2 * Copyright (c) 2011 Patrick McHardy <kaber@trash.net>
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.
8 * Development of IPv6 NAT funded by Astaro.
10 #include <linux/types.h>
11 #include <linux/module.h>
12 #include <linux/skbuff.h>
13 #include <linux/ipv6.h>
14 #include <linux/netfilter.h>
15 #include <linux/netfilter_ipv6.h>
16 #include <net/secure_seq.h>
17 #include <net/checksum.h>
18 #include <net/ip6_checksum.h>
19 #include <net/ip6_route.h>
22 #include <net/netfilter/nf_conntrack_core.h>
23 #include <net/netfilter/nf_conntrack.h>
24 #include <net/netfilter/nf_nat_core.h>
25 #include <net/netfilter/nf_nat_l3proto.h>
26 #include <net/netfilter/nf_nat_l4proto.h>
28 static const struct nf_nat_l3proto nf_nat_l3proto_ipv6;
31 static void nf_nat_ipv6_decode_session(struct sk_buff *skb,
32 const struct nf_conn *ct,
33 enum ip_conntrack_dir dir,
34 unsigned long statusbit,
37 const struct nf_conntrack_tuple *t = &ct->tuplehash[dir].tuple;
38 struct flowi6 *fl6 = &fl->u.ip6;
40 if (ct->status & statusbit) {
41 fl6->daddr = t->dst.u3.in6;
42 if (t->dst.protonum == IPPROTO_TCP ||
43 t->dst.protonum == IPPROTO_UDP ||
44 t->dst.protonum == IPPROTO_UDPLITE ||
45 t->dst.protonum == IPPROTO_DCCP ||
46 t->dst.protonum == IPPROTO_SCTP)
47 fl6->fl6_dport = t->dst.u.all;
50 statusbit ^= IPS_NAT_MASK;
52 if (ct->status & statusbit) {
53 fl6->saddr = t->src.u3.in6;
54 if (t->dst.protonum == IPPROTO_TCP ||
55 t->dst.protonum == IPPROTO_UDP ||
56 t->dst.protonum == IPPROTO_UDPLITE ||
57 t->dst.protonum == IPPROTO_DCCP ||
58 t->dst.protonum == IPPROTO_SCTP)
59 fl6->fl6_sport = t->src.u.all;
64 static bool nf_nat_ipv6_in_range(const struct nf_conntrack_tuple *t,
65 const struct nf_nat_range *range)
67 return ipv6_addr_cmp(&t->src.u3.in6, &range->min_addr.in6) >= 0 &&
68 ipv6_addr_cmp(&t->src.u3.in6, &range->max_addr.in6) <= 0;
71 static u32 nf_nat_ipv6_secure_port(const struct nf_conntrack_tuple *t,
74 return secure_ipv6_port_ephemeral(t->src.u3.ip6, t->dst.u3.ip6, dport);
77 static bool nf_nat_ipv6_manip_pkt(struct sk_buff *skb,
78 unsigned int iphdroff,
79 const struct nf_nat_l4proto *l4proto,
80 const struct nf_conntrack_tuple *target,
81 enum nf_nat_manip_type maniptype)
83 struct ipv6hdr *ipv6h;
88 if (!skb_make_writable(skb, iphdroff + sizeof(*ipv6h)))
91 ipv6h = (void *)skb->data + iphdroff;
92 nexthdr = ipv6h->nexthdr;
93 hdroff = ipv6_skip_exthdr(skb, iphdroff + sizeof(*ipv6h),
98 if ((frag_off & htons(~0x7)) == 0 &&
99 !l4proto->manip_pkt(skb, &nf_nat_l3proto_ipv6, iphdroff, hdroff,
103 if (maniptype == NF_NAT_MANIP_SRC)
104 ipv6h->saddr = target->src.u3.in6;
106 ipv6h->daddr = target->dst.u3.in6;
111 static void nf_nat_ipv6_csum_update(struct sk_buff *skb,
112 unsigned int iphdroff, __sum16 *check,
113 const struct nf_conntrack_tuple *t,
114 enum nf_nat_manip_type maniptype)
116 const struct ipv6hdr *ipv6h = (struct ipv6hdr *)(skb->data + iphdroff);
117 const struct in6_addr *oldip, *newip;
119 if (maniptype == NF_NAT_MANIP_SRC) {
120 oldip = &ipv6h->saddr;
121 newip = &t->src.u3.in6;
123 oldip = &ipv6h->daddr;
124 newip = &t->dst.u3.in6;
126 inet_proto_csum_replace16(check, skb, oldip->s6_addr32,
127 newip->s6_addr32, true);
130 static void nf_nat_ipv6_csum_recalc(struct sk_buff *skb,
131 u8 proto, void *data, __sum16 *check,
132 int datalen, int oldlen)
134 if (skb->ip_summed != CHECKSUM_PARTIAL) {
135 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
137 skb->ip_summed = CHECKSUM_PARTIAL;
138 skb->csum_start = skb_headroom(skb) + skb_network_offset(skb) +
139 (data - (void *)skb->data);
140 skb->csum_offset = (void *)check - data;
141 *check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
144 inet_proto_csum_replace2(check, skb,
145 htons(oldlen), htons(datalen), true);
148 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
149 static int nf_nat_ipv6_nlattr_to_range(struct nlattr *tb[],
150 struct nf_nat_range *range)
152 if (tb[CTA_NAT_V6_MINIP]) {
153 nla_memcpy(&range->min_addr.ip6, tb[CTA_NAT_V6_MINIP],
154 sizeof(struct in6_addr));
155 range->flags |= NF_NAT_RANGE_MAP_IPS;
158 if (tb[CTA_NAT_V6_MAXIP])
159 nla_memcpy(&range->max_addr.ip6, tb[CTA_NAT_V6_MAXIP],
160 sizeof(struct in6_addr));
162 range->max_addr = range->min_addr;
168 static const struct nf_nat_l3proto nf_nat_l3proto_ipv6 = {
169 .l3proto = NFPROTO_IPV6,
170 .secure_port = nf_nat_ipv6_secure_port,
171 .in_range = nf_nat_ipv6_in_range,
172 .manip_pkt = nf_nat_ipv6_manip_pkt,
173 .csum_update = nf_nat_ipv6_csum_update,
174 .csum_recalc = nf_nat_ipv6_csum_recalc,
175 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
176 .nlattr_to_range = nf_nat_ipv6_nlattr_to_range,
179 .decode_session = nf_nat_ipv6_decode_session,
183 int nf_nat_icmpv6_reply_translation(struct sk_buff *skb,
185 enum ip_conntrack_info ctinfo,
186 unsigned int hooknum,
190 struct icmp6hdr icmp6;
193 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
194 enum nf_nat_manip_type manip = HOOK2MANIP(hooknum);
195 const struct nf_nat_l4proto *l4proto;
196 struct nf_conntrack_tuple target;
197 unsigned long statusbit;
199 NF_CT_ASSERT(ctinfo == IP_CT_RELATED || ctinfo == IP_CT_RELATED_REPLY);
201 if (!skb_make_writable(skb, hdrlen + sizeof(*inside)))
203 if (nf_ip6_checksum(skb, hooknum, hdrlen, IPPROTO_ICMPV6))
206 inside = (void *)skb->data + hdrlen;
207 if (inside->icmp6.icmp6_type == NDISC_REDIRECT) {
208 if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK)
210 if (ct->status & IPS_NAT_MASK)
214 if (manip == NF_NAT_MANIP_SRC)
215 statusbit = IPS_SRC_NAT;
217 statusbit = IPS_DST_NAT;
219 /* Invert if this is reply direction */
220 if (dir == IP_CT_DIR_REPLY)
221 statusbit ^= IPS_NAT_MASK;
223 if (!(ct->status & statusbit))
226 l4proto = __nf_nat_l4proto_find(NFPROTO_IPV6, inside->ip6.nexthdr);
227 if (!nf_nat_ipv6_manip_pkt(skb, hdrlen + sizeof(inside->icmp6),
228 l4proto, &ct->tuplehash[!dir].tuple, !manip))
231 if (skb->ip_summed != CHECKSUM_PARTIAL) {
232 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
233 inside = (void *)skb->data + hdrlen;
234 inside->icmp6.icmp6_cksum = 0;
235 inside->icmp6.icmp6_cksum =
236 csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
237 skb->len - hdrlen, IPPROTO_ICMPV6,
238 csum_partial(&inside->icmp6,
239 skb->len - hdrlen, 0));
242 nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple);
243 l4proto = __nf_nat_l4proto_find(NFPROTO_IPV6, IPPROTO_ICMPV6);
244 if (!nf_nat_ipv6_manip_pkt(skb, 0, l4proto, &target, manip))
249 EXPORT_SYMBOL_GPL(nf_nat_icmpv6_reply_translation);
252 nf_nat_ipv6_fn(void *priv, struct sk_buff *skb,
253 const struct nf_hook_state *state,
254 unsigned int (*do_chain)(void *priv,
256 const struct nf_hook_state *state,
260 enum ip_conntrack_info ctinfo;
261 struct nf_conn_nat *nat;
262 enum nf_nat_manip_type maniptype = HOOK2MANIP(state->hook);
267 ct = nf_ct_get(skb, &ctinfo);
268 /* Can't track? It's not due to stress, or conntrack would
269 * have dropped it. Hence it's the user's responsibilty to
270 * packet filter it out, or implement conntrack/NAT for that
276 /* Don't try to NAT if this packet is not conntracked */
277 if (nf_ct_is_untracked(ct))
280 nat = nf_ct_nat_ext_add(ct);
286 case IP_CT_RELATED_REPLY:
287 nexthdr = ipv6_hdr(skb)->nexthdr;
288 hdrlen = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
289 &nexthdr, &frag_off);
291 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
292 if (!nf_nat_icmpv6_reply_translation(skb, ct, ctinfo,
299 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
301 /* Seen it before? This can happen for loopback, retrans,
304 if (!nf_nat_initialized(ct, maniptype)) {
307 ret = do_chain(priv, skb, state, ct);
308 if (ret != NF_ACCEPT)
311 if (nf_nat_initialized(ct, HOOK2MANIP(state->hook)))
314 ret = nf_nat_alloc_null_binding(ct, state->hook);
315 if (ret != NF_ACCEPT)
318 pr_debug("Already setup manip %s for ct %p\n",
319 maniptype == NF_NAT_MANIP_SRC ? "SRC" : "DST",
321 if (nf_nat_oif_changed(state->hook, ctinfo, nat, state->out))
328 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
329 ctinfo == IP_CT_ESTABLISHED_REPLY);
330 if (nf_nat_oif_changed(state->hook, ctinfo, nat, state->out))
334 return nf_nat_packet(ct, ctinfo, state->hook, skb);
337 nf_ct_kill_acct(ct, ctinfo, skb);
340 EXPORT_SYMBOL_GPL(nf_nat_ipv6_fn);
343 nf_nat_ipv6_in(void *priv, struct sk_buff *skb,
344 const struct nf_hook_state *state,
345 unsigned int (*do_chain)(void *priv,
347 const struct nf_hook_state *state,
351 struct in6_addr daddr = ipv6_hdr(skb)->daddr;
353 ret = nf_nat_ipv6_fn(priv, skb, state, do_chain);
354 if (ret != NF_DROP && ret != NF_STOLEN &&
355 ipv6_addr_cmp(&daddr, &ipv6_hdr(skb)->daddr))
360 EXPORT_SYMBOL_GPL(nf_nat_ipv6_in);
363 nf_nat_ipv6_out(void *priv, struct sk_buff *skb,
364 const struct nf_hook_state *state,
365 unsigned int (*do_chain)(void *priv,
367 const struct nf_hook_state *state,
371 const struct nf_conn *ct;
372 enum ip_conntrack_info ctinfo;
377 /* root is playing with raw sockets. */
378 if (skb->len < sizeof(struct ipv6hdr))
381 ret = nf_nat_ipv6_fn(priv, skb, state, do_chain);
383 if (ret != NF_DROP && ret != NF_STOLEN &&
384 !(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
385 (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
386 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
388 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.src.u3,
389 &ct->tuplehash[!dir].tuple.dst.u3) ||
390 (ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
391 ct->tuplehash[dir].tuple.src.u.all !=
392 ct->tuplehash[!dir].tuple.dst.u.all)) {
393 err = nf_xfrm_me_harder(state->net, skb, AF_INET6);
395 ret = NF_DROP_ERR(err);
401 EXPORT_SYMBOL_GPL(nf_nat_ipv6_out);
404 nf_nat_ipv6_local_fn(void *priv, struct sk_buff *skb,
405 const struct nf_hook_state *state,
406 unsigned int (*do_chain)(void *priv,
408 const struct nf_hook_state *state,
411 const struct nf_conn *ct;
412 enum ip_conntrack_info ctinfo;
416 /* root is playing with raw sockets. */
417 if (skb->len < sizeof(struct ipv6hdr))
420 ret = nf_nat_ipv6_fn(priv, skb, state, do_chain);
421 if (ret != NF_DROP && ret != NF_STOLEN &&
422 (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
423 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
425 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.dst.u3,
426 &ct->tuplehash[!dir].tuple.src.u3)) {
427 err = ip6_route_me_harder(state->net, skb);
429 ret = NF_DROP_ERR(err);
432 else if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
433 ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
434 ct->tuplehash[dir].tuple.dst.u.all !=
435 ct->tuplehash[!dir].tuple.src.u.all) {
436 err = nf_xfrm_me_harder(state->net, skb, AF_INET6);
438 ret = NF_DROP_ERR(err);
444 EXPORT_SYMBOL_GPL(nf_nat_ipv6_local_fn);
446 static int __init nf_nat_l3proto_ipv6_init(void)
450 err = nf_nat_l4proto_register(NFPROTO_IPV6, &nf_nat_l4proto_icmpv6);
453 err = nf_nat_l3proto_register(&nf_nat_l3proto_ipv6);
459 nf_nat_l4proto_unregister(NFPROTO_IPV6, &nf_nat_l4proto_icmpv6);
464 static void __exit nf_nat_l3proto_ipv6_exit(void)
466 nf_nat_l3proto_unregister(&nf_nat_l3proto_ipv6);
467 nf_nat_l4proto_unregister(NFPROTO_IPV6, &nf_nat_l4proto_icmpv6);
470 MODULE_LICENSE("GPL");
471 MODULE_ALIAS("nf-nat-" __stringify(AF_INET6));
473 module_init(nf_nat_l3proto_ipv6_init);
474 module_exit(nf_nat_l3proto_ipv6_exit);