ip_gre.c

/*
 *	Linux NET3:	GRE over IP protocol decoder.
 *
 *	Authors: Alexey Kuznetsov (kuznet@ms2.inr.ac.ru)
 *
 *	This program is free software; you can redistribute it and/or
 *	modify it under the terms of the GNU General Public License
 *	as published by the Free Software Foundation; either version
 *	2 of the License, or (at your option) any later version.
 *
 */

#include <linux/capability.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <asm/uaccess.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/in.h>
#include <linux/tcp.h>
#include <linux/udp.h>
#include <linux/if_arp.h>
#include <linux/mroute.h>
#include <linux/init.h>
#include <linux/in6.h>
#include <linux/inetdevice.h>
#include <linux/igmp.h>
#include <linux/netfilter_ipv4.h>
#include <linux/if_ether.h>

#include <net/sock.h>
#include <net/ip.h>
#include <net/icmp.h>
#include <net/protocol.h>
#include <net/ipip.h>
#include <net/arp.h>
#include <net/checksum.h>
#include <net/dsfield.h>
#include <net/inet_ecn.h>
#include <net/xfrm.h>
#include <net/net_namespace.h>
#include <net/netns/generic.h>

#ifdef CONFIG_IPV6
#include <net/ipv6.h>
#include <net/ip6_fib.h>
#include <net/ip6_route.h>
#endif

/*
   Problems & solutions
   --------------------

   1. The most important issue is detecting local dead loops.
   They would cause complete host lockup in transmit, which
   would be "resolved" by stack overflow or, if queueing is enabled,
   with infinite looping in net_bh.

   We cannot track such dead loops during route installation,
   it is infeasible task. The most general solutions would be
   to keep skb->encapsulation counter (sort of local ttl),
   and silently drop packet when it expires. It is the best
   solution, but it supposes maintaing new variable in ALL
   skb, even if no tunneling is used.

   Current solution: t->recursion lock breaks dead loops. It looks
   like dev->tbusy flag, but I preferred new variable, because
   the semantics is different. One day, when hard_start_xmit
   will be multithreaded we will have to use skb->encapsulation.



   2. Networking dead loops would not kill routers, but would really
   kill network. IP hop limit plays role of "t->recursion" in this case,
   if we copy it from packet being encapsulated to upper header.
   It is very good solution, but it introduces two problems:

   - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
     do not work over tunnels.
   - traceroute does not work. I planned to relay ICMP from tunnel,
     so that this problem would be solved and traceroute output
     would even more informative. This idea appeared to be wrong:
     only Linux complies to rfc1812 now (yes, guys, Linux is the only
     true router now :-)), all routers (at least, in neighbourhood of mine)
     return only 8 bytes of payload. It is the end.

   Hence, if we want that OSPF worked or traceroute said something reasonable,
   we should search for another solution.

   One of them is to parse packet trying to detect inner encapsulation
   made by our node. It is difficult or even impossible, especially,
   taking into account fragmentation. TO be short, tt is not solution at all.

   Current solution: The solution was UNEXPECTEDLY SIMPLE.
   We force DF flag on tunnels with preconfigured hop limit,
   that is ALL. :-) Well, it does not remove the problem completely,
   but exponential growth of network traffic is changed to linear
   (branches, that exceed pmtu are pruned) and tunnel mtu
   fastly degrades to value <68, where looping stops.
   Yes, it is not good if there exists a router in the loop,
   which does not force DF, even when encapsulating packets have DF set.
   But it is not our problem! Nobody could accuse us, we made
   all that we could make. Even if it is your gated who injected
   fatal route to network, even if it were you who configured
   fatal static route: you are innocent. :-)



   3. Really, ipv4/ipip.c, ipv4/ip_gre.c and ipv6/sit.c contain
   practically identical code. It would be good to glue them
   together, but it is not very evident, how to make them modular.
   sit is integral part of IPv6, ipip and gre are naturally modular.
   We could extract common parts (hash table, ioctl etc)
   to a separate module (ip_tunnel.c).

   Alexey Kuznetsov.
 */

static int ipgre_tunnel_init(struct net_device *dev);
static void ipgre_tunnel_setup(struct net_device *dev);

/* Fallback tunnel: no source, no destination, no key, no options */

static int ipgre_fb_tunnel_init(struct net_device *dev);

#define HASH_SIZE  16

static int ipgre_net_id;
struct ipgre_net {
	struct ip_tunnel *tunnels[4][HASH_SIZE];

	struct net_device *fb_tunnel_dev;
};

/* Tunnel hash table */

/*
   4 hash tables:

   3: (remote,local)
   2: (remote,*)
   1: (*,local)
   0: (*,*)

   We require exact key match i.e. if a key is present in packet
   it will match only tunnel with the same key; if it is not present,
   it will match only keyless tunnel.

   All keysless packets, if not matched configured keyless tunnels
   will match fallback tunnel.
 */

#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)

#define tunnels_r_l	tunnels[3]
#define tunnels_r	tunnels[2]
#define tunnels_l	tunnels[1]
#define tunnels_wc	tunnels[0]

static DEFINE_RWLOCK(ipgre_lock);

/* Given src, dst and key, find appropriate for input tunnel. */

static struct ip_tunnel * ipgre_tunnel_lookup(struct net *net,
		__be32 remote, __be32 local, __be32 key)
{
	unsigned h0 = HASH(remote);
	unsigned h1 = HASH(key);
	struct ip_tunnel *t;
	struct ipgre_net *ign = net_generic(net, ipgre_net_id);

	for (t = ign->tunnels_r_l[h0^h1]; t; t = t->next) {
		if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr) {
			if (t->parms.i_key == key && (t->dev->flags&IFF_UP))
				return t;
		}
	}
	for (t = ign->tunnels_r[h0^h1]; t; t = t->next) {
		if (remote == t->parms.iph.daddr) {
			if (t->parms.i_key == key && (t->dev->flags&IFF_UP))
				return t;
		}
	}
	for (t = ign->tunnels_l[h1]; t; t = t->next) {
		if (local == t->parms.iph.saddr ||
		     (local == t->parms.iph.daddr &&
		      ipv4_is_multicast(local))) {
			if (t->parms.i_key == key && (t->dev->flags&IFF_UP))
				return t;
		}
	}
	for (t = ign->tunnels_wc[h1]; t; t = t->next) {
		if (t->parms.i_key == key && (t->dev->flags&IFF_UP))
			return t;
	}

	if (ign->fb_tunnel_dev->flags&IFF_UP)
		return netdev_priv(ign->fb_tunnel_dev);
	return NULL;
}

static struct ip_tunnel **__ipgre_bucket(struct ipgre_net *ign,
		struct ip_tunnel_parm *parms)
{
	__be32 remote = parms->iph.daddr;
	__be32 local = parms->iph.saddr;
	__be32 key = parms->i_key;
	unsigned h = HASH(key);
	int prio = 0;

	if (local)
		prio |= 1;
	if (remote && !ipv4_is_multicast(remote)) {
		prio |= 2;
		h ^= HASH(remote);
	}

	return &ign->tunnels[prio][h];
}

static inline struct ip_tunnel **ipgre_bucket(struct ipgre_net *ign,
		struct ip_tunnel *t)
{
	return __ipgre_bucket(ign, &t->parms);
}

static void ipgre_tunnel_link(struct ipgre_net *ign, struct ip_tunnel *t)
{
	struct ip_tunnel **tp = ipgre_bucket(ign, t);

	t->next = *tp;
	write_lock_bh(&ipgre_lock);
	*tp = t;
	write_unlock_bh(&ipgre_lock);
}

static void ipgre_tunnel_unlink(struct ipgre_net *ign, struct ip_tunnel *t)
{
	struct ip_tunnel **tp;

	for (tp = ipgre_bucket(ign, t); *tp; tp = &(*tp)->next) {
		if (t == *tp) {
			write_lock_bh(&ipgre_lock);
			*tp = t->next;
			write_unlock_bh(&ipgre_lock);
			break;
		}
	}
}

static struct ip_tunnel * ipgre_tunnel_locate(struct net *net,
		struct ip_tunnel_parm *parms, int create)
{
	__be32 remote = parms->iph.daddr;
	__be32 local = parms->iph.saddr;
	__be32 key = parms->i_key;
	struct ip_tunnel *t, **tp, *nt;
	struct net_device *dev;
	char name[IFNAMSIZ];
	struct ipgre_net *ign = net_generic(net, ipgre_net_id);

	for (tp = __ipgre_bucket(ign, parms); (t = *tp) != NULL; tp = &t->next) {
		if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr) {
			if (key == t->parms.i_key)
				return t;
		}
	}
	if (!create)
		return NULL;

	if (parms->name[0])
		strlcpy(name, parms->name, IFNAMSIZ);
	else
		sprintf(name, "gre%%d");

	dev = alloc_netdev(sizeof(*t), name, ipgre_tunnel_setup);
	if (!dev)
	  return NULL;

	dev_net_set(dev, net);

	if (strchr(name, '%')) {
		if (dev_alloc_name(dev, name) < 0)
			goto failed_free;
	}

	dev->init = ipgre_tunnel_init;
	nt = netdev_priv(dev);
	nt->parms = *parms;

	if (register_netdevice(dev) < 0)
		goto failed_free;

	dev_hold(dev);
	ipgre_tunnel_link(ign, nt);
	return nt;

failed_free:
	free_netdev(dev);
	return NULL;
}

static void ipgre_tunnel_uninit(struct net_device *dev)
{
	struct net *net = dev_net(dev);
	struct ipgre_net *ign = net_generic(net, ipgre_net_id);

	ipgre_tunnel_unlink(ign, netdev_priv(dev));
	dev_put(dev);
}


static void ipgre_err(struct sk_buff *skb, u32 info)
{
#ifndef I_WISH_WORLD_WERE_PERFECT

/* It is not :-( All the routers (except for Linux) return only
   8 bytes of packet payload. It means, that precise relaying of
   ICMP in the real Internet is absolutely infeasible.

   Moreover, Cisco "wise men" put GRE key to the third word
   in GRE header. It makes impossible maintaining even soft state for keyed
   GRE tunnels with enabled checksum. Tell them "thank you".

   Well, I wonder, rfc1812 was written by Cisco employee,
   what the hell these idiots break standrads established
   by themself???
 */

	struct iphdr *iph = (struct iphdr*)skb->data;
	__be16	     *p = (__be16*)(skb->data+(iph->ihl<<2));
	int grehlen = (iph->ihl<<2) + 4;
	const int type = icmp_hdr(skb)->type;
	const int code = icmp_hdr(skb)->code;
	struct ip_tunnel *t;
	__be16 flags;

	flags = p[0];
	if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
		if (flags&(GRE_VERSION|GRE_ROUTING))
			return;
		if (flags&GRE_KEY) {
			grehlen += 4;
			if (flags&GRE_CSUM)
				grehlen += 4;
		}
	}

	/* If only 8 bytes returned, keyed message will be dropped here */
	if (skb_headlen(skb) < grehlen)
		return;

	switch (type) {
	default:
	case ICMP_PARAMETERPROB:
		return;

	case ICMP_DEST_UNREACH:
		switch (code) {
		case ICMP_SR_FAILED:
		case ICMP_PORT_UNREACH:
			/* Impossible event. */
			return;
		case ICMP_FRAG_NEEDED:
			/* Soft state for pmtu is maintained by IP core. */
			return;
		default:
			/* All others are translated to HOST_UNREACH.
			   rfc2003 contains "deep thoughts" about NET_UNREACH,
			   I believe they are just ether pollution. --ANK
			 */
			break;
		}
		break;
	case ICMP_TIME_EXCEEDED:
		if (code != ICMP_EXC_TTL)
			return;
		break;
	}

	read_lock(&ipgre_lock);
	t = ipgre_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr,
			(flags&GRE_KEY) ?
			*(((__be32*)p) + (grehlen>>2) - 1) : 0);
	if (t == NULL || t->parms.iph.daddr == 0 ||
	    ipv4_is_multicast(t->parms.iph.daddr))
		goto out;

	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
		goto out;

	if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
		t->err_count++;
	else
		t->err_count = 1;
	t->err_time = jiffies;
out:
	read_unlock(&ipgre_lock);
	return;
#else
	struct iphdr *iph = (struct iphdr*)dp;
	struct iphdr *eiph;
	__be16	     *p = (__be16*)(dp+(iph->ihl<<2));
	const int type = icmp_hdr(skb)->type;
	const int code = icmp_hdr(skb)->code;
	int rel_type = 0;
	int rel_code = 0;
	__be32 rel_info = 0;
	__u32 n = 0;
	__be16 flags;
	int grehlen = (iph->ihl<<2) + 4;
	struct sk_buff *skb2;
	struct flowi fl;
	struct rtable *rt;

	if (p[1] != htons(ETH_P_IP))
		return;

	flags = p[0];
	if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
		if (flags&(GRE_VERSION|GRE_ROUTING))
			return;
		if (flags&GRE_CSUM)
			grehlen += 4;
		if (flags&GRE_KEY)
			grehlen += 4;
		if (flags&GRE_SEQ)
			grehlen += 4;
	}
	if (len < grehlen + sizeof(struct iphdr))
		return;
	eiph = (struct iphdr*)(dp + grehlen);

	switch (type) {
	default:
		return;
	case ICMP_PARAMETERPROB:
		n = ntohl(icmp_hdr(skb)->un.gateway) >> 24;
		if (n < (iph->ihl<<2))
			return;

		/* So... This guy found something strange INSIDE encapsulated
		   packet. Well, he is fool, but what can we do ?
		 */
		rel_type = ICMP_PARAMETERPROB;
		n -= grehlen;
		rel_info = htonl(n << 24);
		break;

	case ICMP_DEST_UNREACH:
		switch (code) {
		case ICMP_SR_FAILED:
		case ICMP_PORT_UNREACH:
			/* Impossible event. */
			return;
		case ICMP_FRAG_NEEDED:
			/* And it is the only really necessary thing :-) */
			n = ntohs(icmp_hdr(skb)->un.frag.mtu);
			if (n < grehlen+68)
				return;
			n -= grehlen;
			/* BSD 4.2 MORE DOES NOT EXIST IN NATURE. */
			if (n > ntohs(eiph->tot_len))
				return;
			rel_info = htonl(n);
			break;
		default:
			/* All others are translated to HOST_UNREACH.
			   rfc2003 contains "deep thoughts" about NET_UNREACH,
			   I believe, it is just ether pollution. --ANK
			 */
			rel_type = ICMP_DEST_UNREACH;
			rel_code = ICMP_HOST_UNREACH;
			break;
		}
		break;
	case ICMP_TIME_EXCEEDED:
		if (code != ICMP_EXC_TTL)
			return;
		break;
	}

	/* Prepare fake skb to feed it to icmp_send */
	skb2 = skb_clone(skb, GFP_ATOMIC);
	if (skb2 == NULL)
		return;
	dst_release(skb2->dst);
	skb2->dst = NULL;
	skb_pull(skb2, skb->data - (u8*)eiph);
	skb_reset_network_header(skb2);

	/* Try to guess incoming interface */
	memset(&fl, 0, sizeof(fl));
	fl.fl4_dst = eiph->saddr;
	fl.fl4_tos = RT_TOS(eiph->tos);
	fl.proto = IPPROTO_GRE;
	if (ip_route_output_key(dev_net(skb->dev), &rt, &fl)) {
		kfree_skb(skb2);
		return;
	}
	skb2->dev = rt->u.dst.dev;

	/* route "incoming" packet */
	if (rt->rt_flags&RTCF_LOCAL) {
		ip_rt_put(rt);
		rt = NULL;
		fl.fl4_dst = eiph->daddr;
		fl.fl4_src = eiph->saddr;
		fl.fl4_tos = eiph->tos;
		if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
		    rt->u.dst.dev->type != ARPHRD_IPGRE) {
			ip_rt_put(rt);
			kfree_skb(skb2);
			return;
		}
	} else {
		ip_rt_put(rt);
		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, skb2->dev) ||
		    skb2->dst->dev->type != ARPHRD_IPGRE) {
			kfree_skb(skb2);
			return;
		}
	}

	/* change mtu on this route */
	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
		if (n > dst_mtu(skb2->dst)) {
			kfree_skb(skb2);
			return;
		}
		skb2->dst->ops->update_pmtu(skb2->dst, n);
	} else if (type == ICMP_TIME_EXCEEDED) {
		struct ip_tunnel *t = netdev_priv(skb2->dev);
		if (t->parms.iph.ttl) {
			rel_type = ICMP_DEST_UNREACH;
			rel_code = ICMP_HOST_UNREACH;
		}
	}

	icmp_send(skb2, rel_type, rel_code, rel_info);
	kfree_skb(skb2);
#endif
}

static inline void ipgre_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
{
	if (INET_ECN_is_ce(iph->tos)) {
		if (skb->protocol == htons(ETH_P_IP)) {
			IP_ECN_set_ce(ip_hdr(skb));
		} else if (skb->protocol == htons(ETH_P_IPV6)) {
			IP6_ECN_set_ce(ipv6_hdr(skb));
		}
	}
}

static inline u8
ipgre_ecn_encapsulate(u8 tos, struct iphdr *old_iph, struct sk_buff *skb)
{
	u8 inner = 0;
	if (skb->protocol == htons(ETH_P_IP))
		inner = old_iph->tos;
	else if (skb->protocol == htons(ETH_P_IPV6))
		inner = ipv6_get_dsfield((struct ipv6hdr *)old_iph);
	return INET_ECN_encapsulate(tos, inner);
}

static int ipgre_rcv(struct sk_buff *skb)
{
	struct iphdr *iph;
	u8     *h;
	__be16    flags;
	__sum16   csum = 0;
	__be32 key = 0;
	u32    seqno = 0;
	struct ip_tunnel *tunnel;
	int    offset = 4;

	if (!pskb_may_pull(skb, 16))
		goto drop_nolock;

	iph = ip_hdr(skb);
	h = skb->data;
	flags = *(__be16*)h;

	if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
		/* - Version must be 0.
		   - We do not support routing headers.
		 */
		if (flags&(GRE_VERSION|GRE_ROUTING))
			goto drop_nolock;

		if (flags&GRE_CSUM) {
			switch (skb->ip_summed) {
			case CHECKSUM_COMPLETE:
				csum = csum_fold(skb->csum);
				if (!csum)
					break;
				/* fall through */
			case CHECKSUM_NONE:
				skb->csum = 0;
				csum = __skb_checksum_complete(skb);
				skb->ip_summed = CHECKSUM_COMPLETE;
			}
			offset += 4;
		}
		if (flags&GRE_KEY) {
			key = *(__be32*)(h + offset);
			offset += 4;
		}
		if (flags&GRE_SEQ) {
			seqno = ntohl(*(__be32*)(h + offset));
			offset += 4;
		}
	}

	read_lock(&ipgre_lock);
	if ((tunnel = ipgre_tunnel_lookup(dev_net(skb->dev),
					iph->saddr, iph->daddr, key)) != NULL) {
		secpath_reset(skb);

		skb->protocol = *(__be16*)(h + 2);
		/* WCCP version 1 and 2 protocol decoding.
		 * - Change protocol to IP
		 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
		 */
		if (flags == 0 &&
		    skb->protocol == htons(ETH_P_WCCP)) {
			skb->protocol = htons(ETH_P_IP);
			if ((*(h + offset) & 0xF0) != 0x40)
				offset += 4;
		}

		skb->mac_header = skb->network_header;
		__pskb_pull(skb, offset);
		skb_reset_network_header(skb);
		skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
		skb->pkt_type = PACKET_HOST;
#ifdef CONFIG_NET_IPGRE_BROADCAST
		if (ipv4_is_multicast(iph->daddr)) {
			/* Looped back packet, drop it! */
			if (skb->rtable->fl.iif == 0)
				goto drop;
			tunnel->stat.multicast++;
			skb->pkt_type = PACKET_BROADCAST;
		}
#endif

		if (((flags&GRE_CSUM) && csum) ||
		    (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
			tunnel->stat.rx_crc_errors++;
			tunnel->stat.rx_errors++;
			goto drop;
		}
		if (tunnel->parms.i_flags&GRE_SEQ) {
			if (!(flags&GRE_SEQ) ||
			    (tunnel->i_seqno && (s32)(seqno - tunnel->i_seqno) < 0)) {
				tunnel->stat.rx_fifo_errors++;
				tunnel->stat.rx_errors++;
				goto drop;
			}
			tunnel->i_seqno = seqno + 1;
		}
		tunnel->stat.rx_packets++;
		tunnel->stat.rx_bytes += skb->len;
		skb->dev = tunnel->dev;
		dst_release(skb->dst);
		skb->dst = NULL;
		nf_reset(skb);
		ipgre_ecn_decapsulate(iph, skb);
		netif_rx(skb);
		read_unlock(&ipgre_lock);
		return(0);
	}
	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);

drop:
	read_unlock(&ipgre_lock);
drop_nolock:
	kfree_skb(skb);
	return(0);
}

static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
{
	struct ip_tunnel *tunnel = netdev_priv(dev);
	struct net_device_stats *stats = &tunnel->stat;
	struct iphdr  *old_iph = ip_hdr(skb);
	struct iphdr  *tiph;
	u8     tos;
	__be16 df;
	struct rtable *rt;     			/* Route to the other host */
	struct net_device *tdev;			/* Device to other host */
	struct iphdr  *iph;			/* Our new IP header */
	unsigned int max_headroom;		/* The extra header space needed */
	int    gre_hlen;
	__be32 dst;
	int    mtu;

	if (tunnel->recursion++) {
		tunnel->stat.collisions++;
		goto tx_error;
	}

	if (dev->header_ops) {
		gre_hlen = 0;
		tiph = (struct iphdr*)skb->data;
	} else {
		gre_hlen = tunnel->hlen;
		tiph = &tunnel->parms.iph;
	}

	if ((dst = tiph->daddr) == 0) {
		/* NBMA tunnel */

		if (skb->dst == NULL) {
			tunnel->stat.tx_fifo_errors++;
			goto tx_error;
		}

		if (skb->protocol == htons(ETH_P_IP)) {
			rt = skb->rtable;
			if ((dst = rt->rt_gateway) == 0)
				goto tx_error_icmp;
		}
#ifdef CONFIG_IPV6
		else if (skb->protocol == htons(ETH_P_IPV6)) {
			struct in6_addr *addr6;
			int addr_type;
			struct neighbour *neigh = skb->dst->neighbour;

			if (neigh == NULL)
				goto tx_error;

			addr6 = (struct in6_addr*)&neigh->primary_key;
			addr_type = ipv6_addr_type(addr6);

			if (addr_type == IPV6_ADDR_ANY) {
				addr6 = &ipv6_hdr(skb)->daddr;
				addr_type = ipv6_addr_type(addr6);
			}

			if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
				goto tx_error_icmp;

			dst = addr6->s6_addr32[3];
		}
#endif
		else
			goto tx_error;
	}

	tos = tiph->tos;
	if (tos&1) {
		if (skb->protocol == htons(ETH_P_IP))
			tos = old_iph->tos;
		tos &= ~1;
	}

	{
		struct flowi fl = { .oif = tunnel->parms.link,
				    .nl_u = { .ip4_u =
					      { .daddr = dst,
						.saddr = tiph->saddr,
						.tos = RT_TOS(tos) } },
				    .proto = IPPROTO_GRE };
		if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
			tunnel->stat.tx_carrier_errors++;
			goto tx_error;
		}
	}
	tdev = rt->u.dst.dev;

	if (tdev == dev) {
		ip_rt_put(rt);
		tunnel->stat.collisions++;
		goto tx_error;
	}

	df = tiph->frag_off;
	if (df)
		mtu = dst_mtu(&rt->u.dst) - tunnel->hlen;
	else
		mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;

	if (skb->dst)
		skb->dst->ops->update_pmtu(skb->dst, mtu);

	if (skb->protocol == htons(ETH_P_IP)) {
		df |= (old_iph->frag_off&htons(IP_DF));

		if ((old_iph->frag_off&htons(IP_DF)) &&
		    mtu < ntohs(old_iph->tot_len)) {
			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
			ip_rt_put(rt);
			goto tx_error;
		}
	}
#ifdef CONFIG_IPV6
	else if (skb->protocol == htons(ETH_P_IPV6)) {
		struct rt6_info *rt6 = (struct rt6_info*)skb->dst;

		if (rt6 && mtu < dst_mtu(skb->dst) && mtu >= IPV6_MIN_MTU) {
			if ((tunnel->parms.iph.daddr &&
			     !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
			    rt6->rt6i_dst.plen == 128) {
				rt6->rt6i_flags |= RTF_MODIFIED;
				skb->dst->metrics[RTAX_MTU-1] = mtu;
			}
		}

		if (mtu >= IPV6_MIN_MTU && mtu < skb->len - tunnel->hlen + gre_hlen) {
			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
			ip_rt_put(rt);
			goto tx_error;
		}
	}
#endif

	if (tunnel->err_count > 0) {
		if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
			tunnel->err_count--;

			dst_link_failure(skb);
		} else
			tunnel->err_count = 0;
	}

	max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen;

	if (skb_headroom(skb) < max_headroom || skb_shared(skb)||
	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
		if (!new_skb) {
			ip_rt_put(rt);
			stats->tx_dropped++;
			dev_kfree_skb(skb);
			tunnel->recursion--;
			return 0;
		}
		if (skb->sk)
			skb_set_owner_w(new_skb, skb->sk);
		dev_kfree_skb(skb);
		skb = new_skb;
		old_iph = ip_hdr(skb);
	}

	skb->transport_header = skb->network_header;
	skb_push(skb, gre_hlen);
	skb_reset_network_header(skb);
	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
	IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
			      IPSKB_REROUTED);
	dst_release(skb->dst);
	skb->dst = &rt->u.dst;

	/*
	 *	Push down and install the IPIP header.
	 */

	iph 			=	ip_hdr(skb);
	iph->version		=	4;
	iph->ihl		=	sizeof(struct iphdr) >> 2;
	iph->frag_off		=	df;
	iph->protocol		=	IPPROTO_GRE;
	iph->tos		=	ipgre_ecn_encapsulate(tos, old_iph, skb);
	iph->daddr		=	rt->rt_dst;
	iph->saddr		=	rt->rt_src;

	if ((iph->ttl = tiph->ttl) == 0) {
		if (skb->protocol == htons(ETH_P_IP))
			iph->ttl = old_iph->ttl;
#ifdef CONFIG_IPV6
		else if (skb->protocol == htons(ETH_P_IPV6))
			iph->ttl = ((struct ipv6hdr*)old_iph)->hop_limit;
#endif
		else
			iph->ttl = dst_metric(&rt->u.dst, RTAX_HOPLIMIT);
	}

	((__be16*)(iph+1))[0] = tunnel->parms.o_flags;
	((__be16*)(iph+1))[1] = skb->protocol;

	if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
		__be32 *ptr = (__be32*)(((u8*)iph) + tunnel->hlen - 4);

		if (tunnel->parms.o_flags&GRE_SEQ) {
			++tunnel->o_seqno;
			*ptr = htonl(tunnel->o_seqno);
			ptr--;
		}
		if (tunnel->parms.o_flags&GRE_KEY) {
			*ptr = tunnel->parms.o_key;
			ptr--;
		}
		if (tunnel->parms.o_flags&GRE_CSUM) {
			*ptr = 0;
			*(__sum16*)ptr = ip_compute_csum((void*)(iph+1), skb->len - sizeof(struct iphdr));
		}
	}

	nf_reset(skb);

	IPTUNNEL_XMIT();
	tunnel->recursion--;
	return 0;

tx_error_icmp:
	dst_link_failure(skb);

tx_error:
	stats->tx_errors++;
	dev_kfree_skb(skb);
	tunnel->recursion--;
	return 0;
}

static void ipgre_tunnel_bind_dev(struct net_device *dev)
{
	struct net_device *tdev = NULL;
	struct ip_tunnel *tunnel;
	struct iphdr *iph;
	int hlen = LL_MAX_HEADER;
	int mtu = ETH_DATA_LEN;
	int addend = sizeof(struct iphdr) + 4;

	tunnel = netdev_priv(dev);
	iph = &tunnel->parms.iph;

	/* Guess output device to choose reasonable mtu and hard_header_len */

	if (iph->daddr) {
		struct flowi fl = { .oif = tunnel->parms.link,
				    .nl_u = { .ip4_u =
					      { .daddr = iph->daddr,
						.saddr = iph->saddr,
						.tos = RT_TOS(iph->tos) } },
				    .proto = IPPROTO_GRE };
		struct rtable *rt;
		if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
			tdev = rt->u.dst.dev;
			ip_rt_put(rt);
		}
		dev->flags |= IFF_POINTOPOINT;
	}

	if (!tdev && tunnel->parms.link)
		tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);

	if (tdev) {
		hlen = tdev->hard_header_len;
		mtu = tdev->mtu;
	}
	dev->iflink = tunnel->parms.link;

	/* Precalculate GRE options length */
	if (tunnel->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
		if (tunnel->parms.o_flags&GRE_CSUM)
			addend += 4;
		if (tunnel->parms.o_flags&GRE_KEY)
			addend += 4;
		if (tunnel->parms.o_flags&GRE_SEQ)
			addend += 4;
	}
	dev->hard_header_len = hlen + addend;
	dev->mtu = mtu - addend;
	tunnel->hlen = addend;

}

static int
ipgre_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
{
	int err = 0;
	struct ip_tunnel_parm p;
	struct ip_tunnel *t;
	struct net *net = dev_net(dev);
	struct ipgre_net *ign = net_generic(net, ipgre_net_id);

	switch (cmd) {
	case SIOCGETTUNNEL:
		t = NULL;
		if (dev == ign->fb_tunnel_dev) {
			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
				err = -EFAULT;
				break;
			}
			t = ipgre_tunnel_locate(net, &p, 0);
		}
		if (t == NULL)
			t = netdev_priv(dev);
		memcpy(&p, &t->parms, sizeof(p));
		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
			err = -EFAULT;
		break;

	case SIOCADDTUNNEL:
	case SIOCCHGTUNNEL:
		err = -EPERM;
		if (!capable(CAP_NET_ADMIN))