1/*
2 *	"TEE" target extension for Xtables
3 *	Copyright © Sebastian Claßen, 2007
4 *	Jan Engelhardt, 2007-2010
5 *
6 *	based on ipt_ROUTE.c from Cédric de Launois
7 *	<delaunois@info.ucl.be>
8 *
9 *	This program is free software; you can redistribute it and/or
10 *	modify it under the terms of the GNU General Public License
11 *	version 2 or later, as published by the Free Software Foundation.
12 */
13#include <linux/ip.h>
14#include <linux/module.h>
15#include <linux/percpu.h>
16#include <linux/route.h>
17#include <linux/skbuff.h>
18#include <linux/notifier.h>
19#include <net/checksum.h>
20#include <net/icmp.h>
21#include <net/ip.h>
22#include <net/ipv6.h>
23#include <net/ip6_route.h>
24#include <net/route.h>
25#include <linux/netfilter/x_tables.h>
26#include <linux/netfilter/xt_TEE.h>
27
28#if IS_ENABLED(CONFIG_NF_CONNTRACK)
29#	define WITH_CONNTRACK 1
30#	include <net/netfilter/nf_conntrack.h>
31#endif
32
33struct xt_tee_priv {
34	struct notifier_block	notifier;
35	struct xt_tee_tginfo	*tginfo;
36	int			oif;
37};
38
39static const union nf_inet_addr tee_zero_address;
40static DEFINE_PER_CPU(bool, tee_active);
41
42static struct net *pick_net(struct sk_buff *skb)
43{
44#ifdef CONFIG_NET_NS
45	const struct dst_entry *dst;
46
47	if (skb->dev != NULL)
48		return dev_net(skb->dev);
49	dst = skb_dst(skb);
50	if (dst != NULL && dst->dev != NULL)
51		return dev_net(dst->dev);
52#endif
53	return &init_net;
54}
55
56static bool
57tee_tg_route4(struct sk_buff *skb, const struct xt_tee_tginfo *info)
58{
59	const struct iphdr *iph = ip_hdr(skb);
60	struct net *net = pick_net(skb);
61	struct rtable *rt;
62	struct flowi4 fl4;
63
64	memset(&fl4, 0, sizeof(fl4));
65	if (info->priv) {
66		if (info->priv->oif == -1)
67			return false;
68		fl4.flowi4_oif = info->priv->oif;
69	}
70	fl4.daddr = info->gw.ip;
71	fl4.flowi4_tos = RT_TOS(iph->tos);
72	fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
73	fl4.flowi4_flags = FLOWI_FLAG_KNOWN_NH;
74	rt = ip_route_output_key(net, &fl4);
75	if (IS_ERR(rt))
76		return false;
77
78	skb_dst_drop(skb);
79	skb_dst_set(skb, &rt->dst);
80	skb->dev      = rt->dst.dev;
81	skb->protocol = htons(ETH_P_IP);
82	return true;
83}
84
85static unsigned int
86tee_tg4(struct sk_buff *skb, const struct xt_action_param *par)
87{
88	const struct xt_tee_tginfo *info = par->targinfo;
89	struct iphdr *iph;
90
91	if (__this_cpu_read(tee_active))
92		return XT_CONTINUE;
93	/*
94	 * Copy the skb, and route the copy. Will later return %XT_CONTINUE for
95	 * the original skb, which should continue on its way as if nothing has
96	 * happened. The copy should be independently delivered to the TEE
97	 * --gateway.
98	 */
99	skb = pskb_copy(skb, GFP_ATOMIC);
100	if (skb == NULL)
101		return XT_CONTINUE;
102
103#ifdef WITH_CONNTRACK
104	/* Avoid counting cloned packets towards the original connection. */
105	nf_conntrack_put(skb->nfct);
106	skb->nfct     = &nf_ct_untracked_get()->ct_general;
107	skb->nfctinfo = IP_CT_NEW;
108	nf_conntrack_get(skb->nfct);
109#endif
110	/*
111	 * If we are in PREROUTING/INPUT, the checksum must be recalculated
112	 * since the length could have changed as a result of defragmentation.
113	 *
114	 * We also decrease the TTL to mitigate potential TEE loops
115	 * between two hosts.
116	 *
117	 * Set %IP_DF so that the original source is notified of a potentially
118	 * decreased MTU on the clone route. IPv6 does this too.
119	 */
120	iph = ip_hdr(skb);
121	iph->frag_off |= htons(IP_DF);
122	if (par->hooknum == NF_INET_PRE_ROUTING ||
123	    par->hooknum == NF_INET_LOCAL_IN)
124		--iph->ttl;
125	ip_send_check(iph);
126
127	if (tee_tg_route4(skb, info)) {
128		__this_cpu_write(tee_active, true);
129		ip_local_out(skb);
130		__this_cpu_write(tee_active, false);
131	} else {
132		kfree_skb(skb);
133	}
134	return XT_CONTINUE;
135}
136
137#if IS_ENABLED(CONFIG_IPV6)
138static bool
139tee_tg_route6(struct sk_buff *skb, const struct xt_tee_tginfo *info)
140{
141	const struct ipv6hdr *iph = ipv6_hdr(skb);
142	struct net *net = pick_net(skb);
143	struct dst_entry *dst;
144	struct flowi6 fl6;
145
146	memset(&fl6, 0, sizeof(fl6));
147	if (info->priv) {
148		if (info->priv->oif == -1)
149			return false;
150		fl6.flowi6_oif = info->priv->oif;
151	}
152	fl6.daddr = info->gw.in6;
153	fl6.flowlabel = ((iph->flow_lbl[0] & 0xF) << 16) |
154			   (iph->flow_lbl[1] << 8) | iph->flow_lbl[2];
155	dst = ip6_route_output(net, NULL, &fl6);
156	if (dst->error) {
157		dst_release(dst);
158		return false;
159	}
160	skb_dst_drop(skb);
161	skb_dst_set(skb, dst);
162	skb->dev      = dst->dev;
163	skb->protocol = htons(ETH_P_IPV6);
164	return true;
165}
166
167static unsigned int
168tee_tg6(struct sk_buff *skb, const struct xt_action_param *par)
169{
170	const struct xt_tee_tginfo *info = par->targinfo;
171
172	if (__this_cpu_read(tee_active))
173		return XT_CONTINUE;
174	skb = pskb_copy(skb, GFP_ATOMIC);
175	if (skb == NULL)
176		return XT_CONTINUE;
177
178#ifdef WITH_CONNTRACK
179	nf_conntrack_put(skb->nfct);
180	skb->nfct     = &nf_ct_untracked_get()->ct_general;
181	skb->nfctinfo = IP_CT_NEW;
182	nf_conntrack_get(skb->nfct);
183#endif
184	if (par->hooknum == NF_INET_PRE_ROUTING ||
185	    par->hooknum == NF_INET_LOCAL_IN) {
186		struct ipv6hdr *iph = ipv6_hdr(skb);
187		--iph->hop_limit;
188	}
189	if (tee_tg_route6(skb, info)) {
190		__this_cpu_write(tee_active, true);
191		ip6_local_out(skb);
192		__this_cpu_write(tee_active, false);
193	} else {
194		kfree_skb(skb);
195	}
196	return XT_CONTINUE;
197}
198#endif
199
200static int tee_netdev_event(struct notifier_block *this, unsigned long event,
201			    void *ptr)
202{
203	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
204	struct xt_tee_priv *priv;
205
206	priv = container_of(this, struct xt_tee_priv, notifier);
207	switch (event) {
208	case NETDEV_REGISTER:
209		if (!strcmp(dev->name, priv->tginfo->oif))
210			priv->oif = dev->ifindex;
211		break;
212	case NETDEV_UNREGISTER:
213		if (dev->ifindex == priv->oif)
214			priv->oif = -1;
215		break;
216	case NETDEV_CHANGENAME:
217		if (!strcmp(dev->name, priv->tginfo->oif))
218			priv->oif = dev->ifindex;
219		else if (dev->ifindex == priv->oif)
220			priv->oif = -1;
221		break;
222	}
223
224	return NOTIFY_DONE;
225}
226
227static int tee_tg_check(const struct xt_tgchk_param *par)
228{
229	struct xt_tee_tginfo *info = par->targinfo;
230	struct xt_tee_priv *priv;
231
232	/* 0.0.0.0 and :: not allowed */
233	if (memcmp(&info->gw, &tee_zero_address,
234		   sizeof(tee_zero_address)) == 0)
235		return -EINVAL;
236
237	if (info->oif[0]) {
238		if (info->oif[sizeof(info->oif)-1] != '\0')
239			return -EINVAL;
240
241		priv = kzalloc(sizeof(*priv), GFP_KERNEL);
242		if (priv == NULL)
243			return -ENOMEM;
244
245		priv->tginfo  = info;
246		priv->oif     = -1;
247		priv->notifier.notifier_call = tee_netdev_event;
248		info->priv    = priv;
249
250		register_netdevice_notifier(&priv->notifier);
251	} else
252		info->priv = NULL;
253
254	return 0;
255}
256
257static void tee_tg_destroy(const struct xt_tgdtor_param *par)
258{
259	struct xt_tee_tginfo *info = par->targinfo;
260
261	if (info->priv) {
262		unregister_netdevice_notifier(&info->priv->notifier);
263		kfree(info->priv);
264	}
265}
266
267static struct xt_target tee_tg_reg[] __read_mostly = {
268	{
269		.name       = "TEE",
270		.revision   = 1,
271		.family     = NFPROTO_IPV4,
272		.target     = tee_tg4,
273		.targetsize = sizeof(struct xt_tee_tginfo),
274		.checkentry = tee_tg_check,
275		.destroy    = tee_tg_destroy,
276		.me         = THIS_MODULE,
277	},
278#if IS_ENABLED(CONFIG_IPV6)
279	{
280		.name       = "TEE",
281		.revision   = 1,
282		.family     = NFPROTO_IPV6,
283		.target     = tee_tg6,
284		.targetsize = sizeof(struct xt_tee_tginfo),
285		.checkentry = tee_tg_check,
286		.destroy    = tee_tg_destroy,
287		.me         = THIS_MODULE,
288	},
289#endif
290};
291
292static int __init tee_tg_init(void)
293{
294	return xt_register_targets(tee_tg_reg, ARRAY_SIZE(tee_tg_reg));
295}
296
297static void __exit tee_tg_exit(void)
298{
299	xt_unregister_targets(tee_tg_reg, ARRAY_SIZE(tee_tg_reg));
300}
301
302module_init(tee_tg_init);
303module_exit(tee_tg_exit);
304MODULE_AUTHOR("Sebastian Claßen <sebastian.classen@freenet.ag>");
305MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>");
306MODULE_DESCRIPTION("Xtables: Reroute packet copy");
307MODULE_LICENSE("GPL");
308MODULE_ALIAS("ipt_TEE");
309MODULE_ALIAS("ip6t_TEE");
310