xref: /openbmc/linux/net/sched/act_nat.c (revision f9a82c48)
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(tb, TCA_NAT_MAX, nla, nat_policy, NULL);
56 	if (err < 0)
57 		return err;
58 
59 	if (tb[TCA_NAT_PARMS] == NULL)
60 		return -EINVAL;
61 	parm = nla_data(tb[TCA_NAT_PARMS]);
62 
63 	err = tcf_idr_check_alloc(tn, &parm->index, a, bind);
64 	if (!err) {
65 		ret = tcf_idr_create(tn, parm->index, est, a,
66 				     &act_nat_ops, bind, false);
67 		if (ret) {
68 			tcf_idr_cleanup(tn, parm->index);
69 			return ret;
70 		}
71 		ret = ACT_P_CREATED;
72 	} else if (err > 0) {
73 		if (bind)
74 			return 0;
75 		if (!ovr) {
76 			tcf_idr_release(*a, bind);
77 			return -EEXIST;
78 		}
79 	} else {
80 		return err;
81 	}
82 	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
83 	if (err < 0)
84 		goto release_idr;
85 	p = to_tcf_nat(*a);
86 
87 	spin_lock_bh(&p->tcf_lock);
88 	p->old_addr = parm->old_addr;
89 	p->new_addr = parm->new_addr;
90 	p->mask = parm->mask;
91 	p->flags = parm->flags;
92 
93 	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
94 	spin_unlock_bh(&p->tcf_lock);
95 	if (goto_ch)
96 		tcf_chain_put_by_act(goto_ch);
97 
98 	if (ret == ACT_P_CREATED)
99 		tcf_idr_insert(tn, *a);
100 
101 	return ret;
102 release_idr:
103 	tcf_idr_release(*a, bind);
104 	return err;
105 }
106 
107 static int tcf_nat_act(struct sk_buff *skb, const struct tc_action *a,
108 		       struct tcf_result *res)
109 {
110 	struct tcf_nat *p = to_tcf_nat(a);
111 	struct iphdr *iph;
112 	__be32 old_addr;
113 	__be32 new_addr;
114 	__be32 mask;
115 	__be32 addr;
116 	int egress;
117 	int action;
118 	int ihl;
119 	int noff;
120 
121 	spin_lock(&p->tcf_lock);
122 
123 	tcf_lastuse_update(&p->tcf_tm);
124 	old_addr = p->old_addr;
125 	new_addr = p->new_addr;
126 	mask = p->mask;
127 	egress = p->flags & TCA_NAT_FLAG_EGRESS;
128 	action = p->tcf_action;
129 
130 	bstats_update(&p->tcf_bstats, skb);
131 
132 	spin_unlock(&p->tcf_lock);
133 
134 	if (unlikely(action == TC_ACT_SHOT))
135 		goto drop;
136 
137 	noff = skb_network_offset(skb);
138 	if (!pskb_may_pull(skb, sizeof(*iph) + noff))
139 		goto drop;
140 
141 	iph = ip_hdr(skb);
142 
143 	if (egress)
144 		addr = iph->saddr;
145 	else
146 		addr = iph->daddr;
147 
148 	if (!((old_addr ^ addr) & mask)) {
149 		if (skb_try_make_writable(skb, sizeof(*iph) + noff))
150 			goto drop;
151 
152 		new_addr &= mask;
153 		new_addr |= addr & ~mask;
154 
155 		/* Rewrite IP header */
156 		iph = ip_hdr(skb);
157 		if (egress)
158 			iph->saddr = new_addr;
159 		else
160 			iph->daddr = new_addr;
161 
162 		csum_replace4(&iph->check, addr, new_addr);
163 	} else if ((iph->frag_off & htons(IP_OFFSET)) ||
164 		   iph->protocol != IPPROTO_ICMP) {
165 		goto out;
166 	}
167 
168 	ihl = iph->ihl * 4;
169 
170 	/* It would be nice to share code with stateful NAT. */
171 	switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
172 	case IPPROTO_TCP:
173 	{
174 		struct tcphdr *tcph;
175 
176 		if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
177 		    skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
178 			goto drop;
179 
180 		tcph = (void *)(skb_network_header(skb) + ihl);
181 		inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
182 					 true);
183 		break;
184 	}
185 	case IPPROTO_UDP:
186 	{
187 		struct udphdr *udph;
188 
189 		if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
190 		    skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
191 			goto drop;
192 
193 		udph = (void *)(skb_network_header(skb) + ihl);
194 		if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
195 			inet_proto_csum_replace4(&udph->check, skb, addr,
196 						 new_addr, true);
197 			if (!udph->check)
198 				udph->check = CSUM_MANGLED_0;
199 		}
200 		break;
201 	}
202 	case IPPROTO_ICMP:
203 	{
204 		struct icmphdr *icmph;
205 
206 		if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
207 			goto drop;
208 
209 		icmph = (void *)(skb_network_header(skb) + ihl);
210 
211 		if ((icmph->type != ICMP_DEST_UNREACH) &&
212 		    (icmph->type != ICMP_TIME_EXCEEDED) &&
213 		    (icmph->type != ICMP_PARAMETERPROB))
214 			break;
215 
216 		if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
217 					noff))
218 			goto drop;
219 
220 		icmph = (void *)(skb_network_header(skb) + ihl);
221 		iph = (void *)(icmph + 1);
222 		if (egress)
223 			addr = iph->daddr;
224 		else
225 			addr = iph->saddr;
226 
227 		if ((old_addr ^ addr) & mask)
228 			break;
229 
230 		if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
231 					  sizeof(*iph) + noff))
232 			goto drop;
233 
234 		icmph = (void *)(skb_network_header(skb) + ihl);
235 		iph = (void *)(icmph + 1);
236 
237 		new_addr &= mask;
238 		new_addr |= addr & ~mask;
239 
240 		/* XXX Fix up the inner checksums. */
241 		if (egress)
242 			iph->daddr = new_addr;
243 		else
244 			iph->saddr = new_addr;
245 
246 		inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
247 					 false);
248 		break;
249 	}
250 	default:
251 		break;
252 	}
253 
254 out:
255 	return action;
256 
257 drop:
258 	spin_lock(&p->tcf_lock);
259 	p->tcf_qstats.drops++;
260 	spin_unlock(&p->tcf_lock);
261 	return TC_ACT_SHOT;
262 }
263 
264 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
265 			int bind, int ref)
266 {
267 	unsigned char *b = skb_tail_pointer(skb);
268 	struct tcf_nat *p = to_tcf_nat(a);
269 	struct tc_nat opt = {
270 		.index    = p->tcf_index,
271 		.refcnt   = refcount_read(&p->tcf_refcnt) - ref,
272 		.bindcnt  = atomic_read(&p->tcf_bindcnt) - bind,
273 	};
274 	struct tcf_t t;
275 
276 	spin_lock_bh(&p->tcf_lock);
277 	opt.old_addr = p->old_addr;
278 	opt.new_addr = p->new_addr;
279 	opt.mask = p->mask;
280 	opt.flags = p->flags;
281 	opt.action = p->tcf_action;
282 
283 	if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
284 		goto nla_put_failure;
285 
286 	tcf_tm_dump(&t, &p->tcf_tm);
287 	if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
288 		goto nla_put_failure;
289 	spin_unlock_bh(&p->tcf_lock);
290 
291 	return skb->len;
292 
293 nla_put_failure:
294 	spin_unlock_bh(&p->tcf_lock);
295 	nlmsg_trim(skb, b);
296 	return -1;
297 }
298 
299 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
300 			  struct netlink_callback *cb, int type,
301 			  const struct tc_action_ops *ops,
302 			  struct netlink_ext_ack *extack)
303 {
304 	struct tc_action_net *tn = net_generic(net, nat_net_id);
305 
306 	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
307 }
308 
309 static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
310 {
311 	struct tc_action_net *tn = net_generic(net, nat_net_id);
312 
313 	return tcf_idr_search(tn, a, index);
314 }
315 
316 static struct tc_action_ops act_nat_ops = {
317 	.kind		=	"nat",
318 	.id		=	TCA_ID_NAT,
319 	.owner		=	THIS_MODULE,
320 	.act		=	tcf_nat_act,
321 	.dump		=	tcf_nat_dump,
322 	.init		=	tcf_nat_init,
323 	.walk		=	tcf_nat_walker,
324 	.lookup		=	tcf_nat_search,
325 	.size		=	sizeof(struct tcf_nat),
326 };
327 
328 static __net_init int nat_init_net(struct net *net)
329 {
330 	struct tc_action_net *tn = net_generic(net, nat_net_id);
331 
332 	return tc_action_net_init(tn, &act_nat_ops);
333 }
334 
335 static void __net_exit nat_exit_net(struct list_head *net_list)
336 {
337 	tc_action_net_exit(net_list, nat_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   = &nat_net_id,
344 	.size = sizeof(struct tc_action_net),
345 };
346 
347 MODULE_DESCRIPTION("Stateless NAT actions");
348 MODULE_LICENSE("GPL");
349 
350 static int __init nat_init_module(void)
351 {
352 	return tcf_register_action(&act_nat_ops, &nat_net_ops);
353 }
354 
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