9a3cb1d16d19a5e458e27b86b2fb72102471a687
[deliverable/linux.git] / net / sched / act_nat.c
1 /*
2 * Stateless NAT actions
3 *
4 * Copyright (c) 2007 Herbert Xu <herbert@gondor.apana.org.au>
5 *
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)
9 * any later version.
10 */
11
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>
24 #include <net/icmp.h>
25 #include <net/ip.h>
26 #include <net/netlink.h>
27 #include <net/tc_act/tc_nat.h>
28 #include <net/tcp.h>
29 #include <net/udp.h>
30
31
32 #define NAT_TAB_MASK 15
33
34 static struct tcf_hashinfo nat_hash_info;
35
36 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
37 [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
38 };
39
40 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
41 struct tc_action *a, int ovr, int bind)
42 {
43 struct nlattr *tb[TCA_NAT_MAX + 1];
44 struct tc_nat *parm;
45 int ret = 0, err;
46 struct tcf_nat *p;
47
48 if (nla == NULL)
49 return -EINVAL;
50
51 err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy);
52 if (err < 0)
53 return err;
54
55 if (tb[TCA_NAT_PARMS] == NULL)
56 return -EINVAL;
57 parm = nla_data(tb[TCA_NAT_PARMS]);
58
59 if (!tcf_hash_check(parm->index, a, bind)) {
60 ret = tcf_hash_create(parm->index, est, a, sizeof(*p), bind);
61 if (ret)
62 return ret;
63 ret = ACT_P_CREATED;
64 } else {
65 if (bind)
66 return 0;
67 tcf_hash_release(a, bind);
68 if (!ovr)
69 return -EEXIST;
70 }
71 p = to_tcf_nat(a);
72
73 spin_lock_bh(&p->tcf_lock);
74 p->old_addr = parm->old_addr;
75 p->new_addr = parm->new_addr;
76 p->mask = parm->mask;
77 p->flags = parm->flags;
78
79 p->tcf_action = parm->action;
80 spin_unlock_bh(&p->tcf_lock);
81
82 if (ret == ACT_P_CREATED)
83 tcf_hash_insert(a);
84
85 return ret;
86 }
87
88 static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
89 struct tcf_result *res)
90 {
91 struct tcf_nat *p = a->priv;
92 struct iphdr *iph;
93 __be32 old_addr;
94 __be32 new_addr;
95 __be32 mask;
96 __be32 addr;
97 int egress;
98 int action;
99 int ihl;
100 int noff;
101
102 spin_lock(&p->tcf_lock);
103
104 p->tcf_tm.lastuse = jiffies;
105 old_addr = p->old_addr;
106 new_addr = p->new_addr;
107 mask = p->mask;
108 egress = p->flags & TCA_NAT_FLAG_EGRESS;
109 action = p->tcf_action;
110
111 bstats_update(&p->tcf_bstats, skb);
112
113 spin_unlock(&p->tcf_lock);
114
115 if (unlikely(action == TC_ACT_SHOT))
116 goto drop;
117
118 noff = skb_network_offset(skb);
119 if (!pskb_may_pull(skb, sizeof(*iph) + noff))
120 goto drop;
121
122 iph = ip_hdr(skb);
123
124 if (egress)
125 addr = iph->saddr;
126 else
127 addr = iph->daddr;
128
129 if (!((old_addr ^ addr) & mask)) {
130 if (skb_cloned(skb) &&
131 !skb_clone_writable(skb, sizeof(*iph) + noff) &&
132 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
133 goto drop;
134
135 new_addr &= mask;
136 new_addr |= addr & ~mask;
137
138 /* Rewrite IP header */
139 iph = ip_hdr(skb);
140 if (egress)
141 iph->saddr = new_addr;
142 else
143 iph->daddr = new_addr;
144
145 csum_replace4(&iph->check, addr, new_addr);
146 } else if ((iph->frag_off & htons(IP_OFFSET)) ||
147 iph->protocol != IPPROTO_ICMP) {
148 goto out;
149 }
150
151 ihl = iph->ihl * 4;
152
153 /* It would be nice to share code with stateful NAT. */
154 switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
155 case IPPROTO_TCP:
156 {
157 struct tcphdr *tcph;
158
159 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
160 (skb_cloned(skb) &&
161 !skb_clone_writable(skb, ihl + sizeof(*tcph) + noff) &&
162 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
163 goto drop;
164
165 tcph = (void *)(skb_network_header(skb) + ihl);
166 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, 1);
167 break;
168 }
169 case IPPROTO_UDP:
170 {
171 struct udphdr *udph;
172
173 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
174 (skb_cloned(skb) &&
175 !skb_clone_writable(skb, ihl + sizeof(*udph) + noff) &&
176 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
177 goto drop;
178
179 udph = (void *)(skb_network_header(skb) + ihl);
180 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
181 inet_proto_csum_replace4(&udph->check, skb, addr,
182 new_addr, 1);
183 if (!udph->check)
184 udph->check = CSUM_MANGLED_0;
185 }
186 break;
187 }
188 case IPPROTO_ICMP:
189 {
190 struct icmphdr *icmph;
191
192 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
193 goto drop;
194
195 icmph = (void *)(skb_network_header(skb) + ihl);
196
197 if ((icmph->type != ICMP_DEST_UNREACH) &&
198 (icmph->type != ICMP_TIME_EXCEEDED) &&
199 (icmph->type != ICMP_PARAMETERPROB))
200 break;
201
202 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
203 noff))
204 goto drop;
205
206 icmph = (void *)(skb_network_header(skb) + ihl);
207 iph = (void *)(icmph + 1);
208 if (egress)
209 addr = iph->daddr;
210 else
211 addr = iph->saddr;
212
213 if ((old_addr ^ addr) & mask)
214 break;
215
216 if (skb_cloned(skb) &&
217 !skb_clone_writable(skb, ihl + sizeof(*icmph) +
218 sizeof(*iph) + noff) &&
219 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
220 goto drop;
221
222 icmph = (void *)(skb_network_header(skb) + ihl);
223 iph = (void *)(icmph + 1);
224
225 new_addr &= mask;
226 new_addr |= addr & ~mask;
227
228 /* XXX Fix up the inner checksums. */
229 if (egress)
230 iph->daddr = new_addr;
231 else
232 iph->saddr = new_addr;
233
234 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
235 0);
236 break;
237 }
238 default:
239 break;
240 }
241
242 out:
243 return action;
244
245 drop:
246 spin_lock(&p->tcf_lock);
247 p->tcf_qstats.drops++;
248 spin_unlock(&p->tcf_lock);
249 return TC_ACT_SHOT;
250 }
251
252 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
253 int bind, int ref)
254 {
255 unsigned char *b = skb_tail_pointer(skb);
256 struct tcf_nat *p = a->priv;
257 struct tc_nat opt = {
258 .old_addr = p->old_addr,
259 .new_addr = p->new_addr,
260 .mask = p->mask,
261 .flags = p->flags,
262
263 .index = p->tcf_index,
264 .action = p->tcf_action,
265 .refcnt = p->tcf_refcnt - ref,
266 .bindcnt = p->tcf_bindcnt - bind,
267 };
268 struct tcf_t t;
269
270 if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
271 goto nla_put_failure;
272 t.install = jiffies_to_clock_t(jiffies - p->tcf_tm.install);
273 t.lastuse = jiffies_to_clock_t(jiffies - p->tcf_tm.lastuse);
274 t.expires = jiffies_to_clock_t(p->tcf_tm.expires);
275 if (nla_put(skb, TCA_NAT_TM, sizeof(t), &t))
276 goto nla_put_failure;
277
278 return skb->len;
279
280 nla_put_failure:
281 nlmsg_trim(skb, b);
282 return -1;
283 }
284
285 static struct tc_action_ops act_nat_ops = {
286 .kind = "nat",
287 .hinfo = &nat_hash_info,
288 .type = TCA_ACT_NAT,
289 .owner = THIS_MODULE,
290 .act = tcf_nat,
291 .dump = tcf_nat_dump,
292 .init = tcf_nat_init,
293 };
294
295 MODULE_DESCRIPTION("Stateless NAT actions");
296 MODULE_LICENSE("GPL");
297
298 static int __init nat_init_module(void)
299 {
300 int err = tcf_hashinfo_init(&nat_hash_info, NAT_TAB_MASK);
301 if (err)
302 return err;
303 return tcf_register_action(&act_nat_ops);
304 }
305
306 static void __exit nat_cleanup_module(void)
307 {
308 tcf_unregister_action(&act_nat_ops);
309 tcf_hashinfo_destroy(&nat_hash_info);
310 }
311
312 module_init(nat_init_module);
313 module_exit(nat_cleanup_module);
This page took 0.037787 seconds and 4 git commands to generate.