gre: Add kernel GRE support.
authorJesse Gross <jesse@nicira.com>
Sun, 6 Dec 2009 23:10:35 +0000 (15:10 -0800)
committerJesse Gross <jesse@nicira.com>
Mon, 7 Dec 2009 20:48:03 +0000 (12:48 -0800)
This implements the kernel portion of GRE on Linux.  It consists
of a backported module that provides the GRE capabilities of 2.6.32
plus bug fixes to kernels 2.6.18+.

acinclude.m4
datapath/linux-2.6/Makefile.main.in
datapath/linux-2.6/Modules.mk
datapath/linux-2.6/compat-2.6/ip_gre.c [new file with mode: 0644]
debian/openvswitch-switch.init
include/openvswitch/automake.mk
include/openvswitch/gre.h [new file with mode: 0644]
xenserver/etc_init.d_vswitch
xenserver/vswitch-xen.spec

index e37a316dffa97a99d069ed2bd0687f591deca7d9..e4fa5e6210538a52a5db5103fa4e3301f187c77c 100644 (file)
@@ -111,6 +111,16 @@ AC_DEFUN([OVS_CHECK_VETH], [
   fi
 ])
 
+AC_DEFUN([OVS_CHECK_GRE], [
+  AC_MSG_CHECKING([whether to build gre module])
+  if test "$sublevel" -ge 18; then
+    AC_MSG_RESULT([yes])
+    AC_SUBST([BUILD_GRE], 1)
+  else
+    AC_MSG_RESULT([no])
+  fi
+])
+
 AC_DEFUN([OVS_CHECK_LOG2_H], [
   AC_MSG_CHECKING([for $KSRC26/include/linux/log2.h])
   if test -e $KSRC26/include/linux/log2.h; then
@@ -153,6 +163,7 @@ AC_DEFUN([OVS_CHECK_LINUX26_COMPAT], [
                   [OVS_DEFINE([HAVE_PROTO_DATA_VALID])])
   OVS_CHECK_LOG2_H
   OVS_CHECK_VETH
+  OVS_CHECK_GRE
   if cmp -s datapath/linux-2.6/kcompat.h.new \
             datapath/linux-2.6/kcompat.h >/dev/null 2>&1; then
     rm datapath/linux-2.6/kcompat.h.new
index 967e2f7bcff1953fffbcb3d5c1e3c2ebdbc3b2f1..fd231e6968c731dd4df1910fdcc253a80dac3ad1 100644 (file)
@@ -5,6 +5,7 @@ export top_srcdir = @abs_top_srcdir@
 export KSRC = @KBUILD26@
 export VERSION = @VERSION@
 export BUILD_VETH = @BUILD_VETH@
+export BUILD_GRE = @BUILD_GRE@
 
 include $(srcdir)/../Modules.mk
 include $(srcdir)/Modules.mk
index 7892583c2af099696d9e05a191bb509dc8880f31..8d2f969661503156a0633603b6a01bde59de637c 100644 (file)
@@ -47,3 +47,26 @@ dist_modules += veth
 build_modules += $(if $(BUILD_VETH),veth)
 veth_sources = linux-2.6/compat-2.6/veth.c
 veth_headers = 
+
+dist_modules += ip_gre
+build_modules += $(if $(BUILD_GRE),ip_gre)
+ip_gre_sources = \
+       linux-2.6/compat-2.6/ip_gre.c \
+       linux-2.6/compat-2.6/net_namespace-ip_gre.c
+ip_gre_headers = \
+       linux-2.6/compat-2.6/compat26.h \
+       linux-2.6/compat-2.6/include/linux/if.h \
+       linux-2.6/compat-2.6/include/linux/in.h \
+       linux-2.6/compat-2.6/include/linux/inetdevice.h \
+       linux-2.6/compat-2.6/include/linux/if_ether.h \
+       linux-2.6/compat-2.6/include/linux/ip.h \
+       linux-2.6/compat-2.6/include/linux/ipv6.h \
+       linux-2.6/compat-2.6/include/linux/netdevice.h \
+       linux-2.6/compat-2.6/include/linux/skbuff.h \
+       linux-2.6/compat-2.6/include/linux/tcp.h \
+       linux-2.6/compat-2.6/include/linux/types.h \
+       linux-2.6/compat-2.6/include/net/dst.h \
+       linux-2.6/compat-2.6/include/net/ipip.h \
+       linux-2.6/compat-2.6/include/net/netns/generic.h \
+       linux-2.6/compat-2.6/include/net/net_namespace.h \
+       linux-2.6/compat-2.6/include/net/route.h
diff --git a/datapath/linux-2.6/compat-2.6/ip_gre.c b/datapath/linux-2.6/compat-2.6/ip_gre.c
new file mode 100644 (file)
index 0000000..246e864
--- /dev/null
@@ -0,0 +1,1877 @@
+/* ip_gre driver port to Linux 2.6.18 and greater */
+
+#include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
+#define HAVE_NETDEV_STATS
+#endif
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,24)
+#define HAVE_NETDEV_HEADER_OPS
+#endif
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
+#define HAVE_NETDEV_NEEDED_HEADROOM
+#endif
+
+/*
+ *     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/etherdevice.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
+
+#include "compat.h"
+#include "openvswitch/gre.h"
+
+#ifndef GRE_IOCTL_ONLY
+#include <net/rtnetlink.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: HARD_TX_LOCK lock breaks dead loops.
+
+
+
+   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.
+ */
+
+#ifndef GRE_IOCTL_ONLY
+static struct rtnl_link_ops ipgre_link_ops __read_mostly;
+static struct rtnl_link_ops ipgre_tap_ops __read_mostly;
+#endif
+static int ipgre_tunnel_init(struct net_device *dev);
+static void ipgre_tunnel_setup(struct net_device *dev);
+static void ipgre_tap_setup(struct net_device *dev);
+static int ipgre_tunnel_bind_dev(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_device *dev,
+                                             __be32 remote, __be32 local,
+                                             __be32 key, __be16 gre_proto)
+{
+       struct net *net = dev_net(dev);
+       int link = dev->ifindex;
+       unsigned h0 = HASH(remote);
+       unsigned h1 = HASH(key);
+       struct ip_tunnel *t, *cand = NULL;
+       struct ipgre_net *ign = net_generic(net, ipgre_net_id);
+       int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
+                      ARPHRD_ETHER : ARPHRD_IPGRE;
+       int score, cand_score = 4;
+
+       for (t = ign->tunnels_r_l[h0^h1]; t; t = t->next) {
+               if (local != t->parms.iph.saddr ||
+                   remote != t->parms.iph.daddr ||
+                   key != t->parms.i_key ||
+                   !(t->dev->flags & IFF_UP))
+                       continue;
+
+               if (t->dev->type != ARPHRD_IPGRE &&
+                   t->dev->type != dev_type)
+                       continue;
+
+               score = 0;
+               if (t->parms.link != link)
+                       score |= 1;
+               if (t->dev->type != dev_type)
+                       score |= 2;
+               if (score == 0)
+                       return t;
+
+               if (score < cand_score) {
+                       cand = t;
+                       cand_score = score;
+               }
+       }
+
+       for (t = ign->tunnels_r[h0^h1]; t; t = t->next) {
+               if (remote != t->parms.iph.daddr ||
+                   key != t->parms.i_key ||
+                   !(t->dev->flags & IFF_UP))
+                       continue;
+
+               if (t->dev->type != ARPHRD_IPGRE &&
+                   t->dev->type != dev_type)
+                       continue;
+
+               score = 0;
+               if (t->parms.link != link)
+                       score |= 1;
+               if (t->dev->type != dev_type)
+                       score |= 2;
+               if (score == 0)
+                       return t;
+
+               if (score < cand_score) {
+                       cand = t;
+                       cand_score = score;
+               }
+       }
+
+       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))) ||
+                   key != t->parms.i_key ||
+                   !(t->dev->flags & IFF_UP))
+                       continue;
+
+               if (t->dev->type != ARPHRD_IPGRE &&
+                   t->dev->type != dev_type)
+                       continue;
+
+               score = 0;
+               if (t->parms.link != link)
+                       score |= 1;
+               if (t->dev->type != dev_type)
+                       score |= 2;
+               if (score == 0)
+                       return t;
+
+               if (score < cand_score) {
+                       cand = t;
+                       cand_score = score;
+               }
+       }
+
+       for (t = ign->tunnels_wc[h1]; t; t = t->next) {
+               if (t->parms.i_key != key ||
+                   !(t->dev->flags & IFF_UP))
+                       continue;
+
+               if (t->dev->type != ARPHRD_IPGRE &&
+                   t->dev->type != dev_type)
+                       continue;
+
+               score = 0;
+               if (t->parms.link != link)
+                       score |= 1;
+               if (t->dev->type != dev_type)
+                       score |= 2;
+               if (score == 0)
+                       return t;
+
+               if (score < cand_score) {
+                       cand = t;
+                       cand_score = score;
+               }
+       }
+
+       if (cand != NULL)
+               return cand;
+
+       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_find(struct net *net,
+                                          struct ip_tunnel_parm *parms,
+                                          int type)
+{
+       __be32 remote = parms->iph.daddr;
+       __be32 local = parms->iph.saddr;
+       __be32 key = parms->i_key;
+       int link = parms->link;
+       struct ip_tunnel *t, **tp;
+       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 &&
+                   key == t->parms.i_key &&
+                   link == t->parms.link &&
+                   type == t->dev->type)
+                       break;
+
+       return t;
+}
+
+static struct ip_tunnel * ipgre_tunnel_locate(struct net *net,
+               struct ip_tunnel_parm *parms, int gretap, int create)
+{
+       struct ip_tunnel *t, *nt;
+       struct net_device *dev;
+       char name[IFNAMSIZ];
+       struct ipgre_net *ign = net_generic(net, ipgre_net_id);
+
+       t = ipgre_tunnel_find(net, parms, gretap ? ARPHRD_ETHER : ARPHRD_IPGRE);
+       if (t || !create)
+               return t;
+
+       if (parms->name[0])
+               strlcpy(name, parms->name, IFNAMSIZ);
+       else
+               sprintf(name, "gre%%d");
+
+       dev = alloc_netdev(sizeof(*t), name, gretap ? ipgre_tap_setup
+                                                   : 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;
+       }
+
+       if (gretap)
+               random_ether_addr(dev->dev_addr);
+
+#ifndef GRE_IOCTL_ONLY
+       dev->rtnl_link_ops = gretap ? &ipgre_tap_ops : &ipgre_link_ops;
+#endif
+       nt = netdev_priv(dev);
+       nt->parms = *parms;
+
+       dev->mtu = ipgre_tunnel_bind_dev(dev);
+
+       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)
+{
+
+/* 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(skb->dev, iph->daddr, iph->saddr,
+                               flags & GRE_KEY ?
+                               *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
+                               p[1]);
+       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 (time_before(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;
+}
+
+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;
+       __be16 gre_proto;
+       unsigned int len;
+
+       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;
+               }
+       }
+
+       gre_proto = *(__be16 *)(h + 2);
+
+       read_lock(&ipgre_lock);
+       if ((tunnel = ipgre_tunnel_lookup(skb->dev,
+                                         iph->saddr, iph->daddr, key,
+                                         gre_proto))) {
+               struct net_device_stats *stats;
+#ifdef HAVE_NETDEV_STATS
+               stats = &tunnel->dev->stats;
+#else
+               stats = &tunnel->stat;
+#endif
+
+               secpath_reset(skb);
+
+               skb->protocol = gre_proto;
+               /* 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 && gre_proto == 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_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(skb)->fl.iif == 0)
+                               goto drop;
+                       stats->multicast++;
+                       skb->pkt_type = PACKET_BROADCAST;
+               }
+#endif
+
+               if (((flags&GRE_CSUM) && csum) ||
+                   (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
+                       stats->rx_crc_errors++;
+                       stats->rx_errors++;
+                       goto drop;
+               }
+               if (tunnel->parms.i_flags&GRE_SEQ) {
+                       if (!(flags&GRE_SEQ) ||
+                           (tunnel->i_seqno && (s32)(seqno - tunnel->i_seqno) < 0)) {
+                               stats->rx_fifo_errors++;
+                               stats->rx_errors++;
+                               goto drop;
+                       }
+                       tunnel->i_seqno = seqno + 1;
+               }
+
+               len = skb->len;
+
+               /* Warning: All skb pointers will be invalidated! */
+               if (tunnel->dev->type == ARPHRD_ETHER) {
+                       if (!pskb_may_pull(skb, ETH_HLEN)) {
+                               stats->rx_length_errors++;
+                               stats->rx_errors++;
+                               goto drop;
+                       }
+
+                       iph = ip_hdr(skb);
+                       skb->protocol = eth_type_trans(skb, tunnel->dev);
+                       skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
+               }
+
+               stats->rx_packets++;
+               stats->rx_bytes += len;
+               skb->dev = tunnel->dev;
+               skb_dst_drop(skb);
+               nf_reset(skb);
+
+               skb_reset_network_header(skb);
+               ipgre_ecn_decapsulate(iph, skb);
+
+#ifdef CHECKSUM_HW
+               /* XXX: Temporary workaround to avoid a panic when doing
+                * bridging due to multiple meanings of CHECKSUM_HW. */
+               if (skb->ip_summed == CHECKSUM_HW)
+                       skb->ip_summed = CHECKSUM_NONE;
+#endif
+
+               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 netdev_tx_t ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+       struct ip_tunnel *tunnel = netdev_priv(dev);
+       struct net_device_stats *stats;
+       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;
+
+#ifdef HAVE_NETDEV_STATS
+       stats = &tunnel->dev->stats;
+#else
+       stats = &tunnel->stat;
+#endif
+
+       if (dev->type == ARPHRD_ETHER)
+               IPCB(skb)->flags = 0;
+
+#ifdef HAVE_NETDEV_HEADER_OPS
+       if (dev->header_ops && dev->type == ARPHRD_IPGRE) {
+#else
+       if (dev->hard_header && dev->type == ARPHRD_IPGRE) {
+#endif
+               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(skb) == NULL) {
+                       stats->tx_fifo_errors++;
+                       goto tx_error;
+               }
+
+               if (skb->protocol == htons(ETH_P_IP)) {
+                       rt = skb_rtable(skb);
+                       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(skb)->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) {
+               tos = 0;
+               if (skb->protocol == htons(ETH_P_IP))
+                       tos = old_iph->tos;
+       }
+
+       {
+               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)) {
+                       stats->tx_carrier_errors++;
+                       goto tx_error;
+               }
+       }
+       tdev = rt->u.dst.dev;
+
+       if (tdev == dev) {
+               ip_rt_put(rt);
+               stats->collisions++;
+               goto tx_error;
+       }
+
+       df = tiph->frag_off;
+       if (df)
+#ifdef HAVE_NETDEV_NEEDED_HEADROOM
+               mtu = dst_mtu(&rt->u.dst) - dev->hard_header_len - tunnel->hlen;
+#else
+               mtu = dst_mtu(&rt->u.dst) - tunnel->hlen;
+#endif
+       else
+               mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
+
+       if (skb_dst(skb))
+               skb_dst(skb)->ops->update_pmtu(skb_dst(skb), 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(skb);
+
+               if (rt6 && mtu < dst_mtu(skb_dst(skb)) && 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(skb)->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 (time_before(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);
+                       return NETDEV_TX_OK;
+               }
+               if (skb->sk)
+                       skb_set_owner_w(new_skb, skb->sk);
+               dev_kfree_skb(skb);
+               skb = new_skb;
+               old_iph = ip_hdr(skb);
+       }
+
+       skb_reset_transport_header(skb);
+       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);
+       skb_dst_drop(skb);
+       skb_dst_set(skb, &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] = (dev->type == ARPHRD_ETHER) ?
+                                  htons(ETH_P_TEB) : 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();
+       return NETDEV_TX_OK;
+
+tx_error_icmp:
+       dst_link_failure(skb);
+
+tx_error:
+       stats->tx_errors++;
+       dev_kfree_skb(skb);
+       return NETDEV_TX_OK;
+}
+
+static int 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 needed_headroom */
+
+       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);
+               }
+
+               if (dev->type != ARPHRD_ETHER)
+                       dev->flags |= IFF_POINTOPOINT;
+       }
+
+       if (!tdev && tunnel->parms.link)
+               tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
+
+       if (tdev) {
+#ifdef HAVE_NETDEV_NEEDED_HEADROOM
+               hlen = tdev->hard_header_len + tdev->needed_headroom;
+#else
+               hlen = tdev->hard_header_len;
+#endif
+               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;
+       }
+#ifdef HAVE_NETDEV_NEEDED_HEADROOM
+       dev->needed_headroom = hlen + addend;
+       mtu -= dev->hard_header_len + addend;
+#else
+       dev->hard_header_len = hlen + addend;
+       mtu -= addend;
+#endif
+       tunnel->hlen = addend;
+
+       if (mtu < 68)
+               mtu = 68;
+
+       return mtu;
+}
+
+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);
+       int add_tunnel, gretap;
+
+       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, false, 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:
+       case SIOCADDGRETAP:
+       case SIOCCHGGRETAP:
+               err = -EPERM;
+               if (!capable(CAP_NET_ADMIN))
+                       goto done;
+
+               err = -EFAULT;
+               if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
+                       goto done;
+
+               err = -EINVAL;
+               if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE ||
+                   p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) ||
+                   ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)))
+                       goto done;
+
+               add_tunnel = (cmd == SIOCADDTUNNEL || cmd == SIOCADDGRETAP);
+               gretap = (cmd == SIOCADDGRETAP || cmd == SIOCCHGGRETAP);
+
+               if (p.iph.ttl)
+                       p.iph.frag_off |= htons(IP_DF);
+
+               if (!(p.i_flags&GRE_KEY))
+                       p.i_key = 0;
+               if (!(p.o_flags&GRE_KEY))
+                       p.o_key = 0;
+
+               t = ipgre_tunnel_locate(net, &p, gretap, add_tunnel);
+
+               if (dev != ign->fb_tunnel_dev && !add_tunnel) {
+                       if (t != NULL) {
+                               if (t->dev != dev) {
+                                       err = -EEXIST;
+                                       break;
+                               }
+                       } else {
+                               unsigned nflags = 0;
+
+                               t = netdev_priv(dev);
+
+                               if (ipv4_is_multicast(p.iph.daddr))
+                                       nflags = IFF_BROADCAST;
+                               else if (p.iph.daddr)
+                                       nflags = IFF_POINTOPOINT;
+
+                               if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
+                                       err = -EINVAL;
+                                       break;
+                               }
+                               ipgre_tunnel_unlink(ign, t);
+                               t->parms.iph.saddr = p.iph.saddr;
+                               t->parms.iph.daddr = p.iph.daddr;
+                               t->parms.i_key = p.i_key;
+                               t->parms.o_key = p.o_key;
+                               memcpy(dev->dev_addr, &p.iph.saddr, 4);
+                               memcpy(dev->broadcast, &p.iph.daddr, 4);
+                               ipgre_tunnel_link(ign, t);
+                               netdev_state_change(dev);
+                       }
+               }
+
+               if (t) {
+                       err = 0;
+                       if (!add_tunnel) {
+                               t->parms.iph.ttl = p.iph.ttl;
+                               t->parms.iph.tos = p.iph.tos;
+                               t->parms.iph.frag_off = p.iph.frag_off;
+                               if (t->parms.link != p.link) {
+                                       t->parms.link = p.link;
+                                       dev->mtu = ipgre_tunnel_bind_dev(dev);
+                                       netdev_state_change(dev);
+                               }
+                       }
+                       if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
+                               err = -EFAULT;
+               } else
+                       err = (add_tunnel ? -ENOBUFS : -ENOENT);
+               break;
+
+       case SIOCDELTUNNEL:
+               err = -EPERM;
+               if (!capable(CAP_NET_ADMIN))
+                       goto done;
+
+               if (dev == ign->fb_tunnel_dev) {
+                       err = -EFAULT;
+                       if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
+                               goto done;
+                       err = -ENOENT;
+                       if ((t = ipgre_tunnel_locate(net, &p, false, 0)) == NULL)
+                               goto done;
+                       err = -EPERM;
+                       if (t == netdev_priv(ign->fb_tunnel_dev))
+                               goto done;
+                       dev = t->dev;
+               }
+               unregister_netdevice(dev);
+               err = 0;
+               break;
+
+       default:
+               err = -EINVAL;
+       }
+
+done:
+       return err;
+}
+
+#ifndef HAVE_NETDEV_STATS
+static struct net_device_stats *ipgre_tunnel_get_stats(struct net_device *dev)
+{
+       return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
+}
+#endif
+
+static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
+{
+       struct ip_tunnel *tunnel = netdev_priv(dev);
+       if (new_mtu < 68 ||
+#ifdef HAVE_NETDEV_NEEDED_HEADROOM
+       new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
+#else
+       new_mtu > 0xFFF8 - tunnel->hlen)
+#endif
+               return -EINVAL;
+       dev->mtu = new_mtu;
+       return 0;
+}
+
+/* Nice toy. Unfortunately, useless in real life :-)
+   It allows to construct virtual multiprotocol broadcast "LAN"
+   over the Internet, provided multicast routing is tuned.
+
+
+   I have no idea was this bicycle invented before me,
+   so that I had to set ARPHRD_IPGRE to a random value.
+   I have an impression, that Cisco could make something similar,
+   but this feature is apparently missing in IOS<=11.2(8).
+
+   I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
+   with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
+
+   ping -t 255 224.66.66.66
+
+   If nobody answers, mbone does not work.
+
+   ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
+   ip addr add 10.66.66.<somewhat>/24 dev Universe
+   ifconfig Universe up
+   ifconfig Universe add fe80::<Your_real_addr>/10
+   ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
+   ftp 10.66.66.66
+   ...
+   ftp fec0:6666:6666::193.233.7.65
+   ...
+
+ */
+
+#ifdef HAVE_NETDEV_HEADER_OPS
+static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
+                       unsigned short type,
+                       const void *daddr, const void *saddr, unsigned len)
+#else
+static int ipgre_header(struct sk_buff *skb, struct net_device *dev, unsigned short type,
+                       void *daddr, void *saddr, unsigned len)
+#endif
+{
+       struct ip_tunnel *t = netdev_priv(dev);
+       struct iphdr *iph = (struct iphdr *)skb_push(skb, t->hlen);
+       __be16 *p = (__be16*)(iph+1);
+
+       memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
+       p[0]            = t->parms.o_flags;
+       p[1]            = htons(type);
+
+       /*
+        *      Set the source hardware address.
+        */
+
+       if (saddr)
+               memcpy(&iph->saddr, saddr, 4);
+
+       if (daddr) {
+               memcpy(&iph->daddr, daddr, 4);
+               return t->hlen;
+       }
+       if (iph->daddr && !ipv4_is_multicast(iph->daddr))
+               return t->hlen;
+
+       return -t->hlen;
+}
+
+#ifdef HAVE_NETDEV_HEADER_OPS
+static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
+#else
+static int ipgre_header_parse(struct sk_buff *skb, unsigned char *haddr)
+#endif
+{
+       struct iphdr *iph = (struct iphdr *) skb_mac_header(skb);
+       memcpy(haddr, &iph->saddr, 4);
+       return 4;
+}
+
+#ifdef HAVE_NETDEV_HEADER_OPS
+static const struct header_ops ipgre_header_ops = {
+       .create = ipgre_header,
+       .parse  = ipgre_header_parse,
+};
+#endif
+
+#ifdef CONFIG_NET_IPGRE_BROADCAST
+static int ipgre_open(struct net_device *dev)
+{
+       struct ip_tunnel *t = netdev_priv(dev);
+
+       if (ipv4_is_multicast(t->parms.iph.daddr)) {
+               struct flowi fl = { .oif = t->parms.link,
+                                   .nl_u = { .ip4_u =
+                                             { .daddr = t->parms.iph.daddr,
+                                               .saddr = t->parms.iph.saddr,
+                                               .tos = RT_TOS(t->parms.iph.tos) } },
+                                   .proto = IPPROTO_GRE };
+               struct rtable *rt;
+               if (ip_route_output_key(dev_net(dev), &rt, &fl))
+                       return -EADDRNOTAVAIL;
+               dev = rt->u.dst.dev;
+               ip_rt_put(rt);
+               if (__in_dev_get_rtnl(dev) == NULL)
+                       return -EADDRNOTAVAIL;
+               t->mlink = dev->ifindex;
+               ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
+       }
+       return 0;
+}
+
+static int ipgre_close(struct net_device *dev)
+{
+       struct ip_tunnel *t = netdev_priv(dev);
+
+       if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
+               struct in_device *in_dev;
+               in_dev = inetdev_by_index(dev_net(dev), t->mlink);
+               if (in_dev) {
+                       ip_mc_dec_group(in_dev, t->parms.iph.daddr);
+                       in_dev_put(in_dev);
+               }
+       }
+       return 0;
+}
+
+#endif
+
+#ifdef HAVE_NET_DEVICE_OPS
+static const struct net_device_ops ipgre_netdev_ops = {
+       .ndo_init               = ipgre_tunnel_init,
+       .ndo_uninit             = ipgre_tunnel_uninit,
+#ifdef CONFIG_NET_IPGRE_BROADCAST
+       .ndo_open               = ipgre_open,
+       .ndo_stop               = ipgre_close,
+#endif
+       .ndo_start_xmit         = ipgre_tunnel_xmit,
+       .ndo_do_ioctl           = ipgre_tunnel_ioctl,
+       .ndo_change_mtu         = ipgre_tunnel_change_mtu,
+};
+#endif
+
+static void ipgre_tunnel_setup(struct net_device *dev)
+{
+#ifdef HAVE_NET_DEVICE_OPS
+       dev->netdev_ops         = &ipgre_netdev_ops;
+#else
+       dev->init               = ipgre_tunnel_init;
+       dev->uninit             = ipgre_tunnel_uninit;
+       dev->hard_start_xmit    = ipgre_tunnel_xmit;
+#ifndef HAVE_NETDEV_STATS
+       dev->get_stats          = ipgre_tunnel_get_stats;
+#endif
+       dev->do_ioctl           = ipgre_tunnel_ioctl;
+       dev->change_mtu         = ipgre_tunnel_change_mtu;
+#endif /* HAVE_NET_DEVICE_OPS */
+       dev->destructor         = free_netdev;
+
+       dev->type               = ARPHRD_IPGRE;
+#ifdef HAVE_NETDEV_NEEDED_HEADROOM
+       dev->needed_headroom    = LL_MAX_HEADER + sizeof(struct iphdr) + 4;
+#else
+       dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr) + 4;
+#endif
+       dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr) - 4;
+       dev->flags              = IFF_NOARP;
+       dev->iflink             = 0;
+       dev->addr_len           = 4;
+       dev->features           |= NETIF_F_NETNS_LOCAL;
+       dev->priv_flags         &= ~IFF_XMIT_DST_RELEASE;
+}
+
+static int ipgre_tunnel_init(struct net_device *dev)
+{
+       struct ip_tunnel *tunnel;
+       struct iphdr *iph;
+
+       tunnel = netdev_priv(dev);
+       iph = &tunnel->parms.iph;
+
+       tunnel->dev = dev;
+       strcpy(tunnel->parms.name, dev->name);
+
+       memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
+       memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
+
+       if (iph->daddr) {
+#ifdef CONFIG_NET_IPGRE_BROADCAST
+               if (ipv4_is_multicast(iph->daddr)) {
+                       if (!iph->saddr)
+                               return -EINVAL;
+                       dev->flags = IFF_BROADCAST;
+#ifdef HAVE_NETDEV_HEADER_OPS
+                       dev->header_ops = &ipgre_header_ops;
+#else
+                       dev->hard_header = ipgre_header;
+                       dev->hard_header_parse = ipgre_header_parse;
+#endif
+#ifndef HAVE_NET_DEVICE_OPS
+                       dev->open = ipgre_open;
+                       dev->stop = ipgre_close;
+#endif
+               }
+#endif
+       } else {
+#ifdef HAVE_NETDEV_HEADER_OPS
+               dev->header_ops = &ipgre_header_ops;
+#else
+               dev->hard_header = ipgre_header;
+               dev->hard_header_parse = ipgre_header_parse;
+#endif
+       }
+
+       return 0;
+}
+
+#ifdef HAVE_NET_DEVICE_OPS
+static void ipgre_fb_tunnel_init(struct net_device *dev)
+#else
+static int ipgre_fb_tunnel_init(struct net_device *dev)
+#endif
+{
+       struct ip_tunnel *tunnel = netdev_priv(dev);
+       struct iphdr *iph = &tunnel->parms.iph;
+       struct ipgre_net *ign = net_generic(dev_net(dev), ipgre_net_id);
+
+       tunnel->dev = dev;
+       strcpy(tunnel->parms.name, dev->name);
+
+       iph->version            = 4;
+       iph->protocol           = IPPROTO_GRE;
+       iph->ihl                = 5;
+       tunnel->hlen            = sizeof(struct iphdr) + 4;
+
+       dev_hold(dev);
+       ign->tunnels_wc[0]      = tunnel;
+
+#ifndef HAVE_NET_DEVICE_OPS
+       return 0;
+#endif
+}
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,32)
+static struct net_protocol ipgre_protocol = {
+#else
+static const struct net_protocol ipgre_protocol = {
+#endif
+       .handler        =       ipgre_rcv,
+       .err_handler    =       ipgre_err,
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
+       .netns_ok       =       1,
+#endif
+};
+
+static void ipgre_destroy_tunnels(struct ipgre_net *ign)
+{
+       int prio;
+
+       for (prio = 0; prio < 4; prio++) {
+               int h;
+               for (h = 0; h < HASH_SIZE; h++) {
+                       struct ip_tunnel *t;
+                       while ((t = ign->tunnels[prio][h]) != NULL)
+                               unregister_netdevice(t->dev);
+               }
+       }
+}
+
+static int ipgre_init_net(struct net *net)
+{
+       int err;
+       struct ipgre_net *ign;
+
+       err = -ENOMEM;
+       ign = kzalloc(sizeof(struct ipgre_net), GFP_KERNEL);
+       if (ign == NULL)
+               goto err_alloc;
+
+       err = net_assign_generic(net, ipgre_net_id, ign);
+       if (err < 0)
+               goto err_assign;
+
+       ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), GRE_IOCTL_DEVICE,
+                                          ipgre_tunnel_setup);
+       if (!ign->fb_tunnel_dev) {
+               err = -ENOMEM;
+               goto err_alloc_dev;
+       }
+       dev_net_set(ign->fb_tunnel_dev, net);
+
+#ifdef HAVE_NET_DEVICE_OPS
+       ipgre_fb_tunnel_init(ign->fb_tunnel_dev);
+#else
+       ign->fb_tunnel_dev->init = ipgre_fb_tunnel_init;
+#endif
+#ifndef GRE_IOCTL_ONLY
+       ign->fb_tunnel_dev->rtnl_link_ops = &ipgre_link_ops;
+#endif
+
+       if ((err = register_netdev(ign->fb_tunnel_dev)))
+               goto err_reg_dev;
+
+       return 0;
+
+err_reg_dev:
+       free_netdev(ign->fb_tunnel_dev);
+err_alloc_dev:
+       /* nothing */
+err_assign:
+       kfree(ign);
+err_alloc:
+       return err;
+}
+
+static void ipgre_exit_net(struct net *net)
+{
+       struct ipgre_net *ign;
+
+       ign = net_generic(net, ipgre_net_id);
+       rtnl_lock();
+       ipgre_destroy_tunnels(ign);
+       rtnl_unlock();
+       kfree(ign);
+}
+
+static struct pernet_operations ipgre_net_ops = {
+       .init = ipgre_init_net,
+       .exit = ipgre_exit_net,
+};
+
+static int ipgre_tap_init(struct net_device *dev)
+{
+       struct ip_tunnel *tunnel;
+
+       tunnel = netdev_priv(dev);
+
+       tunnel->dev = dev;
+       strcpy(tunnel->parms.name, dev->name);
+
+       ipgre_tunnel_bind_dev(dev);
+
+       return 0;
+}
+
+#ifdef HAVE_NET_DEVICE_OPS
+static const struct net_device_ops ipgre_tap_netdev_ops = {
+       .ndo_init               = ipgre_tap_init,
+       .ndo_uninit             = ipgre_tunnel_uninit,
+       .ndo_start_xmit         = ipgre_tunnel_xmit,
+       .ndo_set_mac_address    = eth_mac_addr,
+       .ndo_validate_addr      = eth_validate_addr,
+       .ndo_do_ioctl           = ipgre_tunnel_ioctl,
+       .ndo_change_mtu         = ipgre_tunnel_change_mtu,
+};
+#endif
+
+static void ipgre_tap_setup(struct net_device *dev)
+{
+       ether_setup(dev);
+
+#ifdef HAVE_NET_DEVICE_OPS
+       dev->netdev_ops         = &ipgre_tap_netdev_ops;
+#else
+       dev->init               = ipgre_tap_init;
+       dev->uninit             = ipgre_tunnel_uninit;
+       dev->hard_start_xmit    = ipgre_tunnel_xmit;
+#ifndef HAVE_NETDEV_STATS
+       dev->get_stats          = ipgre_tunnel_get_stats;
+#endif
+       dev->do_ioctl           = ipgre_tunnel_ioctl;
+       dev->change_mtu         = ipgre_tunnel_change_mtu;
+#endif /* HAVE_NET_DEVICE_OPS */
+       dev->destructor         = free_netdev;
+
+       dev->iflink             = 0;
+       dev->features           |= NETIF_F_NETNS_LOCAL;
+}
+
+#ifndef GRE_IOCTL_ONLY
+static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
+{
+       __be16 flags;
+
+       if (!data)
+               return 0;
+
+       flags = 0;
+       if (data[IFLA_GRE_IFLAGS])
+               flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
+       if (data[IFLA_GRE_OFLAGS])
+               flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
+       if (flags & (GRE_VERSION|GRE_ROUTING))
+               return -EINVAL;
+
+       return 0;
+}
+
+static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
+{
+       __be32 daddr;
+
+       if (tb[IFLA_ADDRESS]) {
+               if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
+                       return -EINVAL;
+               if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
+                       return -EADDRNOTAVAIL;
+       }
+
+       if (!data)
+               goto out;
+
+       if (data[IFLA_GRE_REMOTE]) {
+               memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
+               if (!daddr)
+                       return -EINVAL;
+       }
+
+out:
+       return ipgre_tunnel_validate(tb, data);
+}
+
+static void ipgre_netlink_parms(struct nlattr *data[],
+                               struct ip_tunnel_parm *parms)
+{
+       memset(parms, 0, sizeof(*parms));
+
+       parms->iph.protocol = IPPROTO_GRE;
+
+       if (!data)
+               return;
+
+       if (data[IFLA_GRE_LINK])
+               parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
+
+       if (data[IFLA_GRE_IFLAGS])
+               parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
+
+       if (data[IFLA_GRE_OFLAGS])
+               parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
+
+       if (data[IFLA_GRE_IKEY])
+               parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
+
+       if (data[IFLA_GRE_OKEY])
+               parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
+
+       if (data[IFLA_GRE_LOCAL])
+               parms->iph.saddr = nla_get_be32(data[IFLA_GRE_LOCAL]);
+
+       if (data[IFLA_GRE_REMOTE])
+               parms->iph.daddr = nla_get_be32(data[IFLA_GRE_REMOTE]);
+
+       if (data[IFLA_GRE_TTL])
+               parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
+
+       if (data[IFLA_GRE_TOS])
+               parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
+
+       if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC]))
+               parms->iph.frag_off = htons(IP_DF);
+}
+
+static int ipgre_newlink(struct net_device *dev, struct nlattr *tb[],
+                        struct nlattr *data[])
+{
+       struct ip_tunnel *nt;
+       struct net *net = dev_net(dev);
+       struct ipgre_net *ign = net_generic(net, ipgre_net_id);
+       int mtu;
+       int err;
+
+       nt = netdev_priv(dev);
+       ipgre_netlink_parms(data, &nt->parms);
+
+       if (ipgre_tunnel_find(net, &nt->parms, dev->type))
+               return -EEXIST;
+
+       if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
+               random_ether_addr(dev->dev_addr);
+
+       mtu = ipgre_tunnel_bind_dev(dev);
+       if (!tb[IFLA_MTU])
+               dev->mtu = mtu;
+
+       err = register_netdevice(dev);
+       if (err)
+               goto out;
+
+       dev_hold(dev);
+       ipgre_tunnel_link(ign, nt);
+
+out:
+       return err;
+}
+
+static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
+                           struct nlattr *data[])
+{
+       struct ip_tunnel *t, *nt;
+       struct net *net = dev_net(dev);
+       struct ipgre_net *ign = net_generic(net, ipgre_net_id);
+       struct ip_tunnel_parm p;
+       int mtu;
+
+       if (dev == ign->fb_tunnel_dev)
+               return -EINVAL;
+
+       nt = netdev_priv(dev);
+       ipgre_netlink_parms(data, &p);
+
+       t = ipgre_tunnel_locate(net, &p, false, 0);
+
+       if (t) {
+               if (t->dev != dev)
+                       return -EEXIST;
+       } else {
+               t = nt;
+
+               if (dev->type != ARPHRD_ETHER) {
+                       unsigned nflags = 0;
+
+                       if (ipv4_is_multicast(p.iph.daddr))
+                               nflags = IFF_BROADCAST;
+                       else if (p.iph.daddr)
+                               nflags = IFF_POINTOPOINT;
+
+                       if ((dev->flags ^ nflags) &
+                           (IFF_POINTOPOINT | IFF_BROADCAST))
+                               return -EINVAL;
+               }
+
+               ipgre_tunnel_unlink(ign, t);
+               t->parms.iph.saddr = p.iph.saddr;
+               t->parms.iph.daddr = p.iph.daddr;
+               t->parms.i_key = p.i_key;
+               if (dev->type != ARPHRD_ETHER) {
+                       memcpy(dev->dev_addr, &p.iph.saddr, 4);
+                       memcpy(dev->broadcast, &p.iph.daddr, 4);
+               }
+               ipgre_tunnel_link(ign, t);
+               netdev_state_change(dev);
+       }
+
+       t->parms.o_key = p.o_key;
+       t->parms.iph.ttl = p.iph.ttl;
+       t->parms.iph.tos = p.iph.tos;
+       t->parms.iph.frag_off = p.iph.frag_off;
+
+       if (t->parms.link != p.link) {
+               t->parms.link = p.link;
+               mtu = ipgre_tunnel_bind_dev(dev);
+               if (!tb[IFLA_MTU])
+                       dev->mtu = mtu;
+               netdev_state_change(dev);
+       }
+
+       return 0;
+}
+
+static size_t ipgre_get_size(const struct net_device *dev)
+{
+       return
+               /* IFLA_GRE_LINK */
+               nla_total_size(4) +
+               /* IFLA_GRE_IFLAGS */
+               nla_total_size(2) +
+               /* IFLA_GRE_OFLAGS */
+               nla_total_size(2) +
+               /* IFLA_GRE_IKEY */
+               nla_total_size(4) +
+               /* IFLA_GRE_OKEY */
+               nla_total_size(4) +
+               /* IFLA_GRE_LOCAL */
+               nla_total_size(4) +
+               /* IFLA_GRE_REMOTE */
+               nla_total_size(4) +
+               /* IFLA_GRE_TTL */
+               nla_total_size(1) +
+               /* IFLA_GRE_TOS */
+               nla_total_size(1) +
+               /* IFLA_GRE_PMTUDISC */
+               nla_total_size(1) +
+               0;
+}
+
+static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
+{
+       struct ip_tunnel *t = netdev_priv(dev);
+       struct ip_tunnel_parm *p = &t->parms;
+
+       NLA_PUT_U32(skb, IFLA_GRE_LINK, p->link);
+       NLA_PUT_BE16(skb, IFLA_GRE_IFLAGS, p->i_flags);
+       NLA_PUT_BE16(skb, IFLA_GRE_OFLAGS, p->o_flags);
+       NLA_PUT_BE32(skb, IFLA_GRE_IKEY, p->i_key);
+       NLA_PUT_BE32(skb, IFLA_GRE_OKEY, p->o_key);
+       NLA_PUT_BE32(skb, IFLA_GRE_LOCAL, p->iph.saddr);
+       NLA_PUT_BE32(skb, IFLA_GRE_REMOTE, p->iph.daddr);
+       NLA_PUT_U8(skb, IFLA_GRE_TTL, p->iph.ttl);
+       NLA_PUT_U8(skb, IFLA_GRE_TOS, p->iph.tos);
+       NLA_PUT_U8(skb, IFLA_GRE_PMTUDISC, !!(p->iph.frag_off & htons(IP_DF)));
+
+       return 0;
+
+nla_put_failure:
+       return -EMSGSIZE;
+}
+
+static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
+       [IFLA_GRE_LINK]         = { .type = NLA_U32 },
+       [IFLA_GRE_IFLAGS]       = { .type = NLA_U16 },
+       [IFLA_GRE_OFLAGS]       = { .type = NLA_U16 },
+       [IFLA_GRE_IKEY]         = { .type = NLA_U32 },
+       [IFLA_GRE_OKEY]         = { .type = NLA_U32 },
+       [IFLA_GRE_LOCAL]        = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
+       [IFLA_GRE_REMOTE]       = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
+       [IFLA_GRE_TTL]          = { .type = NLA_U8 },
+       [IFLA_GRE_TOS]          = { .type = NLA_U8 },
+       [IFLA_GRE_PMTUDISC]     = { .type = NLA_U8 },
+};
+
+static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
+       .kind           = "gre",
+       .maxtype        = IFLA_GRE_MAX,
+       .policy         = ipgre_policy,
+       .priv_size      = sizeof(struct ip_tunnel),
+       .setup          = ipgre_tunnel_setup,
+       .validate       = ipgre_tunnel_validate,
+       .newlink        = ipgre_newlink,
+       .changelink     = ipgre_changelink,
+       .get_size       = ipgre_get_size,
+       .fill_info      = ipgre_fill_info,
+};
+
+static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
+       .kind           = "gretap",
+       .maxtype        = IFLA_GRE_MAX,
+       .policy         = ipgre_policy,
+       .priv_size      = sizeof(struct ip_tunnel),
+       .setup          = ipgre_tap_setup,
+       .validate       = ipgre_tap_validate,
+       .newlink        = ipgre_newlink,
+       .changelink     = ipgre_changelink,
+       .get_size       = ipgre_get_size,
+       .fill_info      = ipgre_fill_info,
+};
+#endif
+
+/*
+ *     And now the modules code and kernel interface.
+ */
+
+static int __init ipgre_init(void)
+{
+       int err;
+
+       printk(KERN_INFO "GRE over IPv4 tunneling driver\n");
+
+       if (inet_add_protocol(&ipgre_protocol, IPPROTO_GRE) < 0) {
+               printk(KERN_INFO "ipgre init: can't add protocol\n");
+               return -EAGAIN;
+       }
+
+       err = register_pernet_gen_device(&ipgre_net_id, &ipgre_net_ops);
+       if (err < 0)
+               goto gen_device_failed;
+
+#ifndef GRE_IOCTL_ONLY
+       err = rtnl_link_register(&ipgre_link_ops);
+       if (err < 0)
+               goto rtnl_link_failed;
+
+       err = rtnl_link_register(&ipgre_tap_ops);
+       if (err < 0)
+               goto tap_ops_failed;
+#endif
+
+out:
+       return err;
+
+#ifndef GRE_IOCTL_ONLY
+tap_ops_failed:
+       rtnl_link_unregister(&ipgre_link_ops);
+rtnl_link_failed:
+       unregister_pernet_gen_device(ipgre_net_id, &ipgre_net_ops);
+#endif
+gen_device_failed:
+       inet_del_protocol(&ipgre_protocol, IPPROTO_GRE);
+       goto out;
+
+}
+
+static void __exit ipgre_fini(void)
+{
+#ifndef GRE_IOCTL_ONLY
+       rtnl_link_unregister(&ipgre_tap_ops);
+       rtnl_link_unregister(&ipgre_link_ops);
+#endif
+       unregister_pernet_gen_device(ipgre_net_id, &ipgre_net_ops);
+       if (inet_del_protocol(&ipgre_protocol, IPPROTO_GRE) < 0)
+               printk(KERN_INFO "ipgre close: can't remove protocol\n");
+}
+
+module_init(ipgre_init);
+module_exit(ipgre_fini);
+MODULE_DESCRIPTION("GRE over IPv4 tunneling driver");
+MODULE_LICENSE("GPL");
+#ifndef GRE_IOCTL_ONLY
+MODULE_ALIAS_RTNL_LINK("gre");
+MODULE_ALIAS_RTNL_LINK("gretap");
+#endif
+
index da4ec692d202c20f2abb0fa951cf11af512d54b1..dd4bb05965bfd14dea8de92d84518687db6032b3 100755 (executable)
@@ -261,6 +261,18 @@ case "$1" in
             exit 1
         fi
 
+        echo -n "Loading ip_gre_mod: "
+        if grep -q '^ip_gre$' /proc/modules; then
+            modprobe -r ip_gre
+        fi
+        if grep -q '^ip_gre_mod$' /proc/modules; then
+            echo "already loaded, nothing to do."
+        elif modprobe ip_gre_mod; then
+            echo "success."
+        else
+            echo "could not find module."
+        fi
+
         for netdev in $NETDEVS; do
             check_op "Removing IP address from $netdev" ifconfig $netdev 0.0.0.0
         done
@@ -386,6 +398,10 @@ case "$1" in
        echo "$NAME."
 
         check_op "Deleting datapath" ovs-dpctl del-dp of0
+        echo -n "Unloading ip_gre module: "
+        if modprobe -r ip_gre_mod; then
+            echo "success."
+        fi
         check_op "Unloading kernel module" modprobe -r openvswitch_mod
        ;;
     force-stop)
index 889a21f5667114bc1221367982137ab632d953c3..92e071884181568a8c4adbeb6ab84405b91e5368 100644 (file)
@@ -1,4 +1,5 @@
 noinst_HEADERS += \
+       include/openvswitch/gre.h \
        include/openvswitch/brcompat-netlink.h \
        include/openvswitch/datapath-protocol.h
 
diff --git a/include/openvswitch/gre.h b/include/openvswitch/gre.h
new file mode 100644 (file)
index 0000000..2b24cf6
--- /dev/null
@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2008, 2009 Nicira Networks.
+ *
+ * This file is offered under your choice of two licenses: Apache 2.0 or GNU
+ * GPL 2.0 or later.  The permission statements for each of these licenses is
+ * given below.  You may license your modifications to this file under either
+ * of these licenses or both.  If you wish to license your modifications under
+ * only one of these licenses, delete the permission text for the other
+ * license.
+ *
+ * ----------------------------------------------------------------------
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * ----------------------------------------------------------------------
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+ * ----------------------------------------------------------------------
+ */
+
+#ifndef OPENVSWITCH_GRE_H
+#define OPENVSWITCH_GRE_H 1
+
+#include <linux/if_tunnel.h>
+#include <linux/version.h>
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
+#define GRE_IOCTL_ONLY
+#elif LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)
+enum
+{
+       IFLA_GRE_UNSPEC,
+       IFLA_GRE_LINK,
+       IFLA_GRE_IFLAGS,
+       IFLA_GRE_OFLAGS,
+       IFLA_GRE_IKEY,
+       IFLA_GRE_OKEY,
+       IFLA_GRE_LOCAL,
+       IFLA_GRE_REMOTE,
+       IFLA_GRE_TTL,
+       IFLA_GRE_TOS,
+       IFLA_GRE_PMTUDISC,
+       __IFLA_GRE_MAX,
+};
+
+#define IFLA_GRE_MAX   (__IFLA_GRE_MAX - 1)
+#endif
+
+#define GRE_IOCTL_DEVICE "gre0"
+
+#define SIOCGETGRETAP  SIOCGETTUNNEL
+#define SIOCADDGRETAP  (SIOCDEVPRIVATE + 10)
+#define SIOCDELGRETAP  SIOCDELTUNNEL
+#define SIOCCHGGRETAP  (SIOCDEVPRIVATE + 11)
+
+#endif /* openvswitch/gre.h */
index bd621c106363a8739810cb455924f0bf8703c4fd..7299430b8db900f130fb9859202c85271ecdfba5 100755 (executable)
@@ -92,6 +92,9 @@ function insert_modules_if_required {
     if [ -n "$BRCOMPATD_PIDFILE" ] && ! lsmod | grep -q "brcompat_mod"; then
         action "Inserting brcompat module" modprobe brcompat_mod
     fi
+    if [ -f "/lib/modules/`uname -r`/kernel/net/vswitch/ip_gre_mod.ko" ] && ! lsmod | grep -q "ip_gre_mod"; then
+        action "Inserting ip_gre module" modprobe ip_gre_mod
+    fi
 }
 
 function remove_modules {
index 1b984b19f1df5defaa21348725e8931674b2164e..a0899edd18f1508c3d07d5ff782cbb61b9d860c5 100644 (file)
@@ -305,6 +305,9 @@ fi
 /etc/profile.d/vswitch.sh
 /lib/modules/%{xen_version}/kernel/net/vswitch/openvswitch_mod.ko
 /lib/modules/%{xen_version}/kernel/net/vswitch/brcompat_mod.ko
+%if %(echo '%{xen_version}'|awk -F"." '{if ($3>=18) print 1; else print 0;}')
+/lib/modules/%{xen_version}/kernel/net/vswitch/ip_gre_mod.ko
+%endif
 /usr/share/vswitch/scripts/dump-vif-details
 /usr/share/vswitch/scripts/refresh-xs-network-uuids
 /usr/share/vswitch/scripts/interface-reconfigure