Jan Engelhardt | e281b19 | 2010-04-19 14:17:47 +0200 | [diff] [blame] | 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/route.h> |
| 16 | #include <linux/skbuff.h> |
| 17 | #include <net/checksum.h> |
| 18 | #include <net/icmp.h> |
| 19 | #include <net/ip.h> |
| 20 | #include <net/ipv6.h> |
| 21 | #include <net/ip6_route.h> |
| 22 | #include <net/route.h> |
| 23 | #include <linux/netfilter/x_tables.h> |
| 24 | #include <linux/netfilter/xt_TEE.h> |
| 25 | |
| 26 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
| 27 | # define WITH_CONNTRACK 1 |
| 28 | # include <net/netfilter/nf_conntrack.h> |
| 29 | #endif |
| 30 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) |
| 31 | # define WITH_IPV6 1 |
| 32 | #endif |
| 33 | |
| 34 | static const union nf_inet_addr tee_zero_address; |
| 35 | |
| 36 | static struct net *pick_net(struct sk_buff *skb) |
| 37 | { |
| 38 | #ifdef CONFIG_NET_NS |
| 39 | const struct dst_entry *dst; |
| 40 | |
| 41 | if (skb->dev != NULL) |
| 42 | return dev_net(skb->dev); |
| 43 | dst = skb_dst(skb); |
| 44 | if (dst != NULL && dst->dev != NULL) |
| 45 | return dev_net(dst->dev); |
| 46 | #endif |
| 47 | return &init_net; |
| 48 | } |
| 49 | |
| 50 | static bool tee_tg_route_oif(struct flowi *f, struct net *net, |
| 51 | const struct xt_tee_tginfo *info) |
| 52 | { |
| 53 | const struct net_device *dev; |
| 54 | |
| 55 | if (*info->oif != '\0') |
| 56 | return true; |
| 57 | dev = dev_get_by_name(net, info->oif); |
| 58 | if (dev == NULL) |
| 59 | return false; |
| 60 | f->oif = dev->ifindex; |
| 61 | return true; |
| 62 | } |
| 63 | |
| 64 | static bool |
| 65 | tee_tg_route4(struct sk_buff *skb, const struct xt_tee_tginfo *info) |
| 66 | { |
| 67 | const struct iphdr *iph = ip_hdr(skb); |
| 68 | struct net *net = pick_net(skb); |
| 69 | struct rtable *rt; |
| 70 | struct flowi fl; |
| 71 | |
| 72 | memset(&fl, 0, sizeof(fl)); |
| 73 | if (!tee_tg_route_oif(&fl, net, info)) |
| 74 | return false; |
| 75 | fl.nl_u.ip4_u.daddr = info->gw.ip; |
| 76 | fl.nl_u.ip4_u.tos = RT_TOS(iph->tos); |
| 77 | fl.nl_u.ip4_u.scope = RT_SCOPE_UNIVERSE; |
| 78 | if (ip_route_output_key(net, &rt, &fl) != 0) |
| 79 | return false; |
| 80 | |
| 81 | dst_release(skb_dst(skb)); |
| 82 | skb_dst_set(skb, &rt->u.dst); |
| 83 | skb->dev = rt->u.dst.dev; |
| 84 | skb->protocol = htons(ETH_P_IP); |
| 85 | return true; |
| 86 | } |
| 87 | |
| 88 | static unsigned int |
| 89 | tee_tg4(struct sk_buff *skb, const struct xt_target_param *par) |
| 90 | { |
| 91 | const struct xt_tee_tginfo *info = par->targinfo; |
| 92 | struct iphdr *iph; |
| 93 | |
| 94 | /* |
| 95 | * Copy the skb, and route the copy. Will later return %XT_CONTINUE for |
| 96 | * the original skb, which should continue on its way as if nothing has |
| 97 | * happened. The copy should be independently delivered to the TEE |
| 98 | * --gateway. |
| 99 | */ |
| 100 | skb = pskb_copy(skb, GFP_ATOMIC); |
| 101 | if (skb == NULL) |
| 102 | return XT_CONTINUE; |
| 103 | |
| 104 | #ifdef WITH_CONNTRACK |
| 105 | /* Avoid counting cloned packets towards the original connection. */ |
| 106 | nf_conntrack_put(skb->nfct); |
| 107 | skb->nfct = &nf_conntrack_untracked.ct_general; |
| 108 | skb->nfctinfo = IP_CT_NEW; |
| 109 | nf_conntrack_get(skb->nfct); |
| 110 | #endif |
| 111 | /* |
| 112 | * If we are in PREROUTING/INPUT, the checksum must be recalculated |
| 113 | * since the length could have changed as a result of defragmentation. |
| 114 | * |
| 115 | * We also decrease the TTL to mitigate potential TEE loops |
| 116 | * between two hosts. |
| 117 | * |
| 118 | * Set %IP_DF so that the original source is notified of a potentially |
| 119 | * decreased MTU on the clone route. IPv6 does this too. |
| 120 | */ |
| 121 | iph = ip_hdr(skb); |
| 122 | iph->frag_off |= htons(IP_DF); |
| 123 | if (par->hooknum == NF_INET_PRE_ROUTING || |
| 124 | par->hooknum == NF_INET_LOCAL_IN) |
| 125 | --iph->ttl; |
| 126 | ip_send_check(iph); |
| 127 | |
| 128 | /* |
| 129 | * Xtables is not reentrant currently, so a choice has to be made: |
| 130 | * 1. return absolute verdict for the original and let the cloned |
| 131 | * packet travel through the chains |
| 132 | * 2. let the original continue travelling and not pass the clone |
| 133 | * to Xtables. |
| 134 | * #2 is chosen. Normally, we would use ip_local_out for the clone. |
| 135 | * Because iph->check is already correct and we don't pass it to |
| 136 | * Xtables anyway, a shortcut to dst_output [forwards to ip_output] can |
| 137 | * be taken. %IPSKB_REROUTED needs to be set so that ip_output does not |
| 138 | * invoke POSTROUTING on the cloned packet. |
| 139 | */ |
| 140 | IPCB(skb)->flags |= IPSKB_REROUTED; |
| 141 | if (tee_tg_route4(skb, info)) |
| 142 | ip_output(skb); |
| 143 | else |
| 144 | kfree_skb(skb); |
| 145 | |
| 146 | return XT_CONTINUE; |
| 147 | } |
| 148 | |
| 149 | #ifdef WITH_IPV6 |
| 150 | static bool |
| 151 | tee_tg_route6(struct sk_buff *skb, const struct xt_tee_tginfo *info) |
| 152 | { |
| 153 | const struct ipv6hdr *iph = ipv6_hdr(skb); |
| 154 | struct net *net = pick_net(skb); |
| 155 | struct dst_entry *dst; |
| 156 | struct flowi fl; |
| 157 | |
| 158 | memset(&fl, 0, sizeof(fl)); |
| 159 | if (!tee_tg_route_oif(&fl, net, info)) |
| 160 | return false; |
| 161 | fl.nl_u.ip6_u.daddr = info->gw.in6; |
| 162 | fl.nl_u.ip6_u.flowlabel = ((iph->flow_lbl[0] & 0xF) << 16) | |
| 163 | (iph->flow_lbl[1] << 8) | iph->flow_lbl[2]; |
| 164 | dst = ip6_route_output(net, NULL, &fl); |
| 165 | if (dst == NULL) |
| 166 | return false; |
| 167 | |
| 168 | dst_release(skb_dst(skb)); |
| 169 | skb_dst_set(skb, dst); |
| 170 | skb->dev = dst->dev; |
| 171 | skb->protocol = htons(ETH_P_IPV6); |
| 172 | return true; |
| 173 | } |
| 174 | |
| 175 | static unsigned int |
| 176 | tee_tg6(struct sk_buff *skb, const struct xt_target_param *par) |
| 177 | { |
| 178 | const struct xt_tee_tginfo *info = par->targinfo; |
| 179 | |
| 180 | skb = pskb_copy(skb, GFP_ATOMIC); |
| 181 | if (skb == NULL) |
| 182 | return XT_CONTINUE; |
| 183 | |
| 184 | #ifdef WITH_CONNTRACK |
| 185 | nf_conntrack_put(skb->nfct); |
| 186 | skb->nfct = &nf_conntrack_untracked.ct_general; |
| 187 | skb->nfctinfo = IP_CT_NEW; |
| 188 | nf_conntrack_get(skb->nfct); |
| 189 | #endif |
| 190 | if (par->hooknum == NF_INET_PRE_ROUTING || |
| 191 | par->hooknum == NF_INET_LOCAL_IN) { |
| 192 | struct ipv6hdr *iph = ipv6_hdr(skb); |
| 193 | --iph->hop_limit; |
| 194 | } |
| 195 | IP6CB(skb)->flags |= IP6SKB_REROUTED; |
| 196 | if (tee_tg_route6(skb, info)) |
| 197 | ip6_output(skb); |
| 198 | else |
| 199 | kfree_skb(skb); |
| 200 | |
| 201 | return XT_CONTINUE; |
| 202 | } |
| 203 | #endif /* WITH_IPV6 */ |
| 204 | |
| 205 | static int tee_tg_check(const struct xt_tgchk_param *par) |
| 206 | { |
| 207 | const struct xt_tee_tginfo *info = par->targinfo; |
| 208 | |
| 209 | if (info->oif[sizeof(info->oif)-1] != '\0') |
| 210 | return -EINVAL; |
| 211 | /* 0.0.0.0 and :: not allowed */ |
| 212 | return (memcmp(&info->gw, &tee_zero_address, |
| 213 | sizeof(tee_zero_address)) == 0) ? -EINVAL : 0; |
| 214 | } |
| 215 | |
| 216 | static struct xt_target tee_tg_reg[] __read_mostly = { |
| 217 | { |
| 218 | .name = "TEE", |
| 219 | .revision = 1, |
| 220 | .family = NFPROTO_IPV4, |
| 221 | .target = tee_tg4, |
| 222 | .targetsize = sizeof(struct xt_tee_tginfo), |
| 223 | .checkentry = tee_tg_check, |
| 224 | .me = THIS_MODULE, |
| 225 | }, |
| 226 | #ifdef WITH_IPV6 |
| 227 | { |
| 228 | .name = "TEE", |
| 229 | .revision = 1, |
| 230 | .family = NFPROTO_IPV6, |
| 231 | .target = tee_tg6, |
| 232 | .targetsize = sizeof(struct xt_tee_tginfo), |
| 233 | .checkentry = tee_tg_check, |
| 234 | .me = THIS_MODULE, |
| 235 | }, |
| 236 | #endif |
| 237 | }; |
| 238 | |
| 239 | static int __init tee_tg_init(void) |
| 240 | { |
| 241 | return xt_register_targets(tee_tg_reg, ARRAY_SIZE(tee_tg_reg)); |
| 242 | } |
| 243 | |
| 244 | static void __exit tee_tg_exit(void) |
| 245 | { |
| 246 | xt_unregister_targets(tee_tg_reg, ARRAY_SIZE(tee_tg_reg)); |
| 247 | } |
| 248 | |
| 249 | module_init(tee_tg_init); |
| 250 | module_exit(tee_tg_exit); |
| 251 | MODULE_AUTHOR("Sebastian Claßen <sebastian.classen@freenet.ag>"); |
| 252 | MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>"); |
| 253 | MODULE_DESCRIPTION("Xtables: Reroute packet copy"); |
| 254 | MODULE_LICENSE("GPL"); |
| 255 | MODULE_ALIAS("ipt_TEE"); |
| 256 | MODULE_ALIAS("ip6t_TEE"); |