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