Linux Audio

Check our new training course

Yocto / OpenEmbedded training

Mar 24-27, 2025, special US time zones
Register
Loading...
v5.4
  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
 29static unsigned int nat_net_id;
 30static struct tc_action_ops act_nat_ops;
 
 
 
 
 
 
 
 
 31
 32static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
 33	[TCA_NAT_PARMS]	= { .len = sizeof(struct tc_nat) },
 34};
 35
 36static 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	if (ret == ACT_P_CREATED)
 97		tcf_idr_insert(tn, *a);
 98
 99	return ret;
100release_idr:
101	tcf_idr_release(*a, bind);
102	return err;
103}
104
105static int tcf_nat_act(struct sk_buff *skb, const struct tc_action *a,
106		       struct tcf_result *res)
107{
108	struct tcf_nat *p = to_tcf_nat(a);
 
 
 
 
 
 
 
 
109	struct iphdr *iph;
110	__be32 old_addr;
111	__be32 new_addr;
112	__be32 mask;
113	__be32 addr;
114	int egress;
115	int action;
116	int ihl;
117	int noff;
118
119	spin_lock(&p->tcf_lock);
120
121	tcf_lastuse_update(&p->tcf_tm);
122	old_addr = p->old_addr;
123	new_addr = p->new_addr;
124	mask = p->mask;
125	egress = p->flags & TCA_NAT_FLAG_EGRESS;
126	action = p->tcf_action;
127
128	bstats_update(&p->tcf_bstats, skb);
129
130	spin_unlock(&p->tcf_lock);
131
132	if (unlikely(action == TC_ACT_SHOT))
133		goto drop;
134
135	noff = skb_network_offset(skb);
136	if (!pskb_may_pull(skb, sizeof(*iph) + noff))
137		goto drop;
138
139	iph = ip_hdr(skb);
140
141	if (egress)
142		addr = iph->saddr;
143	else
144		addr = iph->daddr;
145
146	if (!((old_addr ^ addr) & mask)) {
147		if (skb_try_make_writable(skb, sizeof(*iph) + noff))
 
 
148			goto drop;
149
150		new_addr &= mask;
151		new_addr |= addr & ~mask;
152
153		/* Rewrite IP header */
154		iph = ip_hdr(skb);
155		if (egress)
156			iph->saddr = new_addr;
157		else
158			iph->daddr = new_addr;
159
160		csum_replace4(&iph->check, addr, new_addr);
161	} else if ((iph->frag_off & htons(IP_OFFSET)) ||
162		   iph->protocol != IPPROTO_ICMP) {
163		goto out;
164	}
165
166	ihl = iph->ihl * 4;
167
168	/* It would be nice to share code with stateful NAT. */
169	switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
170	case IPPROTO_TCP:
171	{
172		struct tcphdr *tcph;
173
174		if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
175		    skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
 
 
176			goto drop;
177
178		tcph = (void *)(skb_network_header(skb) + ihl);
179		inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
180					 true);
181		break;
182	}
183	case IPPROTO_UDP:
184	{
185		struct udphdr *udph;
186
187		if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
188		    skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
 
 
189			goto drop;
190
191		udph = (void *)(skb_network_header(skb) + ihl);
192		if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
193			inet_proto_csum_replace4(&udph->check, skb, addr,
194						 new_addr, true);
195			if (!udph->check)
196				udph->check = CSUM_MANGLED_0;
197		}
198		break;
199	}
200	case IPPROTO_ICMP:
201	{
202		struct icmphdr *icmph;
203
204		if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
205			goto drop;
206
207		icmph = (void *)(skb_network_header(skb) + ihl);
208
209		if ((icmph->type != ICMP_DEST_UNREACH) &&
210		    (icmph->type != ICMP_TIME_EXCEEDED) &&
211		    (icmph->type != ICMP_PARAMETERPROB))
212			break;
213
214		if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
215					noff))
216			goto drop;
217
218		icmph = (void *)(skb_network_header(skb) + ihl);
219		iph = (void *)(icmph + 1);
220		if (egress)
221			addr = iph->daddr;
222		else
223			addr = iph->saddr;
224
225		if ((old_addr ^ addr) & mask)
226			break;
227
228		if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
229					  sizeof(*iph) + noff))
 
 
230			goto drop;
231
232		icmph = (void *)(skb_network_header(skb) + ihl);
233		iph = (void *)(icmph + 1);
234
235		new_addr &= mask;
236		new_addr |= addr & ~mask;
237
238		/* XXX Fix up the inner checksums. */
239		if (egress)
240			iph->daddr = new_addr;
241		else
242			iph->saddr = new_addr;
243
244		inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
245					 false);
246		break;
247	}
248	default:
249		break;
250	}
251
252out:
253	return action;
254
255drop:
256	spin_lock(&p->tcf_lock);
257	p->tcf_qstats.drops++;
258	spin_unlock(&p->tcf_lock);
259	return TC_ACT_SHOT;
260}
261
262static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
263			int bind, int ref)
264{
265	unsigned char *b = skb_tail_pointer(skb);
266	struct tcf_nat *p = to_tcf_nat(a);
267	struct tc_nat opt = {
 
 
 
 
 
268		.index    = p->tcf_index,
269		.refcnt   = refcount_read(&p->tcf_refcnt) - ref,
270		.bindcnt  = atomic_read(&p->tcf_bindcnt) - bind,
 
271	};
272	struct tcf_t t;
273
274	spin_lock_bh(&p->tcf_lock);
275	opt.old_addr = p->old_addr;
276	opt.new_addr = p->new_addr;
277	opt.mask = p->mask;
278	opt.flags = p->flags;
279	opt.action = p->tcf_action;
280
281	if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
282		goto nla_put_failure;
283
284	tcf_tm_dump(&t, &p->tcf_tm);
285	if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
286		goto nla_put_failure;
287	spin_unlock_bh(&p->tcf_lock);
288
289	return skb->len;
290
291nla_put_failure:
292	spin_unlock_bh(&p->tcf_lock);
293	nlmsg_trim(skb, b);
294	return -1;
295}
296
297static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
298			  struct netlink_callback *cb, int type,
299			  const struct tc_action_ops *ops,
300			  struct netlink_ext_ack *extack)
301{
302	struct tc_action_net *tn = net_generic(net, nat_net_id);
303
304	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
305}
306
307static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
308{
309	struct tc_action_net *tn = net_generic(net, nat_net_id);
310
311	return tcf_idr_search(tn, a, index);
312}
313
314static struct tc_action_ops act_nat_ops = {
315	.kind		=	"nat",
316	.id		=	TCA_ID_NAT,
 
 
317	.owner		=	THIS_MODULE,
318	.act		=	tcf_nat_act,
319	.dump		=	tcf_nat_dump,
 
 
320	.init		=	tcf_nat_init,
321	.walk		=	tcf_nat_walker,
322	.lookup		=	tcf_nat_search,
323	.size		=	sizeof(struct tcf_nat),
324};
325
326static __net_init int nat_init_net(struct net *net)
327{
328	struct tc_action_net *tn = net_generic(net, nat_net_id);
329
330	return tc_action_net_init(net, tn, &act_nat_ops);
331}
332
333static void __net_exit nat_exit_net(struct list_head *net_list)
334{
335	tc_action_net_exit(net_list, nat_net_id);
336}
337
338static struct pernet_operations nat_net_ops = {
339	.init = nat_init_net,
340	.exit_batch = nat_exit_net,
341	.id   = &nat_net_id,
342	.size = sizeof(struct tc_action_net),
343};
344
345MODULE_DESCRIPTION("Stateless NAT actions");
346MODULE_LICENSE("GPL");
347
348static int __init nat_init_module(void)
349{
350	return tcf_register_action(&act_nat_ops, &nat_net_ops);
351}
352
353static void __exit nat_cleanup_module(void)
354{
355	tcf_unregister_action(&act_nat_ops, &nat_net_ops);
356}
357
358module_init(nat_init_module);
359module_exit(nat_cleanup_module);
v3.1
 
  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
 33static struct tcf_common *tcf_nat_ht[NAT_TAB_MASK + 1];
 34static u32 nat_idx_gen;
 35static DEFINE_RWLOCK(nat_lock);
 36
 37static struct tcf_hashinfo nat_hash_info = {
 38	.htab	=	tcf_nat_ht,
 39	.hmask	=	NAT_TAB_MASK,
 40	.lock	=	&nat_lock,
 41};
 42
 43static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
 44	[TCA_NAT_PARMS]	= { .len = sizeof(struct tc_nat) },
 45};
 46
 47static int tcf_nat_init(struct nlattr *nla, struct nlattr *est,
 48			struct tc_action *a, int ovr, int bind)
 
 
 49{
 
 50	struct nlattr *tb[TCA_NAT_MAX + 1];
 
 51	struct tc_nat *parm;
 52	int ret = 0, err;
 53	struct tcf_nat *p;
 54	struct tcf_common *pc;
 55
 56	if (nla == NULL)
 57		return -EINVAL;
 58
 59	err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy);
 
 60	if (err < 0)
 61		return err;
 62
 63	if (tb[TCA_NAT_PARMS] == NULL)
 64		return -EINVAL;
 65	parm = nla_data(tb[TCA_NAT_PARMS]);
 66
 67	pc = tcf_hash_check(parm->index, a, bind, &nat_hash_info);
 68	if (!pc) {
 69		pc = tcf_hash_create(parm->index, est, a, sizeof(*p), bind,
 70				     &nat_idx_gen, &nat_hash_info);
 71		if (IS_ERR(pc))
 72			return PTR_ERR(pc);
 73		p = to_tcf_nat(pc);
 
 74		ret = ACT_P_CREATED;
 75	} else {
 76		p = to_tcf_nat(pc);
 
 77		if (!ovr) {
 78			tcf_hash_release(pc, bind, &nat_hash_info);
 79			return -EEXIST;
 80		}
 
 
 81	}
 
 
 
 
 82
 83	spin_lock_bh(&p->tcf_lock);
 84	p->old_addr = parm->old_addr;
 85	p->new_addr = parm->new_addr;
 86	p->mask = parm->mask;
 87	p->flags = parm->flags;
 88
 89	p->tcf_action = parm->action;
 90	spin_unlock_bh(&p->tcf_lock);
 
 
 91
 92	if (ret == ACT_P_CREATED)
 93		tcf_hash_insert(pc, &nat_hash_info);
 94
 95	return ret;
 
 
 
 96}
 97
 98static int tcf_nat_cleanup(struct tc_action *a, int bind)
 
 99{
100	struct tcf_nat *p = a->priv;
101
102	return tcf_hash_release(&p->common, bind, &nat_hash_info);
103}
104
105static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
106		   struct tcf_result *res)
107{
108	struct tcf_nat *p = a->priv;
109	struct iphdr *iph;
110	__be32 old_addr;
111	__be32 new_addr;
112	__be32 mask;
113	__be32 addr;
114	int egress;
115	int action;
116	int ihl;
117	int noff;
118
119	spin_lock(&p->tcf_lock);
120
121	p->tcf_tm.lastuse = jiffies;
122	old_addr = p->old_addr;
123	new_addr = p->new_addr;
124	mask = p->mask;
125	egress = p->flags & TCA_NAT_FLAG_EGRESS;
126	action = p->tcf_action;
127
128	bstats_update(&p->tcf_bstats, skb);
129
130	spin_unlock(&p->tcf_lock);
131
132	if (unlikely(action == TC_ACT_SHOT))
133		goto drop;
134
135	noff = skb_network_offset(skb);
136	if (!pskb_may_pull(skb, sizeof(*iph) + noff))
137		goto drop;
138
139	iph = ip_hdr(skb);
140
141	if (egress)
142		addr = iph->saddr;
143	else
144		addr = iph->daddr;
145
146	if (!((old_addr ^ addr) & mask)) {
147		if (skb_cloned(skb) &&
148		    !skb_clone_writable(skb, sizeof(*iph) + noff) &&
149		    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
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_cloned(skb) &&
178		     !skb_clone_writable(skb, ihl + sizeof(*tcph) + noff) &&
179		     pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
180			goto drop;
181
182		tcph = (void *)(skb_network_header(skb) + ihl);
183		inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, 1);
 
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_cloned(skb) &&
192		     !skb_clone_writable(skb, ihl + sizeof(*udph) + noff) &&
193		     pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
194			goto drop;
195
196		udph = (void *)(skb_network_header(skb) + ihl);
197		if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
198			inet_proto_csum_replace4(&udph->check, skb, addr,
199						 new_addr, 1);
200			if (!udph->check)
201				udph->check = CSUM_MANGLED_0;
202		}
203		break;
204	}
205	case IPPROTO_ICMP:
206	{
207		struct icmphdr *icmph;
208
209		if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
210			goto drop;
211
212		icmph = (void *)(skb_network_header(skb) + ihl);
213
214		if ((icmph->type != ICMP_DEST_UNREACH) &&
215		    (icmph->type != ICMP_TIME_EXCEEDED) &&
216		    (icmph->type != ICMP_PARAMETERPROB))
217			break;
218
219		if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
220					noff))
221			goto drop;
222
223		icmph = (void *)(skb_network_header(skb) + ihl);
224		iph = (void *)(icmph + 1);
225		if (egress)
226			addr = iph->daddr;
227		else
228			addr = iph->saddr;
229
230		if ((old_addr ^ addr) & mask)
231			break;
232
233		if (skb_cloned(skb) &&
234		    !skb_clone_writable(skb, ihl + sizeof(*icmph) +
235					     sizeof(*iph) + noff) &&
236		    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
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					 0);
253		break;
254	}
255	default:
256		break;
257	}
258
259out:
260	return action;
261
262drop:
263	spin_lock(&p->tcf_lock);
264	p->tcf_qstats.drops++;
265	spin_unlock(&p->tcf_lock);
266	return TC_ACT_SHOT;
267}
268
269static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
270			int bind, int ref)
271{
272	unsigned char *b = skb_tail_pointer(skb);
273	struct tcf_nat *p = a->priv;
274	struct tc_nat opt = {
275		.old_addr = p->old_addr,
276		.new_addr = p->new_addr,
277		.mask     = p->mask,
278		.flags    = p->flags,
279
280		.index    = p->tcf_index,
281		.action   = p->tcf_action,
282		.refcnt   = p->tcf_refcnt - ref,
283		.bindcnt  = p->tcf_bindcnt - bind,
284	};
285	struct tcf_t t;
286
287	NLA_PUT(skb, TCA_NAT_PARMS, sizeof(opt), &opt);
288	t.install = jiffies_to_clock_t(jiffies - p->tcf_tm.install);
289	t.lastuse = jiffies_to_clock_t(jiffies - p->tcf_tm.lastuse);
290	t.expires = jiffies_to_clock_t(p->tcf_tm.expires);
291	NLA_PUT(skb, TCA_NAT_TM, sizeof(t), &t);
 
 
 
 
 
 
 
 
 
292
293	return skb->len;
294
295nla_put_failure:
 
296	nlmsg_trim(skb, b);
297	return -1;
298}
299
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
300static struct tc_action_ops act_nat_ops = {
301	.kind		=	"nat",
302	.hinfo		=	&nat_hash_info,
303	.type		=	TCA_ACT_NAT,
304	.capab		=	TCA_CAP_NONE,
305	.owner		=	THIS_MODULE,
306	.act		=	tcf_nat,
307	.dump		=	tcf_nat_dump,
308	.cleanup	=	tcf_nat_cleanup,
309	.lookup		=	tcf_hash_search,
310	.init		=	tcf_nat_init,
311	.walk		=	tcf_generic_walker
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
312};
313
314MODULE_DESCRIPTION("Stateless NAT actions");
315MODULE_LICENSE("GPL");
316
317static int __init nat_init_module(void)
318{
319	return tcf_register_action(&act_nat_ops);
320}
321
322static void __exit nat_cleanup_module(void)
323{
324	tcf_unregister_action(&act_nat_ops);
325}
326
327module_init(nat_init_module);
328module_exit(nat_cleanup_module);