xref: /openbmc/linux/net/sched/act_nat.c (revision cb1aaebe)
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/pkt_cls.h>
25 #include <net/icmp.h>
26 #include <net/ip.h>
27 #include <net/netlink.h>
28 #include <net/tc_act/tc_nat.h>
29 #include <net/tcp.h>
30 #include <net/udp.h>
31 
32 
33 static unsigned int nat_net_id;
34 static struct tc_action_ops act_nat_ops;
35 
36 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
37 	[TCA_NAT_PARMS]	= { .len = sizeof(struct tc_nat) },
38 };
39 
40 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
41 			struct tc_action **a, int ovr, int bind,
42 			bool rtnl_held,	struct tcf_proto *tp,
43 			struct netlink_ext_ack *extack)
44 {
45 	struct tc_action_net *tn = net_generic(net, nat_net_id);
46 	struct nlattr *tb[TCA_NAT_MAX + 1];
47 	struct tcf_chain *goto_ch = NULL;
48 	struct tc_nat *parm;
49 	int ret = 0, err;
50 	struct tcf_nat *p;
51 
52 	if (nla == NULL)
53 		return -EINVAL;
54 
55 	err = nla_parse_nested_deprecated(tb, TCA_NAT_MAX, nla, nat_policy,
56 					  NULL);
57 	if (err < 0)
58 		return err;
59 
60 	if (tb[TCA_NAT_PARMS] == NULL)
61 		return -EINVAL;
62 	parm = nla_data(tb[TCA_NAT_PARMS]);
63 
64 	err = tcf_idr_check_alloc(tn, &parm->index, a, bind);
65 	if (!err) {
66 		ret = tcf_idr_create(tn, parm->index, est, a,
67 				     &act_nat_ops, bind, false);
68 		if (ret) {
69 			tcf_idr_cleanup(tn, parm->index);
70 			return ret;
71 		}
72 		ret = ACT_P_CREATED;
73 	} else if (err > 0) {
74 		if (bind)
75 			return 0;
76 		if (!ovr) {
77 			tcf_idr_release(*a, bind);
78 			return -EEXIST;
79 		}
80 	} else {
81 		return err;
82 	}
83 	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
84 	if (err < 0)
85 		goto release_idr;
86 	p = to_tcf_nat(*a);
87 
88 	spin_lock_bh(&p->tcf_lock);
89 	p->old_addr = parm->old_addr;
90 	p->new_addr = parm->new_addr;
91 	p->mask = parm->mask;
92 	p->flags = parm->flags;
93 
94 	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
95 	spin_unlock_bh(&p->tcf_lock);
96 	if (goto_ch)
97 		tcf_chain_put_by_act(goto_ch);
98 
99 	if (ret == ACT_P_CREATED)
100 		tcf_idr_insert(tn, *a);
101 
102 	return ret;
103 release_idr:
104 	tcf_idr_release(*a, bind);
105 	return err;
106 }
107 
108 static int tcf_nat_act(struct sk_buff *skb, const struct tc_action *a,
109 		       struct tcf_result *res)
110 {
111 	struct tcf_nat *p = to_tcf_nat(a);
112 	struct iphdr *iph;
113 	__be32 old_addr;
114 	__be32 new_addr;
115 	__be32 mask;
116 	__be32 addr;
117 	int egress;
118 	int action;
119 	int ihl;
120 	int noff;
121 
122 	spin_lock(&p->tcf_lock);
123 
124 	tcf_lastuse_update(&p->tcf_tm);
125 	old_addr = p->old_addr;
126 	new_addr = p->new_addr;
127 	mask = p->mask;
128 	egress = p->flags & TCA_NAT_FLAG_EGRESS;
129 	action = p->tcf_action;
130 
131 	bstats_update(&p->tcf_bstats, skb);
132 
133 	spin_unlock(&p->tcf_lock);
134 
135 	if (unlikely(action == TC_ACT_SHOT))
136 		goto drop;
137 
138 	noff = skb_network_offset(skb);
139 	if (!pskb_may_pull(skb, sizeof(*iph) + noff))
140 		goto drop;
141 
142 	iph = ip_hdr(skb);
143 
144 	if (egress)
145 		addr = iph->saddr;
146 	else
147 		addr = iph->daddr;
148 
149 	if (!((old_addr ^ addr) & mask)) {
150 		if (skb_try_make_writable(skb, sizeof(*iph) + noff))
151 			goto drop;
152 
153 		new_addr &= mask;
154 		new_addr |= addr & ~mask;
155 
156 		/* Rewrite IP header */
157 		iph = ip_hdr(skb);
158 		if (egress)
159 			iph->saddr = new_addr;
160 		else
161 			iph->daddr = new_addr;
162 
163 		csum_replace4(&iph->check, addr, new_addr);
164 	} else if ((iph->frag_off & htons(IP_OFFSET)) ||
165 		   iph->protocol != IPPROTO_ICMP) {
166 		goto out;
167 	}
168 
169 	ihl = iph->ihl * 4;
170 
171 	/* It would be nice to share code with stateful NAT. */
172 	switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
173 	case IPPROTO_TCP:
174 	{
175 		struct tcphdr *tcph;
176 
177 		if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
178 		    skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
179 			goto drop;
180 
181 		tcph = (void *)(skb_network_header(skb) + ihl);
182 		inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
183 					 true);
184 		break;
185 	}
186 	case IPPROTO_UDP:
187 	{
188 		struct udphdr *udph;
189 
190 		if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
191 		    skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
192 			goto drop;
193 
194 		udph = (void *)(skb_network_header(skb) + ihl);
195 		if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
196 			inet_proto_csum_replace4(&udph->check, skb, addr,
197 						 new_addr, true);
198 			if (!udph->check)
199 				udph->check = CSUM_MANGLED_0;
200 		}
201 		break;
202 	}
203 	case IPPROTO_ICMP:
204 	{
205 		struct icmphdr *icmph;
206 
207 		if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
208 			goto drop;
209 
210 		icmph = (void *)(skb_network_header(skb) + ihl);
211 
212 		if ((icmph->type != ICMP_DEST_UNREACH) &&
213 		    (icmph->type != ICMP_TIME_EXCEEDED) &&
214 		    (icmph->type != ICMP_PARAMETERPROB))
215 			break;
216 
217 		if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
218 					noff))
219 			goto drop;
220 
221 		icmph = (void *)(skb_network_header(skb) + ihl);
222 		iph = (void *)(icmph + 1);
223 		if (egress)
224 			addr = iph->daddr;
225 		else
226 			addr = iph->saddr;
227 
228 		if ((old_addr ^ addr) & mask)
229 			break;
230 
231 		if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
232 					  sizeof(*iph) + noff))
233 			goto drop;
234 
235 		icmph = (void *)(skb_network_header(skb) + ihl);
236 		iph = (void *)(icmph + 1);
237 
238 		new_addr &= mask;
239 		new_addr |= addr & ~mask;
240 
241 		/* XXX Fix up the inner checksums. */
242 		if (egress)
243 			iph->daddr = new_addr;
244 		else
245 			iph->saddr = new_addr;
246 
247 		inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
248 					 false);
249 		break;
250 	}
251 	default:
252 		break;
253 	}
254 
255 out:
256 	return action;
257 
258 drop:
259 	spin_lock(&p->tcf_lock);
260 	p->tcf_qstats.drops++;
261 	spin_unlock(&p->tcf_lock);
262 	return TC_ACT_SHOT;
263 }
264 
265 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
266 			int bind, int ref)
267 {
268 	unsigned char *b = skb_tail_pointer(skb);
269 	struct tcf_nat *p = to_tcf_nat(a);
270 	struct tc_nat opt = {
271 		.index    = p->tcf_index,
272 		.refcnt   = refcount_read(&p->tcf_refcnt) - ref,
273 		.bindcnt  = atomic_read(&p->tcf_bindcnt) - bind,
274 	};
275 	struct tcf_t t;
276 
277 	spin_lock_bh(&p->tcf_lock);
278 	opt.old_addr = p->old_addr;
279 	opt.new_addr = p->new_addr;
280 	opt.mask = p->mask;
281 	opt.flags = p->flags;
282 	opt.action = p->tcf_action;
283 
284 	if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
285 		goto nla_put_failure;
286 
287 	tcf_tm_dump(&t, &p->tcf_tm);
288 	if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
289 		goto nla_put_failure;
290 	spin_unlock_bh(&p->tcf_lock);
291 
292 	return skb->len;
293 
294 nla_put_failure:
295 	spin_unlock_bh(&p->tcf_lock);
296 	nlmsg_trim(skb, b);
297 	return -1;
298 }
299 
300 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
301 			  struct netlink_callback *cb, int type,
302 			  const struct tc_action_ops *ops,
303 			  struct netlink_ext_ack *extack)
304 {
305 	struct tc_action_net *tn = net_generic(net, nat_net_id);
306 
307 	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
308 }
309 
310 static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
311 {
312 	struct tc_action_net *tn = net_generic(net, nat_net_id);
313 
314 	return tcf_idr_search(tn, a, index);
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 	.walk		=	tcf_nat_walker,
325 	.lookup		=	tcf_nat_search,
326 	.size		=	sizeof(struct tcf_nat),
327 };
328 
329 static __net_init int nat_init_net(struct net *net)
330 {
331 	struct tc_action_net *tn = net_generic(net, nat_net_id);
332 
333 	return tc_action_net_init(tn, &act_nat_ops);
334 }
335 
336 static void __net_exit nat_exit_net(struct list_head *net_list)
337 {
338 	tc_action_net_exit(net_list, nat_net_id);
339 }
340 
341 static struct pernet_operations nat_net_ops = {
342 	.init = nat_init_net,
343 	.exit_batch = nat_exit_net,
344 	.id   = &nat_net_id,
345 	.size = sizeof(struct tc_action_net),
346 };
347 
348 MODULE_DESCRIPTION("Stateless NAT actions");
349 MODULE_LICENSE("GPL");
350 
351 static int __init nat_init_module(void)
352 {
353 	return tcf_register_action(&act_nat_ops, &nat_net_ops);
354 }
355 
356 static void __exit nat_cleanup_module(void)
357 {
358 	tcf_unregister_action(&act_nat_ops, &nat_net_ops);
359 }
360 
361 module_init(nat_init_module);
362 module_exit(nat_cleanup_module);
363