Merge branch 'x86-xen-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6] / 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 static struct tcf_common *tcf_nat_ht[NAT_TAB_MASK + 1];
34 static u32 nat_idx_gen;
35 static DEFINE_RWLOCK(nat_lock);
36
37 static struct tcf_hashinfo nat_hash_info = {
38         .htab   =       tcf_nat_ht,
39         .hmask  =       NAT_TAB_MASK,
40         .lock   =       &nat_lock,
41 };
42
43 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
44         [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
45 };
46
47 static int tcf_nat_init(struct nlattr *nla, struct nlattr *est,
48                         struct tc_action *a, int ovr, int bind)
49 {
50         struct nlattr *tb[TCA_NAT_MAX + 1];
51         struct tc_nat *parm;
52         int ret = 0, err;
53         struct tcf_nat *p;
54         struct tcf_common *pc;
55
56         if (nla == NULL)
57                 return -EINVAL;
58
59         err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy);
60         if (err < 0)
61                 return err;
62
63         if (tb[TCA_NAT_PARMS] == NULL)
64                 return -EINVAL;
65         parm = nla_data(tb[TCA_NAT_PARMS]);
66
67         pc = tcf_hash_check(parm->index, a, bind, &nat_hash_info);
68         if (!pc) {
69                 pc = tcf_hash_create(parm->index, est, a, sizeof(*p), bind,
70                                      &nat_idx_gen, &nat_hash_info);
71                 if (IS_ERR(pc))
72                     return PTR_ERR(pc);
73                 p = to_tcf_nat(pc);
74                 ret = ACT_P_CREATED;
75         } else {
76                 p = to_tcf_nat(pc);
77                 if (!ovr) {
78                         tcf_hash_release(pc, bind, &nat_hash_info);
79                         return -EEXIST;
80                 }
81         }
82
83         spin_lock_bh(&p->tcf_lock);
84         p->old_addr = parm->old_addr;
85         p->new_addr = parm->new_addr;
86         p->mask = parm->mask;
87         p->flags = parm->flags;
88
89         p->tcf_action = parm->action;
90         spin_unlock_bh(&p->tcf_lock);
91
92         if (ret == ACT_P_CREATED)
93                 tcf_hash_insert(pc, &nat_hash_info);
94
95         return ret;
96 }
97
98 static int tcf_nat_cleanup(struct tc_action *a, int bind)
99 {
100         struct tcf_nat *p = a->priv;
101
102         return tcf_hash_release(&p->common, bind, &nat_hash_info);
103 }
104
105 static int tcf_nat(struct sk_buff *skb, struct tc_action *a,
106                    struct tcf_result *res)
107 {
108         struct tcf_nat *p = a->priv;
109         struct iphdr *iph;
110         __be32 old_addr;
111         __be32 new_addr;
112         __be32 mask;
113         __be32 addr;
114         int egress;
115         int action;
116         int ihl;
117
118         spin_lock(&p->tcf_lock);
119
120         p->tcf_tm.lastuse = jiffies;
121         old_addr = p->old_addr;
122         new_addr = p->new_addr;
123         mask = p->mask;
124         egress = p->flags & TCA_NAT_FLAG_EGRESS;
125         action = p->tcf_action;
126
127         p->tcf_bstats.bytes += qdisc_pkt_len(skb);
128         p->tcf_bstats.packets++;
129
130         spin_unlock(&p->tcf_lock);
131
132         if (unlikely(action == TC_ACT_SHOT))
133                 goto drop;
134
135         if (!pskb_may_pull(skb, sizeof(*iph)))
136                 goto drop;
137
138         iph = ip_hdr(skb);
139
140         if (egress)
141                 addr = iph->saddr;
142         else
143                 addr = iph->daddr;
144
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))
149                         goto drop;
150
151                 new_addr &= mask;
152                 new_addr |= addr & ~mask;
153
154                 /* Rewrite IP header */
155                 iph = ip_hdr(skb);
156                 if (egress)
157                         iph->saddr = new_addr;
158                 else
159                         iph->daddr = new_addr;
160
161                 csum_replace4(&iph->check, addr, new_addr);
162         }
163
164         ihl = iph->ihl * 4;
165
166         /* It would be nice to share code with stateful NAT. */
167         switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
168         case IPPROTO_TCP:
169         {
170                 struct tcphdr *tcph;
171
172                 if (!pskb_may_pull(skb, ihl + sizeof(*tcph)) ||
173                     (skb_cloned(skb) &&
174                      !skb_clone_writable(skb, ihl + sizeof(*tcph)) &&
175                      pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
176                         goto drop;
177
178                 tcph = (void *)(skb_network_header(skb) + ihl);
179                 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, 1);
180                 break;
181         }
182         case IPPROTO_UDP:
183         {
184                 struct udphdr *udph;
185
186                 if (!pskb_may_pull(skb, ihl + sizeof(*udph)) ||
187                     (skb_cloned(skb) &&
188                      !skb_clone_writable(skb, ihl + sizeof(*udph)) &&
189                      pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
190                         goto drop;
191
192                 udph = (void *)(skb_network_header(skb) + ihl);
193                 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
194                         inet_proto_csum_replace4(&udph->check, skb, addr,
195                                                  new_addr, 1);
196                         if (!udph->check)
197                                 udph->check = CSUM_MANGLED_0;
198                 }
199                 break;
200         }
201         case IPPROTO_ICMP:
202         {
203                 struct icmphdr *icmph;
204
205                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph)))
206                         goto drop;
207
208                 icmph = (void *)(skb_network_header(skb) + ihl);
209
210                 if ((icmph->type != ICMP_DEST_UNREACH) &&
211                     (icmph->type != ICMP_TIME_EXCEEDED) &&
212                     (icmph->type != ICMP_PARAMETERPROB))
213                         break;
214
215                 iph = (void *)(icmph + 1);
216                 if (egress)
217                         addr = iph->daddr;
218                 else
219                         addr = iph->saddr;
220
221                 if ((old_addr ^ addr) & mask)
222                         break;
223
224                 if (skb_cloned(skb) &&
225                     !skb_clone_writable(skb,
226                                         ihl + sizeof(*icmph) + sizeof(*iph)) &&
227                     pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
228                         goto drop;
229
230                 icmph = (void *)(skb_network_header(skb) + ihl);
231                 iph = (void *)(icmph + 1);
232
233                 new_addr &= mask;
234                 new_addr |= addr & ~mask;
235
236                 /* XXX Fix up the inner checksums. */
237                 if (egress)
238                         iph->daddr = new_addr;
239                 else
240                         iph->saddr = new_addr;
241
242                 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
243                                          1);
244                 break;
245         }
246         default:
247                 break;
248         }
249
250         return action;
251
252 drop:
253         spin_lock(&p->tcf_lock);
254         p->tcf_qstats.drops++;
255         spin_unlock(&p->tcf_lock);
256         return TC_ACT_SHOT;
257 }
258
259 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
260                         int bind, int ref)
261 {
262         unsigned char *b = skb_tail_pointer(skb);
263         struct tcf_nat *p = a->priv;
264         struct tc_nat *opt;
265         struct tcf_t t;
266         int s;
267
268         s = sizeof(*opt);
269
270         /* netlink spinlocks held above us - must use ATOMIC */
271         opt = kzalloc(s, GFP_ATOMIC);
272         if (unlikely(!opt))
273                 return -ENOBUFS;
274
275         opt->old_addr = p->old_addr;
276         opt->new_addr = p->new_addr;
277         opt->mask = p->mask;
278         opt->flags = p->flags;
279
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;
284
285         NLA_PUT(skb, TCA_NAT_PARMS, s, 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);
290
291         kfree(opt);
292
293         return skb->len;
294
295 nla_put_failure:
296         nlmsg_trim(skb, b);
297         kfree(opt);
298         return -1;
299 }
300
301 static struct tc_action_ops act_nat_ops = {
302         .kind           =       "nat",
303         .hinfo          =       &nat_hash_info,
304         .type           =       TCA_ACT_NAT,
305         .capab          =       TCA_CAP_NONE,
306         .owner          =       THIS_MODULE,
307         .act            =       tcf_nat,
308         .dump           =       tcf_nat_dump,
309         .cleanup        =       tcf_nat_cleanup,
310         .lookup         =       tcf_hash_search,
311         .init           =       tcf_nat_init,
312         .walk           =       tcf_generic_walker
313 };
314
315 MODULE_DESCRIPTION("Stateless NAT actions");
316 MODULE_LICENSE("GPL");
317
318 static int __init nat_init_module(void)
319 {
320         return tcf_register_action(&act_nat_ops);
321 }
322
323 static void __exit nat_cleanup_module(void)
324 {
325         tcf_unregister_action(&act_nat_ops);
326 }
327
328 module_init(nat_init_module);
329 module_exit(nat_cleanup_module);