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
33 static struct tcf_common *tcf_nat_ht[NAT_TAB_MASK + 1];
34 static u32 nat_idx_gen;
35 static DEFINE_RWLOCK(nat_lock);
37 static struct tcf_hashinfo nat_hash_info = {
39 .hmask = NAT_TAB_MASK,
43 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
44 [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
47 static int tcf_nat_init(struct nlattr *nla, struct nlattr *est,
48 struct tc_action *a, int ovr, int bind)
50 struct nlattr *tb[TCA_NAT_MAX + 1];
54 struct tcf_common *pc;
59 err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy);
63 if (tb[TCA_NAT_PARMS] == NULL)
65 parm = nla_data(tb[TCA_NAT_PARMS]);
67 pc = tcf_hash_check(parm->index, a, bind, &nat_hash_info);
69 pc = tcf_hash_create(parm->index, est, a, sizeof(*p), bind,
70 &nat_idx_gen, &nat_hash_info);
78 tcf_hash_release(pc, bind, &nat_hash_info);
83 spin_lock_bh(&p->tcf_lock);
84 p->old_addr = parm->old_addr;
85 p->new_addr = parm->new_addr;
87 p->flags = parm->flags;
89 p->tcf_action = parm->action;
90 spin_unlock_bh(&p->tcf_lock);
92 if (ret == ACT_P_CREATED)
93 tcf_hash_insert(pc, &nat_hash_info);
98 static int tcf_nat_cleanup(struct tc_action *a, int bind)
100 struct tcf_nat *p = a->priv;
102 return tcf_hash_release(&p->common, bind, &nat_hash_info);
105 static int tcf_nat(struct sk_buff *skb, struct tc_action *a,
106 struct tcf_result *res)
108 struct tcf_nat *p = a->priv;
118 spin_lock(&p->tcf_lock);
120 p->tcf_tm.lastuse = jiffies;
121 old_addr = p->old_addr;
122 new_addr = p->new_addr;
124 egress = p->flags & TCA_NAT_FLAG_EGRESS;
125 action = p->tcf_action;
127 p->tcf_bstats.bytes += qdisc_pkt_len(skb);
128 p->tcf_bstats.packets++;
130 spin_unlock(&p->tcf_lock);
132 if (unlikely(action == TC_ACT_SHOT))
135 if (!pskb_may_pull(skb, sizeof(*iph)))
145 if (!((old_addr ^ addr) & mask)) {
146 if (skb_cloned(skb) &&
147 !skb_clone_writable(skb, sizeof(*iph)) &&
148 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
152 new_addr |= addr & ~mask;
154 /* Rewrite IP header */
157 iph->saddr = new_addr;
159 iph->daddr = new_addr;
161 csum_replace4(&iph->check, addr, new_addr);
162 } else if ((iph->frag_off & htons(IP_OFFSET)) ||
163 iph->protocol != IPPROTO_ICMP) {
169 /* It would be nice to share code with stateful NAT. */
170 switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
175 if (!pskb_may_pull(skb, ihl + sizeof(*tcph)) ||
177 !skb_clone_writable(skb, ihl + sizeof(*tcph)) &&
178 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
181 tcph = (void *)(skb_network_header(skb) + ihl);
182 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, 1);
189 if (!pskb_may_pull(skb, ihl + sizeof(*udph)) ||
191 !skb_clone_writable(skb, ihl + sizeof(*udph)) &&
192 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
195 udph = (void *)(skb_network_header(skb) + ihl);
196 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
197 inet_proto_csum_replace4(&udph->check, skb, addr,
200 udph->check = CSUM_MANGLED_0;
206 struct icmphdr *icmph;
208 if (!pskb_may_pull(skb, ihl + sizeof(*icmph)))
211 icmph = (void *)(skb_network_header(skb) + ihl);
213 if ((icmph->type != ICMP_DEST_UNREACH) &&
214 (icmph->type != ICMP_TIME_EXCEEDED) &&
215 (icmph->type != ICMP_PARAMETERPROB))
218 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph)))
221 icmph = (void *)(skb_network_header(skb) + ihl);
222 iph = (void *)(icmph + 1);
228 if ((old_addr ^ addr) & mask)
231 if (skb_cloned(skb) &&
232 !skb_clone_writable(skb,
233 ihl + sizeof(*icmph) + sizeof(*iph)) &&
234 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
237 icmph = (void *)(skb_network_header(skb) + ihl);
238 iph = (void *)(icmph + 1);
241 new_addr |= addr & ~mask;
243 /* XXX Fix up the inner checksums. */
245 iph->daddr = new_addr;
247 iph->saddr = new_addr;
249 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
261 spin_lock(&p->tcf_lock);
262 p->tcf_qstats.drops++;
263 spin_unlock(&p->tcf_lock);
267 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
270 unsigned char *b = skb_tail_pointer(skb);
271 struct tcf_nat *p = a->priv;
275 opt.old_addr = p->old_addr;
276 opt.new_addr = p->new_addr;
278 opt.flags = p->flags;
280 opt.index = p->tcf_index;
281 opt.action = p->tcf_action;
282 opt.refcnt = p->tcf_refcnt - ref;
283 opt.bindcnt = p->tcf_bindcnt - bind;
285 NLA_PUT(skb, TCA_NAT_PARMS, sizeof(opt), &opt);
286 t.install = jiffies_to_clock_t(jiffies - p->tcf_tm.install);
287 t.lastuse = jiffies_to_clock_t(jiffies - p->tcf_tm.lastuse);
288 t.expires = jiffies_to_clock_t(p->tcf_tm.expires);
289 NLA_PUT(skb, TCA_NAT_TM, sizeof(t), &t);
298 static struct tc_action_ops act_nat_ops = {
300 .hinfo = &nat_hash_info,
302 .capab = TCA_CAP_NONE,
303 .owner = THIS_MODULE,
305 .dump = tcf_nat_dump,
306 .cleanup = tcf_nat_cleanup,
307 .lookup = tcf_hash_search,
308 .init = tcf_nat_init,
309 .walk = tcf_generic_walker
312 MODULE_DESCRIPTION("Stateless NAT actions");
313 MODULE_LICENSE("GPL");
315 static int __init nat_init_module(void)
317 return tcf_register_action(&act_nat_ops);
320 static void __exit nat_cleanup_module(void)
322 tcf_unregister_action(&act_nat_ops);
325 module_init(nat_init_module);
326 module_exit(nat_cleanup_module);