_icmpv6;
502 }
503 return ret;
504
505 cleanup_icmpv6:
506 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
507 cleanup_udp6:
508 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp6);
509 cleanup_tcp6:
510 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
511 cleanup_hooks:
512 nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
513 cleanup_pernet:
514 unregister_pernet_subsys(&ipv6_net_ops);
515 cleanup_sockopt:
516 nf_unregister_sockopt(&so_getorigdst6);
517 return ret;
518 }
519
520 static void __exit nf_conntrack_l3proto_ipv6_fini(void)
521 {
522 synchronize_net();
523 nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
524 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
525 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_udp6);
526 nf_ct_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
527 nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
528 unregister_pernet_subsys(&ipv6_net_ops);
529 nf_unregister_sockopt(&so_getorigdst6);
530 }
531
532 module_init(nf_conntrack_l3proto_ipv6_init);
533 module_exit(nf_conntrack_l3proto_ipv6_fini);
This page took 0.069707 seconds and 5 git commands to generate.