1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Stateless NAT actions
4 *
5 * Copyright (c) 2007 Herbert Xu <herbert@gondor.apana.org.au>
6 */
7
8 #include <linux/errno.h>
9 #include <linux/init.h>
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/netfilter.h>
13 #include <linux/rtnetlink.h>
14 #include <linux/skbuff.h>
15 #include <linux/slab.h>
16 #include <linux/spinlock.h>
17 #include <linux/string.h>
18 #include <linux/tc_act/tc_nat.h>
19 #include <net/act_api.h>
20 #include <net/pkt_cls.h>
21 #include <net/icmp.h>
22 #include <net/ip.h>
23 #include <net/netlink.h>
24 #include <net/tc_act/tc_nat.h>
25 #include <net/tcp.h>
26 #include <net/udp.h>
27
28
29 static unsigned int nat_net_id;
30 static struct tc_action_ops act_nat_ops;
31
32 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
33 [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
34 };
35
tcf_nat_init(struct net * net,struct nlattr * nla,struct nlattr * est,struct tc_action ** a,int ovr,int bind,bool rtnl_held,struct tcf_proto * tp,struct netlink_ext_ack * extack)36 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
37 struct tc_action **a, int ovr, int bind,
38 bool rtnl_held, struct tcf_proto *tp,
39 struct netlink_ext_ack *extack)
40 {
41 struct tc_action_net *tn = net_generic(net, nat_net_id);
42 struct nlattr *tb[TCA_NAT_MAX + 1];
43 struct tcf_chain *goto_ch = NULL;
44 struct tc_nat *parm;
45 int ret = 0, err;
46 struct tcf_nat *p;
47 u32 index;
48
49 if (nla == NULL)
50 return -EINVAL;
51
52 err = nla_parse_nested_deprecated(tb, TCA_NAT_MAX, nla, nat_policy,
53 NULL);
54 if (err < 0)
55 return err;
56
57 if (tb[TCA_NAT_PARMS] == NULL)
58 return -EINVAL;
59 parm = nla_data(tb[TCA_NAT_PARMS]);
60 index = parm->index;
61 err = tcf_idr_check_alloc(tn, &index, a, bind);
62 if (!err) {
63 ret = tcf_idr_create(tn, index, est, a,
64 &act_nat_ops, bind, false);
65 if (ret) {
66 tcf_idr_cleanup(tn, index);
67 return ret;
68 }
69 ret = ACT_P_CREATED;
70 } else if (err > 0) {
71 if (bind)
72 return 0;
73 if (!ovr) {
74 tcf_idr_release(*a, bind);
75 return -EEXIST;
76 }
77 } else {
78 return err;
79 }
80 err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
81 if (err < 0)
82 goto release_idr;
83 p = to_tcf_nat(*a);
84
85 spin_lock_bh(&p->tcf_lock);
86 p->old_addr = parm->old_addr;
87 p->new_addr = parm->new_addr;
88 p->mask = parm->mask;
89 p->flags = parm->flags;
90
91 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
92 spin_unlock_bh(&p->tcf_lock);
93 if (goto_ch)
94 tcf_chain_put_by_act(goto_ch);
95
96 return ret;
97 release_idr:
98 tcf_idr_release(*a, bind);
99 return err;
100 }
101
tcf_nat_act(struct sk_buff * skb,const struct tc_action * a,struct tcf_result * res)102 static int tcf_nat_act(struct sk_buff *skb, const struct tc_action *a,
103 struct tcf_result *res)
104 {
105 struct tcf_nat *p = to_tcf_nat(a);
106 struct iphdr *iph;
107 __be32 old_addr;
108 __be32 new_addr;
109 __be32 mask;
110 __be32 addr;
111 int egress;
112 int action;
113 int ihl;
114 int noff;
115
116 spin_lock(&p->tcf_lock);
117
118 tcf_lastuse_update(&p->tcf_tm);
119 old_addr = p->old_addr;
120 new_addr = p->new_addr;
121 mask = p->mask;
122 egress = p->flags & TCA_NAT_FLAG_EGRESS;
123 action = p->tcf_action;
124
125 bstats_update(&p->tcf_bstats, skb);
126
127 spin_unlock(&p->tcf_lock);
128
129 if (unlikely(action == TC_ACT_SHOT))
130 goto drop;
131
132 noff = skb_network_offset(skb);
133 if (!pskb_may_pull(skb, sizeof(*iph) + noff))
134 goto drop;
135
136 iph = ip_hdr(skb);
137
138 if (egress)
139 addr = iph->saddr;
140 else
141 addr = iph->daddr;
142
143 if (!((old_addr ^ addr) & mask)) {
144 if (skb_try_make_writable(skb, sizeof(*iph) + noff))
145 goto drop;
146
147 new_addr &= mask;
148 new_addr |= addr & ~mask;
149
150 /* Rewrite IP header */
151 iph = ip_hdr(skb);
152 if (egress)
153 iph->saddr = new_addr;
154 else
155 iph->daddr = new_addr;
156
157 csum_replace4(&iph->check, addr, new_addr);
158 } else if ((iph->frag_off & htons(IP_OFFSET)) ||
159 iph->protocol != IPPROTO_ICMP) {
160 goto out;
161 }
162
163 ihl = iph->ihl * 4;
164
165 /* It would be nice to share code with stateful NAT. */
166 switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
167 case IPPROTO_TCP:
168 {
169 struct tcphdr *tcph;
170
171 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
172 skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
173 goto drop;
174
175 tcph = (void *)(skb_network_header(skb) + ihl);
176 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
177 true);
178 break;
179 }
180 case IPPROTO_UDP:
181 {
182 struct udphdr *udph;
183
184 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
185 skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
186 goto drop;
187
188 udph = (void *)(skb_network_header(skb) + ihl);
189 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
190 inet_proto_csum_replace4(&udph->check, skb, addr,
191 new_addr, true);
192 if (!udph->check)
193 udph->check = CSUM_MANGLED_0;
194 }
195 break;
196 }
197 case IPPROTO_ICMP:
198 {
199 struct icmphdr *icmph;
200
201 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
202 goto drop;
203
204 icmph = (void *)(skb_network_header(skb) + ihl);
205
206 if ((icmph->type != ICMP_DEST_UNREACH) &&
207 (icmph->type != ICMP_TIME_EXCEEDED) &&
208 (icmph->type != ICMP_PARAMETERPROB))
209 break;
210
211 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
212 noff))
213 goto drop;
214
215 icmph = (void *)(skb_network_header(skb) + ihl);
216 iph = (void *)(icmph + 1);
217 if (egress)
218 addr = iph->daddr;
219 else
220 addr = iph->saddr;
221
222 if ((old_addr ^ addr) & mask)
223 break;
224
225 if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
226 sizeof(*iph) + noff))
227 goto drop;
228
229 icmph = (void *)(skb_network_header(skb) + ihl);
230 iph = (void *)(icmph + 1);
231
232 new_addr &= mask;
233 new_addr |= addr & ~mask;
234
235 /* XXX Fix up the inner checksums. */
236 if (egress)
237 iph->daddr = new_addr;
238 else
239 iph->saddr = new_addr;
240
241 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
242 false);
243 break;
244 }
245 default:
246 break;
247 }
248
249 out:
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
tcf_nat_dump(struct sk_buff * skb,struct tc_action * a,int bind,int ref)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 = to_tcf_nat(a);
264 struct tc_nat opt = {
265 .index = p->tcf_index,
266 .refcnt = refcount_read(&p->tcf_refcnt) - ref,
267 .bindcnt = atomic_read(&p->tcf_bindcnt) - bind,
268 };
269 struct tcf_t t;
270
271 spin_lock_bh(&p->tcf_lock);
272 opt.old_addr = p->old_addr;
273 opt.new_addr = p->new_addr;
274 opt.mask = p->mask;
275 opt.flags = p->flags;
276 opt.action = p->tcf_action;
277
278 if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
279 goto nla_put_failure;
280
281 tcf_tm_dump(&t, &p->tcf_tm);
282 if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
283 goto nla_put_failure;
284 spin_unlock_bh(&p->tcf_lock);
285
286 return skb->len;
287
288 nla_put_failure:
289 spin_unlock_bh(&p->tcf_lock);
290 nlmsg_trim(skb, b);
291 return -1;
292 }
293
tcf_nat_walker(struct net * net,struct sk_buff * skb,struct netlink_callback * cb,int type,const struct tc_action_ops * ops,struct netlink_ext_ack * extack)294 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
295 struct netlink_callback *cb, int type,
296 const struct tc_action_ops *ops,
297 struct netlink_ext_ack *extack)
298 {
299 struct tc_action_net *tn = net_generic(net, nat_net_id);
300
301 return tcf_generic_walker(tn, skb, cb, type, ops, extack);
302 }
303
tcf_nat_search(struct net * net,struct tc_action ** a,u32 index)304 static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
305 {
306 struct tc_action_net *tn = net_generic(net, nat_net_id);
307
308 return tcf_idr_search(tn, a, index);
309 }
310
311 static struct tc_action_ops act_nat_ops = {
312 .kind = "nat",
313 .id = TCA_ID_NAT,
314 .owner = THIS_MODULE,
315 .act = tcf_nat_act,
316 .dump = tcf_nat_dump,
317 .init = tcf_nat_init,
318 .walk = tcf_nat_walker,
319 .lookup = tcf_nat_search,
320 .size = sizeof(struct tcf_nat),
321 };
322
nat_init_net(struct net * net)323 static __net_init int nat_init_net(struct net *net)
324 {
325 struct tc_action_net *tn = net_generic(net, nat_net_id);
326
327 return tc_action_net_init(net, tn, &act_nat_ops);
328 }
329
nat_exit_net(struct list_head * net_list)330 static void __net_exit nat_exit_net(struct list_head *net_list)
331 {
332 tc_action_net_exit(net_list, nat_net_id);
333 }
334
335 static struct pernet_operations nat_net_ops = {
336 .init = nat_init_net,
337 .exit_batch = nat_exit_net,
338 .id = &nat_net_id,
339 .size = sizeof(struct tc_action_net),
340 };
341
342 MODULE_DESCRIPTION("Stateless NAT actions");
343 MODULE_LICENSE("GPL");
344
nat_init_module(void)345 static int __init nat_init_module(void)
346 {
347 return tcf_register_action(&act_nat_ops, &nat_net_ops);
348 }
349
nat_cleanup_module(void)350 static void __exit nat_cleanup_module(void)
351 {
352 tcf_unregister_action(&act_nat_ops, &nat_net_ops);
353 }
354
355 module_init(nat_init_module);
356 module_exit(nat_cleanup_module);
357