netfilter: xtables: inclusion of xt_TEE
Jan Engelhardt [Mon, 19 Apr 2010 12:17:47 +0000 (14:17 +0200)]
xt_TEE can be used to clone and reroute a packet. This can for
example be used to copy traffic at a router for logging purposes
to another dedicated machine.

References: http://www.gossamer-threads.com/lists/iptables/devel/68781
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>

include/linux/netfilter/Kbuild
include/linux/netfilter/xt_TEE.h [new file with mode: 0644]
net/ipv4/ip_output.c
net/ipv6/ip6_output.c
net/netfilter/Kconfig
net/netfilter/Makefile
net/netfilter/xt_TEE.c [new file with mode: 0644]

index a5a63e4..48767cd 100644 (file)
@@ -16,6 +16,7 @@ header-y += xt_RATEEST.h
 header-y += xt_SECMARK.h
 header-y += xt_TCPMSS.h
 header-y += xt_TCPOPTSTRIP.h
+header-y += xt_TEE.h
 header-y += xt_TPROXY.h
 header-y += xt_comment.h
 header-y += xt_connbytes.h
diff --git a/include/linux/netfilter/xt_TEE.h b/include/linux/netfilter/xt_TEE.h
new file mode 100644 (file)
index 0000000..55d4a50
--- /dev/null
@@ -0,0 +1,9 @@
+#ifndef _XT_TEE_TARGET_H
+#define _XT_TEE_TARGET_H
+
+struct xt_tee_tginfo {
+       union nf_inet_addr gw;
+       char oif[16];
+};
+
+#endif /* _XT_TEE_TARGET_H */
index f09135e..0abfdde 100644 (file)
@@ -309,6 +309,7 @@ int ip_output(struct sk_buff *skb)
                            ip_finish_output,
                            !(IPCB(skb)->flags & IPSKB_REROUTED));
 }
+EXPORT_SYMBOL_GPL(ip_output);
 
 int ip_queue_xmit(struct sk_buff *skb, int ipfragok)
 {
index c10a38a..d09be7f 100644 (file)
@@ -176,6 +176,7 @@ int ip6_output(struct sk_buff *skb)
                            ip6_finish_output,
                            !(IP6CB(skb)->flags & IP6SKB_REROUTED));
 }
+EXPORT_SYMBOL_GPL(ip6_output);
 
 /*
  *     xmit an sk_buff (used by TCP)
index 8055786..673a6c8 100644 (file)
@@ -502,6 +502,13 @@ config NETFILTER_XT_TARGET_RATEEST
 
          To compile it as a module, choose M here.  If unsure, say N.
 
+config NETFILTER_XT_TARGET_TEE
+       tristate '"TEE" - packet cloning to alternate destiantion'
+       depends on NETFILTER_ADVANCED
+       ---help---
+       This option adds a "TEE" target with which a packet can be cloned and
+       this clone be rerouted to another nexthop.
+
 config NETFILTER_XT_TARGET_TPROXY
        tristate '"TPROXY" target support (EXPERIMENTAL)'
        depends on EXPERIMENTAL
index cd31afe..14e3a8f 100644 (file)
@@ -59,6 +59,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_SECMARK) += xt_SECMARK.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_TPROXY) += xt_TPROXY.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_TCPMSS) += xt_TCPMSS.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_TCPOPTSTRIP) += xt_TCPOPTSTRIP.o
+obj-$(CONFIG_NETFILTER_XT_TARGET_TEE) += xt_TEE.o
 obj-$(CONFIG_NETFILTER_XT_TARGET_TRACE) += xt_TRACE.o
 
 # matches
diff --git a/net/netfilter/xt_TEE.c b/net/netfilter/xt_TEE.c
new file mode 100644 (file)
index 0000000..b3d7301
--- /dev/null
@@ -0,0 +1,256 @@
+/*
+ *     "TEE" target extension for Xtables
+ *     Copyright © Sebastian Claßen, 2007
+ *     Jan Engelhardt, 2007-2010
+ *
+ *     based on ipt_ROUTE.c from Cédric de Launois
+ *     <delaunois@info.ucl.be>
+ *
+ *     This program is free software; you can redistribute it and/or
+ *     modify it under the terms of the GNU General Public License
+ *     version 2 or later, as published by the Free Software Foundation.
+ */
+#include <linux/ip.h>
+#include <linux/module.h>
+#include <linux/route.h>
+#include <linux/skbuff.h>
+#include <net/checksum.h>
+#include <net/icmp.h>
+#include <net/ip.h>
+#include <net/ipv6.h>
+#include <net/ip6_route.h>
+#include <net/route.h>
+#include <linux/netfilter/x_tables.h>
+#include <linux/netfilter/xt_TEE.h>
+
+#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
+#      define WITH_CONNTRACK 1
+#      include <net/netfilter/nf_conntrack.h>
+#endif
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+#      define WITH_IPV6 1
+#endif
+
+static const union nf_inet_addr tee_zero_address;
+
+static struct net *pick_net(struct sk_buff *skb)
+{
+#ifdef CONFIG_NET_NS
+       const struct dst_entry *dst;
+
+       if (skb->dev != NULL)
+               return dev_net(skb->dev);
+       dst = skb_dst(skb);
+       if (dst != NULL && dst->dev != NULL)
+               return dev_net(dst->dev);
+#endif
+       return &init_net;
+}
+
+static bool tee_tg_route_oif(struct flowi *f, struct net *net,
+                            const struct xt_tee_tginfo *info)
+{
+       const struct net_device *dev;
+
+       if (*info->oif != '\0')
+               return true;
+       dev = dev_get_by_name(net, info->oif);
+       if (dev == NULL)
+               return false;
+       f->oif = dev->ifindex;
+       return true;
+}
+
+static bool
+tee_tg_route4(struct sk_buff *skb, const struct xt_tee_tginfo *info)
+{
+       const struct iphdr *iph = ip_hdr(skb);
+       struct net *net = pick_net(skb);
+       struct rtable *rt;
+       struct flowi fl;
+
+       memset(&fl, 0, sizeof(fl));
+       if (!tee_tg_route_oif(&fl, net, info))
+               return false;
+       fl.nl_u.ip4_u.daddr = info->gw.ip;
+       fl.nl_u.ip4_u.tos   = RT_TOS(iph->tos);
+       fl.nl_u.ip4_u.scope = RT_SCOPE_UNIVERSE;
+       if (ip_route_output_key(net, &rt, &fl) != 0)
+               return false;
+
+       dst_release(skb_dst(skb));
+       skb_dst_set(skb, &rt->u.dst);
+       skb->dev      = rt->u.dst.dev;
+       skb->protocol = htons(ETH_P_IP);
+       return true;
+}
+
+static unsigned int
+tee_tg4(struct sk_buff *skb, const struct xt_target_param *par)
+{
+       const struct xt_tee_tginfo *info = par->targinfo;
+       struct iphdr *iph;
+
+       /*
+        * Copy the skb, and route the copy. Will later return %XT_CONTINUE for
+        * the original skb, which should continue on its way as if nothing has
+        * happened. The copy should be independently delivered to the TEE
+        * --gateway.
+        */
+       skb = pskb_copy(skb, GFP_ATOMIC);
+       if (skb == NULL)
+               return XT_CONTINUE;
+
+#ifdef WITH_CONNTRACK
+       /* Avoid counting cloned packets towards the original connection. */
+       nf_conntrack_put(skb->nfct);
+       skb->nfct     = &nf_conntrack_untracked.ct_general;
+       skb->nfctinfo = IP_CT_NEW;
+       nf_conntrack_get(skb->nfct);
+#endif
+       /*
+        * If we are in PREROUTING/INPUT, the checksum must be recalculated
+        * since the length could have changed as a result of defragmentation.
+        *
+        * We also decrease the TTL to mitigate potential TEE loops
+        * between two hosts.
+        *
+        * Set %IP_DF so that the original source is notified of a potentially
+        * decreased MTU on the clone route. IPv6 does this too.
+        */
+       iph = ip_hdr(skb);
+       iph->frag_off |= htons(IP_DF);
+       if (par->hooknum == NF_INET_PRE_ROUTING ||
+           par->hooknum == NF_INET_LOCAL_IN)
+               --iph->ttl;
+       ip_send_check(iph);
+
+       /*
+        * Xtables is not reentrant currently, so a choice has to be made:
+        * 1. return absolute verdict for the original and let the cloned
+        *    packet travel through the chains
+        * 2. let the original continue travelling and not pass the clone
+        *    to Xtables.
+        * #2 is chosen. Normally, we would use ip_local_out for the clone.
+        * Because iph->check is already correct and we don't pass it to
+        * Xtables anyway, a shortcut to dst_output [forwards to ip_output] can
+        * be taken. %IPSKB_REROUTED needs to be set so that ip_output does not
+        * invoke POSTROUTING on the cloned packet.
+        */
+       IPCB(skb)->flags |= IPSKB_REROUTED;
+       if (tee_tg_route4(skb, info))
+               ip_output(skb);
+       else
+               kfree_skb(skb);
+
+       return XT_CONTINUE;
+}
+
+#ifdef WITH_IPV6
+static bool
+tee_tg_route6(struct sk_buff *skb, const struct xt_tee_tginfo *info)
+{
+       const struct ipv6hdr *iph = ipv6_hdr(skb);
+       struct net *net = pick_net(skb);
+       struct dst_entry *dst;
+       struct flowi fl;
+
+       memset(&fl, 0, sizeof(fl));
+       if (!tee_tg_route_oif(&fl, net, info))
+               return false;
+       fl.nl_u.ip6_u.daddr = info->gw.in6;
+       fl.nl_u.ip6_u.flowlabel = ((iph->flow_lbl[0] & 0xF) << 16) |
+                                 (iph->flow_lbl[1] << 8) | iph->flow_lbl[2];
+       dst = ip6_route_output(net, NULL, &fl);
+       if (dst == NULL)
+               return false;
+
+       dst_release(skb_dst(skb));
+       skb_dst_set(skb, dst);
+       skb->dev      = dst->dev;
+       skb->protocol = htons(ETH_P_IPV6);
+       return true;
+}
+
+static unsigned int
+tee_tg6(struct sk_buff *skb, const struct xt_target_param *par)
+{
+       const struct xt_tee_tginfo *info = par->targinfo;
+
+       skb = pskb_copy(skb, GFP_ATOMIC);
+       if (skb == NULL)
+               return XT_CONTINUE;
+
+#ifdef WITH_CONNTRACK
+       nf_conntrack_put(skb->nfct);
+       skb->nfct     = &nf_conntrack_untracked.ct_general;
+       skb->nfctinfo = IP_CT_NEW;
+       nf_conntrack_get(skb->nfct);
+#endif
+       if (par->hooknum == NF_INET_PRE_ROUTING ||
+           par->hooknum == NF_INET_LOCAL_IN) {
+               struct ipv6hdr *iph = ipv6_hdr(skb);
+               --iph->hop_limit;
+       }
+       IP6CB(skb)->flags |= IP6SKB_REROUTED;
+       if (tee_tg_route6(skb, info))
+               ip6_output(skb);
+       else
+               kfree_skb(skb);
+
+       return XT_CONTINUE;
+}
+#endif /* WITH_IPV6 */
+
+static int tee_tg_check(const struct xt_tgchk_param *par)
+{
+       const struct xt_tee_tginfo *info = par->targinfo;
+
+       if (info->oif[sizeof(info->oif)-1] != '\0')
+               return -EINVAL;
+       /* 0.0.0.0 and :: not allowed */
+       return (memcmp(&info->gw, &tee_zero_address,
+              sizeof(tee_zero_address)) == 0) ? -EINVAL : 0;
+}
+
+static struct xt_target tee_tg_reg[] __read_mostly = {
+       {
+               .name       = "TEE",
+               .revision   = 1,
+               .family     = NFPROTO_IPV4,
+               .target     = tee_tg4,
+               .targetsize = sizeof(struct xt_tee_tginfo),
+               .checkentry = tee_tg_check,
+               .me         = THIS_MODULE,
+       },
+#ifdef WITH_IPV6
+       {
+               .name       = "TEE",
+               .revision   = 1,
+               .family     = NFPROTO_IPV6,
+               .target     = tee_tg6,
+               .targetsize = sizeof(struct xt_tee_tginfo),
+               .checkentry = tee_tg_check,
+               .me         = THIS_MODULE,
+       },
+#endif
+};
+
+static int __init tee_tg_init(void)
+{
+       return xt_register_targets(tee_tg_reg, ARRAY_SIZE(tee_tg_reg));
+}
+
+static void __exit tee_tg_exit(void)
+{
+       xt_unregister_targets(tee_tg_reg, ARRAY_SIZE(tee_tg_reg));
+}
+
+module_init(tee_tg_init);
+module_exit(tee_tg_exit);
+MODULE_AUTHOR("Sebastian Claßen <sebastian.classen@freenet.ag>");
+MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>");
+MODULE_DESCRIPTION("Xtables: Reroute packet copy");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("ipt_TEE");
+MODULE_ALIAS("ip6t_TEE");