1
2
3
4
5
6
7
8
9
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
32static int nat_net_id;
33static struct tc_action_ops act_nat_ops;
34
35static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
36 [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
37};
38
39static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
40 struct tc_action **a, int ovr, int bind)
41{
42 struct tc_action_net *tn = net_generic(net, nat_net_id);
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_idr_check(tn, parm->index, a, bind)) {
60 ret = tcf_idr_create(tn, parm->index, est, a,
61 &act_nat_ops, bind, false);
62 if (ret)
63 return ret;
64 ret = ACT_P_CREATED;
65 } else {
66 if (bind)
67 return 0;
68 if (!ovr) {
69 tcf_idr_release(*a, bind);
70 return -EEXIST;
71 }
72 }
73 p = to_tcf_nat(*a);
74
75 spin_lock_bh(&p->tcf_lock);
76 p->old_addr = parm->old_addr;
77 p->new_addr = parm->new_addr;
78 p->mask = parm->mask;
79 p->flags = parm->flags;
80
81 p->tcf_action = parm->action;
82 spin_unlock_bh(&p->tcf_lock);
83
84 if (ret == ACT_P_CREATED)
85 tcf_idr_insert(tn, *a);
86
87 return ret;
88}
89
90static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
91 struct tcf_result *res)
92{
93 struct tcf_nat *p = to_tcf_nat(a);
94 struct iphdr *iph;
95 __be32 old_addr;
96 __be32 new_addr;
97 __be32 mask;
98 __be32 addr;
99 int egress;
100 int action;
101 int ihl;
102 int noff;
103
104 spin_lock(&p->tcf_lock);
105
106 tcf_lastuse_update(&p->tcf_tm);
107 old_addr = p->old_addr;
108 new_addr = p->new_addr;
109 mask = p->mask;
110 egress = p->flags & TCA_NAT_FLAG_EGRESS;
111 action = p->tcf_action;
112
113 bstats_update(&p->tcf_bstats, skb);
114
115 spin_unlock(&p->tcf_lock);
116
117 if (unlikely(action == TC_ACT_SHOT))
118 goto drop;
119
120 noff = skb_network_offset(skb);
121 if (!pskb_may_pull(skb, sizeof(*iph) + noff))
122 goto drop;
123
124 iph = ip_hdr(skb);
125
126 if (egress)
127 addr = iph->saddr;
128 else
129 addr = iph->daddr;
130
131 if (!((old_addr ^ addr) & mask)) {
132 if (skb_try_make_writable(skb, sizeof(*iph) + noff))
133 goto drop;
134
135 new_addr &= mask;
136 new_addr |= addr & ~mask;
137
138
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
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_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
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_try_make_writable(skb, ihl + sizeof(*udph) + noff))
173 goto drop;
174
175 udph = (void *)(skb_network_header(skb) + ihl);
176 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
177 inet_proto_csum_replace4(&udph->check, skb, addr,
178 new_addr, 1);
179 if (!udph->check)
180 udph->check = CSUM_MANGLED_0;
181 }
182 break;
183 }
184 case IPPROTO_ICMP:
185 {
186 struct icmphdr *icmph;
187
188 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
189 goto drop;
190
191 icmph = (void *)(skb_network_header(skb) + ihl);
192
193 if ((icmph->type != ICMP_DEST_UNREACH) &&
194 (icmph->type != ICMP_TIME_EXCEEDED) &&
195 (icmph->type != ICMP_PARAMETERPROB))
196 break;
197
198 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
199 noff))
200 goto drop;
201
202 icmph = (void *)(skb_network_header(skb) + ihl);
203 iph = (void *)(icmph + 1);
204 if (egress)
205 addr = iph->daddr;
206 else
207 addr = iph->saddr;
208
209 if ((old_addr ^ addr) & mask)
210 break;
211
212 if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
213 sizeof(*iph) + noff))
214 goto drop;
215
216 icmph = (void *)(skb_network_header(skb) + ihl);
217 iph = (void *)(icmph + 1);
218
219 new_addr &= mask;
220 new_addr |= addr & ~mask;
221
222
223 if (egress)
224 iph->daddr = new_addr;
225 else
226 iph->saddr = new_addr;
227
228 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
229 0);
230 break;
231 }
232 default:
233 break;
234 }
235
236out:
237 return action;
238
239drop:
240 spin_lock(&p->tcf_lock);
241 p->tcf_qstats.drops++;
242 spin_unlock(&p->tcf_lock);
243 return TC_ACT_SHOT;
244}
245
246static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
247 int bind, int ref)
248{
249 unsigned char *b = skb_tail_pointer(skb);
250 struct tcf_nat *p = to_tcf_nat(a);
251 struct tc_nat opt = {
252 .old_addr = p->old_addr,
253 .new_addr = p->new_addr,
254 .mask = p->mask,
255 .flags = p->flags,
256
257 .index = p->tcf_index,
258 .action = p->tcf_action,
259 .refcnt = p->tcf_refcnt - ref,
260 .bindcnt = p->tcf_bindcnt - bind,
261 };
262 struct tcf_t t;
263
264 if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
265 goto nla_put_failure;
266
267 tcf_tm_dump(&t, &p->tcf_tm);
268 if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
269 goto nla_put_failure;
270
271 return skb->len;
272
273nla_put_failure:
274 nlmsg_trim(skb, b);
275 return -1;
276}
277
278static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
279 struct netlink_callback *cb, int type,
280 const struct tc_action_ops *ops)
281{
282 struct tc_action_net *tn = net_generic(net, nat_net_id);
283
284 return tcf_generic_walker(tn, skb, cb, type, ops);
285}
286
287static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
288{
289 struct tc_action_net *tn = net_generic(net, nat_net_id);
290
291 return tcf_idr_search(tn, a, index);
292}
293
294static struct tc_action_ops act_nat_ops = {
295 .kind = "nat",
296 .type = TCA_ACT_NAT,
297 .owner = THIS_MODULE,
298 .act = tcf_nat,
299 .dump = tcf_nat_dump,
300 .init = tcf_nat_init,
301 .walk = tcf_nat_walker,
302 .lookup = tcf_nat_search,
303 .size = sizeof(struct tcf_nat),
304};
305
306static __net_init int nat_init_net(struct net *net)
307{
308 struct tc_action_net *tn = net_generic(net, nat_net_id);
309
310 return tc_action_net_init(tn, &act_nat_ops);
311}
312
313static void __net_exit nat_exit_net(struct net *net)
314{
315 struct tc_action_net *tn = net_generic(net, nat_net_id);
316
317 tc_action_net_exit(tn);
318}
319
320static struct pernet_operations nat_net_ops = {
321 .init = nat_init_net,
322 .exit = nat_exit_net,
323 .id = &nat_net_id,
324 .size = sizeof(struct tc_action_net),
325};
326
327MODULE_DESCRIPTION("Stateless NAT actions");
328MODULE_LICENSE("GPL");
329
330static int __init nat_init_module(void)
331{
332 return tcf_register_action(&act_nat_ops, &nat_net_ops);
333}
334
335static void __exit nat_cleanup_module(void)
336{
337 tcf_unregister_action(&act_nat_ops, &nat_net_ops);
338}
339
340module_init(nat_init_module);
341module_exit(nat_cleanup_module);
342
