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_try_make_writable(skb
, sizeof(*iph
) + noff
))
133 new_addr
|= addr
& ~mask
;
135 /* Rewrite IP header */
138 iph
->saddr
= new_addr
;
140 iph
->daddr
= new_addr
;
142 csum_replace4(&iph
->check
, addr
, new_addr
);
143 } else if ((iph
->frag_off
& htons(IP_OFFSET
)) ||
144 iph
->protocol
!= IPPROTO_ICMP
) {
150 /* It would be nice to share code with stateful NAT. */
151 switch (iph
->frag_off
& htons(IP_OFFSET
) ? 0 : iph
->protocol
) {
156 if (!pskb_may_pull(skb
, ihl
+ sizeof(*tcph
) + noff
) ||
157 skb_try_make_writable(skb
, ihl
+ sizeof(*tcph
) + noff
))
160 tcph
= (void *)(skb_network_header(skb
) + ihl
);
161 inet_proto_csum_replace4(&tcph
->check
, skb
, addr
, new_addr
,
169 if (!pskb_may_pull(skb
, ihl
+ sizeof(*udph
) + noff
) ||
170 skb_try_make_writable(skb
, ihl
+ sizeof(*udph
) + noff
))
173 udph
= (void *)(skb_network_header(skb
) + ihl
);
174 if (udph
->check
|| skb
->ip_summed
== CHECKSUM_PARTIAL
) {
175 inet_proto_csum_replace4(&udph
->check
, skb
, addr
,
178 udph
->check
= CSUM_MANGLED_0
;
184 struct icmphdr
*icmph
;
186 if (!pskb_may_pull(skb
, ihl
+ sizeof(*icmph
) + noff
))
189 icmph
= (void *)(skb_network_header(skb
) + ihl
);
191 if ((icmph
->type
!= ICMP_DEST_UNREACH
) &&
192 (icmph
->type
!= ICMP_TIME_EXCEEDED
) &&
193 (icmph
->type
!= ICMP_PARAMETERPROB
))
196 if (!pskb_may_pull(skb
, ihl
+ sizeof(*icmph
) + sizeof(*iph
) +
200 icmph
= (void *)(skb_network_header(skb
) + ihl
);
201 iph
= (void *)(icmph
+ 1);
207 if ((old_addr
^ addr
) & mask
)
210 if (skb_try_make_writable(skb
, ihl
+ sizeof(*icmph
) +
211 sizeof(*iph
) + noff
))
214 icmph
= (void *)(skb_network_header(skb
) + ihl
);
215 iph
= (void *)(icmph
+ 1);
218 new_addr
|= addr
& ~mask
;
220 /* XXX Fix up the inner checksums. */
222 iph
->daddr
= new_addr
;
224 iph
->saddr
= new_addr
;
226 inet_proto_csum_replace4(&icmph
->checksum
, skb
, addr
, new_addr
,
238 spin_lock(&p
->tcf_lock
);
239 p
->tcf_qstats
.drops
++;
240 spin_unlock(&p
->tcf_lock
);
244 static int tcf_nat_dump(struct sk_buff
*skb
, struct tc_action
*a
,
247 unsigned char *b
= skb_tail_pointer(skb
);
248 struct tcf_nat
*p
= a
->priv
;
249 struct tc_nat opt
= {
250 .old_addr
= p
->old_addr
,
251 .new_addr
= p
->new_addr
,
255 .index
= p
->tcf_index
,
256 .action
= p
->tcf_action
,
257 .refcnt
= p
->tcf_refcnt
- ref
,
258 .bindcnt
= p
->tcf_bindcnt
- bind
,
262 if (nla_put(skb
, TCA_NAT_PARMS
, sizeof(opt
), &opt
))
263 goto nla_put_failure
;
264 t
.install
= jiffies_to_clock_t(jiffies
- p
->tcf_tm
.install
);
265 t
.lastuse
= jiffies_to_clock_t(jiffies
- p
->tcf_tm
.lastuse
);
266 t
.expires
= jiffies_to_clock_t(p
->tcf_tm
.expires
);
267 if (nla_put(skb
, TCA_NAT_TM
, sizeof(t
), &t
))
268 goto nla_put_failure
;
277 static struct tc_action_ops act_nat_ops
= {
280 .owner
= THIS_MODULE
,
282 .dump
= tcf_nat_dump
,
283 .init
= tcf_nat_init
,
286 MODULE_DESCRIPTION("Stateless NAT actions");
287 MODULE_LICENSE("GPL");
289 static int __init
nat_init_module(void)
291 return tcf_register_action(&act_nat_ops
, NAT_TAB_MASK
);
294 static void __exit
nat_cleanup_module(void)
296 tcf_unregister_action(&act_nat_ops
);
299 module_init(nat_init_module
);
300 module_exit(nat_cleanup_module
);