xref: /openbmc/linux/net/netfilter/xt_TEE.c (revision 7fe2f639)
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 defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
29 #	define WITH_CONNTRACK 1
30 #	include <net/netfilter/nf_conntrack.h>
31 #endif
32 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
33 #	define WITH_IPV6 1
34 #endif
35 
36 struct xt_tee_priv {
37 	struct notifier_block	notifier;
38 	struct xt_tee_tginfo	*tginfo;
39 	int			oif;
40 };
41 
42 static const union nf_inet_addr tee_zero_address;
43 static DEFINE_PER_CPU(bool, tee_active);
44 
45 static struct net *pick_net(struct sk_buff *skb)
46 {
47 #ifdef CONFIG_NET_NS
48 	const struct dst_entry *dst;
49 
50 	if (skb->dev != NULL)
51 		return dev_net(skb->dev);
52 	dst = skb_dst(skb);
53 	if (dst != NULL && dst->dev != NULL)
54 		return dev_net(dst->dev);
55 #endif
56 	return &init_net;
57 }
58 
59 static bool
60 tee_tg_route4(struct sk_buff *skb, const struct xt_tee_tginfo *info)
61 {
62 	const struct iphdr *iph = ip_hdr(skb);
63 	struct net *net = pick_net(skb);
64 	struct rtable *rt;
65 	struct flowi4 fl4;
66 
67 	memset(&fl4, 0, sizeof(fl4));
68 	if (info->priv) {
69 		if (info->priv->oif == -1)
70 			return false;
71 		fl4.flowi4_oif = info->priv->oif;
72 	}
73 	fl4.daddr = info->gw.ip;
74 	fl4.flowi4_tos = RT_TOS(iph->tos);
75 	fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
76 	rt = ip_route_output_key(net, &fl4);
77 	if (IS_ERR(rt))
78 		return false;
79 
80 	skb_dst_drop(skb);
81 	skb_dst_set(skb, &rt->dst);
82 	skb->dev      = rt->dst.dev;
83 	skb->protocol = htons(ETH_P_IP);
84 	return true;
85 }
86 
87 static unsigned int
88 tee_tg4(struct sk_buff *skb, const struct xt_action_param *par)
89 {
90 	const struct xt_tee_tginfo *info = par->targinfo;
91 	struct iphdr *iph;
92 
93 	if (percpu_read(tee_active))
94 		return XT_CONTINUE;
95 	/*
96 	 * Copy the skb, and route the copy. Will later return %XT_CONTINUE for
97 	 * the original skb, which should continue on its way as if nothing has
98 	 * happened. The copy should be independently delivered to the TEE
99 	 * --gateway.
100 	 */
101 	skb = pskb_copy(skb, GFP_ATOMIC);
102 	if (skb == NULL)
103 		return XT_CONTINUE;
104 
105 #ifdef WITH_CONNTRACK
106 	/* Avoid counting cloned packets towards the original connection. */
107 	nf_conntrack_put(skb->nfct);
108 	skb->nfct     = &nf_ct_untracked_get()->ct_general;
109 	skb->nfctinfo = IP_CT_NEW;
110 	nf_conntrack_get(skb->nfct);
111 #endif
112 	/*
113 	 * If we are in PREROUTING/INPUT, the checksum must be recalculated
114 	 * since the length could have changed as a result of defragmentation.
115 	 *
116 	 * We also decrease the TTL to mitigate potential TEE loops
117 	 * between two hosts.
118 	 *
119 	 * Set %IP_DF so that the original source is notified of a potentially
120 	 * decreased MTU on the clone route. IPv6 does this too.
121 	 */
122 	iph = ip_hdr(skb);
123 	iph->frag_off |= htons(IP_DF);
124 	if (par->hooknum == NF_INET_PRE_ROUTING ||
125 	    par->hooknum == NF_INET_LOCAL_IN)
126 		--iph->ttl;
127 	ip_send_check(iph);
128 
129 	if (tee_tg_route4(skb, info)) {
130 		percpu_write(tee_active, true);
131 		ip_local_out(skb);
132 		percpu_write(tee_active, false);
133 	} else {
134 		kfree_skb(skb);
135 	}
136 	return XT_CONTINUE;
137 }
138 
139 #ifdef WITH_IPV6
140 static bool
141 tee_tg_route6(struct sk_buff *skb, const struct xt_tee_tginfo *info)
142 {
143 	const struct ipv6hdr *iph = ipv6_hdr(skb);
144 	struct net *net = pick_net(skb);
145 	struct dst_entry *dst;
146 	struct flowi6 fl6;
147 
148 	memset(&fl6, 0, sizeof(fl6));
149 	if (info->priv) {
150 		if (info->priv->oif == -1)
151 			return false;
152 		fl6.flowi6_oif = info->priv->oif;
153 	}
154 	fl6.daddr = info->gw.in6;
155 	fl6.flowlabel = ((iph->flow_lbl[0] & 0xF) << 16) |
156 			   (iph->flow_lbl[1] << 8) | iph->flow_lbl[2];
157 	dst = ip6_route_output(net, NULL, &fl6);
158 	if (dst == NULL)
159 		return false;
160 
161 	skb_dst_drop(skb);
162 	skb_dst_set(skb, dst);
163 	skb->dev      = dst->dev;
164 	skb->protocol = htons(ETH_P_IPV6);
165 	return true;
166 }
167 
168 static unsigned int
169 tee_tg6(struct sk_buff *skb, const struct xt_action_param *par)
170 {
171 	const struct xt_tee_tginfo *info = par->targinfo;
172 
173 	if (percpu_read(tee_active))
174 		return XT_CONTINUE;
175 	skb = pskb_copy(skb, GFP_ATOMIC);
176 	if (skb == NULL)
177 		return XT_CONTINUE;
178 
179 #ifdef WITH_CONNTRACK
180 	nf_conntrack_put(skb->nfct);
181 	skb->nfct     = &nf_ct_untracked_get()->ct_general;
182 	skb->nfctinfo = IP_CT_NEW;
183 	nf_conntrack_get(skb->nfct);
184 #endif
185 	if (par->hooknum == NF_INET_PRE_ROUTING ||
186 	    par->hooknum == NF_INET_LOCAL_IN) {
187 		struct ipv6hdr *iph = ipv6_hdr(skb);
188 		--iph->hop_limit;
189 	}
190 	if (tee_tg_route6(skb, info)) {
191 		percpu_write(tee_active, true);
192 		ip6_local_out(skb);
193 		percpu_write(tee_active, false);
194 	} else {
195 		kfree_skb(skb);
196 	}
197 	return XT_CONTINUE;
198 }
199 #endif /* WITH_IPV6 */
200 
201 static int tee_netdev_event(struct notifier_block *this, unsigned long event,
202 			    void *ptr)
203 {
204 	struct net_device *dev = ptr;
205 	struct xt_tee_priv *priv;
206 
207 	priv = container_of(this, struct xt_tee_priv, notifier);
208 	switch (event) {
209 	case NETDEV_REGISTER:
210 		if (!strcmp(dev->name, priv->tginfo->oif))
211 			priv->oif = dev->ifindex;
212 		break;
213 	case NETDEV_UNREGISTER:
214 		if (dev->ifindex == priv->oif)
215 			priv->oif = -1;
216 		break;
217 	case NETDEV_CHANGENAME:
218 		if (!strcmp(dev->name, priv->tginfo->oif))
219 			priv->oif = dev->ifindex;
220 		else if (dev->ifindex == priv->oif)
221 			priv->oif = -1;
222 		break;
223 	}
224 
225 	return NOTIFY_DONE;
226 }
227 
228 static int tee_tg_check(const struct xt_tgchk_param *par)
229 {
230 	struct xt_tee_tginfo *info = par->targinfo;
231 	struct xt_tee_priv *priv;
232 
233 	/* 0.0.0.0 and :: not allowed */
234 	if (memcmp(&info->gw, &tee_zero_address,
235 		   sizeof(tee_zero_address)) == 0)
236 		return -EINVAL;
237 
238 	if (info->oif[0]) {
239 		if (info->oif[sizeof(info->oif)-1] != '\0')
240 			return -EINVAL;
241 
242 		priv = kzalloc(sizeof(*priv), GFP_KERNEL);
243 		if (priv == NULL)
244 			return -ENOMEM;
245 
246 		priv->tginfo  = info;
247 		priv->oif     = -1;
248 		priv->notifier.notifier_call = tee_netdev_event;
249 		info->priv    = priv;
250 
251 		register_netdevice_notifier(&priv->notifier);
252 	} else
253 		info->priv = NULL;
254 
255 	return 0;
256 }
257 
258 static void tee_tg_destroy(const struct xt_tgdtor_param *par)
259 {
260 	struct xt_tee_tginfo *info = par->targinfo;
261 
262 	if (info->priv) {
263 		unregister_netdevice_notifier(&info->priv->notifier);
264 		kfree(info->priv);
265 	}
266 }
267 
268 static struct xt_target tee_tg_reg[] __read_mostly = {
269 	{
270 		.name       = "TEE",
271 		.revision   = 1,
272 		.family     = NFPROTO_IPV4,
273 		.target     = tee_tg4,
274 		.targetsize = sizeof(struct xt_tee_tginfo),
275 		.checkentry = tee_tg_check,
276 		.destroy    = tee_tg_destroy,
277 		.me         = THIS_MODULE,
278 	},
279 #ifdef WITH_IPV6
280 	{
281 		.name       = "TEE",
282 		.revision   = 1,
283 		.family     = NFPROTO_IPV6,
284 		.target     = tee_tg6,
285 		.targetsize = sizeof(struct xt_tee_tginfo),
286 		.checkentry = tee_tg_check,
287 		.destroy    = tee_tg_destroy,
288 		.me         = THIS_MODULE,
289 	},
290 #endif
291 };
292 
293 static int __init tee_tg_init(void)
294 {
295 	return xt_register_targets(tee_tg_reg, ARRAY_SIZE(tee_tg_reg));
296 }
297 
298 static void __exit tee_tg_exit(void)
299 {
300 	xt_unregister_targets(tee_tg_reg, ARRAY_SIZE(tee_tg_reg));
301 }
302 
303 module_init(tee_tg_init);
304 module_exit(tee_tg_exit);
305 MODULE_AUTHOR("Sebastian Claßen <sebastian.classen@freenet.ag>");
306 MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>");
307 MODULE_DESCRIPTION("Xtables: Reroute packet copy");
308 MODULE_LICENSE("GPL");
309 MODULE_ALIAS("ipt_TEE");
310 MODULE_ALIAS("ip6t_TEE");
311