Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / 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 const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
35         [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
36 };
37
38 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
39                         struct tc_action *a, int ovr, int bind)
40 {
41         struct nlattr *tb[TCA_NAT_MAX + 1];
42         struct tc_nat *parm;
43         int ret = 0, err;
44         struct tcf_nat *p;
45
46         if (nla == NULL)
47                 return -EINVAL;
48
49         err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy);
50         if (err < 0)
51                 return err;
52
53         if (tb[TCA_NAT_PARMS] == NULL)
54                 return -EINVAL;
55         parm = nla_data(tb[TCA_NAT_PARMS]);
56
57         if (!tcf_hash_check(parm->index, a, bind)) {
58                 ret = tcf_hash_create(parm->index, est, a, sizeof(*p), bind);
59                 if (ret)
60                         return ret;
61                 ret = ACT_P_CREATED;
62         } else {
63                 if (bind)
64                         return 0;
65                 tcf_hash_release(a, bind);
66                 if (!ovr)
67                         return -EEXIST;
68         }
69         p = to_tcf_nat(a);
70
71         spin_lock_bh(&p->tcf_lock);
72         p->old_addr = parm->old_addr;
73         p->new_addr = parm->new_addr;
74         p->mask = parm->mask;
75         p->flags = parm->flags;
76
77         p->tcf_action = parm->action;
78         spin_unlock_bh(&p->tcf_lock);
79
80         if (ret == ACT_P_CREATED)
81                 tcf_hash_insert(a);
82
83         return ret;
84 }
85
86 static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
87                    struct tcf_result *res)
88 {
89         struct tcf_nat *p = a->priv;
90         struct iphdr *iph;
91         __be32 old_addr;
92         __be32 new_addr;
93         __be32 mask;
94         __be32 addr;
95         int egress;
96         int action;
97         int ihl;
98         int noff;
99
100         spin_lock(&p->tcf_lock);
101
102         p->tcf_tm.lastuse = jiffies;
103         old_addr = p->old_addr;
104         new_addr = p->new_addr;
105         mask = p->mask;
106         egress = p->flags & TCA_NAT_FLAG_EGRESS;
107         action = p->tcf_action;
108
109         bstats_update(&p->tcf_bstats, skb);
110
111         spin_unlock(&p->tcf_lock);
112
113         if (unlikely(action == TC_ACT_SHOT))
114                 goto drop;
115
116         noff = skb_network_offset(skb);
117         if (!pskb_may_pull(skb, sizeof(*iph) + noff))
118                 goto drop;
119
120         iph = ip_hdr(skb);
121
122         if (egress)
123                 addr = iph->saddr;
124         else
125                 addr = iph->daddr;
126
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))
131                         goto drop;
132
133                 new_addr &= mask;
134                 new_addr |= addr & ~mask;
135
136                 /* Rewrite IP header */
137                 iph = ip_hdr(skb);
138                 if (egress)
139                         iph->saddr = new_addr;
140                 else
141                         iph->daddr = new_addr;
142
143                 csum_replace4(&iph->check, addr, new_addr);
144         } else if ((iph->frag_off & htons(IP_OFFSET)) ||
145                    iph->protocol != IPPROTO_ICMP) {
146                 goto out;
147         }
148
149         ihl = iph->ihl * 4;
150
151         /* It would be nice to share code with stateful NAT. */
152         switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
153         case IPPROTO_TCP:
154         {
155                 struct tcphdr *tcph;
156
157                 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
158                     (skb_cloned(skb) &&
159                      !skb_clone_writable(skb, ihl + sizeof(*tcph) + noff) &&
160                      pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
161                         goto drop;
162
163                 tcph = (void *)(skb_network_header(skb) + ihl);
164                 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, 1);
165                 break;
166         }
167         case IPPROTO_UDP:
168         {
169                 struct udphdr *udph;
170
171                 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
172                     (skb_cloned(skb) &&
173                      !skb_clone_writable(skb, ihl + sizeof(*udph) + noff) &&
174                      pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
175                         goto drop;
176
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,
180                                                  new_addr, 1);
181                         if (!udph->check)
182                                 udph->check = CSUM_MANGLED_0;
183                 }
184                 break;
185         }
186         case IPPROTO_ICMP:
187         {
188                 struct icmphdr *icmph;
189
190                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
191                         goto drop;
192
193                 icmph = (void *)(skb_network_header(skb) + ihl);
194
195                 if ((icmph->type != ICMP_DEST_UNREACH) &&
196                     (icmph->type != ICMP_TIME_EXCEEDED) &&
197                     (icmph->type != ICMP_PARAMETERPROB))
198                         break;
199
200                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
201                                         noff))
202                         goto drop;
203
204                 icmph = (void *)(skb_network_header(skb) + ihl);
205                 iph = (void *)(icmph + 1);
206                 if (egress)
207                         addr = iph->daddr;
208                 else
209                         addr = iph->saddr;
210
211                 if ((old_addr ^ addr) & mask)
212                         break;
213
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))
218                         goto drop;
219
220                 icmph = (void *)(skb_network_header(skb) + ihl);
221                 iph = (void *)(icmph + 1);
222
223                 new_addr &= mask;
224                 new_addr |= addr & ~mask;
225
226                 /* XXX Fix up the inner checksums. */
227                 if (egress)
228                         iph->daddr = new_addr;
229                 else
230                         iph->saddr = new_addr;
231
232                 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
233                                          0);
234                 break;
235         }
236         default:
237                 break;
238         }
239
240 out:
241         return action;
242
243 drop:
244         spin_lock(&p->tcf_lock);
245         p->tcf_qstats.drops++;
246         spin_unlock(&p->tcf_lock);
247         return TC_ACT_SHOT;
248 }
249
250 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
251                         int bind, int ref)
252 {
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,
258                 .mask     = p->mask,
259                 .flags    = p->flags,
260
261                 .index    = p->tcf_index,
262                 .action   = p->tcf_action,
263                 .refcnt   = p->tcf_refcnt - ref,
264                 .bindcnt  = p->tcf_bindcnt - bind,
265         };
266         struct tcf_t t;
267
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;
275
276         return skb->len;
277
278 nla_put_failure:
279         nlmsg_trim(skb, b);
280         return -1;
281 }
282
283 static struct tc_action_ops act_nat_ops = {
284         .kind           =       "nat",
285         .type           =       TCA_ACT_NAT,
286         .owner          =       THIS_MODULE,
287         .act            =       tcf_nat,
288         .dump           =       tcf_nat_dump,
289         .init           =       tcf_nat_init,
290 };
291
292 MODULE_DESCRIPTION("Stateless NAT actions");
293 MODULE_LICENSE("GPL");
294
295 static int __init nat_init_module(void)
296 {
297         return tcf_register_action(&act_nat_ops, NAT_TAB_MASK);
298 }
299
300 static void __exit nat_cleanup_module(void)
301 {
302         tcf_unregister_action(&act_nat_ops);
303 }
304
305 module_init(nat_init_module);
306 module_exit(nat_cleanup_module);