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 #include <net/tc_wrapper.h>
28
29 static struct tc_action_ops act_nat_ops;
30
31 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
32 [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
33 };
34
tcf_nat_init(struct net * net,struct nlattr * nla,struct nlattr * est,struct tc_action ** a,struct tcf_proto * tp,u32 flags,struct netlink_ext_ack * extack)35 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
36 struct tc_action **a, struct tcf_proto *tp,
37 u32 flags, struct netlink_ext_ack *extack)
38 {
39 struct tc_action_net *tn = net_generic(net, act_nat_ops.net_id);
40 bool bind = flags & TCA_ACT_FLAGS_BIND;
41 struct tcf_nat_parms *nparm, *oparm;
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_from_flags(tn, index, est, a, &act_nat_ops,
64 bind, flags);
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 (!(flags & TCA_ACT_FLAGS_REPLACE)) {
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
84 nparm = kzalloc(sizeof(*nparm), GFP_KERNEL);
85 if (!nparm) {
86 err = -ENOMEM;
87 goto release_idr;
88 }
89
90 nparm->old_addr = parm->old_addr;
91 nparm->new_addr = parm->new_addr;
92 nparm->mask = parm->mask;
93 nparm->flags = parm->flags;
94
95 p = to_tcf_nat(*a);
96
97 spin_lock_bh(&p->tcf_lock);
98 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
99 oparm = rcu_replace_pointer(p->parms, nparm, lockdep_is_held(&p->tcf_lock));
100 spin_unlock_bh(&p->tcf_lock);
101
102 if (goto_ch)
103 tcf_chain_put_by_act(goto_ch);
104
105 if (oparm)
106 kfree_rcu(oparm, rcu);
107
108 return ret;
109 release_idr:
110 tcf_idr_release(*a, bind);
111 return err;
112 }
113
tcf_nat_act(struct sk_buff * skb,const struct tc_action * a,struct tcf_result * res)114 TC_INDIRECT_SCOPE int tcf_nat_act(struct sk_buff *skb,
115 const struct tc_action *a,
116 struct tcf_result *res)
117 {
118 struct tcf_nat *p = to_tcf_nat(a);
119 struct tcf_nat_parms *parms;
120 struct iphdr *iph;
121 __be32 old_addr;
122 __be32 new_addr;
123 __be32 mask;
124 __be32 addr;
125 int egress;
126 int action;
127 int ihl;
128 int noff;
129
130 tcf_lastuse_update(&p->tcf_tm);
131 tcf_action_update_bstats(&p->common, skb);
132
133 action = READ_ONCE(p->tcf_action);
134
135 parms = rcu_dereference_bh(p->parms);
136 old_addr = parms->old_addr;
137 new_addr = parms->new_addr;
138 mask = parms->mask;
139 egress = parms->flags & TCA_NAT_FLAG_EGRESS;
140
141 if (unlikely(action == TC_ACT_SHOT))
142 goto drop;
143
144 noff = skb_network_offset(skb);
145 if (!pskb_may_pull(skb, sizeof(*iph) + noff))
146 goto drop;
147
148 iph = ip_hdr(skb);
149
150 if (egress)
151 addr = iph->saddr;
152 else
153 addr = iph->daddr;
154
155 if (!((old_addr ^ addr) & mask)) {
156 if (skb_try_make_writable(skb, sizeof(*iph) + noff))
157 goto drop;
158
159 new_addr &= mask;
160 new_addr |= addr & ~mask;
161
162 /* Rewrite IP header */
163 iph = ip_hdr(skb);
164 if (egress)
165 iph->saddr = new_addr;
166 else
167 iph->daddr = new_addr;
168
169 csum_replace4(&iph->check, addr, new_addr);
170 } else if ((iph->frag_off & htons(IP_OFFSET)) ||
171 iph->protocol != IPPROTO_ICMP) {
172 goto out;
173 }
174
175 ihl = iph->ihl * 4;
176
177 /* It would be nice to share code with stateful NAT. */
178 switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
179 case IPPROTO_TCP:
180 {
181 struct tcphdr *tcph;
182
183 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
184 skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
185 goto drop;
186
187 tcph = (void *)(skb_network_header(skb) + ihl);
188 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
189 true);
190 break;
191 }
192 case IPPROTO_UDP:
193 {
194 struct udphdr *udph;
195
196 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
197 skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
198 goto drop;
199
200 udph = (void *)(skb_network_header(skb) + ihl);
201 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
202 inet_proto_csum_replace4(&udph->check, skb, addr,
203 new_addr, true);
204 if (!udph->check)
205 udph->check = CSUM_MANGLED_0;
206 }
207 break;
208 }
209 case IPPROTO_ICMP:
210 {
211 struct icmphdr *icmph;
212
213 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
214 goto drop;
215
216 icmph = (void *)(skb_network_header(skb) + ihl);
217
218 if (!icmp_is_err(icmph->type))
219 break;
220
221 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
222 noff))
223 goto drop;
224
225 icmph = (void *)(skb_network_header(skb) + ihl);
226 iph = (void *)(icmph + 1);
227 if (egress)
228 addr = iph->daddr;
229 else
230 addr = iph->saddr;
231
232 if ((old_addr ^ addr) & mask)
233 break;
234
235 if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
236 sizeof(*iph) + noff))
237 goto drop;
238
239 icmph = (void *)(skb_network_header(skb) + ihl);
240 iph = (void *)(icmph + 1);
241
242 new_addr &= mask;
243 new_addr |= addr & ~mask;
244
245 /* XXX Fix up the inner checksums. */
246 if (egress)
247 iph->daddr = new_addr;
248 else
249 iph->saddr = new_addr;
250
251 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
252 false);
253 break;
254 }
255 default:
256 break;
257 }
258
259 out:
260 return action;
261
262 drop:
263 tcf_action_inc_drop_qstats(&p->common);
264 return TC_ACT_SHOT;
265 }
266
tcf_nat_dump(struct sk_buff * skb,struct tc_action * a,int bind,int ref)267 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
268 int bind, int ref)
269 {
270 unsigned char *b = skb_tail_pointer(skb);
271 struct tcf_nat *p = to_tcf_nat(a);
272 struct tc_nat opt = {
273 .index = p->tcf_index,
274 .refcnt = refcount_read(&p->tcf_refcnt) - ref,
275 .bindcnt = atomic_read(&p->tcf_bindcnt) - bind,
276 };
277 struct tcf_nat_parms *parms;
278 struct tcf_t t;
279
280 spin_lock_bh(&p->tcf_lock);
281
282 opt.action = p->tcf_action;
283
284 parms = rcu_dereference_protected(p->parms, lockdep_is_held(&p->tcf_lock));
285
286 opt.old_addr = parms->old_addr;
287 opt.new_addr = parms->new_addr;
288 opt.mask = parms->mask;
289 opt.flags = parms->flags;
290
291 if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
292 goto nla_put_failure;
293
294 tcf_tm_dump(&t, &p->tcf_tm);
295 if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
296 goto nla_put_failure;
297 spin_unlock_bh(&p->tcf_lock);
298
299 return skb->len;
300
301 nla_put_failure:
302 spin_unlock_bh(&p->tcf_lock);
303 nlmsg_trim(skb, b);
304 return -1;
305 }
306
tcf_nat_cleanup(struct tc_action * a)307 static void tcf_nat_cleanup(struct tc_action *a)
308 {
309 struct tcf_nat *p = to_tcf_nat(a);
310 struct tcf_nat_parms *parms;
311
312 parms = rcu_dereference_protected(p->parms, 1);
313 if (parms)
314 kfree_rcu(parms, rcu);
315 }
316
317 static struct tc_action_ops act_nat_ops = {
318 .kind = "nat",
319 .id = TCA_ID_NAT,
320 .owner = THIS_MODULE,
321 .act = tcf_nat_act,
322 .dump = tcf_nat_dump,
323 .init = tcf_nat_init,
324 .cleanup = tcf_nat_cleanup,
325 .size = sizeof(struct tcf_nat),
326 };
327
nat_init_net(struct net * net)328 static __net_init int nat_init_net(struct net *net)
329 {
330 struct tc_action_net *tn = net_generic(net, act_nat_ops.net_id);
331
332 return tc_action_net_init(net, tn, &act_nat_ops);
333 }
334
nat_exit_net(struct list_head * net_list)335 static void __net_exit nat_exit_net(struct list_head *net_list)
336 {
337 tc_action_net_exit(net_list, act_nat_ops.net_id);
338 }
339
340 static struct pernet_operations nat_net_ops = {
341 .init = nat_init_net,
342 .exit_batch = nat_exit_net,
343 .id = &act_nat_ops.net_id,
344 .size = sizeof(struct tc_action_net),
345 };
346
347 MODULE_DESCRIPTION("Stateless NAT actions");
348 MODULE_LICENSE("GPL");
349
nat_init_module(void)350 static int __init nat_init_module(void)
351 {
352 return tcf_register_action(&act_nat_ops, &nat_net_ops);
353 }
354
nat_cleanup_module(void)355 static void __exit nat_cleanup_module(void)
356 {
357 tcf_unregister_action(&act_nat_ops, &nat_net_ops);
358 }
359
360 module_init(nat_init_module);
361 module_exit(nat_cleanup_module);
362