VPP43 - NSH refactoring: Added nsh-map nodes

- removed vnet/vnet/nsh-gre
- removed all nsh from vnet/vnet/nsh_vxlan_gpe to
  vnet/vnet/nsh
- moved vnet/vnet/nsh_vxlan_gpe to vnet/vnet/vxlan_gpe
- added cli and binary api for VXLAN GPE tunnels
- plan to move vnet/vnet/nsh to new repo (sfc_nsh) and make plugin
- added cli for NSH (binary API will be done in sfc_nsh)
- vnet/vnet/gre will be extended in VPP-54

Change-Id: I1d27def916532321577ccd68cb982ae0d0a07e6f
Signed-off-by: Keith Burns (alagalah) <alagalah@gmail.com>

19 files changed, 1672 insertions, 2440 deletions

diff --git a/vnet/Makefile.am b/vnet/Makefile.am
index 1239a8b8df4..8a9f214ca70 100644
--- a/vnet/Makefile.am
+++ b/vnet/Makefile.am
@@ -399,37 +399,40 @@ nobase_include_HEADERS +=			\
 ########################################
 # NSH Map: nsh
 ########################################
+ libvnet_la_SOURCES +=				\
+  vnet/nsh/nsh.c
 
-nobase_include_HEADERS +=			\
- vnet/nsh/nsh_packet.h				\
- vnet/nsh/nsh_error.def
+ nobase_include_HEADERS +=			\
+  vnet/nsh/nsh_packet.h				\
+  vnet/nsh/nsh.h				\
+  vnet/nsh/nsh_error.def
 
 
 ########################################
 # Tunnel protocol: nsh-gre
 ########################################
 
-libvnet_la_SOURCES +=				\
- vnet/nsh-gre/nsh_gre.c				\
- vnet/nsh-gre/encap.c				\
- vnet/nsh-gre/decap.c
+# libvnet_la_SOURCES +=				\
+#   vnet/nsh-gre/nsh_gre.c			\
+#   vnet/nsh-gre/encap.c				\
+#   vnet/nsh-gre/decap.c
 
-nobase_include_HEADERS +=			\
- vnet/nsh-gre/nsh_gre.h				
+# nobase_include_HEADERS +=			\
+#   vnet/nsh-gre/nsh_gre.h				
 
 ########################################
-# Tunnel protocol: nsh-vxlan-gpe
+# Tunnel protocol: vxlan-gpe
 ########################################
 
 libvnet_la_SOURCES +=				\
- vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.c		\
- vnet/nsh-vxlan-gpe/encap.c			\
- vnet/nsh-vxlan-gpe/decap.c
+ vnet/vxlan-gpe/vxlan_gpe.c			\
+ vnet/vxlan-gpe/encap.c	        		\
+ vnet/vxlan-gpe/decap.c
 
 nobase_include_HEADERS +=			\
- vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.h		\
- vnet/nsh-vxlan-gpe/vxlan_gpe_packet.h		\
- vnet/nsh-vxlan-gpe/nsh_vxlan_gpe_error.def
+ vnet/vxlan-gpe/vxlan_gpe.h			\
+ vnet/vxlan-gpe/vxlan_gpe_packet.h		\
+ vnet/vxlan-gpe/vxlan_gpe_error.def
 
 ########################################
 # LISP control plane: lisp-cp
diff --git a/vnet/vnet/gre/node.c b/vnet/vnet/gre/node.c
index 7d07223fc71..5809c5d3e6f 100644
--- a/vnet/vnet/gre/node.c
+++ b/vnet/vnet/gre/node.c
@@ -24,7 +24,8 @@
 _(PUNT, "error-punt")                           \
 _(DROP, "error-drop")                           \
 _(IP4_INPUT, "ip4-input")                       \
-_(IP6_INPUT, "ip6-input")
+_(IP6_INPUT, "ip6-input")			\
+_(NSH_INPUT, "nsh-input")
 
 typedef enum {
 #define _(s,n) GRE_INPUT_NEXT_##s,
diff --git a/vnet/vnet/nsh-gre/encap.c b/vnet/vnet/nsh-gre/encap.c
deleted file mode 100644
index 78b02178263..00000000000
--- a/vnet/vnet/nsh-gre/encap.c
+++ /dev/null
@@ -1,376 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * 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.
- */
-#include <vppinfra/error.h>
-#include <vppinfra/hash.h>
-#include <vnet/vnet.h>
-#include <vnet/ip/ip.h>
-#include <vnet/ethernet/ethernet.h>
-#include <vnet/nsh-gre/nsh_gre.h>
-
-/* Statistics (not really errors) */
-#define foreach_nsh_gre_encap_error             \
-_(ENCAPSULATED, "good packets encapsulated")
-
-static char * nsh_gre_encap_error_strings[] = {
-#define _(sym,string) string,
-  foreach_nsh_gre_encap_error
-#undef _
-};
-
-typedef enum {
-#define _(sym,str) NSH_GRE_ENCAP_ERROR_##sym,
-    foreach_nsh_gre_encap_error
-#undef _
-    NSH_GRE_ENCAP_N_ERROR,
-} nsh_gre_encap_error_t;
-
-typedef enum {
-    NSH_GRE_ENCAP_NEXT_IP4_LOOKUP,
-    NSH_GRE_ENCAP_NEXT_DROP,
-    NSH_GRE_ENCAP_N_NEXT,
-} nsh_gre_encap_next_t;
-
-typedef struct {
-  u32 tunnel_index;
-} nsh_gre_encap_trace_t;
-
-u8 * format_nsh_gre_encap_trace (u8 * s, va_list * args)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  nsh_gre_encap_trace_t * t = va_arg (*args, nsh_gre_encap_trace_t *);
-
-  s = format (s, "NSH-GRE-ENCAP: tunnel %d", t->tunnel_index);
-  return s;
-}
-
-static uword
-nsh_gre_encap (vlib_main_t * vm,
-               vlib_node_runtime_t * node,
-               vlib_frame_t * from_frame)
-{
-  u32 n_left_from, next_index, * from, * to_next;
-  nsh_gre_main_t * ngm = &nsh_gre_main;
-  vnet_main_t * vnm = ngm->vnet_main;
-  vnet_interface_main_t * im = &vnm->interface_main;
-  u32 pkts_encapsulated = 0;
-  u16 old_l0 = 0, old_l1 = 0;
-  u32 cpu_index = os_get_cpu_number();
-  u32 stats_sw_if_index, stats_n_packets, stats_n_bytes;
-
-  from = vlib_frame_vector_args (from_frame);
-  n_left_from = from_frame->n_vectors;
-
-  next_index = node->cached_next_index;
-  stats_sw_if_index = node->runtime_data[0];
-  stats_n_packets = stats_n_bytes = 0;
-
-  while (n_left_from > 0)
-    {
-      u32 n_left_to_next;
-
-      vlib_get_next_frame (vm, node, next_index,
-			   to_next, n_left_to_next);
-
-      while (n_left_from >= 4 && n_left_to_next >= 2)
-	{
-          u32 bi0, bi1;
-	  vlib_buffer_t * b0, * b1;
-	  u32 next0 = NSH_GRE_ENCAP_NEXT_IP4_LOOKUP;
-	  u32 next1 = NSH_GRE_ENCAP_NEXT_IP4_LOOKUP;
-          u32 sw_if_index0, sw_if_index1, len0, len1;
-          vnet_hw_interface_t * hi0, * hi1;
-          ip4_header_t * ip0, * ip1;
-          u64 * copy_src0, * copy_dst0;
-          u64 * copy_src1, * copy_dst1;
-          nsh_gre_tunnel_t * t0, * t1;
-          u16 new_l0, new_l1;
-          ip_csum_t sum0, sum1;
-
-	  /* Prefetch next iteration. */
-	  {
-	    vlib_buffer_t * p2, * p3;
-
-	    p2 = vlib_get_buffer (vm, from[2]);
-	    p3 = vlib_get_buffer (vm, from[3]);
-
-	    vlib_prefetch_buffer_header (p2, LOAD);
-	    vlib_prefetch_buffer_header (p3, LOAD);
-
-	    CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
-	    CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
-	  }
-
-	  bi0 = from[0];
-	  bi1 = from[1];
-	  to_next[0] = bi0;
-	  to_next[1] = bi1;
-	  from += 2;
-	  to_next += 2;
-	  n_left_to_next -= 2;
-	  n_left_from -= 2;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-	  b1 = vlib_get_buffer (vm, bi1);
-
-          /* 1-wide cache? */
-          sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX];
-          sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_TX];
-          hi0 = vnet_get_sup_hw_interface
-            (vnm, vnet_buffer(b0)->sw_if_index[VLIB_TX]);
-          hi1 = vnet_get_sup_hw_interface 
-            (vnm, vnet_buffer(b1)->sw_if_index[VLIB_TX]);
-
-          t0 = pool_elt_at_index (ngm->tunnels, hi0->dev_instance);
-          t1 = pool_elt_at_index (ngm->tunnels, hi1->dev_instance);
-
-          ASSERT(vec_len(t0->rewrite) >= 24);
-          ASSERT(vec_len(t1->rewrite) >= 24);
-
-          /* Apply the rewrite string. $$$$ vnet_rewrite? */
-          vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite));
-          vlib_buffer_advance (b1, -(word)_vec_len(t1->rewrite));
-
-          ip0 = vlib_buffer_get_current(b0);
-          ip1 = vlib_buffer_get_current(b1);
-          /* Copy the fixed header */
-          copy_dst0 = (u64 *) ip0;
-          copy_src0 = (u64 *) t0->rewrite;
-          copy_dst1 = (u64 *) ip1;
-          copy_src1 = (u64 *) t1->rewrite;
-
-          copy_dst0[0] = copy_src0[0];
-          copy_dst0[1] = copy_src0[1];
-          copy_dst0[2] = copy_src0[2];
-
-          copy_dst1[0] = copy_src1[0];
-          copy_dst1[1] = copy_src1[1];
-          copy_dst1[2] = copy_src1[2];
-          
-          /* If there are TLVs to copy, do so */
-          if (PREDICT_FALSE (_vec_len(t0->rewrite) > 24))
-            clib_memcpy (&copy_dst0[3], t0->rewrite + 24 , 
-                    _vec_len (t0->rewrite)-24);
-
-          if (PREDICT_FALSE (_vec_len(t1->rewrite) > 24))
-            clib_memcpy (&copy_dst1[3], t1->rewrite + 24 , 
-                    _vec_len (t1->rewrite)-24);
-
-          /* fix the <bleep>ing outer-IP checksums */
-          sum0 = ip0->checksum;
-          /* old_l0 always 0, see the rewrite setup */
-          new_l0 = 
-            clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
-          
-          sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
-                                 length /* changed member */);
-          ip0->checksum = ip_csum_fold (sum0);
-          ip0->length = new_l0;
-
-          sum1 = ip1->checksum;
-          /* old_l1 always 1, see the rewrite setup */
-          new_l1 = 
-            clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1));
-          
-          sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t,
-                                 length /* changed member */);
-          ip1->checksum = ip_csum_fold (sum1);
-          ip1->length = new_l1;
-
-          /* Reset to look up tunnel partner in the configured FIB */
-          vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->encap_fib_index;
-          vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->encap_fib_index;
-          vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0;
-          vnet_buffer(b1)->sw_if_index[VLIB_RX] = sw_if_index1;
-          pkts_encapsulated += 2;
-
-          len0 = vlib_buffer_length_in_chain(vm, b0);
-          len1 = vlib_buffer_length_in_chain(vm, b0);
-          stats_n_packets += 2;
-          stats_n_bytes += len0 + len1;
-
-          /* Batch stats increment on the same vxlan tunnel so counter is not
-           incremented per packet. Note stats are still incremented for deleted
-           and admin-down tunnel where packets are dropped. It is not worthwhile
-           to check for this rare case and affect normal path performance. */
-          if (PREDICT_FALSE(
-              (sw_if_index0 != stats_sw_if_index)
-                  || (sw_if_index1 != stats_sw_if_index))) {
-            stats_n_packets -= 2;
-            stats_n_bytes -= len0 + len1;
-            if (sw_if_index0 == sw_if_index1) {
-              if (stats_n_packets)
-                vlib_increment_combined_counter(
-                    im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
-                    cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
-              stats_sw_if_index = sw_if_index0;
-              stats_n_packets = 2;
-              stats_n_bytes = len0 + len1;
-            } else {
-              vlib_increment_combined_counter(
-                  im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
-                  cpu_index, sw_if_index0, 1, len0);
-              vlib_increment_combined_counter(
-                  im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
-                  cpu_index, sw_if_index1, 1, len1);
-            }
-          }
-
-          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
-            {
-              nsh_gre_encap_trace_t *tr = 
-                vlib_add_trace (vm, node, b0, sizeof (*tr));
-              tr->tunnel_index = t0 - ngm->tunnels;
-            }
-
-          if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED)) 
-            {
-              nsh_gre_encap_trace_t *tr = 
-                vlib_add_trace (vm, node, b1, sizeof (*tr));
-              tr->tunnel_index = t1 - ngm->tunnels;
-            }
-
-	  vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
-					   to_next, n_left_to_next,
-					   bi0, bi1, next0, next1);
-	}
-    
-      while (n_left_from > 0 && n_left_to_next > 0)
-	{
-	  u32 bi0;
-	  vlib_buffer_t * b0;
-	  u32 next0 = NSH_GRE_ENCAP_NEXT_IP4_LOOKUP;
-      u32 sw_if_index0, len0;
-          vnet_hw_interface_t * hi0;
-          ip4_header_t * ip0;
-          u64 * copy_src0, * copy_dst0;
-          nsh_gre_tunnel_t * t0;
-          u16 new_l0;
-          ip_csum_t sum0;
-
-	  bi0 = from[0];
-	  to_next[0] = bi0;
-	  from += 1;
-	  to_next += 1;
-	  n_left_from -= 1;
-	  n_left_to_next -= 1;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-
-          /* 1-wide cache? */
-          sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX];
-          hi0 = vnet_get_sup_hw_interface
-            (vnm, vnet_buffer(b0)->sw_if_index[VLIB_TX]);
-
-          t0 = pool_elt_at_index (ngm->tunnels, hi0->dev_instance);
-
-          ASSERT(vec_len(t0->rewrite) >= 24);
-
-          /* Apply the rewrite string. $$$$ vnet_rewrite? */
-          vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite));
-
-          ip0 = vlib_buffer_get_current(b0);
-          /* Copy the fixed header */
-          copy_dst0 = (u64 *) ip0;
-          copy_src0 = (u64 *) t0->rewrite;
-          copy_dst0[0] = copy_src0[0];
-          copy_dst0[1] = copy_src0[1];
-          copy_dst0[2] = copy_src0[2];
-          
-          /* If there are TLVs to copy, do so */
-          if (PREDICT_FALSE (_vec_len(t0->rewrite) > 24))
-            clib_memcpy (&copy_dst0[3], t0->rewrite + 24 , 
-                    _vec_len (t0->rewrite)-24);
-
-          /* fix the <bleep>ing outer-IP checksum */
-          sum0 = ip0->checksum;
-          /* old_l0 always 0, see the rewrite setup */
-          new_l0 = 
-            clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
-          
-          sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
-                                 length /* changed member */);
-          ip0->checksum = ip_csum_fold (sum0);
-          ip0->length = new_l0;
-
-          /* Reset to look up tunnel partner in the configured FIB */
-          vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->encap_fib_index;
-          vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0;
-          pkts_encapsulated ++;
-
-          len0 = vlib_buffer_length_in_chain(vm, b0);
-          stats_n_packets += 1;
-          stats_n_bytes += len0;
-
-          /* Batch stats increment on the same vxlan tunnel so counter is not
-           incremented per packet. Note stats are still incremented for deleted
-           and admin-down tunnel where packets are dropped. It is not worthwhile
-           to check for this rare case and affect normal path performance. */
-          if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index)) {
-            stats_n_packets -= 1;
-            stats_n_bytes -= len0;
-            if (stats_n_packets)
-              vlib_increment_combined_counter(
-                  im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
-                  cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
-            stats_n_packets = 1;
-            stats_n_bytes = len0;
-            stats_sw_if_index = sw_if_index0;
-          }
-          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
-            {
-              nsh_gre_encap_trace_t *tr = 
-                vlib_add_trace (vm, node, b0, sizeof (*tr));
-              tr->tunnel_index = t0 - ngm->tunnels;
-            }
-	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
-					   to_next, n_left_to_next,
-					   bi0, next0);
-	}
-
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-  vlib_node_increment_counter (vm, node->node_index, 
-                               NSH_GRE_ENCAP_ERROR_ENCAPSULATED, 
-                               pkts_encapsulated);
-  /* Increment any remaining batch stats */
-  if (stats_n_packets) {
-    vlib_increment_combined_counter(
-        im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, cpu_index,
-        stats_sw_if_index, stats_n_packets, stats_n_bytes);
-    node->runtime_data[0] = stats_sw_if_index;
-  }
-
-  return from_frame->n_vectors;
-}
-
-VLIB_REGISTER_NODE (nsh_gre_encap_node) = {
-  .function = nsh_gre_encap,
-  .name = "nsh-gre-encap",
-  .vector_size = sizeof (u32),
-  .format_trace = format_nsh_gre_encap_trace,
-  .type = VLIB_NODE_TYPE_INTERNAL,
-
-  .n_errors = ARRAY_LEN(nsh_gre_encap_error_strings),
-  .error_strings = nsh_gre_encap_error_strings,
-
-  .n_next_nodes = NSH_GRE_ENCAP_N_NEXT,
-
-  //  add dispositions here
-  .next_nodes = {
-        [NSH_GRE_ENCAP_NEXT_IP4_LOOKUP] = "ip4-lookup",
-        [NSH_GRE_ENCAP_NEXT_DROP] = "error-drop",
-  },
-};
diff --git a/vnet/vnet/nsh-gre/nsh_gre.c b/vnet/vnet/nsh-gre/nsh_gre.c
deleted file mode 100644
index e75ed9dd862..00000000000
--- a/vnet/vnet/nsh-gre/nsh_gre.c
+++ /dev/null
@@ -1,552 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * 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.
- */
-#include <vppinfra/error.h>
-#include <vppinfra/hash.h>
-#include <vnet/vnet.h>
-#include <vnet/ip/ip.h>
-#include <vnet/l2/l2_input.h>
-#include <vnet/ethernet/ethernet.h>
-#include <vnet/nsh-gre/nsh_gre.h>
-
-nsh_gre_main_t nsh_gre_main;
-
-static u8 * format_decap_next (u8 * s, va_list * args)
-{
-  u32 next_index = va_arg (*args, u32);
-
-  switch (next_index)
-    {
-    case NSH_GRE_INPUT_NEXT_DROP:
-      return format (s, "drop");
-    case NSH_GRE_INPUT_NEXT_IP4_INPUT:
-      return format (s, "ip4");
-    case NSH_GRE_INPUT_NEXT_IP6_INPUT:
-      return format (s, "ip6");
-    case NSH_GRE_INPUT_NEXT_ETHERNET_INPUT:
-      return format (s, "ethernet");
-    default:
-      return format (s, "index %d", next_index);
-    }
-  return s;
-}
-
-
-u8 * format_nsh_gre_tunnel (u8 * s, va_list * args)
-{
-  nsh_gre_tunnel_t * t = va_arg (*args, nsh_gre_tunnel_t *);
-  nsh_gre_main_t * ngm = &nsh_gre_main;
-
-  s = format (s, "[%d] %U (src) %U (dst) fibs: (encap %d, decap %d)",
-              t - ngm->tunnels,
-              format_ip4_address, &t->src,
-              format_ip4_address, &t->dst,
-              t->encap_fib_index,
-              t->decap_fib_index);
-
-  s = format (s, " decap-next %U\n", format_decap_next, t->decap_next_index);
-
-  s = format (s, "  ver %d ", (t->nsh_hdr.ver_o_c>>6));
-  if (t->nsh_hdr.ver_o_c & NSH_O_BIT)
-      s = format (s, "O-set ");
-
-  if (t->nsh_hdr.ver_o_c & NSH_C_BIT)
-      s = format (s, "C-set ");
-
-  s = format (s, "len %d (%d bytes) md_type %d next_protocol %d\n",
-              t->nsh_hdr.length, t->nsh_hdr.length * 4, t->nsh_hdr.md_type, t->nsh_hdr.next_protocol);
-  
-  s = format (s, "  service path %d service index %d\n",
-              (t->nsh_hdr.spi_si>>NSH_SPI_SHIFT) & NSH_SPI_MASK,
-              t->nsh_hdr.spi_si & NSH_SINDEX_MASK);
-
-  s = format (s, "  c1 %d c2 %d c3 %d c4 %d\n",
-              t->nsh_hdr.c1, t->nsh_hdr.c2, t->nsh_hdr.c3, t->nsh_hdr.c4);
-
-  return s;
-}
-
-static u8 * format_nsh_gre_name (u8 * s, va_list * args)
-{
-  nsh_gre_main_t * ngm = &nsh_gre_main;
-  u32 i = va_arg (*args, u32);
-  u32 show_dev_instance = ~0;
-
-  if (i < vec_len (ngm->dev_inst_by_real))
-    show_dev_instance = ngm->dev_inst_by_real[i];
-
-  if (show_dev_instance != ~0)
-    i = show_dev_instance;
-
-  return format (s, "nsh_gre_tunnel%d", i);
-}
-
-static int nsh_gre_name_renumber (vnet_hw_interface_t * hi,
-                                  u32 new_dev_instance)
-{
-  nsh_gre_main_t * ngm = &nsh_gre_main;
-
-  vec_validate_init_empty (ngm->dev_inst_by_real, hi->dev_instance, ~0);
-
-  ngm->dev_inst_by_real [hi->dev_instance] = new_dev_instance;
-
-  return 0;
-}
-
-static uword dummy_interface_tx (vlib_main_t * vm,
-                                 vlib_node_runtime_t * node,
-                                 vlib_frame_t * frame)
-{
-  clib_warning ("you shouldn't be here, leaking buffers...");
-  return frame->n_vectors;
-}
-
-VNET_DEVICE_CLASS (nsh_gre_device_class,static) = {
-  .name = "NSH_GRE",
-  .format_device_name = format_nsh_gre_name,
-  .format_tx_trace = format_nsh_gre_encap_trace,
-  .tx_function = dummy_interface_tx,
-  .name_renumber = nsh_gre_name_renumber,
-};
-
-static uword dummy_set_rewrite (vnet_main_t * vnm,
-                                u32 sw_if_index,
-                                u32 l3_type,
-                                void * dst_address,
-                                void * rewrite,
-                                uword max_rewrite_bytes)
-{
-  return 0;
-}
-
-static u8 * format_nsh_gre_header_with_length (u8 * s, va_list * args)
-{
-  u32 dev_instance = va_arg (*args, u32);
-  s = format (s, "unimplemented dev %u", dev_instance);
-  return s;
-}
-
-VNET_HW_INTERFACE_CLASS (nsh_gre_hw_class) = {
-  .name = "NSH_GRE",
-  .format_header = format_nsh_gre_header_with_length,
-  .set_rewrite = dummy_set_rewrite,
-};
-
-#define foreach_copy_field                      \
-_(src.as_u32)                                   \
-_(dst.as_u32)                                   \
-_(encap_fib_index)                              \
-_(decap_fib_index)                              \
-_(decap_next_index)
-
-
-#define foreach_copy_nshhdr_field               \
-_(ver_o_c)					\
-_(length)					\
-_(md_type)					\
-_(next_protocol)				\
-_(spi_si)					\
-_(c1)						\
-_(c2)						\
-_(c3)						\
-_(c4)						\
-_(tlvs)
-
-#define foreach_32bit_field			\
-_(spi_si)                                       \
-_(c1)                                           \
-_(c2)                                           \
-_(c3)                                           \
-_(c4)
-
-static int nsh_gre_rewrite (nsh_gre_tunnel_t * t)
-{
-  u8 *rw = 0;
-  ip4_header_t * ip0;
-  nsh_header_t * nsh0;
-  ip4_gre_and_nsh_header_t * h0;
-  int len;
-
-  len = sizeof (*h0) + vec_len(t->nsh_hdr.tlvs)*4;
-
-  vec_validate_aligned (rw, len-1, CLIB_CACHE_LINE_BYTES);
-
-  h0 = (ip4_gre_and_nsh_header_t *) rw;
-
-  /* Fixed portion of the (outer) ip4 header */
-  ip0 = &h0->ip4;
-  ip0->ip_version_and_header_length = 0x45;
-  ip0->ttl = 254;
-  ip0->protocol = IP_PROTOCOL_GRE;
-  /* we fix up the ip4 header length and checksum after-the-fact */
-  ip0->src_address.as_u32 = t->src.as_u32;
-  ip0->dst_address.as_u32 = t->dst.as_u32;
-  ip0->checksum = ip4_header_checksum (ip0);
-
-  /* GRE header, zero execpt for the NSH ethertype */
-  h0->gre.protocol = clib_host_to_net_u16(GRE_PROTOCOL_nsh);
-
-  /* NSH header */
-  nsh0 = &h0->nsh;
-  nsh0->ver_o_c = t->nsh_hdr.ver_o_c;
-  nsh0->md_type = t->nsh_hdr.md_type;
-  nsh0->next_protocol = t->nsh_hdr.next_protocol;
-  nsh0->spi_si = t->nsh_hdr.spi_si;
-  nsh0->c1 = t->nsh_hdr.c1;
-  nsh0->c2 = t->nsh_hdr.c2;
-  nsh0->c3 = t->nsh_hdr.c3;
-  nsh0->c4 = t->nsh_hdr.c4;
-  
-  /* Endian swap 32-bit fields */
-#define _(x) nsh0->x = clib_host_to_net_u32(nsh0->x);
-  foreach_32bit_field;
-#undef _
-
-  /* fix nsh header length */
-  t->nsh_hdr.length = 6 + vec_len(t->nsh_hdr.tlvs);
-  nsh0->length = t->nsh_hdr.length;
-
-  /* Copy any TLVs */
-  if (vec_len(t->nsh_hdr.tlvs))
-    clib_memcpy (nsh0->tlvs, t->nsh_hdr.tlvs, 4*vec_len(t->nsh_hdr.tlvs));
-
-  t->rewrite = rw;
-  return (0);
-}
-
-int vnet_nsh_gre_add_del_tunnel (vnet_nsh_gre_add_del_tunnel_args_t *a,
-                                 u32 * sw_if_indexp)
-{
-  nsh_gre_main_t * ngm = &nsh_gre_main;
-  nsh_gre_tunnel_t *t = 0;
-  vnet_main_t * vnm = ngm->vnet_main;
-  vnet_hw_interface_t * hi;
-  uword * p;
-  u32 hw_if_index = ~0;
-  u32 sw_if_index = ~0;
-  int rv;
-  u64 key;
-  u32 spi_si_net_byte_order;
-
-  spi_si_net_byte_order = clib_host_to_net_u32(a->nsh_hdr.spi_si);
-
-  key = (((u64)(a->src.as_u32))<<32) | spi_si_net_byte_order;
-
-  p = hash_get (ngm->nsh_gre_tunnel_by_src_address, key);
-  
-  if (a->is_add)
-    {
-      /* adding a tunnel: tunnel must not already exist */
-      if (p) 
-        return VNET_API_ERROR_INVALID_VALUE;
-      
-      if (a->decap_next_index >= NSH_GRE_INPUT_N_NEXT)
-        return VNET_API_ERROR_INVALID_DECAP_NEXT;
-      
-      pool_get_aligned (ngm->tunnels, t, CLIB_CACHE_LINE_BYTES);
-      memset (t, 0, sizeof (*t));
-      
-      /* copy from arg structure */
-#define _(x) t->x = a->x;
-      foreach_copy_field;
-#undef _
-      
-      /* copy from arg structure */
-#define _(x) t->nsh_hdr.x = a->nsh_hdr.x;
-      foreach_copy_nshhdr_field;
-#undef _
-
-      rv = nsh_gre_rewrite (t);
-
-      if (rv)
-        {
-          pool_put (ngm->tunnels, t);
-          return rv;
-        }
-
-      hash_set (ngm->nsh_gre_tunnel_by_src_address, key, t - ngm->tunnels);
-      
-      if (vec_len (ngm->free_nsh_gre_tunnel_hw_if_indices) > 0)
-        {
-          hw_if_index = ngm->free_nsh_gre_tunnel_hw_if_indices
-            [vec_len (ngm->free_nsh_gre_tunnel_hw_if_indices)-1];
-          _vec_len (ngm->free_nsh_gre_tunnel_hw_if_indices) -= 1;
-          
-          hi = vnet_get_hw_interface (vnm, hw_if_index);
-          hi->dev_instance = t - ngm->tunnels;
-          hi->hw_instance = hi->dev_instance;
-        }
-      else 
-        {
-          hw_if_index = vnet_register_interface
-            (vnm, nsh_gre_device_class.index, t - ngm->tunnels,
-             nsh_gre_hw_class.index, t - ngm->tunnels);
-          hi = vnet_get_hw_interface (vnm, hw_if_index);
-          hi->output_node_index = nsh_gre_encap_node.index;
-        }
-      
-      t->hw_if_index = hw_if_index;
-      t->sw_if_index = sw_if_index = hi->sw_if_index;
-      
-      vnet_sw_interface_set_flags (vnm, hi->sw_if_index, 
-                                   VNET_SW_INTERFACE_FLAG_ADMIN_UP);
-    }
-  else
-    {
-      /* deleting a tunnel: tunnel must exist */
-      if (!p) 
-        return VNET_API_ERROR_NO_SUCH_ENTRY;
-
-      t = pool_elt_at_index (ngm->tunnels, p[0]);
-
-      vnet_sw_interface_set_flags (vnm, t->sw_if_index, 0 /* down */);
-      vec_add1 (ngm->free_nsh_gre_tunnel_hw_if_indices, t->hw_if_index);
-
-      hash_unset (ngm->nsh_gre_tunnel_by_src_address, key);
-      vec_free (t->rewrite);
-      pool_put (ngm->tunnels, t);
-    }
-
-  if (sw_if_indexp)
-      *sw_if_indexp = sw_if_index;
-
-  return 0;
-}
-
-static u32 fib_index_from_fib_id (u32 fib_id)
-{
-  ip4_main_t * im = &ip4_main;
-  uword * p;
-
-  p = hash_get (im->fib_index_by_table_id, fib_id);
-  if (!p)
-    return ~0;
-
-  return p[0];
-}
-
-static uword unformat_decap_next (unformat_input_t * input, va_list * args)
-{
-  u32 * result = va_arg (*args, u32 *);
-  u32 tmp;
-  
-  if (unformat (input, "drop"))
-    *result = NSH_GRE_INPUT_NEXT_DROP;
-  else if (unformat (input, "ip4"))
-    *result = NSH_GRE_INPUT_NEXT_IP4_INPUT;
-  else if (unformat (input, "ip6"))
-    *result = NSH_GRE_INPUT_NEXT_IP6_INPUT;
-  else if (unformat (input, "ethernet"))
-    *result = NSH_GRE_INPUT_NEXT_ETHERNET_INPUT;
-  else if (unformat (input, "%d", &tmp))
-    *result = tmp;
-  else
-    return 0;
-  return 1;
-}
-
-static clib_error_t *
-nsh_gre_add_del_tunnel_command_fn (vlib_main_t * vm,
-                                   unformat_input_t * input,
-                                   vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, * line_input = &_line_input;
-  ip4_address_t src, dst;
-  u8 is_add = 1;
-  u8 src_set = 0;
-  u8 dst_set = 0;
-  u32 encap_fib_index = 0;
-  u32 decap_fib_index = 0;
-  u8 ver_o_c = 0;
-  u8 length = 0;
-  u8 md_type = 0;
-  u8 next_protocol = 1; /* ip4 */
-  u32 spi;
-  u8 spi_set = 0;
-  u32 si;
-  u8 si_set = 0;
-  u32 spi_si;
-  u32 c1 = 0;
-  u32 c2 = 0;
-  u32 c3 = 0;
-  u32 c4 = 0;
-  u32 decap_next_index = 1; /* ip4_input */
-  u32 *tlvs = 0;
-  u32 tmp;
-  int rv;
-  vnet_nsh_gre_add_del_tunnel_args_t _a, * a = &_a;
-  
-  /* Get a line of input. */
-  if (! unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) {
-    if (unformat (line_input, "del"))
-      is_add = 0;
-    else if (unformat (line_input, "src %U", 
-                       unformat_ip4_address, &src))
-      src_set = 1;
-    else if (unformat (line_input, "dst %U",
-                       unformat_ip4_address, &dst))
-      dst_set = 1;
-    else if (unformat (line_input, "encap-vrf-id %d", &tmp))
-      {
-        encap_fib_index = fib_index_from_fib_id (tmp);
-        if (encap_fib_index == ~0)
-          return clib_error_return (0, "nonexistent encap fib id %d", tmp);
-      }
-    else if (unformat (line_input, "decap-vrf-id %d", &tmp))
-      {
-        decap_fib_index = fib_index_from_fib_id (tmp);
-        if (decap_fib_index == ~0)
-          return clib_error_return (0, "nonexistent decap fib id %d", tmp);
-      }
-    else if (unformat (line_input, "decap-next %U", unformat_decap_next, 
-                       &decap_next_index))
-        ;
-    else if (unformat (line_input, "version %d", &tmp))
-      ver_o_c |= (tmp & 3) << 6;
-    else if (unformat (line_input, "o-bit %d", &tmp))
-      ver_o_c |= (tmp & 1) << 5;
-    else if (unformat (line_input, "c-bit %d", &tmp))
-      ver_o_c |= (tmp & 1) << 4;
-    else if (unformat (line_input, "md-type %d", &tmp))
-      md_type = tmp;
-    else if (unformat(line_input, "next-ip4"))
-      next_protocol = 1;
-    else if (unformat(line_input, "next-ip6"))
-      next_protocol = 2;
-    else if (unformat(line_input, "next-ethernet"))
-      next_protocol = 3;
-    else if (unformat (line_input, "c1 %d", &c1))
-      ;
-    else if (unformat (line_input, "c2 %d", &c2))
-      ;
-    else if (unformat (line_input, "c3 %d", &c3))
-      ;
-    else if (unformat (line_input, "c4 %d", &c4))
-      ;
-    else if (unformat (line_input, "spi %d", &spi))
-      spi_set = 1;
-    else if (unformat (line_input, "si %d", &si))
-      si_set = 1;
-    else if (unformat (line_input, "tlv %x"))
-        vec_add1 (tlvs, tmp);
-    else 
-      return clib_error_return (0, "parse error: '%U'", 
-                                format_unformat_error, line_input);
-  }
-
-  unformat_free (line_input);
-
-  if (src_set == 0)
-    return clib_error_return (0, "tunnel src address not specified");
-
-  if (dst_set == 0)
-    return clib_error_return (0, "tunnel dst address not specified");
-
-  if (spi_set == 0)
-    return clib_error_return (0, "spi not specified");
-  
-  if (si_set == 0)
-    return clib_error_return (0, "si not specified");
-
-  spi_si = (spi<<8) | si;
-  
-  memset (a, 0, sizeof (*a));
-
-  a->is_add = is_add;
-
-#define _(x) a->x = x;
-  foreach_copy_field;
-#undef _
-
-      /* copy from arg structure */
-#define _(x) a->nsh_hdr.x = x;
-      foreach_copy_nshhdr_field;
-#undef _
-
-  rv = vnet_nsh_gre_add_del_tunnel (a, 0 /* hw_if_indexp */);
-
-  switch(rv)
-    {
-    case 0:
-      break;
-    case VNET_API_ERROR_INVALID_DECAP_NEXT:
-      return clib_error_return (0, "invalid decap-next...");
-
-    case VNET_API_ERROR_TUNNEL_EXIST:
-      return clib_error_return (0, "tunnel already exists...");
-
-    case VNET_API_ERROR_NO_SUCH_ENTRY:
-      return clib_error_return (0, "session does not exist...");
-
-    default:
-      return clib_error_return 
-        (0, "vnet_nsh_gre_add_del_tunnel returned %d", rv);
-    }
-
-  return 0;
-}
-
-VLIB_CLI_COMMAND (create_nsh_gre_tunnel_command, static) = {
-  .path = "nsh gre tunnel",
-  .short_help = 
-  "nsh gre tunnel src <ip4-addr> dst <ip4-addr>" 
-  "    c1 <nn> c2 <nn> c3 <nn> c4 <nn> spi <nn> si <nn>\n"
-  "    [encap-vrf-id <nn>] [decap-vrf-id <nn>] [o-bit <1|0>] [c-bit <1|0>]\n"
-  "    [md-type <nn>][next-ip4][next-ip6][next-ethernet]\n"
-  "    [tlv <xx>][decap-next [ip4|ip6|ethernet]][del]\n",
-  .function = nsh_gre_add_del_tunnel_command_fn,
-};
-
-static clib_error_t *
-show_nsh_gre_tunnel_command_fn (vlib_main_t * vm,
-                                unformat_input_t * input,
-                                vlib_cli_command_t * cmd)
-{
-  nsh_gre_main_t * ngm = &nsh_gre_main;
-  nsh_gre_tunnel_t * t;
-  
-  if (pool_elts (ngm->tunnels) == 0)
-    vlib_cli_output (vm, "No nsh-gre tunnels configured...");
-
-  pool_foreach (t, ngm->tunnels,
-  ({
-    vlib_cli_output (vm, "%U", format_nsh_gre_tunnel, t);
-  }));
-  
-  return 0;
-}
-
-VLIB_CLI_COMMAND (show_nsh_gre_tunnel_command, static) = {
-    .path = "show nsh gre tunnel",
-    .function = show_nsh_gre_tunnel_command_fn,
-};
-
-clib_error_t *nsh_gre_init (vlib_main_t *vm)
-{
-  nsh_gre_main_t *ngm = &nsh_gre_main;
-  
-  ngm->vnet_main = vnet_get_main();
-  ngm->vlib_main = vm;
-  
-  ngm->nsh_gre_tunnel_by_src_address = hash_create (0, sizeof (uword));
-  gre_register_input_protocol (vm, GRE_PROTOCOL_nsh, 
-                               nsh_gre_input_node.index);
-  return 0;
-}
-
-VLIB_INIT_FUNCTION(nsh_gre_init);
-  
diff --git a/vnet/vnet/nsh-gre/nsh_gre.h b/vnet/vnet/nsh-gre/nsh_gre.h
deleted file mode 100644
index abe115580cb..00000000000
--- a/vnet/vnet/nsh-gre/nsh_gre.h
+++ /dev/null
@@ -1,110 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * 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.
- */
-#ifndef included_vnet_nsh_gre_h
-#define included_vnet_nsh_gre_h
-
-#include <vnet/vnet.h>
-#include <vnet/gre/gre.h>
-#include <vnet/nsh/nsh_packet.h>
-#include <vnet/ip/ip4_packet.h>
-
-typedef CLIB_PACKED (struct {
-  ip4_header_t ip4;             /* 20 bytes */
-  gre_header_t gre;             /* 4 bytes */
-  nsh_header_t nsh;   		/* 28 bytes */
-}) ip4_gre_and_nsh_header_t;
-
-typedef struct {
-  /* Rewrite string. $$$$ embed vnet_rewrite header */
-  u8 * rewrite;
-
-  /* tunnel src and dst addresses */
-  ip4_address_t src;
-  ip4_address_t dst;
-
-  /* FIB indices */
-  u32 encap_fib_index;          /* tunnel partner lookup here */
-  u32 decap_fib_index;          /* inner IP lookup here */
-
-  /* when decapsulating, send pkts here */
-  u32 decap_next_index;
-
-  /* vnet intfc hw/sw_if_index */
-  u32 hw_if_index;
-  u32 sw_if_index;
-
-  /* NSH header fields in HOST byte order */
-  nsh_header_t nsh_hdr;
-
-} nsh_gre_tunnel_t;
-
-#define foreach_nsh_gre_input_next              \
-  _ (DROP, "error-drop")                        \
-  _ (IP4_INPUT, "ip4-input")                    \
-  _ (IP6_INPUT, "ip6-input")                    \
-  _ (ETHERNET_INPUT, "ethernet-input")
-
-typedef enum {
-#define _(s,n) NSH_GRE_INPUT_NEXT_##s,
-  foreach_nsh_gre_input_next
-#undef _
-  NSH_GRE_INPUT_N_NEXT,
-} nsh_gre_input_next_t;
-
-typedef enum {
-#define nsh_gre_error(n,s) NSH_GRE_ERROR_##n,
-#include <vnet/nsh/nsh_error.def>
-#undef nsh_gre_error
-  NSH_GRE_N_ERROR,
-} nsh_gre_input_error_t;
-
-typedef struct {
-  /* vector of encap tunnel instances */
-  nsh_gre_tunnel_t *tunnels;
-
-  /* lookup tunnel by tunnel partner src address */
-  uword * nsh_gre_tunnel_by_src_address;
-
-  /* Free vlib hw_if_indices */
-  u32 * free_nsh_gre_tunnel_hw_if_indices;
-
-  /* show device instance by real device instance */
-  u32 * dev_inst_by_real;
-
-  /* convenience */
-  vlib_main_t * vlib_main;
-  vnet_main_t * vnet_main;
-} nsh_gre_main_t;
-
-nsh_gre_main_t nsh_gre_main;
-
-extern vlib_node_registration_t nsh_gre_input_node;
-extern vlib_node_registration_t nsh_gre_encap_node;
-
-u8 * format_nsh_gre_encap_trace (u8 * s, va_list * args);
-
-typedef struct {
-  u8 is_add;
-  ip4_address_t src, dst;
-  u32 encap_fib_index;
-  u32 decap_fib_index;
-  u32 decap_next_index;
-  nsh_header_t nsh_hdr;
-} vnet_nsh_gre_add_del_tunnel_args_t;
-
-int vnet_nsh_gre_add_del_tunnel (vnet_nsh_gre_add_del_tunnel_args_t *a, 
-                                 u32 * sw_if_indexp);
-
-#endif /* included_vnet_nsh_gre_h */
diff --git a/vnet/vnet/nsh-vxlan-gpe/decap.c b/vnet/vnet/nsh-vxlan-gpe/decap.c
deleted file mode 100644
index 76003e6003f..00000000000
--- a/vnet/vnet/nsh-vxlan-gpe/decap.c
+++ /dev/null
@@ -1,537 +0,0 @@
-/*
- * nsh.c: nsh packet processing
- *
- * Copyright (c) 2013 Cisco and/or its affiliates.
- * 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.
- */
-
-#include <vlib/vlib.h>
-#include <vnet/pg/pg.h>
-#include <vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.h>
-
-vlib_node_registration_t nsh_vxlan_gpe_input_node;
-
-/* From nsh-gre */
-u8 * format_nsh_header_with_length (u8 * s, va_list * args);
-
-typedef struct {
-  u32 next_index;
-  u32 tunnel_index;
-  u32 error;
-  nsh_header_t h;
-} nsh_vxlan_gpe_rx_trace_t;
-
-static u8 * format_nsh_vxlan_gpe_rx_trace (u8 * s, va_list * args)
-{
-  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
-  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  nsh_vxlan_gpe_rx_trace_t * t = va_arg (*args, nsh_vxlan_gpe_rx_trace_t *);
-
-  if (t->tunnel_index != ~0)
-    {
-      s = format (s, "NSH-VXLAN: tunnel %d next %d error %d", t->tunnel_index, 
-                  t->next_index, t->error);
-    }
-  else
-    {
-      s = format (s, "NSH-VXLAN: no tunnel next %d error %d\n", t->next_index, 
-                  t->error);
-    }
-  s = format (s, "\n  %U", format_nsh_header_with_length, &t->h, 
-              (u32) sizeof (t->h) /* max size */);
-  return s;
-}
-
-static uword
-nsh_vxlan_gpe_input (vlib_main_t * vm,
-                     vlib_node_runtime_t * node,
-                     vlib_frame_t * from_frame)
-{
-  u32 n_left_from, next_index, * from, * to_next;
-  nsh_vxlan_gpe_main_t * ngm = &nsh_vxlan_gpe_main;
-  vnet_main_t * vnm = ngm->vnet_main;
-  vnet_interface_main_t * im = &vnm->interface_main;
-  u32 last_tunnel_index = ~0;
-  nsh_vxlan_gpe_tunnel_key_t last_key;
-  u32 pkts_decapsulated = 0;
-  u32 cpu_index = os_get_cpu_number();
-  u32 stats_sw_if_index, stats_n_packets, stats_n_bytes;
-
-  memset (&last_key, 0xff, sizeof (last_key));
-
-  from = vlib_frame_vector_args (from_frame);
-  n_left_from = from_frame->n_vectors;
-
-  next_index = node->cached_next_index;
-  stats_sw_if_index = node->runtime_data[0];
-  stats_n_packets = stats_n_bytes = 0;
-
-  while (n_left_from > 0)
-    {
-      u32 n_left_to_next;
-
-      vlib_get_next_frame (vm, node, next_index,
-			   to_next, n_left_to_next);
-
-      while (n_left_from >= 4 && n_left_to_next >= 2)
-	{
-          u32 bi0, bi1;
-	  vlib_buffer_t * b0, * b1;
-	  u32 next0, next1;
-          ip4_vxlan_gpe_and_nsh_header_t * iuvn0, * iuvn1;
-	  uword * p0, * p1;
-          u32 tunnel_index0, tunnel_index1;
-          nsh_vxlan_gpe_tunnel_t * t0, * t1;
-          nsh_vxlan_gpe_tunnel_key_t key0, key1;
-          u32 error0, error1;
-          u32 sw_if_index0, sw_if_index1, len0, len1;
-
-	  /* Prefetch next iteration. */
-	  {
-	    vlib_buffer_t * p2, * p3;
-
-	    p2 = vlib_get_buffer (vm, from[2]);
-	    p3 = vlib_get_buffer (vm, from[3]);
-
-	    vlib_prefetch_buffer_header (p2, LOAD);
-	    vlib_prefetch_buffer_header (p3, LOAD);
-
-	    CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
-	    CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
-	  }
-
-	  bi0 = from[0];
-	  bi1 = from[1];
-	  to_next[0] = bi0;
-	  to_next[1] = bi1;
-	  from += 2;
-	  to_next += 2;
-	  n_left_to_next -= 2;
-	  n_left_from -= 2;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-	  b1 = vlib_get_buffer (vm, bi1);
-
-          /* udp leaves current_data pointing at the vxlan header */
-          vlib_buffer_advance 
-            (b0, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
-          vlib_buffer_advance 
-            (b1, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
-
-          iuvn0 = vlib_buffer_get_current (b0);
-          iuvn1 = vlib_buffer_get_current (b1);
-
-          /* pop (ip, udp, vxlan, nsh) */
-          vlib_buffer_advance (b0, sizeof (*iuvn0));
-          vlib_buffer_advance (b1, sizeof (*iuvn1));
-
-          tunnel_index0 = ~0;
-          error0 = 0;
-          next0 = NSH_VXLAN_GPE_INPUT_NEXT_DROP;
-
-          tunnel_index1 = ~0;
-          error1 = 0;
-          next1 = NSH_VXLAN_GPE_INPUT_NEXT_DROP;
-
-          key0.src = iuvn0->ip4.src_address.as_u32;
-          key0.vni = iuvn0->vxlan.vni_res;
-          key0.spi_si = iuvn0->nsh.spi_si;
-          key0.pad = 0;
-
-          if (PREDICT_FALSE ((key0.as_u64[0] != last_key.as_u64[0])
-                             || (key0.as_u64[1] != last_key.as_u64[1])))
-            {
-              p0 = hash_get_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, &key0);
-
-              if (p0 == 0)
-                {
-                  error0 = NSH_VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
-                  goto trace0;
-                }
-
-              last_key.as_u64[0] = key0.as_u64[0];
-              last_key.as_u64[1] = key0.as_u64[1];
-              tunnel_index0 = last_tunnel_index = p0[0];
-            }
-          else
-            tunnel_index0 = last_tunnel_index;
-
-          t0 = pool_elt_at_index (ngm->tunnels, tunnel_index0);
-
-          next0 = t0->decap_next_index;
-          sw_if_index0 = t0->sw_if_index;
-          len0 = vlib_buffer_length_in_chain(vm, b0);
-
-          /* Required to make the l2 tag push / pop code work on l2 subifs */
-          vnet_update_l2_len (b0);
-
-          if (next0 == NSH_VXLAN_GPE_INPUT_NEXT_NSH_VXLAN_GPE_ENCAP)
-            {
-              /*
-               * Functioning as SFF (ie "half NSH tunnel mode")
-               * If ingress (we are in decap.c) with NSH header, and 'decap next nsh-vxlan-gpe' then "NSH switch"
-               * 1. Take DST, remap to SRC, remap other keys in place
-               * 2. Look up new t0 as per above
-               * 3. Set sw_if_index[VLIB_TX] to be t0->sw_if_index
-               */
-              uword * next_p0;
-              nsh_vxlan_gpe_tunnel_t  * next_t0;
-              nsh_vxlan_gpe_tunnel_key_t next_key0;
-
-              next_key0.src = iuvn0->ip4.dst_address.as_u32;
-              next_key0.vni = iuvn0->vxlan.vni_res;
-              next_key0.spi_si = iuvn0->nsh.spi_si;
-              next_key0.pad = 0;
-
-              next_p0 = hash_get_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, &next_key0);
-
-              if (next_p0 == 0)
-                {
-                  error0 = NSH_VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
-                  goto trace0;
-                }
-              next_t0 = pool_elt_at_index (ngm->tunnels, next_p0[0]);
-              vnet_buffer(b0)->sw_if_index[VLIB_TX] = next_t0->sw_if_index;
-
-            }
-          else
-            {
-              /*
-               * ip[46] lookup in the configured FIB
-               * nsh-vxlan-gpe-encap, here's the encap tunnel sw_if_index
-               */
-              vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index;
-            }
-
-          pkts_decapsulated++;
-          stats_n_packets += 1;
-          stats_n_bytes += len0;
-
-          if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index))
-          {
-            stats_n_packets -= 1;
-            stats_n_bytes -= len0;
-            if (stats_n_packets)
-              vlib_increment_combined_counter(
-                  im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX,
-                  cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
-            stats_n_packets = 1;
-            stats_n_bytes = len0;
-            stats_sw_if_index = sw_if_index0;
-          }
-
-        trace0:
-          b0->error = error0 ? node->errors[error0] : 0;
-
-          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
-            {
-              nsh_vxlan_gpe_rx_trace_t *tr
-                = vlib_add_trace (vm, node, b0, sizeof (*tr));
-              tr->next_index = next0;
-              tr->error = error0;
-              tr->tunnel_index = tunnel_index0;
-              tr->h = iuvn0->nsh;
-            }
-
-          key1.src = iuvn1->ip4.src_address.as_u32;
-          key1.vni = iuvn1->vxlan.vni_res;
-          key1.spi_si = iuvn1->nsh.spi_si;
-          key1.pad = 0;
-
-          if (PREDICT_FALSE ((key1.as_u64[0] != last_key.as_u64[0])
-                             || (key1.as_u64[1] != last_key.as_u64[1])))
-            {
-              p1 = hash_get_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, &key1);
-
-              if (p1 == 0)
-                {
-                  error1 = NSH_VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
-                  goto trace1;
-                }
-
-              last_key.as_u64[0] = key1.as_u64[0];
-              last_key.as_u64[1] = key1.as_u64[1];
-              tunnel_index1 = last_tunnel_index = p1[0];
-            }
-          else
-            tunnel_index1 = last_tunnel_index;
-
-          t1 = pool_elt_at_index (ngm->tunnels, tunnel_index1);
-
-          next1 = t1->decap_next_index;
-          sw_if_index1 = t1->sw_if_index;
-          len1 = vlib_buffer_length_in_chain(vm, b1);
-
-          /* Required to make the l2 tag push / pop code work on l2 subifs */
-          vnet_update_l2_len (b1);
-
-          if (next1 == NSH_VXLAN_GPE_INPUT_NEXT_NSH_VXLAN_GPE_ENCAP)
-            {
-              /*
-               * Functioning as SFF (ie "half NSH tunnel mode")
-               * If ingress (we are in decap.c) with NSH header, and 'decap next nsh-vxlan-gpe' then "NSH switch"
-               * 1. Take DST, remap to SRC, remap other keys in place
-               * 2. Look up new t0 as per above
-               * 3. Set sw_if_index[VLIB_TX] to be t0->sw_if_index
-               */
-              uword * next_p1;
-              nsh_vxlan_gpe_tunnel_t  * next_t1;
-              nsh_vxlan_gpe_tunnel_key_t next_key1;
-
-              next_key1.src = iuvn1->ip4.dst_address.as_u32;
-              next_key1.vni = iuvn1->vxlan.vni_res;
-              next_key1.spi_si = iuvn1->nsh.spi_si;
-              next_key1.pad = 0;
-
-              next_p1 = hash_get_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, &next_key1);
-
-              if (next_p1 == 0)
-                {
-                  error1 = NSH_VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
-                  goto trace1;
-                }
-              next_t1 = pool_elt_at_index (ngm->tunnels, next_p1[0]);
-              vnet_buffer(b1)->sw_if_index[VLIB_TX] = next_t1->sw_if_index;
-
-            }
-          else
-            {
-              /*
-               * ip[46] lookup in the configured FIB
-               * nsh-vxlan-gpe-encap, here's the encap tunnel sw_if_index
-               */
-              vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->decap_fib_index;
-            }
-
-          pkts_decapsulated++;
-          stats_n_packets += 1;
-          stats_n_bytes += len1;
-          /* Batch stats increment on the same vxlan tunnel so counter
-           is not incremented per packet */
-          if (PREDICT_FALSE(sw_if_index1 != stats_sw_if_index))
-          {
-            stats_n_packets -= 1;
-            stats_n_bytes -= len1;
-            if (stats_n_packets)
-              vlib_increment_combined_counter(
-                  im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX,
-                  cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
-            stats_n_packets = 1;
-            stats_n_bytes = len1;
-            stats_sw_if_index = sw_if_index1;
-          }
-          vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->decap_fib_index;
-
-        trace1:
-          b1->error = error1 ? node->errors[error1] : 0;
-
-          if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
-            {
-              nsh_vxlan_gpe_rx_trace_t *tr
-                = vlib_add_trace (vm, node, b1, sizeof (*tr));
-              tr->next_index = next1;
-              tr->error = error1;
-              tr->tunnel_index = tunnel_index1;
-              tr->h = iuvn1->nsh;
-            }
-
-	  vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
-					   to_next, n_left_to_next,
-					   bi0, bi1, next0, next1);
-	}
-    
-      while (n_left_from > 0 && n_left_to_next > 0)
-	{
-	  u32 bi0;
-	  vlib_buffer_t * b0;
-	  u32 next0;
-          ip4_vxlan_gpe_and_nsh_header_t * iuvn0;
-	  uword * p0;
-          u32 tunnel_index0;
-          nsh_vxlan_gpe_tunnel_t * t0;
-          nsh_vxlan_gpe_tunnel_key_t key0;
-          u32 error0;
-          u32 sw_if_index0, len0;
-
-	  bi0 = from[0];
-	  to_next[0] = bi0;
-	  from += 1;
-	  to_next += 1;
-	  n_left_from -= 1;
-	  n_left_to_next -= 1;
-
-	  b0 = vlib_get_buffer (vm, bi0);
-
-          /* udp leaves current_data pointing at the vxlan header */
-          vlib_buffer_advance 
-            (b0, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
-
-          iuvn0 = vlib_buffer_get_current (b0);
-
-          /* pop (ip, udp, vxlan, nsh) */
-          vlib_buffer_advance (b0, sizeof (*iuvn0));
-
-          tunnel_index0 = ~0;
-          error0 = 0;
-          next0 = NSH_VXLAN_GPE_INPUT_NEXT_DROP;
-
-          key0.src = iuvn0->ip4.src_address.as_u32;
-          key0.vni = iuvn0->vxlan.vni_res;
-          key0.spi_si = iuvn0->nsh.spi_si;
-          key0.pad = 0;
-
-          if (PREDICT_FALSE ((key0.as_u64[0] != last_key.as_u64[0])
-                             || (key0.as_u64[1] != last_key.as_u64[1])))
-            {
-              p0 = hash_get_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, &key0);
-          
-              if (p0 == 0)
-                {
-                  error0 = NSH_VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
-                  goto trace00;
-                }
-
-              last_key.as_u64[0] = key0.as_u64[0];
-              last_key.as_u64[1] = key0.as_u64[1];
-              tunnel_index0 = last_tunnel_index = p0[0];
-            }
-          else
-            tunnel_index0 = last_tunnel_index;
-
-          t0 = pool_elt_at_index (ngm->tunnels, tunnel_index0);
-
-          next0 = t0->decap_next_index;
-          sw_if_index0 = t0->sw_if_index;
-          len0 = vlib_buffer_length_in_chain(vm, b0);
-
-          /* Required to make the l2 tag push / pop code work on l2 subifs */
-          vnet_update_l2_len (b0);
-
-          if (next0 == NSH_VXLAN_GPE_INPUT_NEXT_NSH_VXLAN_GPE_ENCAP)
-            {
-              /*
-               * Functioning as SFF (ie "half NSH tunnel mode")
-               * If ingress (we are in decap.c) with NSH header, and 'decap next nsh-vxlan-gpe' then "NSH switch"
-               * 1. Take DST, remap to SRC, remap other keys in place
-               * 2. Look up new t0 as per above
-               * 3. Set sw_if_index[VLIB_TX] to be t0->sw_if_index
-               */
-              uword * next_p0;
-              nsh_vxlan_gpe_tunnel_t  * next_t0;
-              nsh_vxlan_gpe_tunnel_key_t next_key0;
-
-              next_key0.src = iuvn0->ip4.dst_address.as_u32;
-              next_key0.vni = iuvn0->vxlan.vni_res;
-              next_key0.spi_si = iuvn0->nsh.spi_si;
-              next_key0.pad = 0;
-
-              next_p0 = hash_get_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, &next_key0);
-
-              if (next_p0 == 0)
-                {
-                  error0 = NSH_VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
-                  goto trace00;
-                }
-              next_t0 = pool_elt_at_index (ngm->tunnels, next_p0[0]);
-              vnet_buffer(b0)->sw_if_index[VLIB_TX] = next_t0->sw_if_index;
-              
-            } 
-          else 
-            {
-              /*
-               * ip[46] lookup in the configured FIB
-               * nsh-vxlan-gpe-encap, here's the encap tunnel sw_if_index
-               */
-              vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index;
-            }
-
-          pkts_decapsulated ++;
-
-          stats_n_packets += 1;
-          stats_n_bytes += len0;
-
-          /* Batch stats increment on the same nsh-vxlan-gpe tunnel so counter
-           is not incremented per packet */
-          if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index))
-          {
-            stats_n_packets -= 1;
-            stats_n_bytes -= len0;
-            if (stats_n_packets)
-              vlib_increment_combined_counter(
-                  im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX,
-                  cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
-            stats_n_packets = 1;
-            stats_n_bytes = len0;
-            stats_sw_if_index = sw_if_index0;
-          }
-
-        trace00:
-          b0->error = error0 ? node->errors[error0] : 0;
-
-          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
-            {
-              nsh_vxlan_gpe_rx_trace_t *tr
-                = vlib_add_trace (vm, node, b0, sizeof (*tr));
-              tr->next_index = next0;
-              tr->error = error0;
-              tr->tunnel_index = tunnel_index0;
-              tr->h = iuvn0->nsh;
-            }
-	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
-					   to_next, n_left_to_next,
-					   bi0, next0);
-	}
-
-      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-    }
-  vlib_node_increment_counter (vm, nsh_vxlan_gpe_input_node.index,
-                               NSH_VXLAN_GPE_ERROR_DECAPSULATED,
-                               pkts_decapsulated);
-  /* Increment any remaining batch stats */
-  if (stats_n_packets)
-  {
-    vlib_increment_combined_counter(
-        im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX, cpu_index,
-        stats_sw_if_index, stats_n_packets, stats_n_bytes);
-    node->runtime_data[0] = stats_sw_if_index;
-  }
-  return from_frame->n_vectors;
-}
-
-static char * nsh_vxlan_gpe_error_strings[] = {
-#define nsh_vxlan_gpe_error(n,s) s,
-#include <vnet/nsh-vxlan-gpe/nsh_vxlan_gpe_error.def>
-#undef nsh_vxlan_gpe_error
-#undef _
-};
-
-VLIB_REGISTER_NODE (nsh_vxlan_gpe_input_node) = {
-  .function = nsh_vxlan_gpe_input,
-  .name = "nsh-vxlan-gpe-input",
-  /* Takes a vector of packets. */
-  .vector_size = sizeof (u32),
-
-  .n_errors = NSH_VXLAN_GPE_N_ERROR,
-  .error_strings = nsh_vxlan_gpe_error_strings,
-
-  .n_next_nodes = NSH_VXLAN_GPE_INPUT_N_NEXT,
-  .next_nodes = {
-#define _(s,n) [NSH_VXLAN_GPE_INPUT_NEXT_##s] = n,
-    foreach_nsh_vxlan_gpe_input_next
-#undef _
-  },
-
-  .format_buffer = format_nsh_header_with_length,
-  .format_trace = format_nsh_vxlan_gpe_rx_trace,
-  // $$$$ .unformat_buffer = unformat_nsh_vxlan_gpe_header,
-};
diff --git a/vnet/vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.c b/vnet/vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.c
deleted file mode 100644
index 88945cd8762..00000000000
--- a/vnet/vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.c
+++ /dev/null
@@ -1,577 +0,0 @@
-/*
- * Copyright (c) 2015 Cisco and/or its affiliates.
- * 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.
- */
-#include <vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.h>
-
-nsh_vxlan_gpe_main_t nsh_vxlan_gpe_main;
-
-static u8 * format_decap_next (u8 * s, va_list * args)
-{
-  u32 next_index = va_arg (*args, u32);
-
-  switch (next_index)
-    {
-    case NSH_VXLAN_GPE_INPUT_NEXT_DROP:
-      return format (s, "drop");
-    case NSH_VXLAN_GPE_INPUT_NEXT_IP4_INPUT:
-      return format (s, "ip4");
-    case NSH_VXLAN_GPE_INPUT_NEXT_IP6_INPUT:
-      return format (s, "ip6");
-    case NSH_VXLAN_GPE_INPUT_NEXT_ETHERNET_INPUT:
-      return format (s, "ethernet");
-    case NSH_VXLAN_GPE_INPUT_NEXT_NSH_VXLAN_GPE_ENCAP:
-      return format (s, "nsh-vxlan-gpe");
-    default:
-      return format (s, "unknown %d", next_index);
-    }
-  return s;
-}
-
-u8 * format_nsh_vxlan_gpe_tunnel (u8 * s, va_list * args)
-{
-  nsh_vxlan_gpe_tunnel_t * t = va_arg (*args, nsh_vxlan_gpe_tunnel_t *);
-  nsh_vxlan_gpe_main_t * ngm = &nsh_vxlan_gpe_main;
-
-  s = format (s, 
-              "[%d] %U (src) %U (dst) fibs: encap %d, decap %d",
-              t - ngm->tunnels,
-              format_ip4_address, &t->src,
-              format_ip4_address, &t->dst,
-              t->encap_fib_index,
-              t->decap_fib_index);
-  s = format (s, " decap next %U\n", format_decap_next, t->decap_next_index);
-  s = format (s, "  vxlan VNI %d ", t->vni);
-  s = format (s, "nsh ver %d ", (t->nsh_hdr.ver_o_c>>6));
-  if (t->nsh_hdr.ver_o_c & NSH_O_BIT)
-      s = format (s, "O-set ");
-
-  if (t->nsh_hdr.ver_o_c & NSH_C_BIT)
-      s = format (s, "C-set ");
-
-  s = format (s, "len %d (%d bytes) md_type %d next_protocol %d\n",
-              t->nsh_hdr.length, t->nsh_hdr.length * 4, t->nsh_hdr.md_type, t->nsh_hdr.next_protocol);
-  
-  s = format (s, "  service path %d service index %d\n",
-              (t->nsh_hdr.spi_si>>NSH_SPI_SHIFT) & NSH_SPI_MASK,
-              t->nsh_hdr.spi_si & NSH_SINDEX_MASK);
-
-  s = format (s, "  c1 %d c2 %d c3 %d c4 %d\n",
-              t->nsh_hdr.c1, t->nsh_hdr.c2, t->nsh_hdr.c3, t->nsh_hdr.c4);
-
-  return s;
-}
-
-static u8 * format_nsh_vxlan_gpe_name (u8 * s, va_list * args)
-{
-  nsh_vxlan_gpe_main_t * ngm = &nsh_vxlan_gpe_main;
-  u32 i = va_arg (*args, u32);
-  u32 show_dev_instance = ~0;
-
-  if (i < vec_len (ngm->dev_inst_by_real))
-    show_dev_instance = ngm->dev_inst_by_real[i];
-
-  if (show_dev_instance != ~0)
-    i = show_dev_instance;
-
-  return format (s, "nsh_vxlan_gpe_tunnel%d", i);
-}
-
-static int nsh_vxlan_gpe_name_renumber (vnet_hw_interface_t * hi,
-                                        u32 new_dev_instance)
-{
-  nsh_vxlan_gpe_main_t * ngm = &nsh_vxlan_gpe_main;
-
-  vec_validate_init_empty (ngm->dev_inst_by_real, hi->dev_instance, ~0);
-
-  ngm->dev_inst_by_real [hi->dev_instance] = new_dev_instance;
-
-  return 0;
-}
-
-static uword dummy_interface_tx (vlib_main_t * vm,
-                                 vlib_node_runtime_t * node,
-                                 vlib_frame_t * frame)
-{
-  clib_warning ("you shouldn't be here, leaking buffers...");
-  return frame->n_vectors;
-}
-
-VNET_DEVICE_CLASS (nsh_vxlan_gpe_device_class,static) = {
-  .name = "NSH_VXLAN_GPE",
-  .format_device_name = format_nsh_vxlan_gpe_name,
-  .format_tx_trace = format_nsh_vxlan_gpe_encap_trace,
-  .tx_function = dummy_interface_tx,
-  .name_renumber = nsh_vxlan_gpe_name_renumber,
-};
-
-static uword dummy_set_rewrite (vnet_main_t * vnm,
-                                u32 sw_if_index,
-                                u32 l3_type,
-                                void * dst_address,
-                                void * rewrite,
-                                uword max_rewrite_bytes)
-{
-  return 0;
-}
-
-static u8 * format_nsh_vxlan_gpe_header_with_length (u8 * s, va_list * args)
-{
-  u32 dev_instance = va_arg (*args, u32);
-  s = format (s, "unimplemented dev %u", dev_instance);
-  return s;
-}
-
-VNET_HW_INTERFACE_CLASS (nsh_vxlan_gpe_hw_class) = {
-  .name = "NSH_VXLAN_GPE",
-  .format_header = format_nsh_vxlan_gpe_header_with_length,
-  .set_rewrite = dummy_set_rewrite,
-};
-
-#define foreach_copy_field                      \
-_(src.as_u32)                                   \
-_(dst.as_u32)                                   \
-_(vni)                                          \
-_(encap_fib_index)                              \
-_(decap_fib_index)                              \
-_(decap_next_index)
-
-
-#define foreach_copy_nshhdr_field               \
-_(ver_o_c)					\
-_(length)					\
-_(md_type)					\
-_(next_protocol)				\
-_(spi_si)					\
-_(c1)						\
-_(c2)						\
-_(c3)						\
-_(c4)						\
-_(tlvs)
-
-#define foreach_32bit_field			\
-_(spi_si)                                       \
-_(c1)                                           \
-_(c2)                                           \
-_(c3)                                           \
-_(c4)
-
-static int nsh_vxlan_gpe_rewrite (nsh_vxlan_gpe_tunnel_t * t)
-{
-  u8 *rw = 0;
-  ip4_header_t * ip0;
-  nsh_header_t * nsh0;
-  ip4_vxlan_gpe_and_nsh_header_t * h0;
-  int len;
-
-  len = sizeof (*h0) + vec_len(t->nsh_hdr.tlvs)*4;
-
-  vec_validate_aligned (rw, len-1, CLIB_CACHE_LINE_BYTES);
-
-  h0 = (ip4_vxlan_gpe_and_nsh_header_t *) rw;
-
-  /* Fixed portion of the (outer) ip4 header */
-  ip0 = &h0->ip4;
-  ip0->ip_version_and_header_length = 0x45;
-  ip0->ttl = 254;
-  ip0->protocol = IP_PROTOCOL_UDP;
-
-  /* we fix up the ip4 header length and checksum after-the-fact */
-  ip0->src_address.as_u32 = t->src.as_u32;
-  ip0->dst_address.as_u32 = t->dst.as_u32;
-  ip0->checksum = ip4_header_checksum (ip0);
-
-  /* UDP header, randomize src port on something, maybe? */
-  h0->udp.src_port = clib_host_to_net_u16 (4790);
-  h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_vxlan_gpe);
-
-  /* VXLAN header. Are we having fun yet? */
-  h0->vxlan.flags = VXLAN_GPE_FLAGS_I | VXLAN_GPE_FLAGS_P;
-  h0->vxlan.ver_res = VXLAN_GPE_VERSION;
-  h0->vxlan.next_protocol = VXLAN_NEXT_PROTOCOL_NSH;
-  h0->vxlan.vni_res = clib_host_to_net_u32 (t->vni<<8);
-
-  /* NSH header */
-  nsh0 = &h0->nsh;
-  nsh0->ver_o_c = t->nsh_hdr.ver_o_c;
-  nsh0->md_type = t->nsh_hdr.md_type;
-  nsh0->next_protocol = t->nsh_hdr.next_protocol;
-  nsh0->spi_si = t->nsh_hdr.spi_si;
-  nsh0->c1 = t->nsh_hdr.c1;
-  nsh0->c2 = t->nsh_hdr.c2;
-  nsh0->c3 = t->nsh_hdr.c3;
-  nsh0->c4 = t->nsh_hdr.c4;
-  
-  /* Endian swap 32-bit fields */
-#define _(x) nsh0->x = clib_host_to_net_u32(nsh0->x);
-  foreach_32bit_field;
-#undef _
-
-  /* fix nsh header length */
-  t->nsh_hdr.length = 6 + vec_len(t->nsh_hdr.tlvs);
-  nsh0->length = t->nsh_hdr.length;
-
-  /* Copy any TLVs */
-  if (vec_len(t->nsh_hdr.tlvs))
-    clib_memcpy (nsh0->tlvs, t->nsh_hdr.tlvs, 4*vec_len(t->nsh_hdr.tlvs));
-
-  t->rewrite = rw;
-  return (0);
-}
-
-int vnet_nsh_vxlan_gpe_add_del_tunnel 
-(vnet_nsh_vxlan_gpe_add_del_tunnel_args_t *a, u32 * sw_if_indexp)
-{
-  nsh_vxlan_gpe_main_t * ngm = &nsh_vxlan_gpe_main;
-  nsh_vxlan_gpe_tunnel_t *t = 0;
-  vnet_main_t * vnm = ngm->vnet_main;
-  vnet_hw_interface_t * hi;
-  uword * p;
-  u32 hw_if_index = ~0;
-  u32 sw_if_index = ~0;
-  int rv;
-  nsh_vxlan_gpe_tunnel_key_t key, *key_copy;
-  hash_pair_t *hp;
-  
-  key.src = a->dst.as_u32; /* decap src in key is encap dst in config */
-  key.vni = clib_host_to_net_u32 (a->vni << 8);
-  key.spi_si = clib_host_to_net_u32(a->nsh_hdr.spi_si);
-  key.pad = 0;
-
-  p = hash_get_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, &key);
-  
-  if (a->is_add)
-    {
-      /* adding a tunnel: tunnel must not already exist */
-      if (p) 
-        return VNET_API_ERROR_INVALID_VALUE;
-      
-      if (a->decap_next_index >= NSH_VXLAN_GPE_INPUT_N_NEXT)
-        return VNET_API_ERROR_INVALID_DECAP_NEXT;
-      
-      pool_get_aligned (ngm->tunnels, t, CLIB_CACHE_LINE_BYTES);
-      memset (t, 0, sizeof (*t));
-      
-      /* copy from arg structure */
-#define _(x) t->x = a->x;
-      foreach_copy_field;
-#undef _
-
-      /* copy from arg structure */
-#define _(x) t->nsh_hdr.x = a->nsh_hdr.x;
-      foreach_copy_nshhdr_field;
-#undef _
-      
-      rv = nsh_vxlan_gpe_rewrite (t);
-
-      if (rv)
-        {
-          pool_put (ngm->tunnels, t);
-          return rv;
-        }
-
-      key_copy = clib_mem_alloc (sizeof (*key_copy));
-      clib_memcpy (key_copy, &key, sizeof (*key_copy));
-
-      hash_set_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, key_copy, 
-                    t - ngm->tunnels);
-      
-      if (vec_len (ngm->free_nsh_vxlan_gpe_tunnel_hw_if_indices) > 0)
-        {
-          hw_if_index = ngm->free_nsh_vxlan_gpe_tunnel_hw_if_indices
-            [vec_len (ngm->free_nsh_vxlan_gpe_tunnel_hw_if_indices)-1];
-          _vec_len (ngm->free_nsh_vxlan_gpe_tunnel_hw_if_indices) -= 1;
-          
-          hi = vnet_get_hw_interface (vnm, hw_if_index);
-          hi->dev_instance = t - ngm->tunnels;
-          hi->hw_instance = hi->dev_instance;
-        }
-      else 
-        {
-          hw_if_index = vnet_register_interface
-            (vnm, nsh_vxlan_gpe_device_class.index, t - ngm->tunnels,
-             nsh_vxlan_gpe_hw_class.index, t - ngm->tunnels);
-          hi = vnet_get_hw_interface (vnm, hw_if_index);
-          hi->output_node_index = nsh_vxlan_gpe_encap_node.index;
-        }
-      
-      t->hw_if_index = hw_if_index;
-      t->sw_if_index = sw_if_index = hi->sw_if_index;
-      
-      vnet_sw_interface_set_flags (vnm, hi->sw_if_index, 
-                                   VNET_SW_INTERFACE_FLAG_ADMIN_UP);
-    }
-  else
-    {
-      /* deleting a tunnel: tunnel must exist */
-      if (!p) 
-        return VNET_API_ERROR_NO_SUCH_ENTRY;
-
-      t = pool_elt_at_index (ngm->tunnels, p[0]);
-
-      vnet_sw_interface_set_flags (vnm, t->sw_if_index, 0 /* down */);
-      vec_add1 (ngm->free_nsh_vxlan_gpe_tunnel_hw_if_indices, t->hw_if_index);
-
-      hp = hash_get_pair (ngm->nsh_vxlan_gpe_tunnel_by_key, &key);
-      key_copy = (void *)(hp->key);
-      hash_unset_mem (ngm->nsh_vxlan_gpe_tunnel_by_key, &key);
-      clib_mem_free (key_copy);
-
-      vec_free (t->rewrite);
-      pool_put (ngm->tunnels, t);
-    }
-
-  if (sw_if_indexp)
-      *sw_if_indexp = sw_if_index;
-
-  return 0;
-}
-
-static u32 fib_index_from_fib_id (u32 fib_id)
-{
-  ip4_main_t * im = &ip4_main;
-  uword * p;
-
-  p = hash_get (im->fib_index_by_table_id, fib_id);
-  if (!p)
-    return ~0;
-
-  return p[0];
-}
-
-static uword unformat_decap_next (unformat_input_t * input, va_list * args)
-{
-  u32 * result = va_arg (*args, u32 *);
-  u32 tmp;
-  
-  if (unformat (input, "drop"))
-    *result = NSH_VXLAN_GPE_INPUT_NEXT_DROP;
-  else if (unformat (input, "ip4"))
-    *result = NSH_VXLAN_GPE_INPUT_NEXT_IP4_INPUT;
-  else if (unformat (input, "ip6"))
-    *result = NSH_VXLAN_GPE_INPUT_NEXT_IP6_INPUT;
-  else if (unformat (input, "ethernet"))
-    *result = NSH_VXLAN_GPE_INPUT_NEXT_ETHERNET_INPUT;
-  else if (unformat (input, "nsh-vxlan-gpe"))
-    *result = NSH_VXLAN_GPE_INPUT_NEXT_NSH_VXLAN_GPE_ENCAP;
-  else if (unformat (input, "%d", &tmp))
-    *result = tmp;
-  else
-    return 0;
-  return 1;
-}
-
-static clib_error_t *
-nsh_vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm,
-                                   unformat_input_t * input,
-                                   vlib_cli_command_t * cmd)
-{
-  unformat_input_t _line_input, * line_input = &_line_input;
-  ip4_address_t src, dst;
-  u8 is_add = 1;
-  u8 src_set = 0;
-  u8 dst_set = 0;
-  u32 encap_fib_index = 0;
-  u32 decap_fib_index = 0;
-  u8 ver_o_c = 0;
-  u8 length = 0;
-  u8 md_type = 0;
-  u8 next_protocol = 1; /* default: ip4 */
-  u32 decap_next_index = NSH_VXLAN_GPE_INPUT_NEXT_IP4_INPUT;
-  u32 spi;
-  u8 spi_set = 0;
-  u32 si;
-  u32 vni;
-  u8 vni_set = 0;
-  u8 si_set = 0;
-  u32 spi_si;
-  u32 c1 = 0;
-  u32 c2 = 0;
-  u32 c3 = 0;
-  u32 c4 = 0;
-  u32 *tlvs = 0;
-  u32 tmp;
-  int rv;
-  vnet_nsh_vxlan_gpe_add_del_tunnel_args_t _a, * a = &_a;
-  
-  /* Get a line of input. */
-  if (! unformat_user (input, unformat_line_input, line_input))
-    return 0;
-
-  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) {
-    if (unformat (line_input, "del"))
-      is_add = 0;
-    else if (unformat (line_input, "src %U", 
-                       unformat_ip4_address, &src))
-      src_set = 1;
-    else if (unformat (line_input, "dst %U",
-                       unformat_ip4_address, &dst))
-      dst_set = 1;
-    else if (unformat (line_input, "encap-vrf-id %d", &tmp))
-      {
-        encap_fib_index = fib_index_from_fib_id (tmp);
-        if (encap_fib_index == ~0)
-          return clib_error_return (0, "nonexistent encap fib id %d", tmp);
-      }
-    else if (unformat (line_input, "decap-vrf-id %d", &tmp))
-      {
-        decap_fib_index = fib_index_from_fib_id (tmp);
-        if (decap_fib_index == ~0)
-          return clib_error_return (0, "nonexistent decap fib id %d", tmp);
-      }
-    else if (unformat (line_input, "decap-next %U", unformat_decap_next, 
-                       &decap_next_index))
-      ;
-    else if (unformat (line_input, "vni %d", &vni))
-      vni_set = 1;
-    else if (unformat (line_input, "version %d", &tmp))
-      ver_o_c |= (tmp & 3) << 6;
-    else if (unformat (line_input, "o-bit %d", &tmp))
-      ver_o_c |= (tmp & 1) << 5;
-    else if (unformat (line_input, "c-bit %d", &tmp))
-      ver_o_c |= (tmp & 1) << 4;
-    else if (unformat (line_input, "md-type %d", &tmp))
-      md_type = tmp;
-    else if (unformat(line_input, "next-ip4"))
-      next_protocol = 1;
-    else if (unformat(line_input, "next-ip6"))
-      next_protocol = 2;
-    else if (unformat(line_input, "next-ethernet"))
-      next_protocol = 3;
-    else if (unformat(line_input, "next-nsh"))
-      next_protocol = 4;
-    else if (unformat (line_input, "c1 %d", &c1))
-      ;
-    else if (unformat (line_input, "c2 %d", &c2))
-      ;
-    else if (unformat (line_input, "c3 %d", &c3))
-      ;
-    else if (unformat (line_input, "c4 %d", &c4))
-      ;
-    else if (unformat (line_input, "spi %d", &spi))
-      spi_set = 1;
-    else if (unformat (line_input, "si %d", &si))
-      si_set = 1;
-    else if (unformat (line_input, "tlv %x"))
-        vec_add1 (tlvs, tmp);
-    else 
-      return clib_error_return (0, "parse error: '%U'", 
-                                format_unformat_error, line_input);
-  }
-
-  unformat_free (line_input);
-
-  if (src_set == 0)
-    return clib_error_return (0, "tunnel src address not specified");
-
-  if (dst_set == 0)
-    return clib_error_return (0, "tunnel dst address not specified");
-
-  if (vni_set == 0)
-    return clib_error_return (0, "vni not specified");
-
-  if (spi_set == 0)
-    return clib_error_return (0, "spi not specified");
-  
-  if (si_set == 0)
-    return clib_error_return (0, "si not specified");
-
-  spi_si = (spi<<8) | si;
-  
-  memset (a, 0, sizeof (*a));
-
-  a->is_add = is_add;
-
-#define _(x) a->x = x;
-  foreach_copy_field;
-#undef _
-
-#define _(x) a->nsh_hdr.x = x;
-  foreach_copy_nshhdr_field;
-#undef _
-  
-  rv = vnet_nsh_vxlan_gpe_add_del_tunnel (a, 0 /* hw_if_indexp */);
-
-  switch(rv)
-    {
-    case 0:
-      break;
-    case VNET_API_ERROR_INVALID_DECAP_NEXT:
-      return clib_error_return (0, "invalid decap-next...");
-
-    case VNET_API_ERROR_TUNNEL_EXIST:
-      return clib_error_return (0, "tunnel already exists...");
-
-    case VNET_API_ERROR_NO_SUCH_ENTRY:
-      return clib_error_return (0, "tunnel does not exist...");
-
-    default:
-      return clib_error_return 
-        (0, "vnet_nsh_vxlan_gpe_add_del_tunnel returned %d", rv);
-    }
-
-  return 0;
-}
-
-VLIB_CLI_COMMAND (create_nsh_vxlan_gpe_tunnel_command, static) = {
-  .path = "nsh vxlan tunnel",
-  .short_help = 
-  "nsh vxlan tunnel src <ip4-addr> dst <ip4-addr>" 
-  "    c1 <nn> c2 <nn> c3 <nn> c4 <nn> spi <nn> si <nn> vni <nn>\n"
-  "    [encap-vrf-id <nn>] [decap-vrf-id <nn>] [o-bit <1|0>] [c-bit <1|0>]\n"
-  "    [md-type <nn>][next-ip4][next-ip6][next-ethernet][next-nsh]\n"
-  "    [tlv <xx>][decap-next [ip4|ip6|ethernet|nsh-vxlan-gpe]][del]\n",
-  .function = nsh_vxlan_gpe_add_del_tunnel_command_fn,
-};
-
-static clib_error_t *
-show_nsh_vxlan_gpe_tunnel_command_fn (vlib_main_t * vm,
-                                unformat_input_t * input,
-                                vlib_cli_command_t * cmd)
-{
-  nsh_vxlan_gpe_main_t * ngm = &nsh_vxlan_gpe_main;
-  nsh_vxlan_gpe_tunnel_t * t;
-  
-  if (pool_elts (ngm->tunnels) == 0)
-    vlib_cli_output (vm, "No nsh-vxlan-gpe tunnels configured...");
-
-  pool_foreach (t, ngm->tunnels,
-  ({
-    vlib_cli_output (vm, "%U", format_nsh_vxlan_gpe_tunnel, t);
-  }));
-  
-  return 0;
-}
-
-VLIB_CLI_COMMAND (show_nsh_vxlan_gpe_tunnel_command, static) = {
-    .path = "show nsh vxlan tunnel",
-    .function = show_nsh_vxlan_gpe_tunnel_command_fn,
-};
-
-clib_error_t *nsh_vxlan_gpe_init (vlib_main_t *vm)
-{
-  nsh_vxlan_gpe_main_t *ngm = &nsh_vxlan_gpe_main;
-  
-  ngm->vnet_main = vnet_get_main();
-  ngm->vlib_main = vm;
-  
-  ngm->nsh_vxlan_gpe_tunnel_by_key 
-    = hash_create_mem (0, sizeof(nsh_vxlan_gpe_tunnel_key_t), sizeof (uword));
-
-  udp_register_dst_port (vm, UDP_DST_PORT_vxlan_gpe, 
-                         nsh_vxlan_gpe_input_node.index, 1 /* is_ip4 */);
-  return 0;
-}
-
-VLIB_INIT_FUNCTION(nsh_vxlan_gpe_init);
-  
diff --git a/vnet/vnet/nsh/nsh.c b/vnet/vnet/nsh/nsh.c
new file mode 100644
index 00000000000..49edf711038
--- /dev/null
+++ b/vnet/vnet/nsh/nsh.c
@@ -0,0 +1,770 @@
+/*
+ * nsh.c - nsh mapping
+ *
+ * Copyright (c) 2013 Cisco and/or its affiliates.
+ * 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.
+ */
+
+#include <vlib/vlib.h>
+#include <vnet/pg/pg.h>
+#include <vnet/nsh/nsh.h>
+
+
+typedef struct {
+  nsh_header_t nsh_header;
+} nsh_input_trace_t;
+
+u8 * format_nsh_header (u8 * s, va_list * args)
+{
+  nsh_header_t * nsh = va_arg (*args, nsh_header_t *);
+
+  s = format (s, "nsh ver %d ", (nsh->ver_o_c>>6));
+  if (nsh->ver_o_c & NSH_O_BIT)
+      s = format (s, "O-set ");
+
+  if (nsh->ver_o_c & NSH_C_BIT)
+      s = format (s, "C-set ");
+
+  s = format (s, "len %d (%d bytes) md_type %d next_protocol %d\n",
+              nsh->length, nsh->length * 4, nsh->md_type, nsh->next_protocol);
+  
+  s = format (s, "  service path %d service index %d\n",
+              (nsh->nsp_nsi>>NSH_NSP_SHIFT) & NSH_NSP_MASK,
+              nsh->nsp_nsi & NSH_NSI_MASK);
+
+  s = format (s, "  c1 %d c2 %d c3 %d c4 %d\n",
+              nsh->c1, nsh->c2, nsh->c3, nsh->c4);
+
+  return s;
+}
+
+u8 * format_nsh_map (u8 * s, va_list * args)
+{
+  nsh_map_t * map = va_arg (*args, nsh_map_t *);
+
+  s = format (s, "nsh entry nsp: %d nsi: %d ",
+              (map->nsp_nsi>>NSH_NSP_SHIFT) & NSH_NSP_MASK,
+              map->nsp_nsi & NSH_NSI_MASK);
+  s = format (s, "maps to nsp: %d nsi: %d ",
+              (map->mapped_nsp_nsi>>NSH_NSP_SHIFT) & NSH_NSP_MASK,
+              map->mapped_nsp_nsi & NSH_NSI_MASK);
+  
+  switch (map->next_node)
+    {
+    case NSH_INPUT_NEXT_ENCAP_GRE:
+      {
+	s = format (s, "encapped by GRE intf: %d", map->sw_if_index);
+	break;
+      }
+    case NSH_INPUT_NEXT_ENCAP_VXLANGPE:
+      {
+	s = format (s, "encapped by VXLAN GPE intf: %d", map->sw_if_index);
+	break;
+      }
+    default:
+      s = format (s, "only GRE and VXLANGPE support in this rev");
+    }
+
+  return s;
+}
+
+
+#define foreach_copy_nshhdr_field               \
+_(ver_o_c)					\
+_(length)					\
+_(md_type)					\
+_(next_protocol)				\
+_(nsp_nsi)					\
+_(c1)						\
+_(c2)						\
+_(c3)						\
+_(c4)						
+/* Temp killing tlvs as its causing pain - fix in NSH_SFC */
+
+
+#define foreach_32bit_field			\
+_(nsp_nsi)                                      \
+_(c1)                                           \
+_(c2)                                           \
+_(c3)                                           \
+_(c4)
+
+
+u8 * format_nsh_header_with_length (u8 * s, va_list * args)
+{
+  nsh_header_t * h = va_arg (*args, nsh_header_t *);
+  u32 max_header_bytes = va_arg (*args, u32);
+  u32 tmp, header_bytes;
+
+  header_bytes = sizeof (h[0]);
+  if (max_header_bytes != 0 && header_bytes > max_header_bytes)
+    return format (s, "nsh header truncated");
+
+  tmp = clib_net_to_host_u32 (h->nsp_nsi);
+  s = format (s, "  nsp %d nsi %d ",
+              (tmp>>NSH_NSP_SHIFT) & NSH_NSP_MASK,
+              tmp & NSH_NSI_MASK);
+
+  s = format (s, "c1 %u c2 %u c3 %u c4 %u",
+              clib_net_to_host_u32 (h->c1),
+              clib_net_to_host_u32 (h->c2),
+              clib_net_to_host_u32 (h->c3),
+              clib_net_to_host_u32 (h->c4));
+
+  s = format (s, "ver %d ", h->ver_o_c>>6);
+
+  if (h->ver_o_c & NSH_O_BIT)
+      s = format (s, "O-set ");
+
+  if (h->ver_o_c & NSH_C_BIT)
+      s = format (s, "C-set ");
+
+  s = format (s, "len %d (%d bytes) md_type %d next_protocol %d\n",
+              h->length, h->length * 4, h->md_type, h->next_protocol);
+  return s;
+}
+
+u8 * format_nsh_input_map_trace (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+  nsh_input_trace_t * t
+      = va_arg (*args, nsh_input_trace_t *);
+
+  s = format (s, "\n  %U", format_nsh_header, &t->nsh_header, 
+              (u32) sizeof (t->nsh_header) ); 
+
+  return s;
+}
+
+static uword
+nsh_input_map (vlib_main_t * vm,
+               vlib_node_runtime_t * node,
+               vlib_frame_t * from_frame)
+{
+  u32 n_left_from, next_index, * from, * to_next;
+  nsh_main_t * nm = &nsh_main;
+
+  from = vlib_frame_vector_args (from_frame);
+  n_left_from = from_frame->n_vectors;
+
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      u32 n_left_to_next;
+
+      vlib_get_next_frame (vm, node, next_index,
+			   to_next, n_left_to_next);
+
+      while (n_left_from >= 4 && n_left_to_next >= 2)
+	{
+	  u32 bi0, bi1;
+	  vlib_buffer_t * b0, * b1;
+	  u32 next0 = NSH_INPUT_NEXT_DROP, next1 = NSH_INPUT_NEXT_DROP; 
+	  uword * entry0, * entry1;
+	  nsh_header_t * hdr0 = 0, * hdr1 = 0;
+	  u32 nsp_nsi0, nsp_nsi1;
+	  u32 error0, error1;
+	  nsh_map_t * map0 = 0, * map1 = 0;
+
+	  /* Prefetch next iteration. */
+	  {
+	    vlib_buffer_t * p2, * p3;
+
+	    p2 = vlib_get_buffer (vm, from[2]);
+	    p3 = vlib_get_buffer (vm, from[3]);
+
+	    vlib_prefetch_buffer_header (p2, LOAD);
+	    vlib_prefetch_buffer_header (p3, LOAD);
+
+	    CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
+	    CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
+	  }
+
+	  bi0 = from[0];
+	  bi1 = from[1];
+	  to_next[0] = bi0;
+	  to_next[1] = bi1;
+	  from += 2;
+	  to_next += 2;
+	  n_left_from -= 2;
+	  n_left_to_next -= 2;
+
+	  error0 = 0;
+	  error1 = 0;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  hdr0 = vlib_buffer_get_current (b0);
+	  nsp_nsi0 = clib_net_to_host_u32(hdr0->nsp_nsi);
+	  entry0 = hash_get_mem (nm->nsh_mapping_by_key, &nsp_nsi0);
+
+	  b1 = vlib_get_buffer (vm, bi1);
+	  hdr1 = vlib_buffer_get_current (b1);
+	  nsp_nsi1 = clib_net_to_host_u32(hdr1->nsp_nsi);
+	  entry1 = hash_get_mem (nm->nsh_mapping_by_key, &nsp_nsi1);
+
+ 	  if (PREDICT_FALSE(entry0 == 0))
+	    {
+	      error0 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace0;
+	    }
+	  
+	  if (PREDICT_FALSE(entry1 == 0))
+	    {
+	      error1 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace1;
+	    }
+	  
+ 	  /* Entry should point to a mapping ...*/
+	  map0 = pool_elt_at_index (nm->nsh_mappings, entry0[0]);
+	  map1 = pool_elt_at_index (nm->nsh_mappings, entry1[0]);
+
+	  if (PREDICT_FALSE(map0 == 0))
+	    {
+	      error0 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace0;
+	    }
+
+	  if (PREDICT_FALSE(map1 == 0))
+	    {
+	      error1 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace1;
+	    }
+
+	  entry0 = hash_get_mem (nm->nsh_entry_by_key, &map0->mapped_nsp_nsi);
+	  entry1 = hash_get_mem (nm->nsh_entry_by_key, &map1->mapped_nsp_nsi);
+
+	  if (PREDICT_FALSE(entry0 == 0))
+	    {
+	      error0 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace0;
+	    }
+	  if (PREDICT_FALSE(entry1 == 0))
+	    {
+	      error1 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace1;
+	    }
+
+	  hdr0 = pool_elt_at_index (nm->nsh_entries, entry0[0]);
+	  hdr1 = pool_elt_at_index (nm->nsh_entries, entry1[0]);
+
+	  /* set up things for next node to transmit ie which node to handle it and where */
+	  next0 = map0->next_node;
+	  next1 = map1->next_node;
+	  vnet_buffer(b0)->sw_if_index[VLIB_TX] = map0->sw_if_index;
+	  vnet_buffer(b1)->sw_if_index[VLIB_TX] = map1->sw_if_index;
+
+        trace0:
+          b0->error = error0 ? node->errors[error0] : 0;
+
+	  if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      nsh_input_trace_t *tr =
+		vlib_add_trace (vm, node, b0, sizeof (*tr));
+              tr->nsh_header = *hdr0; 
+            }
+
+        trace1:
+          b1->error = error1 ? node->errors[error1] : 0;
+
+	  if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      nsh_input_trace_t *tr =
+		vlib_add_trace (vm, node, b1, sizeof (*tr));
+              tr->nsh_header = *hdr1; 
+            }
+
+
+	  vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, bi1, next0, next1);
+
+	}
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t * b0;
+	  u32 next0 = NSH_INPUT_NEXT_DROP; 
+	  uword * entry0;
+	  nsh_header_t * hdr0 = 0;
+	  u32 nsp_nsi0;
+	  u32 error0;
+	  nsh_map_t * map0 = 0;
+
+	  next_index = next0; 
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+	  error0 = 0;
+
+	  b0 = vlib_get_buffer (vm, bi0);
+	  hdr0 = vlib_buffer_get_current (b0);
+	  nsp_nsi0 = clib_net_to_host_u32(hdr0->nsp_nsi);
+	  entry0 = hash_get_mem (nm->nsh_mapping_by_key, &nsp_nsi0);
+
+	  if (PREDICT_FALSE(entry0 == 0))
+	    {
+	      error0 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace00;
+	    }
+	  
+	  /* Entry should point to a mapping ...*/
+	  map0 = pool_elt_at_index (nm->nsh_mappings, entry0[0]);
+
+	  if (PREDICT_FALSE(map0 == 0))
+	    {
+	      error0 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace00;
+	    }
+
+	  entry0 = hash_get_mem (nm->nsh_entry_by_key, &map0->mapped_nsp_nsi);
+
+	  if (PREDICT_FALSE(entry0 == 0))
+	    {
+	      error0 = NSH_INPUT_ERROR_NO_MAPPING;
+	      goto trace00;
+	    }
+
+	  hdr0 = pool_elt_at_index (nm->nsh_entries, entry0[0]);
+
+	  /* set up things for next node to transmit ie which node to handle it and where */
+	  next0 = map0->next_node;
+	  vnet_buffer(b0)->sw_if_index[VLIB_TX] = map0->sw_if_index;
+
+        trace00:
+          b0->error = error0 ? node->errors[error0] : 0;
+
+	  if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      nsh_input_trace_t *tr =
+		vlib_add_trace (vm, node, b0, sizeof (*tr));
+              tr->nsh_header = *hdr0; 
+            }
+
+
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+	  				   to_next, n_left_to_next,
+	  				   bi0, next0);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+
+    }
+
+
+  return from_frame->n_vectors;
+}
+
+
+int vnet_nsh_add_del_map (vnet_nsh_add_del_map_args_t *a)
+{
+  nsh_main_t * nm = &nsh_main;
+  nsh_map_t *map = 0;
+  u32 key, *key_copy;
+  uword * entry;
+  hash_pair_t *hp;
+
+  key = a->map.nsp_nsi;
+  
+  entry = hash_get_mem (nm->nsh_mapping_by_key, &key);
+
+  if (a->is_add)
+    {
+      /* adding an entry, must not already exist */
+      if (entry)
+        return VNET_API_ERROR_INVALID_VALUE;
+      
+      pool_get_aligned (nm->nsh_mappings, map, CLIB_CACHE_LINE_BYTES);
+      memset (map, 0, sizeof (*map));
+
+      /* copy from arg structure */
+      map->nsp_nsi = a->map.nsp_nsi;
+      map->mapped_nsp_nsi = a->map.mapped_nsp_nsi;
+      map->sw_if_index = a->map.sw_if_index;
+      map->next_node = a->map.next_node;
+      
+
+      key_copy = clib_mem_alloc (sizeof (*key_copy));
+      clib_memcpy (key_copy, &key, sizeof (*key_copy));
+
+      hash_set_mem (nm->nsh_mapping_by_key, key_copy,
+                    map - nm->nsh_mappings);
+    }
+  else
+    {
+      if (!entry)
+	return VNET_API_ERROR_NO_SUCH_ENTRY;
+
+      map = pool_elt_at_index (nm->nsh_mappings, entry[0]);
+      hp = hash_get_pair (nm->nsh_mapping_by_key, &key);
+      key_copy = (void *)(hp->key);
+      hash_unset_mem (nm->nsh_mapping_by_key, &key);
+      clib_mem_free (key_copy);
+
+      pool_put (nm->nsh_mappings, map);
+    }
+
+  return 0;
+}
+
+static clib_error_t *
+nsh_add_del_map_command_fn (vlib_main_t * vm,
+                                   unformat_input_t * input,
+                                   vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, * line_input = &_line_input;
+  u8 is_add = 1;
+  u32 nsp, nsi, mapped_nsp, mapped_nsi;
+  int nsp_set = 0, nsi_set = 0, mapped_nsp_set = 0, mapped_nsi_set = 0;
+  u32 next_node = ~0;
+  u32 sw_if_index = ~0; // temporary requirement to get this moved over to NSHSFC
+  vnet_nsh_add_del_map_args_t _a, * a = &_a;
+  int rv;
+
+  /* Get a line of input. */
+  if (! unformat_user (input, unformat_line_input, line_input))
+    return 0;
+  
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) {
+    if (unformat (line_input, "del"))
+      is_add = 0;
+    else if (unformat (line_input, "nsp %d", &nsp))
+      nsp_set = 1;
+    else if (unformat (line_input, "nsi %d", &nsi))
+      nsi_set = 1;
+    else if (unformat (line_input, "mapped-nsp %d", &mapped_nsp))
+      mapped_nsp_set = 1;
+    else if (unformat (line_input, "mapped-nsi %d", &mapped_nsi))
+      mapped_nsi_set = 1;
+    else if (unformat (line_input, "encap-gre-intf %d", &sw_if_index))
+      next_node = NSH_INPUT_NEXT_ENCAP_GRE;
+    else if (unformat (line_input, "encap-vxlan-gpe-intf %d", &sw_if_index))
+      next_node = NSH_INPUT_NEXT_ENCAP_VXLANGPE;
+    else if (unformat (line_input, "encap-none"))
+      next_node = NSH_INPUT_NEXT_DROP; // Once moved to NSHSFC see nsh.h:foreach_nsh_input_next to handle this case
+    else 
+      return clib_error_return (0, "parse error: '%U'", 
+                                format_unformat_error, line_input);
+  }
+
+  unformat_free (line_input);
+
+  if (nsp_set == 0 || nsi_set == 0)
+    return clib_error_return (0, "nsp nsi pair required. Key: for NSH entry");
+
+  if (mapped_nsp_set == 0 || mapped_nsi_set == 0)
+    return clib_error_return (0, "mapped-nsp mapped-nsi pair required. Key: for NSH entry");
+
+  if (next_node == ~0)
+    return clib_error_return (0, "must specific action: [encap-gre-intf <nn> | encap-vxlan-gpe-intf <nn> | encap-none]");
+
+  memset (a, 0, sizeof (*a));
+
+  /* set args structure */
+  a->is_add = is_add;
+  a->map.nsp_nsi = (nsp<< NSH_NSP_SHIFT) | nsi;
+  a->map.mapped_nsp_nsi = (mapped_nsp<< NSH_NSP_SHIFT) | mapped_nsi;
+  a->map.sw_if_index = sw_if_index;
+  a->map.next_node = next_node;
+
+
+  rv = vnet_nsh_add_del_map (a);
+
+  switch(rv)
+    {
+    case 0:
+      break;
+    case VNET_API_ERROR_INVALID_VALUE:
+      return clib_error_return (0, "mapping already exists. Remove it first.");
+
+    case VNET_API_ERROR_NO_SUCH_ENTRY:
+      return clib_error_return (0, "mapping does not exist.");
+
+    default:
+      return clib_error_return 
+        (0, "vnet_nsh_add_del_map returned %d", rv);
+    }
+  return 0;
+}
+
+
+VLIB_CLI_COMMAND (create_nsh_map_command, static) = {
+  .path = "create nsh map",
+  .short_help = 
+  "create nsh map nsp <nn> nsi <nn> [del] map-nsp <nn> map-nsi <nn> [encap-gre-intf <nn> | encap-vxlan-gpe-intf <nn> | encap-none]\n",
+  .function = nsh_add_del_map_command_fn,
+};
+
+static clib_error_t *
+show_nsh_map_command_fn (vlib_main_t * vm,
+				      unformat_input_t * input,
+				      vlib_cli_command_t * cmd)
+{
+  nsh_main_t * nm = &nsh_main;
+  nsh_map_t * map;
+
+  if (pool_elts (nm->nsh_mappings) == 0)
+    vlib_cli_output (vm, "No nsh maps configured.");
+
+  pool_foreach (map, nm->nsh_mappings,
+		({
+		  vlib_cli_output (vm, "%U", format_nsh_map, map);
+		}));
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (show_nsh_map_command, static) = {
+  .path = "show nsh map",
+  .function = show_nsh_map_command_fn,
+};
+
+
+int vnet_nsh_add_del_entry (vnet_nsh_add_del_entry_args_t *a)
+{
+  nsh_main_t * nm = &nsh_main;
+  nsh_header_t *hdr = 0;
+  u32 key, *key_copy;
+  uword * entry;
+  hash_pair_t *hp;
+
+  key = a->nsh.nsp_nsi;
+  
+  entry = hash_get_mem (nm->nsh_entry_by_key, &key);
+
+  if (a->is_add)
+    {
+      /* adding an entry, must not already exist */
+      if (entry)
+        return VNET_API_ERROR_INVALID_VALUE;
+      
+      pool_get_aligned (nm->nsh_entries, hdr, CLIB_CACHE_LINE_BYTES);
+      memset (hdr, 0, sizeof (*hdr));
+
+      /* copy from arg structure */
+#define _(x) hdr->x = a->nsh.x;
+      foreach_copy_nshhdr_field;
+#undef _
+
+      key_copy = clib_mem_alloc (sizeof (*key_copy));
+      clib_memcpy (key_copy, &key, sizeof (*key_copy));
+
+      hash_set_mem (nm->nsh_entry_by_key, key_copy,
+                    hdr - nm->nsh_entries);
+    }
+  else
+    {
+      if (!entry)
+	return VNET_API_ERROR_NO_SUCH_ENTRY;
+
+      hdr = pool_elt_at_index (nm->nsh_entries, entry[0]);
+      hp = hash_get_pair (nm->nsh_entry_by_key, &key);
+      key_copy = (void *)(hp->key);
+      hash_unset_mem (nm->nsh_entry_by_key, &key);
+      clib_mem_free (key_copy);
+
+      pool_put (nm->nsh_entries, hdr);
+    }
+
+  return 0;
+}
+
+
+static clib_error_t *
+nsh_add_del_entry_command_fn (vlib_main_t * vm,
+                                   unformat_input_t * input,
+                                   vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, * line_input = &_line_input;
+  u8 is_add = 1;
+  u8 ver_o_c = 0;
+  u8 length = 0;
+  u8 md_type = 0;
+  u8 next_protocol = 1; /* default: ip4 */
+  u32 nsp;
+  u8 nsp_set = 0;
+  u32 nsi;
+  u8 nsi_set = 0;
+  u32 nsp_nsi;
+  u32 c1 = 0;
+  u32 c2 = 0;
+  u32 c3 = 0;
+  u32 c4 = 0;
+  u32 *tlvs = 0;
+  u32 tmp;
+  int rv;
+  vnet_nsh_add_del_entry_args_t _a, * a = &_a;
+  
+  /* Get a line of input. */
+  if (! unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) {
+    if (unformat (line_input, "del"))
+      is_add = 0;
+    else if (unformat (line_input, "version %d", &tmp))
+      ver_o_c |= (tmp & 3) << 6;
+    else if (unformat (line_input, "o-bit %d", &tmp))
+      ver_o_c |= (tmp & 1) << 5;
+    else if (unformat (line_input, "c-bit %d", &tmp))
+      ver_o_c |= (tmp & 1) << 4;
+    else if (unformat (line_input, "md-type %d", &tmp))
+      md_type = tmp;
+    else if (unformat(line_input, "next-ip4"))
+      next_protocol = 1;
+    else if (unformat(line_input, "next-ip6"))
+      next_protocol = 2;
+    else if (unformat(line_input, "next-ethernet"))
+      next_protocol = 3;
+    else if (unformat (line_input, "c1 %d", &c1))
+      ;
+    else if (unformat (line_input, "c2 %d", &c2))
+      ;
+    else if (unformat (line_input, "c3 %d", &c3))
+      ;
+    else if (unformat (line_input, "c4 %d", &c4))
+      ;
+    else if (unformat (line_input, "nsp %d", &nsp))
+      nsp_set = 1;
+    else if (unformat (line_input, "nsi %d", &nsi))
+      nsi_set = 1;
+    else if (unformat (line_input, "tlv %x"))
+        vec_add1 (tlvs, tmp);
+    else 
+      return clib_error_return (0, "parse error: '%U'", 
+                                format_unformat_error, line_input);
+  }
+
+  unformat_free (line_input);
+
+  if (nsp_set == 0)
+    return clib_error_return (0, "nsp not specified");
+  
+  if (nsi_set == 0)
+    return clib_error_return (0, "nsi not specified");
+
+  if (md_type != 1)
+    return clib_error_return (0, "md-type 1 only supported at this time");
+
+  md_type = 1;
+  length = 6; 
+
+  nsp_nsi = (nsp<<8) | nsi;
+  
+  memset (a, 0, sizeof (*a));
+
+  a->is_add = is_add;
+
+#define _(x) a->nsh.x = x;
+  foreach_copy_nshhdr_field;
+#undef _
+  
+  a->nsh.tlvs[0] = 0 ; // TODO FIXME this shouldn't be set 0 - in NSH_SFC project
+
+  rv = vnet_nsh_add_del_entry (a);
+
+  switch(rv)
+    {
+    case 0:
+      break;
+    default:
+      return clib_error_return 
+        (0, "vnet_nsh_add_del_entry returned %d", rv);
+    }
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (create_nsh_entry_command, static) = {
+  .path = "create nsh entry",
+  .short_help = 
+  "create nsh entry {nsp <nn> nsi <nn>} c1 <nn> c2 <nn> c3 <nn> c4 <nn>"
+  " [md-type <nn>] [tlv <xx>] [del]\n",
+  .function = nsh_add_del_entry_command_fn,
+};
+
+static clib_error_t *
+show_nsh_entry_command_fn (vlib_main_t * vm,
+				      unformat_input_t * input,
+				      vlib_cli_command_t * cmd)
+{
+  nsh_main_t * nm = &nsh_main;
+  nsh_header_t * hdr;
+
+  if (pool_elts (nm->nsh_entries) == 0)
+    vlib_cli_output (vm, "No nsh entries configured.");
+
+  pool_foreach (hdr, nm->nsh_entries,
+		({
+		  vlib_cli_output (vm, "%U", format_nsh_header, hdr);
+		}));
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (show_nsh_entry_command, static) = {
+  .path = "show nsh entry",
+  .function = show_nsh_entry_command_fn,
+};
+
+static char * nsh_input_error_strings[] = {
+#define _(sym,string) string,
+  foreach_nsh_input_error
+#undef _
+};
+
+VLIB_REGISTER_NODE (nsh_input_node) = {
+  .function = nsh_input_map,
+  .name = "nsh-input",
+  .vector_size = sizeof (u32),
+  .format_trace = format_nsh_input_map_trace,
+  .format_buffer = format_nsh_header_with_length,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+
+  .n_errors = ARRAY_LEN(nsh_input_error_strings),
+  .error_strings = nsh_input_error_strings,
+
+  .n_next_nodes = NSH_INPUT_N_NEXT,
+
+  .next_nodes = {
+#define _(s,n) [NSH_INPUT_NEXT_##s] = n,
+    foreach_nsh_input_next
+#undef _
+  },
+};
+
+clib_error_t *nsh_init (vlib_main_t *vm)
+{
+  nsh_main_t *nm = &nsh_main;
+  
+  nm->vnet_main = vnet_get_main();
+  nm->vlib_main = vm;
+  
+  nm->nsh_mapping_by_key
+    = hash_create_mem (0, sizeof(u32), sizeof (uword));
+
+  nm->nsh_mapping_by_mapped_key
+    = hash_create_mem (0, sizeof(u32), sizeof (uword));
+
+  nm->nsh_entry_by_key
+    = hash_create_mem (0, sizeof(u32), sizeof (uword));
+
+  return 0;
+}
+
+VLIB_INIT_FUNCTION(nsh_init);
diff --git a/vnet/vnet/nsh/nsh.h b/vnet/vnet/nsh/nsh.h
new file mode 100644
index 00000000000..d1c46121d25
--- /dev/null
+++ b/vnet/vnet/nsh/nsh.h
@@ -0,0 +1,108 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * 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.
+ */
+#ifndef included_vnet_nsh_h
+#define included_vnet_nsh_h
+
+#include <vnet/vnet.h>
+#include <vnet/nsh/nsh_packet.h>
+#include <vnet/ip/ip4_packet.h>
+
+typedef struct {
+
+  /** Key for nsh_header_t entry: 24bit NSP 8bit NSI */
+  u32 nsp_nsi;
+
+  /** Key for nsh_header_t entry to map to. : 24bit NSP 8bit NSI 
+   *  This may be ~0 if next action is to decap to NSH next protocol
+   *  Note the following heuristic:
+   *  if nsp_nsi == mapped_nsp_nsi then use-case is like SFC SFF
+   *  if nsp_nsi != mapped_nsp_nsi then use-case is like SFC SF
+   *  Note: these are heuristics. Rules about NSI decrement are out of scope
+   */
+  u32 mapped_nsp_nsi;
+
+  /* vnet intfc sw_if_index */
+  u32 sw_if_index;
+
+  u32 next_node;
+
+} nsh_map_t;
+
+typedef struct {
+  nsh_map_t map;
+  u32 is_add;
+} vnet_nsh_add_del_map_args_t;
+
+typedef struct {
+  u8 is_add;
+  nsh_header_t nsh;
+} vnet_nsh_add_del_entry_args_t;
+
+typedef struct {
+  /* vector of nsh_header entry instances */
+  nsh_header_t *nsh_entries;
+
+  /* hash lookup nsh header by key: {u32: nsp_nsi} */
+  uword * nsh_entry_by_key;
+
+  /* vector of nsh_mappings */
+  nsh_map_t *nsh_mappings;
+
+  /* hash lookup nsh mapping by key: {u32: nsp_nsi} */
+  uword * nsh_mapping_by_key;
+  uword * nsh_mapping_by_mapped_key; // for use in NSHSFC
+
+  /* convenience */
+  vlib_main_t * vlib_main;
+  vnet_main_t * vnet_main;
+} nsh_main_t;
+
+nsh_main_t nsh_main;
+
+u8 * format_nsh_input_map_trace (u8 * s, va_list * args);
+u8 * format_nsh_header_with_length (u8 * s, va_list * args);
+
+/* Statistics (not really errors) */
+#define foreach_nsh_input_error    \
+_(MAPPED, "NSH header found and mapped") \
+_(NO_MAPPING, "no mapping for nsh key") \
+_(INVALID_NEXT_PROTOCOL, "invalid next protocol") \
+
+typedef enum {
+#define _(sym,str) NSH_INPUT_ERROR_##sym,
+    foreach_nsh_input_error
+#undef _
+  NSH_INPUT_N_ERROR,
+
+} nsh_input_error_t;
+
+#define foreach_nsh_input_next        \
+  _(DROP, "error-drop")			\
+  _(ENCAP_GRE, "gre-input" )		\
+  _(ENCAP_VXLANGPE, "vxlan-gpe-encap" ) \
+/* /\* TODO once moved to Project:NSH_SFC *\/ */
+  /* _(ENCAP_ETHERNET, "*** TX TO ETHERNET ***")   \ */
+/*   _(DECAP_ETHERNET_LOOKUP, "ethernet-input" )	\ */
+/*   _(DECAP_IP4_INPUT,  "ip4-input") \ */
+/*   _(DECAP_IP6_INPUT,  "ip6-input" ) \  */
+
+typedef enum {
+#define _(s,n) NSH_INPUT_NEXT_##s,
+  foreach_nsh_input_next
+#undef _
+  NSH_INPUT_N_NEXT,
+} nsh_input_next_t;
+
+#endif /* included_vnet_nsh_h */
diff --git a/vnet/vnet/nsh/nsh_error.def b/vnet/vnet/nsh/nsh_error.def
index 532b02a6e89..c54e3b895c8 100644
--- a/vnet/vnet/nsh/nsh_error.def
+++ b/vnet/vnet/nsh/nsh_error.def
@@ -12,6 +12,6 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-nsh_gre_error (DECAPSULATED, "good packets decapsulated")
-nsh_gre_error (NO_SUCH_TUNNEL, "no such tunnel packets")
-nsh_gre_error (INVALID_NEXT_PROTOCOL, "invalid next protocol")
+nsh_input_error (DECAPSULATED, "good packets decapsulated")
+nsh_input_error (NO_MAPPING, "no mapping for nsh key")
+nsh_input_error (INVALID_NEXT_PROTOCOL, "invalid next protocol")
\ No newline at end of file
diff --git a/vnet/vnet/nsh/nsh_gre_error.def b/vnet/vnet/nsh/nsh_gre_error.def
new file mode 100644
index 00000000000..45d8ef424b5
--- /dev/null
+++ b/vnet/vnet/nsh/nsh_gre_error.def
@@ -0,0 +1,17 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * 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.
+ */
+nsh_gre_error (DECAPSULATED, "good packets decapsulated")
+nsh_gre_error (NO_SUCH_TUNNEL, "no such tunnel packets")
+nsh_gre_error (INVALID_NEXT_PROTOCOL, "invalid next protocol")
\ No newline at end of file
diff --git a/vnet/vnet/nsh/nsh_packet.h b/vnet/vnet/nsh/nsh_packet.h
index 87d46a93b6d..cbe4f1e4193 100644
--- a/vnet/vnet/nsh/nsh_packet.h
+++ b/vnet/vnet/nsh/nsh_packet.h
@@ -73,20 +73,21 @@ typedef CLIB_PACKED(struct {
   u8 length;
   u8 md_type;
   u8 next_protocol;
-  u32 spi_si;
+  u32 nsp_nsi; // nsp 24 bits, nsi 8 bits
   /* Context headers, always present */
   u32 c1; u32 c2; u32 c3; u32 c4;
 
   /* Optional variable length metadata */
-  u32 * tlvs;
+  u32 tlvs[0];
 }) nsh_header_t;
 
+#define NSH_VERSION (0<<6)
 #define NSH_O_BIT (1<<5)
 #define NSH_C_BIT (1<<4)
 
 /* Network byte order shift / mask */
-#define NSH_SINDEX_MASK 0xFF
-#define NSH_SPI_MASK (0x00FFFFFF)
-#define NSH_SPI_SHIFT 8
+#define NSH_NSI_MASK 0xFF
+#define NSH_NSP_MASK (0x00FFFFFF)
+#define NSH_NSP_SHIFT 8
 
 #endif /* included_vnet_nsh_packet_h */
diff --git a/vnet/vnet/nsh-gre/decap.c b/vnet/vnet/vxlan-gpe/decap.c
index c10b11b3070..aed5857d0b1 100644
--- a/vnet/vnet/nsh-gre/decap.c
+++ b/vnet/vnet/vxlan-gpe/decap.c
@@ -1,5 +1,5 @@
 /*
- * nsh.c: nsh packet processing
+ * decap.c - decapsulate VXLAN GPE
  *
  * Copyright (c) 2013 Cisco and/or its affiliates.
  * Licensed under the Apache License, Version 2.0 (the "License");
@@ -17,92 +17,60 @@
 
 #include <vlib/vlib.h>
 #include <vnet/pg/pg.h>
-#include <vnet/nsh-gre/nsh_gre.h>
-#include <vnet/nsh/nsh_packet.h>
-
-vlib_node_registration_t nsh_input_node;
+#include <vnet/vxlan-gpe/vxlan_gpe.h>
 
 typedef struct {
   u32 next_index;
   u32 tunnel_index;
   u32 error;
-  nsh_header_t h;
-} nsh_rx_trace_t;
-
+} vxlan_gpe_rx_trace_t;
 
-u8 * format_nsh_header_with_length (u8 * s, va_list * args)
-{
-  nsh_header_t * h = va_arg (*args, nsh_header_t *);
-  u32 max_header_bytes = va_arg (*args, u32);
-  u32 tmp, header_bytes;
-
-  header_bytes = sizeof (h[0]);
-  if (max_header_bytes != 0 && header_bytes > max_header_bytes)
-    return format (s, "gre-nsh header truncated");
-
-  s = format (s, "ver %d ", h->ver_o_c>>6);
-
-  if (h->ver_o_c & NSH_O_BIT)
-      s = format (s, "O-set ");
-
-  if (h->ver_o_c & NSH_C_BIT)
-      s = format (s, "C-set ");
-
-  s = format (s, "len %d (%d bytes) md_type %d next_protocol %d\n",
-              h->length, h->length * 4, h->md_type, h->next_protocol);
-  
-  tmp = clib_net_to_host_u32 (h->spi_si);
-
-  s = format (s, "  spi %d si %d ",
-              (tmp>>NSH_SPI_SHIFT) & NSH_SPI_MASK,
-              tmp & NSH_SINDEX_MASK);
-
-  s = format (s, "c1 %u c2 %u c3 %u c4 %u",
-              clib_net_to_host_u32 (h->c1),
-              clib_net_to_host_u32 (h->c2),
-              clib_net_to_host_u32 (h->c3),
-              clib_net_to_host_u32 (h->c4));
-
-  return s;
-}
-
-
-u8 * format_nsh_rx_trace (u8 * s, va_list * args)
+static u8 * format_vxlan_gpe_rx_trace (u8 * s, va_list * args)
 {
   CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
   CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  nsh_rx_trace_t * t = va_arg (*args, nsh_rx_trace_t *);
+  vxlan_gpe_rx_trace_t * t = va_arg (*args, vxlan_gpe_rx_trace_t *);
 
   if (t->tunnel_index != ~0)
     {
-      s = format (s, "NSH: tunnel %d next %d error %d", t->tunnel_index, 
+      s = format (s, "VXLAN-GPE: tunnel %d next %d error %d", t->tunnel_index, 
                   t->next_index, t->error);
     }
   else
     {
-      s = format (s, "NSH: no tunnel next %d error %d\n", t->next_index, 
+      s = format (s, "VXLAN-GPE: no tunnel next %d error %d\n", t->next_index, 
                   t->error);
     }
-  s = format (s, "\n  %U", format_nsh_header_with_length, &t->h, 
-              (u32) sizeof (t->h) /* max size */);
+  return s;
+}
+
+
+static u8 * format_vxlan_gpe_with_length (u8 * s, va_list * args)
+{
+  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
+  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
+
+
   return s;
 }
 
 static uword
-nsh_gre_input (vlib_main_t * vm,
-               vlib_node_runtime_t * node,
-               vlib_frame_t * from_frame)
+vxlan_gpe_input (vlib_main_t * vm,
+                     vlib_node_runtime_t * node,
+                     vlib_frame_t * from_frame)
 {
   u32 n_left_from, next_index, * from, * to_next;
-  nsh_gre_main_t * ngm = &nsh_gre_main;
+  vxlan_gpe_main_t * ngm = &vxlan_gpe_main;
   vnet_main_t * vnm = ngm->vnet_main;
   vnet_interface_main_t * im = &vnm->interface_main;
   u32 last_tunnel_index = ~0;
-  u64 last_key = ~0ULL;
+  vxlan_gpe_tunnel_key_t last_key;
   u32 pkts_decapsulated = 0;
   u32 cpu_index = os_get_cpu_number();
   u32 stats_sw_if_index, stats_n_packets, stats_n_bytes;
 
+  memset (&last_key, 0xff, sizeof (last_key));
+
   from = vlib_frame_vector_args (from_frame);
   n_left_from = from_frame->n_vectors;
 
@@ -122,11 +90,11 @@ nsh_gre_input (vlib_main_t * vm,
           u32 bi0, bi1;
 	  vlib_buffer_t * b0, * b1;
 	  u32 next0, next1;
-	  nsh_header_t * h0, * h1;
+          ip4_vxlan_gpe_header_t * iuvn0, * iuvn1;
 	  uword * p0, * p1;
           u32 tunnel_index0, tunnel_index1;
-          nsh_gre_tunnel_t * t0, * t1;
-          u64 key0, key1;
+          vxlan_gpe_tunnel_t * t0, * t1;
+          vxlan_gpe_tunnel_key_t key0, key1;
           u32 error0, error1;
           u32 sw_if_index0, sw_if_index1, len0, len1;
 
@@ -156,35 +124,56 @@ nsh_gre_input (vlib_main_t * vm,
 	  b0 = vlib_get_buffer (vm, bi0);
 	  b1 = vlib_get_buffer (vm, bi1);
 
-          h0 = vlib_buffer_get_current (b0);
-          h1 = vlib_buffer_get_current (b1);
+          /* udp leaves current_data pointing at the vxlan header */
+          vlib_buffer_advance 
+            (b0, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
+          vlib_buffer_advance 
+            (b1, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
 
-          /* gre stashed the src ip4 address for us... */
-          key0 = (((u64)(vnet_buffer(b0)->gre.src))<<32) | h0->spi_si;
-          key1 = (((u64)(vnet_buffer(b1)->gre.src))<<32) | h1->spi_si;
+          iuvn0 = vlib_buffer_get_current (b0);
+          iuvn1 = vlib_buffer_get_current (b1);
 
-          /* "pop" nsh header */
-          vlib_buffer_advance (b0, sizeof (*h0));
-          vlib_buffer_advance (b1, sizeof (*h1));
+          /* pop (ip, udp, vxlan) */
+          vlib_buffer_advance (b0, sizeof (*iuvn0));
+          vlib_buffer_advance (b1, sizeof (*iuvn1));
 
           tunnel_index0 = ~0;
           tunnel_index1 = ~0;
           error0 = 0;
           error1 = 0;
-          next0 = NSH_GRE_INPUT_NEXT_DROP;
-          next1 = NSH_GRE_INPUT_NEXT_DROP;
 
-          if (PREDICT_FALSE(key0 != last_key))
+	  next0 = (iuvn0->vxlan.protocol < VXLAN_GPE_INPUT_N_NEXT) ? iuvn0->vxlan.protocol : VXLAN_GPE_INPUT_NEXT_DROP;
+	  next1 = (iuvn1->vxlan.protocol < VXLAN_GPE_INPUT_N_NEXT) ? iuvn1->vxlan.protocol : VXLAN_GPE_INPUT_NEXT_DROP;
+
+
+
+
+          key0.local = iuvn0->ip4.dst_address.as_u32;
+          key1.local = iuvn1->ip4.dst_address.as_u32;
+
+	  key0.remote = iuvn0->ip4.src_address.as_u32;
+	  key1.remote = iuvn1->ip4.src_address.as_u32;
+
+          key0.vni = iuvn0->vxlan.vni_res;
+          key1.vni = iuvn1->vxlan.vni_res;
+
+          key0.pad = 0;
+          key1.pad = 0;
+
+	  /* Processing for key0 */
+          if (PREDICT_FALSE ((key0.as_u64[0] != last_key.as_u64[0])
+                             || (key0.as_u64[1] != last_key.as_u64[1])))
             {
-              p0 = hash_get (ngm->nsh_gre_tunnel_by_src_address, key0);
+              p0 = hash_get_mem (ngm->vxlan_gpe_tunnel_by_key, &key0);
 
               if (p0 == 0)
                 {
-                  error0 = NSH_GRE_ERROR_NO_SUCH_TUNNEL;
+                  error0 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
                   goto trace0;
                 }
 
-              last_key = key0;
+              last_key.as_u64[0] = key0.as_u64[0];
+              last_key.as_u64[1] = key0.as_u64[1];
               tunnel_index0 = last_tunnel_index = p0[0];
             }
           else
@@ -192,17 +181,18 @@ nsh_gre_input (vlib_main_t * vm,
 
           t0 = pool_elt_at_index (ngm->tunnels, tunnel_index0);
 
-          next0 = t0->decap_next_index;
+          next0 = t0->protocol;
+
           sw_if_index0 = t0->sw_if_index;
           len0 = vlib_buffer_length_in_chain(vm, b0);
 
           /* Required to make the l2 tag push / pop code work on l2 subifs */
           vnet_update_l2_len (b0);
-
-          next0 = t0->decap_next_index;
-
-          /* ip[46] lookup in the configured FIB, otherwise an opaque */
-          vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index;
+	  
+	  /*
+	   * ip[46] lookup in the configured FIB
+	   */
+	  vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index;
 
           pkts_decapsulated++;
           stats_n_packets += 1;
@@ -224,27 +214,30 @@ nsh_gre_input (vlib_main_t * vm,
         trace0:
           b0->error = error0 ? node->errors[error0] : 0;
 
-          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) 
+          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
             {
-              nsh_rx_trace_t *tr = vlib_add_trace (vm, node, 
-                                                   b0, sizeof (*tr));
+              vxlan_gpe_rx_trace_t *tr
+                = vlib_add_trace (vm, node, b0, sizeof (*tr));
               tr->next_index = next0;
               tr->error = error0;
               tr->tunnel_index = tunnel_index0;
-              tr->h = h0[0];
             }
 
-          if (PREDICT_FALSE(key1 != last_key))
+
+	  /* Processing for key1 */
+          if (PREDICT_FALSE ((key1.as_u64[0] != last_key.as_u64[0])
+                             || (key1.as_u64[1] != last_key.as_u64[1])))
             {
-              p1 = hash_get (ngm->nsh_gre_tunnel_by_src_address, key1);
+              p1 = hash_get_mem (ngm->vxlan_gpe_tunnel_by_key, &key1);
 
               if (p1 == 0)
                 {
-                  error1 = NSH_GRE_ERROR_NO_SUCH_TUNNEL;
+                  error1 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
                   goto trace1;
                 }
 
-              last_key = key1;
+              last_key.as_u64[0] = key1.as_u64[0];
+              last_key.as_u64[1] = key1.as_u64[1];
               tunnel_index1 = last_tunnel_index = p1[0];
             }
           else
@@ -252,22 +245,24 @@ nsh_gre_input (vlib_main_t * vm,
 
           t1 = pool_elt_at_index (ngm->tunnels, tunnel_index1);
 
-          next1 = t1->decap_next_index;
+          next1 = t1->protocol;
           sw_if_index1 = t1->sw_if_index;
           len1 = vlib_buffer_length_in_chain(vm, b1);
 
           /* Required to make the l2 tag push / pop code work on l2 subifs */
           vnet_update_l2_len (b1);
 
-          next1 = t1->decap_next_index;
+	  /*
+	   * ip[46] lookup in the configured FIB
+	   */
+	  vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->decap_fib_index;
 
-          /* ip[46] lookup in the configured FIB, otherwise an opaque */
-          vnet_buffer(b1)->sw_if_index[VLIB_TX] = t1->decap_fib_index;
 
           pkts_decapsulated++;
           stats_n_packets += 1;
           stats_n_bytes += len1;
-          /* Batch stats increment on the same nsh-gre tunnel so counter
+          
+	  /* Batch stats increment on the same vxlan tunnel so counter
            is not incremented per packet */
           if (PREDICT_FALSE(sw_if_index1 != stats_sw_if_index))
           {
@@ -286,14 +281,13 @@ nsh_gre_input (vlib_main_t * vm,
         trace1:
           b1->error = error1 ? node->errors[error1] : 0;
 
-          if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED)) 
+          if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
             {
-              nsh_rx_trace_t *tr = vlib_add_trace (vm, node, 
-                                                   b1, sizeof (*tr));
+              vxlan_gpe_rx_trace_t *tr
+                = vlib_add_trace (vm, node, b1, sizeof (*tr));
               tr->next_index = next1;
               tr->error = error1;
               tr->tunnel_index = tunnel_index1;
-              tr->h = h1[0];
             }
 
 	  vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
@@ -306,11 +300,11 @@ nsh_gre_input (vlib_main_t * vm,
 	  u32 bi0;
 	  vlib_buffer_t * b0;
 	  u32 next0;
-	  nsh_header_t * h0;
+          ip4_vxlan_gpe_header_t * iuvn0;
 	  uword * p0;
           u32 tunnel_index0;
-          nsh_gre_tunnel_t * t0;
-          u64 key0;
+          vxlan_gpe_tunnel_t * t0;
+          vxlan_gpe_tunnel_key_t key0;
           u32 error0;
           u32 sw_if_index0, len0;
 
@@ -322,29 +316,38 @@ nsh_gre_input (vlib_main_t * vm,
 	  n_left_to_next -= 1;
 
 	  b0 = vlib_get_buffer (vm, bi0);
-          h0 = vlib_buffer_get_current (b0);
 
-          /* gre stashed the src ip4 address for us... */
-          key0 = (((u64)(vnet_buffer(b0)->gre.src))<<32) | h0->spi_si;
+          /* udp leaves current_data pointing at the vxlan header */
+          vlib_buffer_advance 
+            (b0, -(word)(sizeof(udp_header_t)+sizeof(ip4_header_t)));
 
-          /* "pop" nsh header */
-          vlib_buffer_advance (b0, sizeof (*h0));
+          iuvn0 = vlib_buffer_get_current (b0);
+
+          /* pop (ip, udp, vxlan) */
+          vlib_buffer_advance (b0, sizeof (*iuvn0));
 
           tunnel_index0 = ~0;
           error0 = 0;
-          next0 = NSH_GRE_INPUT_NEXT_DROP;
+          next0 = (iuvn0->vxlan.protocol < VXLAN_GPE_INPUT_N_NEXT) ? iuvn0->vxlan.protocol : VXLAN_GPE_INPUT_NEXT_DROP;
 
-          if (PREDICT_FALSE(key0 != last_key))
-            {
-              p0 = hash_get (ngm->nsh_gre_tunnel_by_src_address, key0);
+          key0.local = iuvn0->ip4.dst_address.as_u32;
+	  key0.remote = iuvn0->ip4.src_address.as_u32;
+          key0.vni = iuvn0->vxlan.vni_res;
+          key0.pad = 0;
 
+          if (PREDICT_FALSE ((key0.as_u64[0] != last_key.as_u64[0])
+                             || (key0.as_u64[1] != last_key.as_u64[1])))
+            {
+              p0 = hash_get_mem (ngm->vxlan_gpe_tunnel_by_key, &key0);
+          
               if (p0 == 0)
                 {
-                  error0 = NSH_GRE_ERROR_NO_SUCH_TUNNEL;
+                  error0 = VXLAN_GPE_ERROR_NO_SUCH_TUNNEL;
                   goto trace00;
                 }
 
-              last_key = key0;
+              last_key.as_u64[0] = key0.as_u64[0];
+              last_key.as_u64[1] = key0.as_u64[1];
               tunnel_index0 = last_tunnel_index = p0[0];
             }
           else
@@ -352,23 +355,24 @@ nsh_gre_input (vlib_main_t * vm,
 
           t0 = pool_elt_at_index (ngm->tunnels, tunnel_index0);
 
-          next0 = t0->decap_next_index;
-          sw_if_index0 = t0->sw_if_index;
+          next0 = t0->protocol;
+	  
+	  sw_if_index0 = t0->sw_if_index;
           len0 = vlib_buffer_length_in_chain(vm, b0);
 
           /* Required to make the l2 tag push / pop code work on l2 subifs */
           vnet_update_l2_len (b0);
 
-          next0 = t0->decap_next_index;
+	  /*
+	   * ip[46] lookup in the configured FIB
+	   */
+	  vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index;
 
-          /* ip[46] lookup in the configured FIB, otherwise an opaque */
-          vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->decap_fib_index;
           pkts_decapsulated ++;
-
           stats_n_packets += 1;
           stats_n_bytes += len0;
 
-          /* Batch stats increment on the same nsh-gre tunnel so counter
+          /* Batch stats increment on the same vxlan-gpe tunnel so counter
            is not incremented per packet */
           if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index))
           {
@@ -386,14 +390,13 @@ nsh_gre_input (vlib_main_t * vm,
         trace00:
           b0->error = error0 ? node->errors[error0] : 0;
 
-          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) 
+          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
             {
-              nsh_rx_trace_t *tr = vlib_add_trace (vm, node, 
-                                                   b0, sizeof (*tr));
+              vxlan_gpe_rx_trace_t *tr
+                = vlib_add_trace (vm, node, b0, sizeof (*tr));
               tr->next_index = next0;
               tr->error = error0;
               tr->tunnel_index = tunnel_index0;
-              tr->h = h0[0];
             }
 	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
 					   to_next, n_left_to_next,
@@ -402,8 +405,8 @@ nsh_gre_input (vlib_main_t * vm,
 
       vlib_put_next_frame (vm, node, next_index, n_left_to_next);
     }
-  vlib_node_increment_counter (vm, nsh_gre_input_node.index,
-                               NSH_GRE_ERROR_DECAPSULATED, 
+  vlib_node_increment_counter (vm, vxlan_gpe_input_node.index,
+                               VXLAN_GPE_ERROR_DECAPSULATED,
                                pkts_decapsulated);
   /* Increment any remaining batch stats */
   if (stats_n_packets)
@@ -416,30 +419,32 @@ nsh_gre_input (vlib_main_t * vm,
   return from_frame->n_vectors;
 }
 
-static char * nsh_error_strings[] = {
-#define nsh_gre_error(n,s) s,
-#include <vnet/nsh/nsh_error.def>
-#undef nsh_gre_error
+static char * vxlan_gpe_error_strings[] = {
+#define vxlan_gpe_error(n,s) s,
+#include <vnet/vxlan-gpe/vxlan_gpe_error.def>
+#undef vxlan_gpe_error
 #undef _
 };
 
-VLIB_REGISTER_NODE (nsh_gre_input_node) = {
-  .function = nsh_gre_input,
-  .name = "nsh-gre-input",
+VLIB_REGISTER_NODE (vxlan_gpe_input_node) = {
+  .function = vxlan_gpe_input,
+  .name = "vxlan-gpe-input",
   /* Takes a vector of packets. */
   .vector_size = sizeof (u32),
+  .type = VLIB_NODE_TYPE_INTERNAL,
+  .n_errors = ARRAY_LEN(vxlan_gpe_error_strings),
+  .error_strings = vxlan_gpe_error_strings,
 
-  .n_errors = NSH_GRE_N_ERROR,
-  .error_strings = nsh_error_strings,
-
-  .n_next_nodes = NSH_GRE_INPUT_N_NEXT,
+  .n_next_nodes = VXLAN_GPE_INPUT_N_NEXT,
   .next_nodes = {
-#define _(s,n) [NSH_GRE_INPUT_NEXT_##s] = n,
-    foreach_nsh_gre_input_next
+#define _(s,n) [VXLAN_GPE_INPUT_NEXT_##s] = n,
+    foreach_vxlan_gpe_input_next
 #undef _
   },
 
-  .format_buffer = format_nsh_header_with_length,
-  .format_trace = format_nsh_rx_trace,
-  // $$$$ .unformat_buffer = unformat_nsh_gre_header,
+  .format_buffer = format_vxlan_gpe_with_length,
+  .format_trace = format_vxlan_gpe_rx_trace,
+  // $$$$ .unformat_buffer = unformat_vxlan_gpe_header,
 };
+
+
diff --git a/vnet/vnet/nsh-vxlan-gpe/encap.c b/vnet/vnet/vxlan-gpe/encap.c
index af520b2f8f0..3ffe2a62db2 100644
--- a/vnet/vnet/nsh-vxlan-gpe/encap.c
+++ b/vnet/vnet/vxlan-gpe/encap.c
@@ -17,43 +17,44 @@
 #include <vnet/vnet.h>
 #include <vnet/ip/ip.h>
 #include <vnet/ethernet/ethernet.h>
-#include <vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.h>
+#include <vnet/vxlan-gpe/vxlan_gpe.h>
 
 /* Statistics (not really errors) */
-#define foreach_nsh_vxlan_gpe_encap_error    \
+#define foreach_vxlan_gpe_encap_error    \
 _(ENCAPSULATED, "good packets encapsulated")
 
-static char * nsh_vxlan_gpe_encap_error_strings[] = {
+static char * vxlan_gpe_encap_error_strings[] = {
 #define _(sym,string) string,
-  foreach_nsh_vxlan_gpe_encap_error
+  foreach_vxlan_gpe_encap_error
 #undef _
 };
 
 typedef enum {
-#define _(sym,str) NSH_VXLAN_GPE_ENCAP_ERROR_##sym,
-    foreach_nsh_vxlan_gpe_encap_error
+#define _(sym,str) VXLAN_GPE_ENCAP_ERROR_##sym,
+    foreach_vxlan_gpe_encap_error
 #undef _
-    NSH_VXLAN_GPE_ENCAP_N_ERROR,
-} nsh_vxlan_gpe_encap_error_t;
+    VXLAN_GPE_ENCAP_N_ERROR,
+} vxlan_gpe_encap_error_t;
 
 typedef enum {
-    NSH_VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP,
-    NSH_VXLAN_GPE_ENCAP_NEXT_DROP,
-    NSH_VXLAN_GPE_ENCAP_N_NEXT,
-} nsh_vxlan_gpe_encap_next_t;
+  VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP,
+  VXLAN_GPE_ENCAP_NEXT_DROP,
+  VXLAN_GPE_ENCAP_N_NEXT
+} vxlan_gpe_encap_next_t;
 
 typedef struct {
   u32 tunnel_index;
-} nsh_vxlan_gpe_encap_trace_t;
+} vxlan_gpe_encap_trace_t;
 
-u8 * format_nsh_vxlan_gpe_encap_trace (u8 * s, va_list * args)
+
+u8 * format_vxlan_gpe_encap_trace (u8 * s, va_list * args)
 {
   CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
   CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
-  nsh_vxlan_gpe_encap_trace_t * t 
-      = va_arg (*args, nsh_vxlan_gpe_encap_trace_t *);
+  vxlan_gpe_encap_trace_t * t 
+      = va_arg (*args, vxlan_gpe_encap_trace_t *);
 
-  s = format (s, "NSH-VXLAN-ENCAP: tunnel %d", t->tunnel_index);
+  s = format (s, "VXLAN-GPE-ENCAP: tunnel %d", t->tunnel_index);
   return s;
 }
 
@@ -61,12 +62,12 @@ u8 * format_nsh_vxlan_gpe_encap_trace (u8 * s, va_list * args)
 _(0) _(1) _(2) _(3) _(4) _(5) _(6)
 
 static uword
-nsh_vxlan_gpe_encap (vlib_main_t * vm,
+vxlan_gpe_encap (vlib_main_t * vm,
                vlib_node_runtime_t * node,
                vlib_frame_t * from_frame)
 {
   u32 n_left_from, next_index, * from, * to_next;
-  nsh_vxlan_gpe_main_t * ngm = &nsh_vxlan_gpe_main;
+  vxlan_gpe_main_t * ngm = &vxlan_gpe_main;
   vnet_main_t * vnm = ngm->vnet_main;
   vnet_interface_main_t * im = &vnm->interface_main;
   u32 pkts_encapsulated = 0;
@@ -92,8 +93,8 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
 	{
           u32 bi0, bi1;
 	  vlib_buffer_t * b0, * b1;
-	  u32 next0 = NSH_VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP;
-          u32 next1 = NSH_VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP;
+	  u32 next0 = VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP;
+          u32 next1 = VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP;
           u32 sw_if_index0, sw_if_index1, len0, len1;
           vnet_hw_interface_t * hi0, * hi1;
           ip4_header_t * ip0, * ip1;
@@ -102,7 +103,7 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
           u64 * copy_src1, * copy_dst1;
           u32 * copy_src_last0, * copy_dst_last0;
           u32 * copy_src_last1, * copy_dst_last1;
-          nsh_vxlan_gpe_tunnel_t * t0, * t1;
+          vxlan_gpe_tunnel_t * t0, * t1;
           u16 new_l0, new_l1;
           ip_csum_t sum0, sum1;
 
@@ -158,9 +159,9 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
           copy_dst1 = (u64 *) ip1;
           copy_src1 = (u64 *) t1->rewrite;
 
-          ASSERT (sizeof (ip4_vxlan_gpe_and_nsh_header_t) == 60);
+          ASSERT (sizeof (ip4_vxlan_gpe_header_t) == 36);
 
-          /* Copy first 56 octets 8-bytes at a time */
+          /* Copy first 36 octets 8-bytes at a time */
 #define _(offs) copy_dst0[offs] = copy_src0[offs];
           foreach_fixed_header_offset;
 #undef _
@@ -259,14 +260,14 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
 
           if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
             {
-              nsh_vxlan_gpe_encap_trace_t *tr =
+              vxlan_gpe_encap_trace_t *tr =
                 vlib_add_trace (vm, node, b0, sizeof (*tr));
               tr->tunnel_index = t0 - ngm->tunnels;
             }
 
           if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
             {
-              nsh_vxlan_gpe_encap_trace_t *tr =
+              vxlan_gpe_encap_trace_t *tr =
                 vlib_add_trace (vm, node, b1, sizeof (*tr));
               tr->tunnel_index = t1 - ngm->tunnels;
             }
@@ -280,14 +281,14 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
 	{
 	  u32 bi0;
 	  vlib_buffer_t * b0;
-	  u32 next0 = NSH_VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP;
-      u32 sw_if_index0, len0;
+	  u32 next0 = VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP;
+	  u32 sw_if_index0, len0;
           vnet_hw_interface_t * hi0;
           ip4_header_t * ip0;
           udp_header_t * udp0;
           u64 * copy_src0, * copy_dst0;
           u32 * copy_src_last0, * copy_dst_last0;
-          nsh_vxlan_gpe_tunnel_t * t0;
+          vxlan_gpe_tunnel_t * t0;
           u16 new_l0;
           ip_csum_t sum0;
 
@@ -301,7 +302,7 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
 	  b0 = vlib_get_buffer (vm, bi0);
 
           /* 1-wide cache? */
-          sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX];
+	  sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX];
           hi0 = vnet_get_sup_hw_interface
             (vnm, vnet_buffer(b0)->sw_if_index[VLIB_TX]);
 
@@ -317,9 +318,9 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
           copy_dst0 = (u64 *) ip0;
           copy_src0 = (u64 *) t0->rewrite;
 
-          ASSERT (sizeof (ip4_vxlan_gpe_and_nsh_header_t) == 60);
+          ASSERT (sizeof (ip4_vxlan_gpe_header_t) == 36);
 
-          /* Copy first 56 octets 8-bytes at a time */
+          /* Copy first 36 octets 8-bytes at a time */
 #define _(offs) copy_dst0[offs] = copy_src0[offs];
           foreach_fixed_header_offset;
 #undef _
@@ -354,32 +355,33 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
 
           /* Reset to look up tunnel partner in the configured FIB */
           vnet_buffer(b0)->sw_if_index[VLIB_TX] = t0->encap_fib_index;
-          vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0;
+	  vnet_buffer(b0)->sw_if_index[VLIB_RX] = sw_if_index0;
           pkts_encapsulated ++;
 
-          len0 = vlib_buffer_length_in_chain(vm, b0);
-          stats_n_packets += 1;
-          stats_n_bytes += len0;
-
-          /* Batch stats increment on the same vxlan tunnel so counter is not
-           incremented per packet. Note stats are still incremented for deleted
-           and admin-down tunnel where packets are dropped. It is not worthwhile
-           to check for this rare case and affect normal path performance. */
-          if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index)) {
-            stats_n_packets -= 1;
-            stats_n_bytes -= len0;
-            if (stats_n_packets)
-              vlib_increment_combined_counter(
-                  im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
-                  cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
-            stats_n_packets = 1;
-            stats_n_bytes = len0;
-            stats_sw_if_index = sw_if_index0;
-          }
-          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
-            {
-              nsh_vxlan_gpe_encap_trace_t *tr =
-                vlib_add_trace (vm, node, b0, sizeof (*tr));
+	  len0 = vlib_buffer_length_in_chain(vm, b0); 
+	  stats_n_packets += 1; 
+	  stats_n_bytes += len0; 
+	  
+          /* Batch stats increment on the same vxlan tunnel so counter is not 
+	   *  incremented per packet. Note stats are still incremented for deleted 
+	   *  and admin-down tunnel where packets are dropped. It is not worthwhile 
+	   *  to check for this rare case and affect normal path performance. */ 
+	  if (PREDICT_FALSE(sw_if_index0 != stats_sw_if_index)) 
+	    { 
+	      stats_n_packets -= 1; 
+	      stats_n_bytes -= len0; 
+	      if (stats_n_packets) 
+		vlib_increment_combined_counter( 
+						im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, 
+						cpu_index, stats_sw_if_index, stats_n_packets, stats_n_bytes); 
+	      stats_n_packets = 1;
+	      stats_n_bytes = len0; 
+	      stats_sw_if_index = sw_if_index0; 
+	    } 
+	  if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+	      vxlan_gpe_encap_trace_t *tr =
+		vlib_add_trace (vm, node, b0, sizeof (*tr));
               tr->tunnel_index = t0 - ngm->tunnels;
             }
 	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
@@ -390,33 +392,34 @@ nsh_vxlan_gpe_encap (vlib_main_t * vm,
       vlib_put_next_frame (vm, node, next_index, n_left_to_next);
     }
   vlib_node_increment_counter (vm, node->node_index,
-                               NSH_VXLAN_GPE_ENCAP_ERROR_ENCAPSULATED,
+                               VXLAN_GPE_ENCAP_ERROR_ENCAPSULATED,
                                pkts_encapsulated);
   /* Increment any remaining batch stats */
-  if (stats_n_packets) {
-    vlib_increment_combined_counter(
-        im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, cpu_index,
-        stats_sw_if_index, stats_n_packets, stats_n_bytes);
-    node->runtime_data[0] = stats_sw_if_index;
-  }
+  if (stats_n_packets) { 
+    vlib_increment_combined_counter( 
+				    im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX, cpu_index, 
+				    stats_sw_if_index, stats_n_packets, stats_n_bytes); 
+    node->runtime_data[0] = stats_sw_if_index; 
+  } 
 
   return from_frame->n_vectors;
 }
 
-VLIB_REGISTER_NODE (nsh_vxlan_gpe_encap_node) = {
-  .function = nsh_vxlan_gpe_encap,
-  .name = "nsh-vxlan-gpe-encap",
+VLIB_REGISTER_NODE (vxlan_gpe_encap_node) = {
+  .function = vxlan_gpe_encap,
+  .name = "vxlan-gpe-encap",
   .vector_size = sizeof (u32),
-  .format_trace = format_nsh_vxlan_gpe_encap_trace,
+  .format_trace = format_vxlan_gpe_encap_trace,
   .type = VLIB_NODE_TYPE_INTERNAL,
 
-  .n_errors = ARRAY_LEN(nsh_vxlan_gpe_encap_error_strings),
-  .error_strings = nsh_vxlan_gpe_encap_error_strings,
+  .n_errors = ARRAY_LEN(vxlan_gpe_encap_error_strings),
+  .error_strings = vxlan_gpe_encap_error_strings,
 
-  .n_next_nodes = NSH_VXLAN_GPE_ENCAP_N_NEXT,
+  .n_next_nodes = VXLAN_GPE_ENCAP_N_NEXT,
 
   .next_nodes = {
-        [NSH_VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP] = "ip4-lookup",
-        [NSH_VXLAN_GPE_ENCAP_NEXT_DROP] = "error-drop",
+        [VXLAN_GPE_ENCAP_NEXT_IP4_LOOKUP] = "ip4-lookup",
+        [VXLAN_GPE_ENCAP_NEXT_DROP] = "error-drop",
   },
 };
+
diff --git a/vnet/vnet/nsh-vxlan-gpe/vxlan-gpe-rfc.txt b/vnet/vnet/vxlan-gpe/vxlan-gpe-rfc.txt
index 35cee50f573..35cee50f573 100644
--- a/vnet/vnet/nsh-vxlan-gpe/vxlan-gpe-rfc.txt
+++ b/vnet/vnet/vxlan-gpe/vxlan-gpe-rfc.txt
diff --git a/vnet/vnet/vxlan-gpe/vxlan_gpe.c b/vnet/vnet/vxlan-gpe/vxlan_gpe.c
new file mode 100644
index 00000000000..ef242d0bb8f
--- /dev/null
+++ b/vnet/vnet/vxlan-gpe/vxlan_gpe.c
@@ -0,0 +1,467 @@
+/*
+ * Copyright (c) 2015 Cisco and/or its affiliates.
+ * 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.
+ */
+#include <vnet/vxlan-gpe/vxlan_gpe.h>
+
+vxlan_gpe_main_t vxlan_gpe_main;
+
+u8 * format_vxlan_gpe_tunnel (u8 * s, va_list * args)
+{
+  vxlan_gpe_tunnel_t * t = va_arg (*args, vxlan_gpe_tunnel_t *);
+  vxlan_gpe_main_t * gm = &vxlan_gpe_main;
+
+  s = format (s, "[%d] local: %U remote: %U ",
+              t - gm->tunnels,
+              format_ip4_address, &t->local,
+              format_ip4_address, &t->remote);    
+
+  switch (t->protocol)
+    {
+    case VXLAN_GPE_PROTOCOL_IP4:
+      s = format (s, "next-protocol ip4");
+    case VXLAN_GPE_PROTOCOL_IP6:
+      s = format (s, "next-protocol ip6");
+    case VXLAN_GPE_PROTOCOL_ETHERNET:
+      s = format (s, "next-protocol ethernet");
+    case VXLAN_GPE_PROTOCOL_NSH:
+      s = format (s, "next-protocol nsh");
+    default:
+      s = format (s, "next-protocol unknown %d", t->protocol);
+    }
+
+  s = format (s, " fibs: (encap %d, decap %d)",
+              t->encap_fib_index,
+              t->decap_fib_index);
+
+  s = format (s, "  vxlan VNI %d ", t->vni);
+
+  return s;
+}
+
+static u8 * format_vxlan_gpe_name (u8 * s, va_list * args)
+{
+  vxlan_gpe_main_t * gm = &vxlan_gpe_main;
+  u32 i = va_arg (*args, u32);
+  u32 show_dev_instance = ~0;
+
+  if (i < vec_len (gm->dev_inst_by_real))
+    show_dev_instance = gm->dev_inst_by_real[i];
+
+  if (show_dev_instance != ~0)
+    i = show_dev_instance;
+
+  return format (s, "vxlan_gpe_tunnel%d", i);
+}
+
+static int vxlan_gpe_name_renumber (vnet_hw_interface_t * hi,
+                                        u32 new_dev_instance)
+{
+  vxlan_gpe_main_t * gm = &vxlan_gpe_main;
+
+  vec_validate_init_empty (gm->dev_inst_by_real, hi->dev_instance, ~0);
+
+  gm->dev_inst_by_real [hi->dev_instance] = new_dev_instance;
+
+  return 0;
+}
+
+static uword dummy_interface_tx (vlib_main_t * vm,
+                                 vlib_node_runtime_t * node,
+                                 vlib_frame_t * frame)
+{
+  clib_warning ("you shouldn't be here, leaking buffers...");
+  return frame->n_vectors;
+}
+
+static uword dummy_set_rewrite (vnet_main_t * vnm,
+                                u32 sw_if_index,
+                                u32 l3_type,
+                                void * dst_address,
+                                void * rewrite,
+                                uword max_rewrite_bytes)
+{
+  return 0;
+}
+
+VNET_DEVICE_CLASS (vxlan_gpe_device_class,static) = {
+  .name = "VXLAN_GPE",
+  .format_device_name = format_vxlan_gpe_name,
+  .format_tx_trace = format_vxlan_gpe_encap_trace,
+  .tx_function = dummy_interface_tx,
+  .name_renumber = vxlan_gpe_name_renumber,
+};
+
+static u8 * format_vxlan_gpe_header_with_length (u8 * s, va_list * args)
+{
+  u32 dev_instance = va_arg (*args, u32);
+  s = format (s, "unimplemented dev %u", dev_instance);
+  return s;
+}
+
+VNET_HW_INTERFACE_CLASS (vxlan_gpe_hw_class) = {
+  .name = "VXLAN_GPE",
+  .format_header = format_vxlan_gpe_header_with_length,
+  .set_rewrite = dummy_set_rewrite,
+};
+
+
+#define foreach_gpe_copy_field                  \
+_(local.as_u32)                                 \
+_(remote.as_u32)				\
+_(vni)                                          \
+_(protocol)                                \
+_(encap_fib_index)                              \
+_(decap_fib_index)                              
+
+#define foreach_copy_field			\
+_(src.as_u32)                                   \
+_(dst.as_u32)                                   \
+_(vni)                                          \
+_(encap_fib_index)                              \
+_(decap_fib_index)                              \
+_(decap_next_index)
+
+
+
+static int vxlan_gpe_rewrite (vxlan_gpe_tunnel_t * t)
+{
+  u8 *rw = 0;
+  ip4_header_t * ip0;
+  ip4_vxlan_gpe_header_t * h0;
+  int len;
+
+  len = sizeof (*h0);
+
+  vec_validate_aligned (rw, len-1, CLIB_CACHE_LINE_BYTES);
+
+  h0 = (ip4_vxlan_gpe_header_t *) rw;
+
+  /* Fixed portion of the (outer) ip4 header */
+  ip0 = &h0->ip4;
+  ip0->ip_version_and_header_length = 0x45;
+  ip0->ttl = 254;
+  ip0->protocol = IP_PROTOCOL_UDP;
+
+  /* we fix up the ip4 header length and checksum after-the-fact */
+  ip0->src_address.as_u32 = t->local.as_u32; 
+  ip0->dst_address.as_u32 = t->remote.as_u32; 
+  ip0->checksum = ip4_header_checksum (ip0);
+
+  /* UDP header, randomize src port on something, maybe? */
+  h0->udp.src_port = clib_host_to_net_u16 (4790);
+  h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_vxlan_gpe);
+
+  /* VXLAN header. Are we having fun yet? */
+  h0->vxlan.flags = VXLAN_GPE_FLAGS_I | VXLAN_GPE_FLAGS_P;
+  h0->vxlan.ver_res = VXLAN_GPE_VERSION;
+  h0->vxlan.protocol = VXLAN_GPE_PROTOCOL_IP4;
+  h0->vxlan.vni_res = clib_host_to_net_u32 (t->vni<<8);
+
+  t->rewrite = rw;
+  return (0);
+}
+
+int vnet_vxlan_gpe_add_del_tunnel 
+(vnet_vxlan_gpe_add_del_tunnel_args_t *a, u32 * sw_if_indexp)
+{
+  vxlan_gpe_main_t * gm = &vxlan_gpe_main;
+  vxlan_gpe_tunnel_t *t = 0;
+  vnet_main_t * vnm = gm->vnet_main;
+  vnet_hw_interface_t * hi;
+  uword * p;
+  u32 hw_if_index = ~0;
+  u32 sw_if_index = ~0;
+  int rv;
+  vxlan_gpe_tunnel_key_t key, *key_copy;
+  hash_pair_t *hp;
+  
+  key.local = a->local.as_u32; 
+  key.remote = a->remote.as_u32; 
+  key.vni = clib_host_to_net_u32 (a->vni << 8);
+  key.pad = 0;
+
+  p = hash_get_mem (gm->vxlan_gpe_tunnel_by_key, &key);
+  
+  if (a->is_add)
+    {
+      /* adding a tunnel: tunnel must not already exist */
+      if (p) 
+        return VNET_API_ERROR_INVALID_VALUE;
+      
+      if (a->decap_next_index >= VXLAN_GPE_INPUT_N_NEXT)
+        return VNET_API_ERROR_INVALID_DECAP_NEXT;
+      
+      pool_get_aligned (gm->tunnels, t, CLIB_CACHE_LINE_BYTES);
+      memset (t, 0, sizeof (*t));
+      
+      /* copy from arg structure */
+#define _(x) t->x = a->x;
+      foreach_gpe_copy_field;
+#undef _
+
+      rv = vxlan_gpe_rewrite (t);
+
+      if (rv)
+        {
+          pool_put (gm->tunnels, t);
+          return rv;
+        }
+
+      key_copy = clib_mem_alloc (sizeof (*key_copy));
+      clib_memcpy (key_copy, &key, sizeof (*key_copy));
+
+      hash_set_mem (gm->vxlan_gpe_tunnel_by_key, key_copy, 
+                    t - gm->tunnels);
+      
+      if (vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices) > 0)
+        {
+          hw_if_index = gm->free_vxlan_gpe_tunnel_hw_if_indices
+            [vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices)-1];
+          _vec_len (gm->free_vxlan_gpe_tunnel_hw_if_indices) -= 1;
+          
+          hi = vnet_get_hw_interface (vnm, hw_if_index);
+          hi->dev_instance = t - gm->tunnels;
+          hi->hw_instance = hi->dev_instance;
+        }
+      else 
+        {
+          hw_if_index = vnet_register_interface
+            (vnm, vxlan_gpe_device_class.index, t - gm->tunnels,
+             vxlan_gpe_hw_class.index, t - gm->tunnels);
+          hi = vnet_get_hw_interface (vnm, hw_if_index);
+          hi->output_node_index = vxlan_gpe_encap_node.index;
+        }
+      
+      t->hw_if_index = hw_if_index;
+      t->sw_if_index = sw_if_index = hi->sw_if_index;
+      
+      vnet_sw_interface_set_flags (vnm, hi->sw_if_index, 
+                                   VNET_SW_INTERFACE_FLAG_ADMIN_UP);
+    }
+  else
+    {
+      /* deleting a tunnel: tunnel must exist */
+      if (!p) 
+        return VNET_API_ERROR_NO_SUCH_ENTRY;
+
+      t = pool_elt_at_index (gm->tunnels, p[0]);
+
+      vnet_sw_interface_set_flags (vnm, t->sw_if_index, 0 /* down */);
+      vec_add1 (gm->free_vxlan_gpe_tunnel_hw_if_indices, t->hw_if_index);
+
+      hp = hash_get_pair (gm->vxlan_gpe_tunnel_by_key, &key);
+      key_copy = (void *)(hp->key);
+      hash_unset_mem (gm->vxlan_gpe_tunnel_by_key, &key);
+      clib_mem_free (key_copy);
+
+      vec_free (t->rewrite);
+      pool_put (gm->tunnels, t);
+    }
+
+  if (sw_if_indexp)
+      *sw_if_indexp = sw_if_index;
+
+  return 0;
+}
+
+static u32 fib_index_from_fib_id (u32 fib_id)
+{
+  ip4_main_t * im = &ip4_main;
+  uword * p;
+
+  p = hash_get (im->fib_index_by_table_id, fib_id);
+  if (!p)
+    return ~0;
+
+  return p[0];
+}
+
+static uword unformat_gpe_decap_next (unformat_input_t * input, va_list * args)
+{
+  u32 * result = va_arg (*args, u32 *);
+  u32 tmp;
+  
+  if (unformat (input, "drop"))
+    *result = VXLAN_GPE_INPUT_NEXT_DROP;
+  else if (unformat (input, "ip4"))
+    *result = VXLAN_GPE_INPUT_NEXT_IP4_INPUT;
+  else if (unformat (input, "ip6"))
+    *result = VXLAN_GPE_INPUT_NEXT_IP6_INPUT;
+  else if (unformat (input, "ethernet"))
+    *result = VXLAN_GPE_INPUT_NEXT_ETHERNET_INPUT;
+  else if (unformat (input, "nsh"))
+    *result = VXLAN_GPE_INPUT_NEXT_NSH_INPUT;
+  else if (unformat (input, "%d", &tmp))
+    *result = tmp;
+  else
+    return 0;
+  return 1;
+}
+
+static clib_error_t *
+vxlan_gpe_add_del_tunnel_command_fn (vlib_main_t * vm,
+                                   unformat_input_t * input,
+                                   vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, * line_input = &_line_input;
+  u8 is_add = 1;
+  ip4_address_t local, remote;
+  u8 local_set = 0;
+  u8 remote_set = 0;
+  u32 encap_fib_index = 0;
+  u32 decap_fib_index = 0;
+  u8 protocol = VXLAN_GPE_PROTOCOL_IP4;
+  u32 decap_next_index = VXLAN_GPE_INPUT_NEXT_IP4_INPUT; 
+  u32 vni;
+  u8 vni_set = 0;
+  int rv;
+  u32 tmp;
+  vnet_vxlan_gpe_add_del_tunnel_args_t _a, * a = &_a;
+  
+  /* Get a line of input. */
+  if (! unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) {
+    if (unformat (line_input, "del"))
+      is_add = 0;
+    else if (unformat (line_input, "local %U", 
+                       unformat_ip4_address, &local))
+      local_set = 1;
+    else if (unformat (line_input, "remote %U",
+                       unformat_ip4_address, &remote))
+      remote_set = 1;
+    else if (unformat (line_input, "encap-vrf-id %d", &tmp))
+      {
+        encap_fib_index = fib_index_from_fib_id (tmp);
+        if (encap_fib_index == ~0)
+          return clib_error_return (0, "nonexistent encap fib id %d", tmp);
+      }
+    else if (unformat (line_input, "decap-vrf-id %d", &tmp))
+      {
+        decap_fib_index = fib_index_from_fib_id (tmp);
+        if (decap_fib_index == ~0)
+          return clib_error_return (0, "nonexistent decap fib id %d", tmp);
+      }
+    else if (unformat (line_input, "decap-next %U", unformat_gpe_decap_next, 
+                       &decap_next_index))
+      ;
+    else if (unformat (line_input, "vni %d", &vni))
+      vni_set = 1;
+    else if (unformat(line_input, "next-ip4"))
+      protocol = VXLAN_GPE_PROTOCOL_IP4;
+    else if (unformat(line_input, "next-ip6"))
+      protocol = VXLAN_GPE_PROTOCOL_IP6;
+    else if (unformat(line_input, "next-ethernet"))
+      protocol = VXLAN_GPE_PROTOCOL_ETHERNET;
+    else if (unformat(line_input, "next-nsh"))
+      protocol = VXLAN_GPE_PROTOCOL_NSH;
+    else 
+      return clib_error_return (0, "parse error: '%U'", 
+                                format_unformat_error, line_input);
+  }
+
+  unformat_free (line_input);
+
+  if (local_set == 0)
+    return clib_error_return (0, "tunnel local address not specified");
+
+  if (remote_set == 0)
+    return clib_error_return (0, "tunnel remote address not specified");
+
+  if (vni_set == 0)
+    return clib_error_return (0, "vni not specified");
+
+  memset (a, 0, sizeof (*a));
+
+  a->is_add = is_add;
+
+#define _(x) a->x = x;
+  foreach_gpe_copy_field;
+#undef _
+
+  rv = vnet_vxlan_gpe_add_del_tunnel (a, 0 /* hw_if_indexp */);
+
+  switch(rv)
+    {
+    case 0:
+      break;
+    case VNET_API_ERROR_INVALID_DECAP_NEXT:
+      return clib_error_return (0, "invalid decap-next...");
+
+    case VNET_API_ERROR_TUNNEL_EXIST:
+      return clib_error_return (0, "tunnel already exists...");
+
+    case VNET_API_ERROR_NO_SUCH_ENTRY:
+      return clib_error_return (0, "tunnel does not exist...");
+
+    default:
+      return clib_error_return 
+        (0, "vnet_vxlan_gpe_add_del_tunnel returned %d", rv);
+    }
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (create_vxlan_gpe_tunnel_command, static) = {
+  .path = "create vxlan-gpe tunnel",
+  .short_help = 
+  "create vxlan-gpe tunnel local <ip4-addr> remote <ip4-addr>"
+  " vni <nn> [next-ip4][next-ip6][next-ethernet][next-nsh]"
+  " [encap-vrf-id <nn>] [decap-vrf-id <nn>]"
+  " [del]\n",
+  .function = vxlan_gpe_add_del_tunnel_command_fn,
+};
+
+static clib_error_t *
+show_vxlan_gpe_tunnel_command_fn (vlib_main_t * vm,
+                                unformat_input_t * input,
+                                vlib_cli_command_t * cmd)
+{
+  vxlan_gpe_main_t * gm = &vxlan_gpe_main;
+  vxlan_gpe_tunnel_t * t;
+  
+  if (pool_elts (gm->tunnels) == 0)
+    vlib_cli_output (vm, "No vxlan-gpe tunnels configured.");
+
+  pool_foreach (t, gm->tunnels,
+  ({
+    vlib_cli_output (vm, "%U", format_vxlan_gpe_tunnel, t);
+  }));
+  
+  return 0;
+}
+
+VLIB_CLI_COMMAND (show_vxlan_gpe_tunnel_command, static) = {
+    .path = "show vxlan-gpe",
+    .function = show_vxlan_gpe_tunnel_command_fn,
+};
+
+clib_error_t *vxlan_gpe_init (vlib_main_t *vm)
+{
+  vxlan_gpe_main_t *gm = &vxlan_gpe_main;
+  
+  gm->vnet_main = vnet_get_main();
+  gm->vlib_main = vm;
+  
+  gm->vxlan_gpe_tunnel_by_key 
+    = hash_create_mem (0, sizeof(vxlan_gpe_tunnel_key_t), sizeof (uword));
+
+  udp_register_dst_port (vm, UDP_DST_PORT_vxlan_gpe, 
+                         vxlan_gpe_input_node.index, 1 /* is_ip4 */);
+  return 0;
+}
+
+VLIB_INIT_FUNCTION(vxlan_gpe_init);
+
diff --git a/vnet/vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.h b/vnet/vnet/vxlan-gpe/vxlan_gpe.h
index 3effd3318cb..4c2ac444e34 100644
--- a/vnet/vnet/nsh-vxlan-gpe/nsh_vxlan_gpe.h
+++ b/vnet/vnet/vxlan-gpe/vxlan_gpe.h
@@ -12,8 +12,8 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-#ifndef included_vnet_nsh_vxlan_gpe_h
-#define included_vnet_nsh_vxlan_gpe_h
+#ifndef included_vnet_vxlan_gpe_h
+#define included_vnet_vxlan_gpe_h
 
 #include <vppinfra/error.h>
 #include <vppinfra/hash.h>
@@ -21,45 +21,43 @@
 #include <vnet/ip/ip.h>
 #include <vnet/l2/l2_input.h>
 #include <vnet/ethernet/ethernet.h>
-#include <vnet/gre/gre.h>
-#include <vnet/nsh/nsh_packet.h>
-#include <vnet/nsh-vxlan-gpe/vxlan_gpe_packet.h>
+#include <vnet/vxlan-gpe/vxlan_gpe_packet.h>
 #include <vnet/ip/ip4_packet.h>
 #include <vnet/ip/udp.h>
 
+
 typedef CLIB_PACKED (struct {
   ip4_header_t ip4;             /* 20 bytes */
   udp_header_t udp;             /* 8 bytes */
   vxlan_gpe_header_t vxlan;     /* 8 bytes */
-  nsh_header_t nsh;   		/* 28 bytes */
-}) ip4_vxlan_gpe_and_nsh_header_t;
+}) ip4_vxlan_gpe_header_t;
 
 typedef CLIB_PACKED(struct {
   /* 
-   * Key fields: ip src, vxlan vni, nsh spi_si 
+   * Key fields: local remote, vni 
    * all fields in NET byte order
    */
   union {
     struct {
-      u32 src;
+      u32 local;
+      u32 remote;
       u32 vni;                      /* shifted 8 bits */
-      u32 spi_si;
       u32 pad;
     };
     u64 as_u64[2];
   };
-}) nsh_vxlan_gpe_tunnel_key_t;
+}) vxlan_gpe_tunnel_key_t;
 
 typedef struct {
   /* Rewrite string. $$$$ embed vnet_rewrite header */
   u8 * rewrite;
 
-  /* decap next index */
-  u32 decap_next_index;
+  /* encapsulated protocol */
+  u8 protocol;
 
   /* tunnel src and dst addresses */
-  ip4_address_t src;
-  ip4_address_t dst;
+  ip4_address_t local;
+  ip4_address_t remote;
 
   /* FIB indices */
   u32 encap_fib_index;          /* tunnel partner lookup here */
@@ -72,40 +70,38 @@ typedef struct {
   u32 hw_if_index;
   u32 sw_if_index;
 
-  /* NSH header fields in HOST byte order */
-  nsh_header_t nsh_hdr;
-} nsh_vxlan_gpe_tunnel_t;
+} vxlan_gpe_tunnel_t;
 
-#define foreach_nsh_vxlan_gpe_input_next        \
+#define foreach_vxlan_gpe_input_next        \
 _(DROP, "error-drop")                           \
 _(IP4_INPUT, "ip4-input")                       \
 _(IP6_INPUT, "ip6-input")                       \
 _(ETHERNET_INPUT, "ethernet-input")             \
-_(NSH_VXLAN_GPE_ENCAP, "nsh-vxlan-gpe-encap")
+_(NSH_INPUT, "nsh-input")
 
 typedef enum {
-#define _(s,n) NSH_VXLAN_GPE_INPUT_NEXT_##s,
-  foreach_nsh_vxlan_gpe_input_next
+#define _(s,n) VXLAN_GPE_INPUT_NEXT_##s,
+  foreach_vxlan_gpe_input_next
 #undef _
-  NSH_VXLAN_GPE_INPUT_N_NEXT,
-} nsh_vxlan_gpe_input_next_t;
+  VXLAN_GPE_INPUT_N_NEXT,
+} vxlan_gpe_input_next_t;
 
 typedef enum {
-#define nsh_vxlan_gpe_error(n,s) NSH_VXLAN_GPE_ERROR_##n,
-#include <vnet/nsh-vxlan-gpe/nsh_vxlan_gpe_error.def>
-#undef nsh_vxlan_gpe_error
-  NSH_VXLAN_GPE_N_ERROR,
-} nsh_vxlan_gpe_input_error_t;
+#define vxlan_gpe_error(n,s) VXLAN_GPE_ERROR_##n,
+#include <vnet/vxlan-gpe/vxlan_gpe_error.def>
+#undef vxlan_gpe_error
+  VXLAN_GPE_N_ERROR,
+} vxlan_gpe_input_error_t;
 
 typedef struct {
   /* vector of encap tunnel instances */
-  nsh_vxlan_gpe_tunnel_t *tunnels;
+  vxlan_gpe_tunnel_t *tunnels;
 
   /* lookup tunnel by key */
-  uword * nsh_vxlan_gpe_tunnel_by_key;
+  uword * vxlan_gpe_tunnel_by_key;
 
   /* Free vlib hw_if_indices */
-  u32 * free_nsh_vxlan_gpe_tunnel_hw_if_indices;
+  u32 * free_vxlan_gpe_tunnel_hw_if_indices;
 
   /* show device instance by real device instance */
   u32 * dev_inst_by_real;
@@ -113,26 +109,31 @@ typedef struct {
   /* convenience */
   vlib_main_t * vlib_main;
   vnet_main_t * vnet_main;
-} nsh_vxlan_gpe_main_t;
+} vxlan_gpe_main_t;
 
-nsh_vxlan_gpe_main_t nsh_vxlan_gpe_main;
+vxlan_gpe_main_t vxlan_gpe_main;
 
-extern vlib_node_registration_t nsh_vxlan_gpe_input_node;
-extern vlib_node_registration_t nsh_vxlan_gpe_encap_node;
+extern vlib_node_registration_t vxlan_gpe_encap_node;
+extern vlib_node_registration_t vxlan_gpe_input_node;
 
-u8 * format_nsh_vxlan_gpe_encap_trace (u8 * s, va_list * args);
+u8 * format_vxlan_gpe_encap_trace (u8 * s, va_list * args);
 
 typedef struct {
   u8 is_add;
-  ip4_address_t src, dst;
+  ip4_address_t local, remote;
+  u8 protocol;
   u32 encap_fib_index;
   u32 decap_fib_index;
   u32 decap_next_index;
   u32 vni;
-  nsh_header_t nsh_hdr;
-} vnet_nsh_vxlan_gpe_add_del_tunnel_args_t;
+} vnet_vxlan_gpe_add_del_tunnel_args_t;
+
+
+int vnet_vxlan_gpe_add_del_tunnel 
+(vnet_vxlan_gpe_add_del_tunnel_args_t *a, u32 * sw_if_indexp);
+
+
+
 
-int vnet_nsh_vxlan_gpe_add_del_tunnel 
-(vnet_nsh_vxlan_gpe_add_del_tunnel_args_t *a, u32 * sw_if_indexp);
 
-#endif /* included_vnet_nsh_vxlan_gpe_h */
+#endif /* included_vnet_vxlan_gpe_h */
diff --git a/vnet/vnet/nsh-vxlan-gpe/nsh_vxlan_gpe_error.def b/vnet/vnet/vxlan-gpe/vxlan_gpe_error.def
index 4ba64fe4dc5..9cf1b1cb656 100644
--- a/vnet/vnet/nsh-vxlan-gpe/nsh_vxlan_gpe_error.def
+++ b/vnet/vnet/vxlan-gpe/vxlan_gpe_error.def
@@ -12,5 +12,5 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-nsh_vxlan_gpe_error (DECAPSULATED, "good packets decapsulated")
-nsh_vxlan_gpe_error (NO_SUCH_TUNNEL, "no such tunnel packets")
+vxlan_gpe_error (DECAPSULATED, "good packets decapsulated")
+vxlan_gpe_error (NO_SUCH_TUNNEL, "no such tunnel packets")
diff --git a/vnet/vnet/nsh-vxlan-gpe/vxlan_gpe_packet.h b/vnet/vnet/vxlan-gpe/vxlan_gpe_packet.h
index efc85c4bb54..3403cc9ebad 100644
--- a/vnet/vnet/nsh-vxlan-gpe/vxlan_gpe_packet.h
+++ b/vnet/vnet/vxlan-gpe/vxlan_gpe_packet.h
@@ -52,23 +52,31 @@
  *      0x4 : Network Service Header [NSH]
  */
 
+#define foreach_vxlan_gpe_protocol   \
+_ (0x01, IP4)                         \
+_ (0x02, IP6)                         \
+_ (0x03, ETHERNET)		     \
+_ (0x04, NSH)						
+
+
+typedef enum {
+#define _(n,f) VXLAN_GPE_PROTOCOL_##f = n,
+  foreach_vxlan_gpe_protocol
+#undef _
+} vxlan_gpe_protocol_t;
+
 typedef struct {
   u8 flags;
   u8 ver_res;
   u8 res;
-  u8 next_protocol;
+  /* see vxlan_gpe_protocol_t */
+  u8 protocol;
   u32 vni_res;
 } vxlan_gpe_header_t;
 
 #define VXLAN_GPE_FLAGS_I 0x08
 #define VXLAN_GPE_FLAGS_P 0x04
 #define VXLAN_GPE_FLAGS_O 0x01
-
 #define VXLAN_GPE_VERSION 0x0
 
-#define VXLAN_NEXT_PROTOCOL_IP4 0x1
-#define VXLAN_NEXT_PROTOCOL_IP6 0x2
-#define VXLAN_NEXT_PROTOCOL_ETHERNET 0x3
-#define VXLAN_NEXT_PROTOCOL_NSH 0x4
-
 #endif /* included_vxlan_gpe_packet_h */