xref: /openbmc/linux/net/sched/act_nat.c (revision 4f6cce39)
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 unsigned int nat_net_id;
35 static struct tc_action_ops act_nat_ops;
36 
37 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
38 	[TCA_NAT_PARMS]	= { .len = sizeof(struct tc_nat) },
39 };
40 
41 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
42 			struct tc_action **a, int ovr, int bind)
43 {
44 	struct tc_action_net *tn = net_generic(net, nat_net_id);
45 	struct nlattr *tb[TCA_NAT_MAX + 1];
46 	struct tc_nat *parm;
47 	int ret = 0, err;
48 	struct tcf_nat *p;
49 
50 	if (nla == NULL)
51 		return -EINVAL;
52 
53 	err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy);
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 
61 	if (!tcf_hash_check(tn, parm->index, a, bind)) {
62 		ret = tcf_hash_create(tn, parm->index, est, a,
63 				      &act_nat_ops, bind, false);
64 		if (ret)
65 			return ret;
66 		ret = ACT_P_CREATED;
67 	} else {
68 		if (bind)
69 			return 0;
70 		tcf_hash_release(*a, bind);
71 		if (!ovr)
72 			return -EEXIST;
73 	}
74 	p = to_tcf_nat(*a);
75 
76 	spin_lock_bh(&p->tcf_lock);
77 	p->old_addr = parm->old_addr;
78 	p->new_addr = parm->new_addr;
79 	p->mask = parm->mask;
80 	p->flags = parm->flags;
81 
82 	p->tcf_action = parm->action;
83 	spin_unlock_bh(&p->tcf_lock);
84 
85 	if (ret == ACT_P_CREATED)
86 		tcf_hash_insert(tn, *a);
87 
88 	return ret;
89 }
90 
91 static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
92 		   struct tcf_result *res)
93 {
94 	struct tcf_nat *p = to_tcf_nat(a);
95 	struct iphdr *iph;
96 	__be32 old_addr;
97 	__be32 new_addr;
98 	__be32 mask;
99 	__be32 addr;
100 	int egress;
101 	int action;
102 	int ihl;
103 	int noff;
104 
105 	spin_lock(&p->tcf_lock);
106 
107 	tcf_lastuse_update(&p->tcf_tm);
108 	old_addr = p->old_addr;
109 	new_addr = p->new_addr;
110 	mask = p->mask;
111 	egress = p->flags & TCA_NAT_FLAG_EGRESS;
112 	action = p->tcf_action;
113 
114 	bstats_update(&p->tcf_bstats, skb);
115 
116 	spin_unlock(&p->tcf_lock);
117 
118 	if (unlikely(action == TC_ACT_SHOT))
119 		goto drop;
120 
121 	noff = skb_network_offset(skb);
122 	if (!pskb_may_pull(skb, sizeof(*iph) + noff))
123 		goto drop;
124 
125 	iph = ip_hdr(skb);
126 
127 	if (egress)
128 		addr = iph->saddr;
129 	else
130 		addr = iph->daddr;
131 
132 	if (!((old_addr ^ addr) & mask)) {
133 		if (skb_try_make_writable(skb, sizeof(*iph) + noff))
134 			goto drop;
135 
136 		new_addr &= mask;
137 		new_addr |= addr & ~mask;
138 
139 		/* Rewrite IP header */
140 		iph = ip_hdr(skb);
141 		if (egress)
142 			iph->saddr = new_addr;
143 		else
144 			iph->daddr = new_addr;
145 
146 		csum_replace4(&iph->check, addr, new_addr);
147 	} else if ((iph->frag_off & htons(IP_OFFSET)) ||
148 		   iph->protocol != IPPROTO_ICMP) {
149 		goto out;
150 	}
151 
152 	ihl = iph->ihl * 4;
153 
154 	/* It would be nice to share code with stateful NAT. */
155 	switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
156 	case IPPROTO_TCP:
157 	{
158 		struct tcphdr *tcph;
159 
160 		if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
161 		    skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
162 			goto drop;
163 
164 		tcph = (void *)(skb_network_header(skb) + ihl);
165 		inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
166 					 true);
167 		break;
168 	}
169 	case IPPROTO_UDP:
170 	{
171 		struct udphdr *udph;
172 
173 		if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
174 		    skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
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, true);
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_try_make_writable(skb, ihl + sizeof(*icmph) +
215 					  sizeof(*iph) + noff))
216 			goto drop;
217 
218 		icmph = (void *)(skb_network_header(skb) + ihl);
219 		iph = (void *)(icmph + 1);
220 
221 		new_addr &= mask;
222 		new_addr |= addr & ~mask;
223 
224 		/* XXX Fix up the inner checksums. */
225 		if (egress)
226 			iph->daddr = new_addr;
227 		else
228 			iph->saddr = new_addr;
229 
230 		inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
231 					 false);
232 		break;
233 	}
234 	default:
235 		break;
236 	}
237 
238 out:
239 	return action;
240 
241 drop:
242 	spin_lock(&p->tcf_lock);
243 	p->tcf_qstats.drops++;
244 	spin_unlock(&p->tcf_lock);
245 	return TC_ACT_SHOT;
246 }
247 
248 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
249 			int bind, int ref)
250 {
251 	unsigned char *b = skb_tail_pointer(skb);
252 	struct tcf_nat *p = to_tcf_nat(a);
253 	struct tc_nat opt = {
254 		.old_addr = p->old_addr,
255 		.new_addr = p->new_addr,
256 		.mask     = p->mask,
257 		.flags    = p->flags,
258 
259 		.index    = p->tcf_index,
260 		.action   = p->tcf_action,
261 		.refcnt   = p->tcf_refcnt - ref,
262 		.bindcnt  = p->tcf_bindcnt - bind,
263 	};
264 	struct tcf_t t;
265 
266 	if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
267 		goto nla_put_failure;
268 
269 	tcf_tm_dump(&t, &p->tcf_tm);
270 	if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
271 		goto nla_put_failure;
272 
273 	return skb->len;
274 
275 nla_put_failure:
276 	nlmsg_trim(skb, b);
277 	return -1;
278 }
279 
280 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
281 			  struct netlink_callback *cb, int type,
282 			  const struct tc_action_ops *ops)
283 {
284 	struct tc_action_net *tn = net_generic(net, nat_net_id);
285 
286 	return tcf_generic_walker(tn, skb, cb, type, ops);
287 }
288 
289 static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
290 {
291 	struct tc_action_net *tn = net_generic(net, nat_net_id);
292 
293 	return tcf_hash_search(tn, a, index);
294 }
295 
296 static struct tc_action_ops act_nat_ops = {
297 	.kind		=	"nat",
298 	.type		=	TCA_ACT_NAT,
299 	.owner		=	THIS_MODULE,
300 	.act		=	tcf_nat,
301 	.dump		=	tcf_nat_dump,
302 	.init		=	tcf_nat_init,
303 	.walk		=	tcf_nat_walker,
304 	.lookup		=	tcf_nat_search,
305 	.size		=	sizeof(struct tcf_nat),
306 };
307 
308 static __net_init int nat_init_net(struct net *net)
309 {
310 	struct tc_action_net *tn = net_generic(net, nat_net_id);
311 
312 	return tc_action_net_init(tn, &act_nat_ops, NAT_TAB_MASK);
313 }
314 
315 static void __net_exit nat_exit_net(struct net *net)
316 {
317 	struct tc_action_net *tn = net_generic(net, nat_net_id);
318 
319 	tc_action_net_exit(tn);
320 }
321 
322 static struct pernet_operations nat_net_ops = {
323 	.init = nat_init_net,
324 	.exit = nat_exit_net,
325 	.id   = &nat_net_id,
326 	.size = sizeof(struct tc_action_net),
327 };
328 
329 MODULE_DESCRIPTION("Stateless NAT actions");
330 MODULE_LICENSE("GPL");
331 
332 static int __init nat_init_module(void)
333 {
334 	return tcf_register_action(&act_nat_ops, &nat_net_ops);
335 }
336 
337 static void __exit nat_cleanup_module(void)
338 {
339 	tcf_unregister_action(&act_nat_ops, &nat_net_ops);
340 }
341 
342 module_init(nat_init_module);
343 module_exit(nat_cleanup_module);
344