2 * Stateless NAT actions
4 * Copyright (c) 2007 Herbert Xu <herbert@gondor.apana.org.au>
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the Free
8 * Software Foundation; either version 2 of the License, or (at your option)
12 #include <linux/errno.h>
13 #include <linux/init.h>
14 #include <linux/kernel.h>
15 #include <linux/module.h>
16 #include <linux/netfilter.h>
17 #include <linux/rtnetlink.h>
18 #include <linux/skbuff.h>
19 #include <linux/slab.h>
20 #include <linux/spinlock.h>
21 #include <linux/string.h>
22 #include <linux/tc_act/tc_nat.h>
23 #include <net/act_api.h>
26 #include <net/netlink.h>
27 #include <net/tc_act/tc_nat.h>
32 #define NAT_TAB_MASK 15
34 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
35 [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
38 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
39 struct tc_action *a, int ovr, int bind)
41 struct nlattr *tb[TCA_NAT_MAX + 1];
49 err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy);
53 if (tb[TCA_NAT_PARMS] == NULL)
55 parm = nla_data(tb[TCA_NAT_PARMS]);
57 if (!tcf_hash_check(parm->index, a, bind)) {
58 ret = tcf_hash_create(parm->index, est, a, sizeof(*p),
66 tcf_hash_release(a, bind);
72 spin_lock_bh(&p->tcf_lock);
73 p->old_addr = parm->old_addr;
74 p->new_addr = parm->new_addr;
76 p->flags = parm->flags;
78 p->tcf_action = parm->action;
79 spin_unlock_bh(&p->tcf_lock);
81 if (ret == ACT_P_CREATED)
87 static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
88 struct tcf_result *res)
90 struct tcf_nat *p = a->priv;
101 spin_lock(&p->tcf_lock);
103 p->tcf_tm.lastuse = jiffies;
104 old_addr = p->old_addr;
105 new_addr = p->new_addr;
107 egress = p->flags & TCA_NAT_FLAG_EGRESS;
108 action = p->tcf_action;
110 bstats_update(&p->tcf_bstats, skb);
112 spin_unlock(&p->tcf_lock);
114 if (unlikely(action == TC_ACT_SHOT))
117 noff = skb_network_offset(skb);
118 if (!pskb_may_pull(skb, sizeof(*iph) + noff))
128 if (!((old_addr ^ addr) & mask)) {
129 if (skb_cloned(skb) &&
130 !skb_clone_writable(skb, sizeof(*iph) + noff) &&
131 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
135 new_addr |= addr & ~mask;
137 /* Rewrite IP header */
140 iph->saddr = new_addr;
142 iph->daddr = new_addr;
144 csum_replace4(&iph->check, addr, new_addr);
145 } else if ((iph->frag_off & htons(IP_OFFSET)) ||
146 iph->protocol != IPPROTO_ICMP) {
152 /* It would be nice to share code with stateful NAT. */
153 switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
158 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
160 !skb_clone_writable(skb, ihl + sizeof(*tcph) + noff) &&
161 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
164 tcph = (void *)(skb_network_header(skb) + ihl);
165 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, 1);
172 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
174 !skb_clone_writable(skb, ihl + sizeof(*udph) + noff) &&
175 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
178 udph = (void *)(skb_network_header(skb) + ihl);
179 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
180 inet_proto_csum_replace4(&udph->check, skb, addr,
183 udph->check = CSUM_MANGLED_0;
189 struct icmphdr *icmph;
191 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
194 icmph = (void *)(skb_network_header(skb) + ihl);
196 if ((icmph->type != ICMP_DEST_UNREACH) &&
197 (icmph->type != ICMP_TIME_EXCEEDED) &&
198 (icmph->type != ICMP_PARAMETERPROB))
201 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
205 icmph = (void *)(skb_network_header(skb) + ihl);
206 iph = (void *)(icmph + 1);
212 if ((old_addr ^ addr) & mask)
215 if (skb_cloned(skb) &&
216 !skb_clone_writable(skb, ihl + sizeof(*icmph) +
217 sizeof(*iph) + noff) &&
218 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
221 icmph = (void *)(skb_network_header(skb) + ihl);
222 iph = (void *)(icmph + 1);
225 new_addr |= addr & ~mask;
227 /* XXX Fix up the inner checksums. */
229 iph->daddr = new_addr;
231 iph->saddr = new_addr;
233 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
245 spin_lock(&p->tcf_lock);
246 p->tcf_qstats.drops++;
247 spin_unlock(&p->tcf_lock);
251 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
254 unsigned char *b = skb_tail_pointer(skb);
255 struct tcf_nat *p = a->priv;
256 struct tc_nat opt = {
257 .old_addr = p->old_addr,
258 .new_addr = p->new_addr,
262 .index = p->tcf_index,
263 .action = p->tcf_action,
264 .refcnt = p->tcf_refcnt - ref,
265 .bindcnt = p->tcf_bindcnt - bind,
269 if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
270 goto nla_put_failure;
271 t.install = jiffies_to_clock_t(jiffies - p->tcf_tm.install);
272 t.lastuse = jiffies_to_clock_t(jiffies - p->tcf_tm.lastuse);
273 t.expires = jiffies_to_clock_t(p->tcf_tm.expires);
274 if (nla_put(skb, TCA_NAT_TM, sizeof(t), &t))
275 goto nla_put_failure;
284 static struct tc_action_ops act_nat_ops = {
287 .owner = THIS_MODULE,
289 .dump = tcf_nat_dump,
290 .init = tcf_nat_init,
293 MODULE_DESCRIPTION("Stateless NAT actions");
294 MODULE_LICENSE("GPL");
296 static int __init nat_init_module(void)
298 return tcf_register_action(&act_nat_ops, NAT_TAB_MASK);
301 static void __exit nat_cleanup_module(void)
303 tcf_unregister_action(&act_nat_ops);
306 module_init(nat_init_module);
307 module_exit(nat_cleanup_module);