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), bind);
65 tcf_hash_release(a, bind);
71 spin_lock_bh(&p->tcf_lock);
72 p->old_addr = parm->old_addr;
73 p->new_addr = parm->new_addr;
75 p->flags = parm->flags;
77 p->tcf_action = parm->action;
78 spin_unlock_bh(&p->tcf_lock);
80 if (ret == ACT_P_CREATED)
86 static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
87 struct tcf_result *res)
89 struct tcf_nat *p = a->priv;
100 spin_lock(&p->tcf_lock);
102 p->tcf_tm.lastuse = jiffies;
103 old_addr = p->old_addr;
104 new_addr = p->new_addr;
106 egress = p->flags & TCA_NAT_FLAG_EGRESS;
107 action = p->tcf_action;
109 bstats_update(&p->tcf_bstats, skb);
111 spin_unlock(&p->tcf_lock);
113 if (unlikely(action == TC_ACT_SHOT))
116 noff = skb_network_offset(skb);
117 if (!pskb_may_pull(skb, sizeof(*iph) + noff))
127 if (!((old_addr ^ addr) & mask)) {
128 if (skb_cloned(skb) &&
129 !skb_clone_writable(skb, sizeof(*iph) + noff) &&
130 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
134 new_addr |= addr & ~mask;
136 /* Rewrite IP header */
139 iph->saddr = new_addr;
141 iph->daddr = new_addr;
143 csum_replace4(&iph->check, addr, new_addr);
144 } else if ((iph->frag_off & htons(IP_OFFSET)) ||
145 iph->protocol != IPPROTO_ICMP) {
151 /* It would be nice to share code with stateful NAT. */
152 switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
157 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
159 !skb_clone_writable(skb, ihl + sizeof(*tcph) + noff) &&
160 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
163 tcph = (void *)(skb_network_header(skb) + ihl);
164 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, 1);
171 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
173 !skb_clone_writable(skb, ihl + sizeof(*udph) + noff) &&
174 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
177 udph = (void *)(skb_network_header(skb) + ihl);
178 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
179 inet_proto_csum_replace4(&udph->check, skb, addr,
182 udph->check = CSUM_MANGLED_0;
188 struct icmphdr *icmph;
190 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
193 icmph = (void *)(skb_network_header(skb) + ihl);
195 if ((icmph->type != ICMP_DEST_UNREACH) &&
196 (icmph->type != ICMP_TIME_EXCEEDED) &&
197 (icmph->type != ICMP_PARAMETERPROB))
200 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
204 icmph = (void *)(skb_network_header(skb) + ihl);
205 iph = (void *)(icmph + 1);
211 if ((old_addr ^ addr) & mask)
214 if (skb_cloned(skb) &&
215 !skb_clone_writable(skb, ihl + sizeof(*icmph) +
216 sizeof(*iph) + noff) &&
217 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
220 icmph = (void *)(skb_network_header(skb) + ihl);
221 iph = (void *)(icmph + 1);
224 new_addr |= addr & ~mask;
226 /* XXX Fix up the inner checksums. */
228 iph->daddr = new_addr;
230 iph->saddr = new_addr;
232 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
244 spin_lock(&p->tcf_lock);
245 p->tcf_qstats.drops++;
246 spin_unlock(&p->tcf_lock);
250 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
253 unsigned char *b = skb_tail_pointer(skb);
254 struct tcf_nat *p = a->priv;
255 struct tc_nat opt = {
256 .old_addr = p->old_addr,
257 .new_addr = p->new_addr,
261 .index = p->tcf_index,
262 .action = p->tcf_action,
263 .refcnt = p->tcf_refcnt - ref,
264 .bindcnt = p->tcf_bindcnt - bind,
268 if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
269 goto nla_put_failure;
270 t.install = jiffies_to_clock_t(jiffies - p->tcf_tm.install);
271 t.lastuse = jiffies_to_clock_t(jiffies - p->tcf_tm.lastuse);
272 t.expires = jiffies_to_clock_t(p->tcf_tm.expires);
273 if (nla_put(skb, TCA_NAT_TM, sizeof(t), &t))
274 goto nla_put_failure;
283 static struct tc_action_ops act_nat_ops = {
286 .owner = THIS_MODULE,
288 .dump = tcf_nat_dump,
289 .init = tcf_nat_init,
292 MODULE_DESCRIPTION("Stateless NAT actions");
293 MODULE_LICENSE("GPL");
295 static int __init nat_init_module(void)
297 return tcf_register_action(&act_nat_ops, NAT_TAB_MASK);
300 static void __exit nat_cleanup_module(void)
302 tcf_unregister_action(&act_nat_ops);
305 module_init(nat_init_module);
306 module_exit(nat_cleanup_module);