ip: add shallow virtual reassembly functionality

Type: feature

Change-Id: Ibc8334e26c7e6f6120696c3e313b6e11d73dab99
Signed-off-by: Klement Sekera <ksekera@cisco.com>

13 files changed, 3158 insertions, 95 deletions

diff --git a/src/vnet/CMakeLists.txt b/src/vnet/CMakeLists.txt
index aaa16d8389b..6ce383350bf 100644
--- a/src/vnet/CMakeLists.txt
+++ b/src/vnet/CMakeLists.txt
@@ -432,6 +432,7 @@ list(APPEND VNET_SOURCES
   ip/ip4_source_and_port_range_check.c
   ip/ip4_source_check.c
   ip/reass/ip4_full_reass.c
+  ip/reass/ip4_sv_reass.c
   ip/ip6_format.c
   ip/ip6_forward.c
   ip/ip6_ll_table.c
@@ -442,6 +443,7 @@ list(APPEND VNET_SOURCES
   ip/ip6_neighbor.c
   ip/ip6_pg.c
   ip/reass/ip6_full_reass.c
+  ip/reass/ip6_sv_reass.c
   ip/rd_cp.c
   ip/ip_neighbor.c
   ip/ip_api.c
@@ -464,6 +466,10 @@ list(APPEND VNET_MULTIARCH_SOURCES
   ip/reass/ip4_full_reass.c
   ip/ip6_hop_by_hop.c
   ip/reass/ip6_full_reass.c
+  ip/reass/ip4_sv_reass.c
+  ip/ip6_hop_by_hop.c
+  ip/reass/ip6_full_reass.c
+  ip/reass/ip6_sv_reass.c
   ip/ip6_input.c
   ip/ip6_punt_drop.c
   ip/punt_node.c
diff --git a/src/vnet/buffer.h b/src/vnet/buffer.h
index 5a3bc3fc279..708e399047d 100644
--- a/src/vnet/buffer.h
+++ b/src/vnet/buffer.h
@@ -185,13 +185,40 @@ typedef struct
 	/* reassembly */
 	union
 	{
-	  /* in/out variables */
+	  /* group input/handoff as handoff is done before input is consumed,
+	   * this way we can handoff while keeping input variables intact */
 	  struct
 	  {
-	    u32 next_index;	/* index of next node - used by custom apps */
-	    u32 error_next_index;	/* index of next node if error - used by custom apps */
-	    u16 estimated_mtu;	/* estimated MTU calculated during reassembly */
-	    u16 owner_thread_index;
+	    /* input variables */
+	    struct
+	    {
+	      u32 next_index;	/* index of next node - used by custom apps */
+	      u32 error_next_index;	/* index of next node if error - used by custom apps */
+	    };
+	    /* handoff variables */
+	    struct
+	    {
+	      u16 owner_thread_index;
+	    };
+	  };
+	  /* output variables */
+	  struct
+	  {
+	    union
+	    {
+	      /* shallow virtual reassembly output variables */
+	      struct
+	      {
+		u8 ip_proto;	/* protocol in ip header */
+		u16 l4_src_port;	/* tcp/udp/icmp src port */
+		u16 l4_dst_port;	/* tcp/udp/icmp dst port */
+	      };
+	      /* full reassembly output variables */
+	      struct
+	      {
+		u16 estimated_mtu;	/* estimated MTU calculated during reassembly */
+	      };
+	    };
 	  };
 	  /* internal variables used during reassembly */
 	  struct
@@ -202,11 +229,9 @@ typedef struct
 	    u16 range_last;
 	    u32 next_range_bi;
 	    u16 ip6_frag_hdr_offset;
-	    u16 owner_feature_thread_index;
 	  };
 	} reass;
       };
-
     } ip;
 
     /*
diff --git a/src/vnet/ip/ip.api b/src/vnet/ip/ip.api
index 5d22bc52fea..69c7b88f2e6 100644
--- a/src/vnet/ip/ip.api
+++ b/src/vnet/ip/ip.api
@@ -1016,6 +1016,12 @@ autoreply define ioam_disable
   u16 id;
 };
 
+enum ip_reass_type
+{
+  IP_REASS_TYPE_FULL = 0,
+  IP_REASS_TYPE_SHALLOW_VIRTUAL = 0x1,
+};
+
 autoreply define ip_reassembly_set
 {
   u32 client_index;
@@ -1025,6 +1031,7 @@ autoreply define ip_reassembly_set
   u32 max_reassembly_length;
   u32 expire_walk_interval_ms;
   u8 is_ip6;
+  vl_api_ip_reass_type_t type;
 };
 
 define ip_reassembly_get
@@ -1032,6 +1039,7 @@ define ip_reassembly_get
   u32 client_index;
   u32 context;
   u8 is_ip6;
+  vl_api_ip_reass_type_t type;
 };
 
 define ip_reassembly_get_reply
@@ -1059,6 +1067,7 @@ autoreply define ip_reassembly_enable_disable
   u32 sw_if_index;
   u8 enable_ip4;
   u8 enable_ip6;
+  vl_api_ip_reass_type_t type;
 };
 
 /*
diff --git a/src/vnet/ip/ip4_error.h b/src/vnet/ip/ip4_error.h
index d3bf6d79714..c368b3b00dc 100644
--- a/src/vnet/ip/ip4_error.h
+++ b/src/vnet/ip/ip4_error.h
@@ -89,7 +89,8 @@
   _ (REASS_FRAGMENT_CHAIN_TOO_LONG, "fragment chain too long (drop)")   \
   _ (REASS_NO_BUF, "out of buffers (drop)")                             \
   _ (REASS_MALFORMED_PACKET, "malformed packets")                       \
-  _ (REASS_INTERNAL_ERROR, "drops due to internal reassembly error")
+  _ (REASS_INTERNAL_ERROR, "drops due to internal reassembly error")    \
+  _ (REASS_UNSUPP_IP_PROT, "unsupported ip protocol")
 
 typedef enum
 {
diff --git a/src/vnet/ip/ip6_error.h b/src/vnet/ip/ip6_error.h
index 3ca2be61a55..a6fb16570b6 100644
--- a/src/vnet/ip/ip6_error.h
+++ b/src/vnet/ip/ip6_error.h
@@ -84,7 +84,8 @@
   _ (REASS_FRAGMENT_CHAIN_TOO_LONG, "fragment chain too long (drop)")   \
   _ (REASS_NO_BUF, "out of buffers (drop)")                             \
   _ (REASS_TIMEOUT, "fragments dropped due to reassembly timeout")      \
-  _ (REASS_INTERNAL_ERROR, "drops due to internal reassembly error")
+  _ (REASS_INTERNAL_ERROR, "drops due to internal reassembly error")    \
+  _ (REASS_UNSUPP_IP_PROTO, "unsupported ip protocol")
 
 typedef enum
 {
diff --git a/src/vnet/ip/ip_api.c b/src/vnet/ip/ip_api.c
index 3e04b9f740c..dd1d1dcaf6c 100644
--- a/src/vnet/ip/ip_api.c
+++ b/src/vnet/ip/ip_api.c
@@ -43,7 +43,9 @@
 #include <vnet/fib/ip6_fib.h>
 #include <vnet/fib/fib_path_list.h>
 #include <vnet/ip/ip6_hop_by_hop.h>
+#include <vnet/ip/reass/ip4_sv_reass.h>
 #include <vnet/ip/reass/ip4_full_reass.h>
+#include <vnet/ip/reass/ip6_sv_reass.h>
 #include <vnet/ip/reass/ip6_full_reass.h>
 #include <vnet/ethernet/arp.h>
 #include <vnet/ip/ip_types_api.h>
@@ -2736,23 +2738,51 @@ vl_api_ip_reassembly_set_t_handler (vl_api_ip_reassembly_set_t * mp)
 {
   vl_api_ip_reassembly_set_reply_t *rmp;
   int rv = 0;
-  if (mp->is_ip6)
-    {
-      rv = ip6_full_reass_set (clib_net_to_host_u32 (mp->timeout_ms),
-			       clib_net_to_host_u32 (mp->max_reassemblies),
-			       clib_net_to_host_u32
-			       (mp->max_reassembly_length),
-			       clib_net_to_host_u32
-			       (mp->expire_walk_interval_ms));
-    }
-  else
+  switch ((vl_api_ip_reass_type_t) clib_net_to_host_u32 (mp->type))
     {
-      rv = ip4_full_reass_set (clib_net_to_host_u32 (mp->timeout_ms),
-			       clib_net_to_host_u32 (mp->max_reassemblies),
-			       clib_net_to_host_u32
-			       (mp->max_reassembly_length),
-			       clib_net_to_host_u32
-			       (mp->expire_walk_interval_ms));
+    case IP_REASS_TYPE_FULL:
+      if (mp->is_ip6)
+	{
+	  rv = ip6_full_reass_set (clib_net_to_host_u32 (mp->timeout_ms),
+				   clib_net_to_host_u32
+				   (mp->max_reassemblies),
+				   clib_net_to_host_u32
+				   (mp->max_reassembly_length),
+				   clib_net_to_host_u32
+				   (mp->expire_walk_interval_ms));
+	}
+      else
+	{
+	  rv = ip4_full_reass_set (clib_net_to_host_u32 (mp->timeout_ms),
+				   clib_net_to_host_u32
+				   (mp->max_reassemblies),
+				   clib_net_to_host_u32
+				   (mp->max_reassembly_length),
+				   clib_net_to_host_u32
+				   (mp->expire_walk_interval_ms));
+	}
+      break;
+    case IP_REASS_TYPE_SHALLOW_VIRTUAL:
+      if (mp->is_ip6)
+	{
+	  rv =
+	    ip6_sv_reass_set (clib_net_to_host_u32 (mp->timeout_ms),
+			      clib_net_to_host_u32 (mp->max_reassemblies),
+			      clib_net_to_host_u32
+			      (mp->max_reassembly_length),
+			      clib_net_to_host_u32
+			      (mp->expire_walk_interval_ms));
+	}
+      else
+	{
+	  rv = ip4_sv_reass_set (clib_net_to_host_u32 (mp->timeout_ms),
+				 clib_net_to_host_u32 (mp->max_reassemblies),
+				 clib_net_to_host_u32
+				 (mp->max_reassembly_length),
+				 clib_net_to_host_u32
+				 (mp->expire_walk_interval_ms));
+	}
+      break;
     }
 
   REPLY_MACRO (VL_API_IP_REASSEMBLY_SET_REPLY);
@@ -2772,19 +2802,42 @@ vl_api_ip_reassembly_get_t_handler (vl_api_ip_reassembly_get_t * mp)
   rmp->_vl_msg_id = ntohs (VL_API_IP_REASSEMBLY_GET_REPLY);
   rmp->context = mp->context;
   rmp->retval = 0;
-  if (mp->is_ip6)
+  u32 timeout_ms;
+  u32 max_reassemblies;
+  u32 max_reassembly_length;
+  u32 expire_walk_interval_ms;
+  switch ((vl_api_ip_reass_type_t) clib_net_to_host_u32 (mp->type))
     {
-      rmp->is_ip6 = 1;
-      ip6_full_reass_get (&rmp->timeout_ms, &rmp->max_reassemblies,
-			  &rmp->max_reassembly_length,
-			  &rmp->expire_walk_interval_ms);
-    }
-  else
-    {
-      rmp->is_ip6 = 0;
-      ip4_full_reass_get (&rmp->timeout_ms, &rmp->max_reassemblies,
-			  &rmp->max_reassembly_length,
-			  &rmp->expire_walk_interval_ms);
+    case IP_REASS_TYPE_FULL:
+      if (mp->is_ip6)
+	{
+	  rmp->is_ip6 = 1;
+	  ip6_full_reass_get (&timeout_ms, &max_reassemblies,
+			      &max_reassembly_length,
+			      &expire_walk_interval_ms);
+	}
+      else
+	{
+	  rmp->is_ip6 = 0;
+	  ip4_full_reass_get (&timeout_ms, &max_reassemblies,
+			      &max_reassembly_length,
+			      &expire_walk_interval_ms);
+	}
+      break;
+    case IP_REASS_TYPE_SHALLOW_VIRTUAL:
+      if (mp->is_ip6)
+	{
+	  rmp->is_ip6 = 1;
+	  ip6_sv_reass_get (&timeout_ms, &max_reassemblies,
+			    &max_reassembly_length, &expire_walk_interval_ms);
+	}
+      else
+	{
+	  rmp->is_ip6 = 0;
+	  ip4_sv_reass_get (&timeout_ms, &max_reassemblies,
+			    &max_reassembly_length, &expire_walk_interval_ms);
+	}
+      break;
     }
   rmp->timeout_ms = clib_host_to_net_u32 (rmp->timeout_ms);
   rmp->max_reassemblies = clib_host_to_net_u32 (rmp->max_reassemblies);
@@ -2799,13 +2852,28 @@ void
 {
   vl_api_ip_reassembly_enable_disable_reply_t *rmp;
   int rv = 0;
-  rv = ip4_full_reass_enable_disable (clib_net_to_host_u32 (mp->sw_if_index),
-				      mp->enable_ip4);
-  if (0 == rv)
+  switch ((vl_api_ip_reass_type_t) clib_net_to_host_u32 (mp->type))
     {
+    case IP_REASS_TYPE_FULL:
       rv =
-	ip6_full_reass_enable_disable (clib_net_to_host_u32 (mp->sw_if_index),
-				       mp->enable_ip6);
+	ip4_full_reass_enable_disable (clib_net_to_host_u32 (mp->sw_if_index),
+				       mp->enable_ip4);
+      if (0 == rv)
+	rv =
+	  ip6_full_reass_enable_disable (clib_net_to_host_u32
+					 (mp->sw_if_index), mp->enable_ip6);
+      break;
+    case IP_REASS_TYPE_SHALLOW_VIRTUAL:
+      rv =
+	ip4_sv_reass_enable_disable (clib_net_to_host_u32 (mp->sw_if_index),
+				     mp->enable_ip4);
+      if (0 == rv)
+	{
+	  rv =
+	    ip6_sv_reass_enable_disable (clib_net_to_host_u32
+					 (mp->sw_if_index), mp->enable_ip6);
+	}
+      break;
     }
 
   REPLY_MACRO (VL_API_IP_REASSEMBLY_ENABLE_DISABLE_REPLY);
diff --git a/src/vnet/ip/reass/ip4_full_reass.c b/src/vnet/ip/reass/ip4_full_reass.c
index 7cf4f43b0a8..18ac4d1b1b0 100644
--- a/src/vnet/ip/reass/ip4_full_reass.c
+++ b/src/vnet/ip/reass/ip4_full_reass.c
@@ -1146,12 +1146,8 @@ ip4_full_reass_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
 	  if (PREDICT_FALSE (do_handoff))
 	    {
 	      next0 = IP4_FULL_REASS_NEXT_HANDOFF;
-	      if (is_feature)
-		vnet_buffer (b0)->ip.reass.owner_feature_thread_index =
-		  kv.v.memory_owner_thread_index;
-	      else
-		vnet_buffer (b0)->ip.reass.owner_thread_index =
-		  kv.v.memory_owner_thread_index;
+	      vnet_buffer (b0)->ip.reass.owner_thread_index =
+		kv.v.memory_owner_thread_index;
 	    }
 	  else if (reass)
 	    {
@@ -1166,12 +1162,8 @@ ip4_full_reass_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
 		case IP4_REASS_RC_HANDOFF:
 		  next0 = IP4_FULL_REASS_NEXT_HANDOFF;
 		  b0 = vlib_get_buffer (vm, bi0);
-		  if (is_feature)
-		    vnet_buffer (b0)->ip.reass.owner_feature_thread_index =
-		      handoff_thread_idx;
-		  else
-		    vnet_buffer (b0)->ip.reass.owner_thread_index =
-		      handoff_thread_idx;
+		  vnet_buffer (b0)->ip.reass.owner_thread_index =
+		    handoff_thread_idx;
 		  break;
 		case IP4_REASS_RC_TOO_MANY_FRAGMENTS:
 		  vlib_node_increment_counter (vm, node->node_index,
@@ -1218,16 +1210,10 @@ ip4_full_reass_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
 		{
 		  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
 		    {
-		      if (is_feature)
-			ip4_full_reass_add_trace (vm, node, rm, NULL,
-						  bi0, HANDOFF, 0,
-						  vnet_buffer (b0)->ip.
-						  reass.owner_feature_thread_index);
-		      else
-			ip4_full_reass_add_trace (vm, node, rm, NULL,
-						  bi0, HANDOFF, 0,
-						  vnet_buffer (b0)->ip.
-						  reass.owner_thread_index);
+		      ip4_full_reass_add_trace (vm, node, rm, NULL, bi0,
+						HANDOFF, 0,
+						vnet_buffer (b0)->ip.
+						reass.owner_thread_index);
 		    }
 		}
 	      else if (is_feature && IP4_ERROR_NONE == error0)
@@ -1726,10 +1712,7 @@ ip4_full_reass_handoff_node_inline (vlib_main_t * vm,
 
   while (n_left_from > 0)
     {
-      ti[0] =
-	(is_feature) ? vnet_buffer (b[0])->ip.
-	reass.owner_feature_thread_index : vnet_buffer (b[0])->ip.
-	reass.owner_thread_index;
+      ti[0] = vnet_buffer (b[0])->ip.reass.owner_thread_index;
 
       if (PREDICT_FALSE
 	  ((node->flags & VLIB_NODE_FLAG_TRACE)
diff --git a/src/vnet/ip/reass/ip4_sv_reass.c b/src/vnet/ip/reass/ip4_sv_reass.c
new file mode 100644
index 00000000000..25119e8d0a7
--- /dev/null
+++ b/src/vnet/ip/reass/ip4_sv_reass.c
@@ -0,0 +1,1199 @@
+/*
+ * Copyright (c) 2017 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.
+ */
+
+/**
+ * @file
+ * @brief IPv4 Shallow Virtual Reassembly.
+ *
+ * This file contains the source code for IPv4 Shallow Virtual reassembly.
+ */
+
+#include <vppinfra/vec.h>
+#include <vnet/vnet.h>
+#include <vnet/ip/ip.h>
+#include <vnet/ip/ip4_to_ip6.h>
+#include <vppinfra/fifo.h>
+#include <vppinfra/bihash_16_8.h>
+#include <vnet/ip/reass/ip4_sv_reass.h>
+
+#define MSEC_PER_SEC 1000
+#define IP4_SV_REASS_TIMEOUT_DEFAULT_MS 100
+#define IP4_SV_REASS_EXPIRE_WALK_INTERVAL_DEFAULT_MS 10000	// 10 seconds default
+#define IP4_SV_REASS_MAX_REASSEMBLIES_DEFAULT 1024
+#define IP4_SV_REASS_MAX_REASSEMBLY_LENGTH_DEFAULT 3
+#define IP4_SV_REASS_HT_LOAD_FACTOR (0.75)
+
+typedef enum
+{
+  IP4_SV_REASS_RC_OK,
+  IP4_SV_REASS_RC_TOO_MANY_FRAGMENTS,
+  IP4_SV_REASS_RC_UNSUPP_IP_PROTO,
+} ip4_sv_reass_rc_t;
+
+typedef struct
+{
+  union
+  {
+    struct
+    {
+      u32 xx_id;
+      ip4_address_t src;
+      ip4_address_t dst;
+      u16 frag_id;
+      u8 proto;
+      u8 unused;
+    };
+    u64 as_u64[2];
+  };
+} ip4_sv_reass_key_t;
+
+typedef union
+{
+  struct
+  {
+    u32 reass_index;
+    u32 thread_index;
+  };
+  u64 as_u64;
+} ip4_sv_reass_val_t;
+
+typedef union
+{
+  struct
+  {
+    ip4_sv_reass_key_t k;
+    ip4_sv_reass_val_t v;
+  };
+  clib_bihash_kv_16_8_t kv;
+} ip4_sv_reass_kv_t;
+
+typedef struct
+{
+  // hash table key
+  ip4_sv_reass_key_t key;
+  // time when last packet was received
+  f64 last_heard;
+  // internal id of this reassembly
+  u64 id;
+  // trace operation counter
+  u32 trace_op_counter;
+  // minimum fragment length for this reassembly - used to estimate MTU
+  u16 min_fragment_length;
+  // buffer indexes of buffers in this reassembly in chronological order -
+  // including overlaps and duplicate fragments
+  u32 *cached_buffers;
+  // set to true when this reassembly is completed
+  bool is_complete;
+  // ip protocol
+  u8 ip_proto;
+  // l4 src port
+  u16 l4_src_port;
+  // l4 dst port
+  u16 l4_dst_port;
+  u32 next_index;
+  // lru indexes
+  u32 lru_prev;
+  u32 lru_next;
+} ip4_sv_reass_t;
+
+typedef struct
+{
+  ip4_sv_reass_t *pool;
+  u32 reass_n;
+  u32 id_counter;
+  clib_spinlock_t lock;
+  // lru indexes
+  u32 lru_first;
+  u32 lru_last;
+
+} ip4_sv_reass_per_thread_t;
+
+typedef struct
+{
+  // IPv4 config
+  u32 timeout_ms;
+  f64 timeout;
+  u32 expire_walk_interval_ms;
+  // maximum number of fragments in one reassembly
+  u32 max_reass_len;
+  // maximum number of reassemblies
+  u32 max_reass_n;
+
+  // IPv4 runtime
+  clib_bihash_16_8_t hash;
+  // per-thread data
+  ip4_sv_reass_per_thread_t *per_thread_data;
+
+  // convenience
+  vlib_main_t *vlib_main;
+  vnet_main_t *vnet_main;
+
+  // node index of ip4-drop node
+  u32 ip4_drop_idx;
+  u32 ip4_sv_reass_expire_node_idx;
+
+  /** Worker handoff */
+  u32 fq_index;
+  u32 fq_feature_index;
+
+  // reference count for enabling/disabling feature
+  u32 feature_use_refcount;
+
+} ip4_sv_reass_main_t;
+
+extern ip4_sv_reass_main_t ip4_sv_reass_main;
+
+#ifndef CLIB_MARCH_VARIANT
+ip4_sv_reass_main_t ip4_sv_reass_main;
+#endif /* CLIB_MARCH_VARIANT */
+
+typedef enum
+{
+  IP4_SV_REASSEMBLY_NEXT_INPUT,
+  IP4_SV_REASSEMBLY_NEXT_DROP,
+  IP4_SV_REASSEMBLY_NEXT_HANDOFF,
+  IP4_SV_REASSEMBLY_N_NEXT,
+} ip4_sv_reass_next_t;
+
+typedef enum
+{
+  REASS_FRAGMENT_CACHE,
+  REASS_FINISH,
+  REASS_FRAGMENT_FORWARD,
+} ip4_sv_reass_trace_operation_e;
+
+typedef struct
+{
+  ip4_sv_reass_trace_operation_e action;
+  u32 reass_id;
+  u32 op_id;
+  u8 ip_proto;
+  u16 l4_src_port;
+  u16 l4_dst_port;
+} ip4_sv_reass_trace_t;
+
+extern vlib_node_registration_t ip4_sv_reass_node;
+extern vlib_node_registration_t ip4_sv_reass_node_feature;
+
+static u8 *
+format_ip4_sv_reass_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 *);
+  ip4_sv_reass_trace_t *t = va_arg (*args, ip4_sv_reass_trace_t *);
+  s = format (s, "reass id: %u, op id: %u ", t->reass_id, t->op_id);
+  switch (t->action)
+    {
+    case REASS_FRAGMENT_CACHE:
+      s = format (s, "[cached]");
+      break;
+    case REASS_FINISH:
+      s =
+	format (s, "[finish, ip proto=%u, src_port=%u, dst_port=%u]",
+		t->ip_proto, clib_net_to_host_u16 (t->l4_src_port),
+		clib_net_to_host_u16 (t->l4_dst_port));
+      break;
+    case REASS_FRAGMENT_FORWARD:
+      s =
+	format (s, "[forward, ip proto=%u, src_port=%u, dst_port=%u]",
+		t->ip_proto, clib_net_to_host_u16 (t->l4_src_port),
+		clib_net_to_host_u16 (t->l4_dst_port));
+      break;
+    }
+  return s;
+}
+
+static void
+ip4_sv_reass_add_trace (vlib_main_t * vm, vlib_node_runtime_t * node,
+			ip4_sv_reass_main_t * rm, ip4_sv_reass_t * reass,
+			u32 bi, ip4_sv_reass_trace_operation_e action,
+			u32 ip_proto, u16 l4_src_port, u16 l4_dst_port)
+{
+  vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+  ip4_sv_reass_trace_t *t = vlib_add_trace (vm, node, b, sizeof (t[0]));
+  t->reass_id = reass->id;
+  t->action = action;
+  t->op_id = reass->trace_op_counter;
+  t->ip_proto = ip_proto;
+  t->l4_src_port = l4_src_port;
+  t->l4_dst_port = l4_dst_port;
+  ++reass->trace_op_counter;
+#if 0
+  static u8 *s = NULL;
+  s = format (s, "%U", format_ip4_sv_reass_trace, NULL, NULL, t);
+  printf ("%.*s\n", vec_len (s), s);
+  fflush (stdout);
+  vec_reset_length (s);
+#endif
+}
+
+
+always_inline void
+ip4_sv_reass_free (vlib_main_t * vm, ip4_sv_reass_main_t * rm,
+		   ip4_sv_reass_per_thread_t * rt, ip4_sv_reass_t * reass)
+{
+  clib_bihash_kv_16_8_t kv;
+  kv.key[0] = reass->key.as_u64[0];
+  kv.key[1] = reass->key.as_u64[1];
+  clib_bihash_add_del_16_8 (&rm->hash, &kv, 0);
+  vlib_buffer_free (vm, reass->cached_buffers,
+		    vec_len (reass->cached_buffers));
+  vec_free (reass->cached_buffers);
+  reass->cached_buffers = NULL;
+  if (~0 != reass->lru_prev)
+    {
+      ip4_sv_reass_t *lru_prev =
+	pool_elt_at_index (rt->pool, reass->lru_prev);
+      lru_prev->lru_next = reass->lru_next;
+    }
+  if (~0 != reass->lru_next)
+    {
+      ip4_sv_reass_t *lru_next =
+	pool_elt_at_index (rt->pool, reass->lru_next);
+      lru_next->lru_prev = reass->lru_prev;
+    }
+  if (rt->lru_first == reass - rt->pool)
+    {
+      rt->lru_first = reass->lru_next;
+    }
+  if (rt->lru_last == reass - rt->pool)
+    {
+      rt->lru_last = reass->lru_prev;
+    }
+  pool_put (rt->pool, reass);
+  --rt->reass_n;
+}
+
+always_inline void
+ip4_sv_reass_init (ip4_sv_reass_t * reass)
+{
+  reass->cached_buffers = NULL;
+  reass->is_complete = false;
+}
+
+always_inline ip4_sv_reass_t *
+ip4_sv_reass_find_or_create (vlib_main_t * vm, ip4_sv_reass_main_t * rm,
+			     ip4_sv_reass_per_thread_t * rt,
+			     ip4_sv_reass_kv_t * kv, u8 * do_handoff)
+{
+  ip4_sv_reass_t *reass = NULL;
+  f64 now = vlib_time_now (rm->vlib_main);
+
+  if (!clib_bihash_search_16_8
+      (&rm->hash, (clib_bihash_kv_16_8_t *) kv, (clib_bihash_kv_16_8_t *) kv))
+    {
+      if (vm->thread_index != kv->v.thread_index)
+	{
+	  *do_handoff = 1;
+	  return NULL;
+	}
+      reass = pool_elt_at_index (rt->pool, kv->v.reass_index);
+
+      if (now > reass->last_heard + rm->timeout)
+	{
+	  ip4_sv_reass_free (vm, rm, rt, reass);
+	  reass = NULL;
+	}
+    }
+
+  if (reass)
+    {
+      reass->last_heard = now;
+      return reass;
+    }
+
+  if (rt->reass_n >= rm->max_reass_n && rm->max_reass_n)
+    {
+      reass = pool_elt_at_index (rt->pool, rt->lru_last);
+      ip4_sv_reass_free (vm, rm, rt, reass);
+    }
+
+  pool_get (rt->pool, reass);
+  clib_memset (reass, 0, sizeof (*reass));
+  reass->id = ((u64) vm->thread_index * 1000000000) + rt->id_counter;
+  ++rt->id_counter;
+  ip4_sv_reass_init (reass);
+  ++rt->reass_n;
+  reass->lru_prev = reass->lru_next = ~0;
+
+  if (~0 != rt->lru_last)
+    {
+      ip4_sv_reass_t *lru_last = pool_elt_at_index (rt->pool, rt->lru_last);
+      reass->lru_prev = rt->lru_last;
+      lru_last->lru_next = rt->lru_last = reass - rt->pool;
+    }
+
+  if (~0 == rt->lru_first)
+    {
+      rt->lru_first = rt->lru_last = reass - rt->pool;
+    }
+
+  reass->key.as_u64[0] = ((clib_bihash_kv_16_8_t *) kv)->key[0];
+  reass->key.as_u64[1] = ((clib_bihash_kv_16_8_t *) kv)->key[1];
+  kv->v.reass_index = (reass - rt->pool);
+  kv->v.thread_index = vm->thread_index;
+  reass->last_heard = now;
+
+  if (clib_bihash_add_del_16_8 (&rm->hash, (clib_bihash_kv_16_8_t *) kv, 1))
+    {
+      ip4_sv_reass_free (vm, rm, rt, reass);
+      reass = NULL;
+    }
+
+  return reass;
+}
+
+always_inline ip4_sv_reass_rc_t
+ip4_sv_reass_update (vlib_main_t * vm, vlib_node_runtime_t * node,
+		     ip4_sv_reass_main_t * rm, ip4_sv_reass_per_thread_t * rt,
+		     ip4_sv_reass_t * reass, u32 bi0)
+{
+  vlib_buffer_t *fb = vlib_get_buffer (vm, bi0);
+  ip4_sv_reass_rc_t rc = IP4_SV_REASS_RC_OK;
+  ip4_header_t *fip = vlib_buffer_get_current (fb);
+  const u32 fragment_first = ip4_get_fragment_offset_bytes (fip);
+  if (0 == fragment_first)
+    {
+      reass->ip_proto = fip->protocol;
+      reass->l4_src_port = ip4_get_port (fip, 1);
+      reass->l4_dst_port = ip4_get_port (fip, 0);
+      if (!reass->l4_src_port || !reass->l4_dst_port)
+	return IP4_SV_REASS_RC_UNSUPP_IP_PROTO;
+      reass->is_complete = true;
+      vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
+      if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	{
+	  ip4_sv_reass_add_trace (vm, node, rm, reass, bi0, REASS_FINISH,
+				  reass->ip_proto, reass->l4_src_port,
+				  reass->l4_dst_port);
+	}
+    }
+  vec_add1 (reass->cached_buffers, bi0);
+  if (!reass->is_complete)
+    {
+      if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+	{
+	  ip4_sv_reass_add_trace (vm, node, rm, reass, bi0,
+				  REASS_FRAGMENT_CACHE, ~0, ~0, ~0);
+	}
+      if (vec_len (reass->cached_buffers) > rm->max_reass_len)
+	{
+	  rc = IP4_SV_REASS_RC_TOO_MANY_FRAGMENTS;
+	}
+    }
+  return rc;
+}
+
+always_inline uword
+ip4_sv_reass_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
+		     vlib_frame_t * frame, bool is_feature, bool is_custom)
+{
+  u32 *from = vlib_frame_vector_args (frame);
+  u32 n_left_from, n_left_to_next, *to_next, next_index;
+  ip4_sv_reass_main_t *rm = &ip4_sv_reass_main;
+  ip4_sv_reass_per_thread_t *rt = &rm->per_thread_data[vm->thread_index];
+  clib_spinlock_lock (&rt->lock);
+
+  n_left_from = frame->n_vectors;
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  u32 next0;
+	  u32 error0 = IP4_ERROR_NONE;
+
+	  bi0 = from[0];
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  ip4_header_t *ip0 = vlib_buffer_get_current (b0);
+	  if (!ip4_get_fragment_more (ip0) && !ip4_get_fragment_offset (ip0))
+	    {
+	      // this is a regular packet - no fragmentation
+	      vnet_buffer (b0)->ip.reass.ip_proto = ip0->protocol;
+	      vnet_buffer (b0)->ip.reass.l4_src_port = ip4_get_port (ip0, 1);
+	      vnet_buffer (b0)->ip.reass.l4_dst_port = ip4_get_port (ip0, 0);
+	      next0 = IP4_SV_REASSEMBLY_NEXT_INPUT;
+	      goto packet_enqueue;
+	    }
+	  const u32 fragment_first = ip4_get_fragment_offset_bytes (ip0);
+	  const u32 fragment_length =
+	    clib_net_to_host_u16 (ip0->length) - ip4_header_bytes (ip0);
+	  const u32 fragment_last = fragment_first + fragment_length - 1;
+	  if (fragment_first > fragment_last || fragment_first + fragment_length > UINT16_MAX - 20 || (fragment_length < 8 && ip4_get_fragment_more (ip0)))	// 8 is minimum frag length per RFC 791
+	    {
+	      next0 = IP4_SV_REASSEMBLY_NEXT_DROP;
+	      error0 = IP4_ERROR_REASS_MALFORMED_PACKET;
+	      goto packet_enqueue;
+	    }
+	  ip4_sv_reass_kv_t kv;
+	  u8 do_handoff = 0;
+
+	  kv.k.as_u64[0] =
+	    (u64) vec_elt (ip4_main.fib_index_by_sw_if_index,
+			   vnet_buffer (b0)->sw_if_index[VLIB_RX]) |
+	    (u64) ip0->src_address.as_u32 << 32;
+	  kv.k.as_u64[1] =
+	    (u64) ip0->dst_address.
+	    as_u32 | (u64) ip0->fragment_id << 32 | (u64) ip0->protocol << 48;
+
+	  ip4_sv_reass_t *reass =
+	    ip4_sv_reass_find_or_create (vm, rm, rt, &kv, &do_handoff);
+
+	  if (PREDICT_FALSE (do_handoff))
+	    {
+	      next0 = IP4_SV_REASSEMBLY_NEXT_HANDOFF;
+	      vnet_buffer (b0)->ip.reass.owner_thread_index =
+		kv.v.thread_index;
+	      goto packet_enqueue;
+	    }
+
+	  if (!reass)
+	    {
+	      next0 = IP4_SV_REASSEMBLY_NEXT_DROP;
+	      error0 = IP4_ERROR_REASS_LIMIT_REACHED;
+	      goto packet_enqueue;
+	    }
+
+	  if (reass->is_complete)
+	    {
+	      vnet_buffer (b0)->ip.reass.ip_proto = reass->ip_proto;
+	      vnet_buffer (b0)->ip.reass.l4_src_port = reass->l4_src_port;
+	      vnet_buffer (b0)->ip.reass.l4_dst_port = reass->l4_dst_port;
+	      next0 = IP4_SV_REASSEMBLY_NEXT_INPUT;
+	      error0 = IP4_ERROR_NONE;
+	      if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+		{
+		  ip4_sv_reass_add_trace (vm, node, rm, reass, bi0,
+					  REASS_FRAGMENT_FORWARD,
+					  reass->ip_proto,
+					  reass->l4_src_port,
+					  reass->l4_dst_port);
+		}
+	      goto packet_enqueue;
+	    }
+
+	  ip4_sv_reass_rc_t rc =
+	    ip4_sv_reass_update (vm, node, rm, rt, reass, bi0);
+	  switch (rc)
+	    {
+	    case IP4_SV_REASS_RC_OK:
+	      /* nothing to do here */
+	      break;
+	    case IP4_SV_REASS_RC_TOO_MANY_FRAGMENTS:
+	      vlib_node_increment_counter (vm, node->node_index,
+					   IP4_ERROR_REASS_FRAGMENT_CHAIN_TOO_LONG,
+					   1);
+	      ip4_sv_reass_free (vm, rm, rt, reass);
+	      goto next_packet;
+	      break;
+	    case IP4_SV_REASS_RC_UNSUPP_IP_PROTO:
+	      vlib_node_increment_counter (vm, node->node_index,
+					   IP4_ERROR_REASS_FRAGMENT_CHAIN_TOO_LONG,
+					   1);
+	      ip4_sv_reass_free (vm, rm, rt, reass);
+	      goto next_packet;
+	      break;
+	    }
+	  if (reass->is_complete)
+	    {
+	      u32 idx;
+	      vec_foreach_index (idx, reass->cached_buffers)
+	      {
+		u32 bi0 = vec_elt (reass->cached_buffers, idx);
+		vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
+		u32 next0 = IP4_SV_REASSEMBLY_NEXT_INPUT;
+		if (is_feature)
+		  {
+		    vnet_feature_next (&next0, b0);
+		  }
+		if (is_custom)
+		  {
+		    next0 = vnet_buffer (b0)->ip.reass.next_index;
+		  }
+		if (0 == n_left_to_next)
+		  {
+		    vlib_put_next_frame (vm, node, next_index,
+					 n_left_to_next);
+		    vlib_get_next_frame (vm, node, next_index, to_next,
+					 n_left_to_next);
+		  }
+		to_next[0] = bi0;
+		to_next += 1;
+		n_left_to_next -= 1;
+		vnet_buffer (b0)->ip.reass.ip_proto = reass->ip_proto;
+		vnet_buffer (b0)->ip.reass.l4_src_port = reass->l4_src_port;
+		vnet_buffer (b0)->ip.reass.l4_dst_port = reass->l4_dst_port;
+		if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+		  {
+		    ip4_sv_reass_add_trace (vm, node, rm, reass, bi0,
+					    REASS_FRAGMENT_FORWARD,
+					    reass->ip_proto,
+					    reass->l4_src_port,
+					    reass->l4_dst_port);
+		  }
+		vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+						 to_next, n_left_to_next, bi0,
+						 next0);
+	      }
+	      _vec_len (reass->cached_buffers) = 0;	// buffers are owned by frame now
+	    }
+	  goto next_packet;
+
+	packet_enqueue:
+	  b0->error = node->errors[error0];
+
+	  to_next[0] = bi0;
+	  to_next += 1;
+	  n_left_to_next -= 1;
+	  if (is_feature && IP4_ERROR_NONE == error0)
+	    {
+	      b0 = vlib_get_buffer (vm, bi0);
+	      vnet_feature_next (&next0, b0);
+	    }
+	  if (is_custom)
+	    {
+	      next0 = vnet_buffer (b0)->ip.reass.next_index;
+	    }
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+					   to_next, n_left_to_next,
+					   bi0, next0);
+
+	next_packet:
+	  from += 1;
+	  n_left_from -= 1;
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  clib_spinlock_unlock (&rt->lock);
+  return frame->n_vectors;
+}
+
+static char *ip4_sv_reass_error_strings[] = {
+#define _(sym, string) string,
+  foreach_ip4_error
+#undef _
+};
+
+VLIB_NODE_FN (ip4_sv_reass_node) (vlib_main_t * vm,
+				  vlib_node_runtime_t * node,
+				  vlib_frame_t * frame)
+{
+  return ip4_sv_reass_inline (vm, node, frame, false /* is_feature */ ,
+			      false /* is_custom */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_sv_reass_node) = {
+    .name = "ip4-sv-reassembly",
+    .vector_size = sizeof (u32),
+    .format_trace = format_ip4_sv_reass_trace,
+    .n_errors = ARRAY_LEN (ip4_sv_reass_error_strings),
+    .error_strings = ip4_sv_reass_error_strings,
+    .n_next_nodes = IP4_SV_REASSEMBLY_N_NEXT,
+    .next_nodes =
+        {
+                [IP4_SV_REASSEMBLY_NEXT_INPUT] = "ip4-input",
+                [IP4_SV_REASSEMBLY_NEXT_DROP] = "ip4-drop",
+                [IP4_SV_REASSEMBLY_NEXT_HANDOFF] = "ip4-sv-reassembly-handoff",
+
+        },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FN (ip4_sv_reass_node_feature) (vlib_main_t * vm,
+					  vlib_node_runtime_t * node,
+					  vlib_frame_t * frame)
+{
+  return ip4_sv_reass_inline (vm, node, frame, true /* is_feature */ ,
+			      false /* is_custom */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_sv_reass_node_feature) = {
+    .name = "ip4-sv-reassembly-feature",
+    .vector_size = sizeof (u32),
+    .format_trace = format_ip4_sv_reass_trace,
+    .n_errors = ARRAY_LEN (ip4_sv_reass_error_strings),
+    .error_strings = ip4_sv_reass_error_strings,
+    .n_next_nodes = IP4_SV_REASSEMBLY_N_NEXT,
+    .next_nodes =
+        {
+                [IP4_SV_REASSEMBLY_NEXT_INPUT] = "ip4-input",
+                [IP4_SV_REASSEMBLY_NEXT_DROP] = "ip4-drop",
+                [IP4_SV_REASSEMBLY_NEXT_HANDOFF] = "ip4-sv-reass-feature-hoff",
+        },
+};
+/* *INDENT-ON* */
+
+/* *INDENT-OFF* */
+VNET_FEATURE_INIT (ip4_sv_reass_feature) = {
+    .arc_name = "ip4-unicast",
+    .node_name = "ip4-sv-reassembly-feature",
+    .runs_before = VNET_FEATURES ("ip4-lookup"),
+    .runs_after = 0,
+};
+/* *INDENT-ON* */
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_sv_reass_custom_node) = {
+    .name = "ip4-sv-reassembly-custom-next",
+    .vector_size = sizeof (u32),
+    .format_trace = format_ip4_sv_reass_trace,
+    .n_errors = ARRAY_LEN (ip4_sv_reass_error_strings),
+    .error_strings = ip4_sv_reass_error_strings,
+    .n_next_nodes = IP4_SV_REASSEMBLY_N_NEXT,
+    .next_nodes =
+        {
+                [IP4_SV_REASSEMBLY_NEXT_INPUT] = "ip4-input",
+                [IP4_SV_REASSEMBLY_NEXT_DROP] = "ip4-drop",
+                [IP4_SV_REASSEMBLY_NEXT_HANDOFF] = "ip4-sv-reassembly-handoff",
+
+        },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FN (ip4_sv_reass_custom_node) (vlib_main_t * vm,
+					 vlib_node_runtime_t * node,
+					 vlib_frame_t * frame)
+{
+  return ip4_sv_reass_inline (vm, node, frame, false /* is_feature */ ,
+			      true /* is_custom */ );
+}
+
+#ifndef CLIB_MARCH_VARIANT
+always_inline u32
+ip4_sv_reass_get_nbuckets ()
+{
+  ip4_sv_reass_main_t *rm = &ip4_sv_reass_main;
+  u32 nbuckets;
+  u8 i;
+
+  nbuckets = (u32) (rm->max_reass_n / IP4_SV_REASS_HT_LOAD_FACTOR);
+
+  for (i = 0; i < 31; i++)
+    if ((1 << i) >= nbuckets)
+      break;
+  nbuckets = 1 << i;
+
+  return nbuckets;
+}
+#endif /* CLIB_MARCH_VARIANT */
+
+typedef enum
+{
+  IP4_EVENT_CONFIG_CHANGED = 1,
+} ip4_sv_reass_event_t;
+
+typedef struct
+{
+  int failure;
+  clib_bihash_16_8_t *new_hash;
+} ip4_rehash_cb_ctx;
+
+#ifndef CLIB_MARCH_VARIANT
+static void
+ip4_rehash_cb (clib_bihash_kv_16_8_t * kv, void *_ctx)
+{
+  ip4_rehash_cb_ctx *ctx = _ctx;
+  if (clib_bihash_add_del_16_8 (ctx->new_hash, kv, 1))
+    {
+      ctx->failure = 1;
+    }
+}
+
+static void
+ip4_sv_reass_set_params (u32 timeout_ms, u32 max_reassemblies,
+			 u32 max_reassembly_length,
+			 u32 expire_walk_interval_ms)
+{
+  ip4_sv_reass_main.timeout_ms = timeout_ms;
+  ip4_sv_reass_main.timeout = (f64) timeout_ms / (f64) MSEC_PER_SEC;
+  ip4_sv_reass_main.max_reass_n = max_reassemblies;
+  ip4_sv_reass_main.max_reass_len = max_reassembly_length;
+  ip4_sv_reass_main.expire_walk_interval_ms = expire_walk_interval_ms;
+}
+
+vnet_api_error_t
+ip4_sv_reass_set (u32 timeout_ms, u32 max_reassemblies,
+		  u32 max_reassembly_length, u32 expire_walk_interval_ms)
+{
+  u32 old_nbuckets = ip4_sv_reass_get_nbuckets ();
+  ip4_sv_reass_set_params (timeout_ms, max_reassemblies,
+			   max_reassembly_length, expire_walk_interval_ms);
+  vlib_process_signal_event (ip4_sv_reass_main.vlib_main,
+			     ip4_sv_reass_main.ip4_sv_reass_expire_node_idx,
+			     IP4_EVENT_CONFIG_CHANGED, 0);
+  u32 new_nbuckets = ip4_sv_reass_get_nbuckets ();
+  if (ip4_sv_reass_main.max_reass_n > 0 && new_nbuckets > old_nbuckets)
+    {
+      clib_bihash_16_8_t new_hash;
+      clib_memset (&new_hash, 0, sizeof (new_hash));
+      ip4_rehash_cb_ctx ctx;
+      ctx.failure = 0;
+      ctx.new_hash = &new_hash;
+      clib_bihash_init_16_8 (&new_hash, "ip4-dr", new_nbuckets,
+			     new_nbuckets * 1024);
+      clib_bihash_foreach_key_value_pair_16_8 (&ip4_sv_reass_main.hash,
+					       ip4_rehash_cb, &ctx);
+      if (ctx.failure)
+	{
+	  clib_bihash_free_16_8 (&new_hash);
+	  return -1;
+	}
+      else
+	{
+	  clib_bihash_free_16_8 (&ip4_sv_reass_main.hash);
+	  clib_memcpy_fast (&ip4_sv_reass_main.hash, &new_hash,
+			    sizeof (ip4_sv_reass_main.hash));
+	  clib_bihash_copied (&ip4_sv_reass_main.hash, &new_hash);
+	}
+    }
+  return 0;
+}
+
+vnet_api_error_t
+ip4_sv_reass_get (u32 * timeout_ms, u32 * max_reassemblies,
+		  u32 * max_reassembly_length, u32 * expire_walk_interval_ms)
+{
+  *timeout_ms = ip4_sv_reass_main.timeout_ms;
+  *max_reassemblies = ip4_sv_reass_main.max_reass_n;
+  *max_reassembly_length = ip4_sv_reass_main.max_reass_len;
+  *expire_walk_interval_ms = ip4_sv_reass_main.expire_walk_interval_ms;
+  return 0;
+}
+
+static clib_error_t *
+ip4_sv_reass_init_function (vlib_main_t * vm)
+{
+  ip4_sv_reass_main_t *rm = &ip4_sv_reass_main;
+  clib_error_t *error = 0;
+  u32 nbuckets;
+  vlib_node_t *node;
+
+  rm->vlib_main = vm;
+  rm->vnet_main = vnet_get_main ();
+
+  vec_validate (rm->per_thread_data, vlib_num_workers ());
+  ip4_sv_reass_per_thread_t *rt;
+  vec_foreach (rt, rm->per_thread_data)
+  {
+    clib_spinlock_init (&rt->lock);
+    pool_alloc (rt->pool, rm->max_reass_n);
+    rt->lru_first = rt->lru_last = ~0;
+  }
+
+  node = vlib_get_node_by_name (vm, (u8 *) "ip4-sv-reassembly-expire-walk");
+  ASSERT (node);
+  rm->ip4_sv_reass_expire_node_idx = node->index;
+
+  ip4_sv_reass_set_params (IP4_SV_REASS_TIMEOUT_DEFAULT_MS,
+			   IP4_SV_REASS_MAX_REASSEMBLIES_DEFAULT,
+			   IP4_SV_REASS_MAX_REASSEMBLY_LENGTH_DEFAULT,
+			   IP4_SV_REASS_EXPIRE_WALK_INTERVAL_DEFAULT_MS);
+
+  nbuckets = ip4_sv_reass_get_nbuckets ();
+  clib_bihash_init_16_8 (&rm->hash, "ip4-dr", nbuckets, nbuckets * 1024);
+
+  node = vlib_get_node_by_name (vm, (u8 *) "ip4-drop");
+  ASSERT (node);
+  rm->ip4_drop_idx = node->index;
+
+  rm->fq_index = vlib_frame_queue_main_init (ip4_sv_reass_node.index, 0);
+  rm->fq_feature_index =
+    vlib_frame_queue_main_init (ip4_sv_reass_node_feature.index, 0);
+
+  return error;
+}
+
+VLIB_INIT_FUNCTION (ip4_sv_reass_init_function);
+#endif /* CLIB_MARCH_VARIANT */
+
+static uword
+ip4_sv_reass_walk_expired (vlib_main_t * vm,
+			   vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+  ip4_sv_reass_main_t *rm = &ip4_sv_reass_main;
+  uword event_type, *event_data = 0;
+
+  while (true)
+    {
+      vlib_process_wait_for_event_or_clock (vm,
+					    (f64)
+					    rm->expire_walk_interval_ms /
+					    (f64) MSEC_PER_SEC);
+      event_type = vlib_process_get_events (vm, &event_data);
+
+      switch (event_type)
+	{
+	case ~0:		/* no events => timeout */
+	  /* nothing to do here */
+	  break;
+	case IP4_EVENT_CONFIG_CHANGED:
+	  break;
+	default:
+	  clib_warning ("BUG: event type 0x%wx", event_type);
+	  break;
+	}
+      f64 now = vlib_time_now (vm);
+
+      ip4_sv_reass_t *reass;
+      int *pool_indexes_to_free = NULL;
+
+      uword thread_index = 0;
+      int index;
+      const uword nthreads = vlib_num_workers () + 1;
+      for (thread_index = 0; thread_index < nthreads; ++thread_index)
+	{
+	  ip4_sv_reass_per_thread_t *rt = &rm->per_thread_data[thread_index];
+	  clib_spinlock_lock (&rt->lock);
+
+	  vec_reset_length (pool_indexes_to_free);
+          /* *INDENT-OFF* */
+          pool_foreach_index (index, rt->pool, ({
+                                reass = pool_elt_at_index (rt->pool, index);
+                                if (now > reass->last_heard + rm->timeout)
+                                  {
+                                    vec_add1 (pool_indexes_to_free, index);
+                                  }
+                              }));
+          /* *INDENT-ON* */
+	  int *i;
+          /* *INDENT-OFF* */
+          vec_foreach (i, pool_indexes_to_free)
+          {
+            ip4_sv_reass_t *reass = pool_elt_at_index (rt->pool, i[0]);
+            ip4_sv_reass_free (vm, rm, rt, reass);
+          }
+          /* *INDENT-ON* */
+
+	  clib_spinlock_unlock (&rt->lock);
+	}
+
+      vec_free (pool_indexes_to_free);
+      if (event_data)
+	{
+	  _vec_len (event_data) = 0;
+	}
+    }
+
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_sv_reass_expire_node) = {
+    .function = ip4_sv_reass_walk_expired,
+    .type = VLIB_NODE_TYPE_PROCESS,
+    .name = "ip4-sv-reassembly-expire-walk",
+    .format_trace = format_ip4_sv_reass_trace,
+    .n_errors = ARRAY_LEN (ip4_sv_reass_error_strings),
+    .error_strings = ip4_sv_reass_error_strings,
+
+};
+/* *INDENT-ON* */
+
+static u8 *
+format_ip4_sv_reass_key (u8 * s, va_list * args)
+{
+  ip4_sv_reass_key_t *key = va_arg (*args, ip4_sv_reass_key_t *);
+  s =
+    format (s,
+	    "xx_id: %u, src: %U, dst: %U, frag_id: %u, proto: %u",
+	    key->xx_id, format_ip4_address, &key->src, format_ip4_address,
+	    &key->dst, clib_net_to_host_u16 (key->frag_id), key->proto);
+  return s;
+}
+
+static u8 *
+format_ip4_sv_reass (u8 * s, va_list * args)
+{
+  vlib_main_t *vm = va_arg (*args, vlib_main_t *);
+  ip4_sv_reass_t *reass = va_arg (*args, ip4_sv_reass_t *);
+
+  s = format (s, "ID: %lu, key: %U trace_op_counter: %u\n",
+	      reass->id, format_ip4_sv_reass_key, &reass->key,
+	      reass->trace_op_counter);
+
+  vlib_buffer_t *b;
+  u32 *bip;
+  u32 counter = 0;
+  vec_foreach (bip, reass->cached_buffers)
+  {
+    u32 bi = *bip;
+    do
+      {
+	b = vlib_get_buffer (vm, bi);
+	s = format (s, "  #%03u: bi: %u, ", counter, bi);
+	++counter;
+	bi = b->next_buffer;
+      }
+    while (b->flags & VLIB_BUFFER_NEXT_PRESENT);
+  }
+  return s;
+}
+
+static clib_error_t *
+show_ip4_reass (vlib_main_t * vm,
+		unformat_input_t * input,
+		CLIB_UNUSED (vlib_cli_command_t * lmd))
+{
+  ip4_sv_reass_main_t *rm = &ip4_sv_reass_main;
+
+  vlib_cli_output (vm, "---------------------");
+  vlib_cli_output (vm, "IP4 reassembly status");
+  vlib_cli_output (vm, "---------------------");
+  bool details = false;
+  if (unformat (input, "details"))
+    {
+      details = true;
+    }
+
+  u32 sum_reass_n = 0;
+  ip4_sv_reass_t *reass;
+  uword thread_index;
+  const uword nthreads = vlib_num_workers () + 1;
+  for (thread_index = 0; thread_index < nthreads; ++thread_index)
+    {
+      ip4_sv_reass_per_thread_t *rt = &rm->per_thread_data[thread_index];
+      clib_spinlock_lock (&rt->lock);
+      if (details)
+	{
+          /* *INDENT-OFF* */
+          pool_foreach (reass, rt->pool, {
+            vlib_cli_output (vm, "%U", format_ip4_sv_reass, vm, reass);
+          });
+          /* *INDENT-ON* */
+	}
+      sum_reass_n += rt->reass_n;
+      clib_spinlock_unlock (&rt->lock);
+    }
+  vlib_cli_output (vm, "---------------------");
+  vlib_cli_output (vm, "Current IP4 reassemblies count: %lu\n",
+		   (long unsigned) sum_reass_n);
+  vlib_cli_output (vm,
+		   "Maximum configured concurrent IP4 reassemblies per worker-thread: %lu\n",
+		   (long unsigned) rm->max_reass_n);
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (show_ip4_sv_reass_cmd, static) = {
+    .path = "show ip4-sv-reassembly",
+    .short_help = "show ip4-sv-reassembly [details]",
+    .function = show_ip4_reass,
+};
+/* *INDENT-ON* */
+
+#ifndef CLIB_MARCH_VARIANT
+vnet_api_error_t
+ip4_sv_reass_enable_disable (u32 sw_if_index, u8 enable_disable)
+{
+  return vnet_feature_enable_disable ("ip4-unicast",
+				      "ip4-sv-reassembly-feature",
+				      sw_if_index, enable_disable, 0, 0);
+}
+#endif /* CLIB_MARCH_VARIANT */
+
+
+#define foreach_ip4_sv_reass_handoff_error                       \
+_(CONGESTION_DROP, "congestion drop")
+
+
+typedef enum
+{
+#define _(sym,str) IP4_SV_REASSEMBLY_HANDOFF_ERROR_##sym,
+  foreach_ip4_sv_reass_handoff_error
+#undef _
+    IP4_SV_REASSEMBLY_HANDOFF_N_ERROR,
+} ip4_sv_reass_handoff_error_t;
+
+static char *ip4_sv_reass_handoff_error_strings[] = {
+#define _(sym,string) string,
+  foreach_ip4_sv_reass_handoff_error
+#undef _
+};
+
+typedef struct
+{
+  u32 next_worker_index;
+} ip4_sv_reass_handoff_trace_t;
+
+static u8 *
+format_ip4_sv_reass_handoff_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 *);
+  ip4_sv_reass_handoff_trace_t *t =
+    va_arg (*args, ip4_sv_reass_handoff_trace_t *);
+
+  s =
+    format (s, "ip4-sv-reassembly-handoff: next-worker %d",
+	    t->next_worker_index);
+
+  return s;
+}
+
+always_inline uword
+ip4_sv_reass_handoff_node_inline (vlib_main_t * vm,
+				  vlib_node_runtime_t * node,
+				  vlib_frame_t * frame, bool is_feature)
+{
+  ip4_sv_reass_main_t *rm = &ip4_sv_reass_main;
+
+  vlib_buffer_t *bufs[VLIB_FRAME_SIZE], **b;
+  u32 n_enq, n_left_from, *from;
+  u16 thread_indices[VLIB_FRAME_SIZE], *ti;
+  u32 fq_index;
+
+  from = vlib_frame_vector_args (frame);
+  n_left_from = frame->n_vectors;
+  vlib_get_buffers (vm, from, bufs, n_left_from);
+
+  b = bufs;
+  ti = thread_indices;
+
+  fq_index = (is_feature) ? rm->fq_feature_index : rm->fq_index;
+
+  while (n_left_from > 0)
+    {
+      ti[0] = vnet_buffer (b[0])->ip.reass.owner_thread_index;
+
+      if (PREDICT_FALSE
+	  ((node->flags & VLIB_NODE_FLAG_TRACE)
+	   && (b[0]->flags & VLIB_BUFFER_IS_TRACED)))
+	{
+	  ip4_sv_reass_handoff_trace_t *t =
+	    vlib_add_trace (vm, node, b[0], sizeof (*t));
+	  t->next_worker_index = ti[0];
+	}
+
+      n_left_from -= 1;
+      ti += 1;
+      b += 1;
+    }
+  n_enq =
+    vlib_buffer_enqueue_to_thread (vm, fq_index, from, thread_indices,
+				   frame->n_vectors, 1);
+
+  if (n_enq < frame->n_vectors)
+    vlib_node_increment_counter (vm, node->node_index,
+				 IP4_SV_REASSEMBLY_HANDOFF_ERROR_CONGESTION_DROP,
+				 frame->n_vectors - n_enq);
+  return frame->n_vectors;
+}
+
+VLIB_NODE_FN (ip4_sv_reass_handoff_node) (vlib_main_t * vm,
+					  vlib_node_runtime_t * node,
+					  vlib_frame_t * frame)
+{
+  return ip4_sv_reass_handoff_node_inline (vm, node, frame,
+					   false /* is_feature */ );
+}
+
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_sv_reass_handoff_node) = {
+  .name = "ip4-sv-reassembly-handoff",
+  .vector_size = sizeof (u32),
+  .n_errors = ARRAY_LEN(ip4_sv_reass_handoff_error_strings),
+  .error_strings = ip4_sv_reass_handoff_error_strings,
+  .format_trace = format_ip4_sv_reass_handoff_trace,
+
+  .n_next_nodes = 1,
+
+  .next_nodes = {
+    [0] = "error-drop",
+  },
+};
+/* *INDENT-ON* */
+
+
+/* *INDENT-OFF* */
+VLIB_NODE_FN (ip4_sv_reass_feature_handoff_node) (vlib_main_t * vm,
+						    vlib_node_runtime_t *
+						    node,
+						    vlib_frame_t * frame)
+{
+  return ip4_sv_reass_handoff_node_inline (vm, node, frame,
+					     true /* is_feature */ );
+}
+/* *INDENT-ON* */
+
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip4_sv_reass_feature_handoff_node) = {
+  .name = "ip4-sv-reass-feature-hoff",
+  .vector_size = sizeof (u32),
+  .n_errors = ARRAY_LEN(ip4_sv_reass_handoff_error_strings),
+  .error_strings = ip4_sv_reass_handoff_error_strings,
+  .format_trace = format_ip4_sv_reass_handoff_trace,
+
+  .n_next_nodes = 1,
+
+  .next_nodes = {
+    [0] = "error-drop",
+  },
+};
+/* *INDENT-ON* */
+
+#ifndef CLIB_MARCH_VARIANT
+int
+ip4_sv_reass_enable_disable_with_refcnt (u32 sw_if_index, int is_enable)
+{
+  ip4_sv_reass_main_t *rm = &ip4_sv_reass_main;
+  if (is_enable)
+    {
+      if (!rm->feature_use_refcount)
+	{
+	  ++rm->feature_use_refcount;
+	  return vnet_feature_enable_disable ("ip4-unicast",
+					      "ip4-sv-reassembly-feature",
+					      sw_if_index, 1, 0, 0);
+	}
+      ++rm->feature_use_refcount;
+    }
+  else
+    {
+      --rm->feature_use_refcount;
+      if (!rm->feature_use_refcount)
+	return vnet_feature_enable_disable ("ip4-unicast",
+					    "ip4-sv-reassembly-feature",
+					    sw_if_index, 0, 0, 0);
+    }
+  return -1;
+}
+
+uword
+ip4_sv_reass_custom_register_next_node (uword node_index)
+{
+  return vlib_node_add_next (vlib_get_main (), ip4_sv_reass_custom_node.index,
+			     node_index);
+}
+#endif
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/ip/reass/ip4_sv_reass.h b/src/vnet/ip/reass/ip4_sv_reass.h
new file mode 100644
index 00000000000..cf9f36502a0
--- /dev/null
+++ b/src/vnet/ip/reass/ip4_sv_reass.h
@@ -0,0 +1,59 @@
+/*
+ * Copyright (c) 2017 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.
+ */
+
+/**
+ * @file
+ * @brief IPv4 shallow virtual reassembly.
+ *
+ * This file contains the source code for IPv4 shallow virtual reassembly.
+ */
+
+#ifndef __included_ip4_sv_reass_h__
+#define __included_ip4_sv_reass_h__
+
+#include <vnet/api_errno.h>
+#include <vnet/vnet.h>
+
+/**
+ * @brief set ip4 reassembly configuration
+ */
+vnet_api_error_t ip4_sv_reass_set (u32 timeout_ms, u32 max_reassemblies,
+				   u32 max_reassembly_length,
+				   u32 expire_walk_interval_ms);
+
+/**
+ * @brief get ip4 reassembly configuration
+ */
+vnet_api_error_t ip4_sv_reass_get (u32 * timeout_ms, u32 * max_reassemblies,
+				   u32 * max_reassembly_length,
+				   u32 * expire_walk_interval_ms);
+
+vnet_api_error_t ip4_sv_reass_enable_disable (u32 sw_if_index,
+					      u8 enable_disable);
+
+
+int ip4_sv_reass_enable_disable_with_refcnt (u32 sw_if_index, int is_enable);
+
+uword ip4_sv_reass_custom_register_next_node (uword node_index);
+
+#endif /* __included_ip4_sv_reass_h__ */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/ip/reass/ip6_full_reass.c b/src/vnet/ip/reass/ip6_full_reass.c
index ef1014922c5..0b41dea5a87 100644
--- a/src/vnet/ip/reass/ip6_full_reass.c
+++ b/src/vnet/ip/reass/ip6_full_reass.c
@@ -1101,12 +1101,8 @@ ip6_full_reassembly_inline (vlib_main_t * vm,
 	  if (PREDICT_FALSE (do_handoff))
 	    {
 	      next0 = IP6_FULL_REASSEMBLY_NEXT_HANDOFF;
-	      if (is_feature)
-		vnet_buffer (b0)->ip.reass.owner_feature_thread_index =
-		  kv.v.memory_owner_thread_index;
-	      else
-		vnet_buffer (b0)->ip.reass.owner_thread_index =
-		  kv.v.memory_owner_thread_index;
+	      vnet_buffer (b0)->ip.reass.owner_thread_index =
+		kv.v.memory_owner_thread_index;
 	    }
 	  else if (reass)
 	    {
@@ -1121,12 +1117,8 @@ ip6_full_reassembly_inline (vlib_main_t * vm,
 		case IP6_FULL_REASS_RC_HANDOFF:
 		  next0 = IP6_FULL_REASSEMBLY_NEXT_HANDOFF;
 		  b0 = vlib_get_buffer (vm, bi0);
-		  if (is_feature)
-		    vnet_buffer (b0)->ip.reass.owner_feature_thread_index =
-		      handoff_thread_idx;
-		  else
-		    vnet_buffer (b0)->ip.reass.owner_thread_index =
-		      handoff_thread_idx;
+		  vnet_buffer (b0)->ip.reass.owner_thread_index =
+		    handoff_thread_idx;
 		  break;
 		case IP6_FULL_REASS_RC_TOO_MANY_FRAGMENTS:
 		  vlib_node_increment_counter (vm, node->node_index,
@@ -1179,16 +1171,10 @@ ip6_full_reassembly_inline (vlib_main_t * vm,
 		{
 		  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
 		    {
-		      if (is_feature)
-			ip6_full_reass_add_trace (vm, node, rm, NULL, bi0,
-						  HANDOFF,
-						  vnet_buffer (b0)->ip.
-						  reass.owner_feature_thread_index);
-		      else
-			ip6_full_reass_add_trace (vm, node, rm, NULL, bi0,
-						  HANDOFF,
-						  vnet_buffer (b0)->ip.
-						  reass.owner_thread_index);
+		      ip6_full_reass_add_trace (vm, node, rm, NULL, bi0,
+						HANDOFF,
+						vnet_buffer (b0)->ip.
+						reass.owner_thread_index);
 		    }
 		}
 	      else if (is_feature && IP6_ERROR_NONE == error0)
@@ -1730,10 +1716,7 @@ ip6_full_reassembly_handoff_inline (vlib_main_t * vm,
 
   while (n_left_from > 0)
     {
-      ti[0] =
-	(is_feature) ? vnet_buffer (b[0])->ip.
-	reass.owner_feature_thread_index : vnet_buffer (b[0])->ip.
-	reass.owner_thread_index;
+      ti[0] = vnet_buffer (b[0])->ip.reass.owner_thread_index;
 
       if (PREDICT_FALSE
 	  ((node->flags & VLIB_NODE_FLAG_TRACE)
diff --git a/src/vnet/ip/reass/ip6_sv_reass.c b/src/vnet/ip/reass/ip6_sv_reass.c
new file mode 100644
index 00000000000..e2f425a6aa8
--- /dev/null
+++ b/src/vnet/ip/reass/ip6_sv_reass.c
@@ -0,0 +1,1332 @@
+/*
+ * Copyright (c) 2017 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.
+ */
+
+/**
+ * @file
+ * @brief IPv6 Shallow Virtual Reassembly.
+ *
+ * This file contains the source code for IPv6 Shallow Virtual reassembly.
+ */
+
+#include <vppinfra/vec.h>
+#include <vnet/vnet.h>
+#include <vnet/ip/ip.h>
+#include <vnet/ip/ip6_to_ip4.h>
+#include <vppinfra/bihash_48_8.h>
+#include <vnet/ip/reass/ip6_sv_reass.h>
+
+#define MSEC_PER_SEC 1000
+#define IP6_SV_REASS_TIMEOUT_DEFAULT_MS 100
+#define IP6_SV_REASS_EXPIRE_WALK_INTERVAL_DEFAULT_MS 10000	// 10 seconds default
+#define IP6_SV_REASS_MAX_REASSEMBLIES_DEFAULT 1024
+#define IP6_SV_REASS_MAX_REASSEMBLY_LENGTH_DEFAULT 3
+#define IP6_SV_REASS_HT_LOAD_FACTOR (0.75)
+
+typedef enum
+{
+  IP6_SV_REASS_RC_OK,
+  IP6_SV_REASS_RC_TOO_MANY_FRAGMENTS,
+  IP6_SV_REASS_RC_INTERNAL_ERROR,
+  IP6_SV_REASS_RC_UNSUPP_IP_PROTO,
+} ip6_sv_reass_rc_t;
+
+typedef struct
+{
+  union
+  {
+    struct
+    {
+      ip6_address_t src;
+      ip6_address_t dst;
+      u32 xx_id;
+      u32 frag_id;
+      u8 unused[7];
+      u8 proto;
+    };
+    u64 as_u64[6];
+  };
+} ip6_sv_reass_key_t;
+
+typedef union
+{
+  struct
+  {
+    u32 reass_index;
+    u32 thread_index;
+  };
+  u64 as_u64;
+} ip6_sv_reass_val_t;
+
+typedef union
+{
+  struct
+  {
+    ip6_sv_reass_key_t k;
+    ip6_sv_reass_val_t v;
+  };
+  clib_bihash_kv_48_8_t kv;
+} ip6_sv_reass_kv_t;
+
+typedef struct
+{
+  // hash table key
+  ip6_sv_reass_key_t key;
+  // time when last packet was received
+  f64 last_heard;
+  // internal id of this reassembly
+  u64 id;
+  // trace operation counter
+  u32 trace_op_counter;
+  // buffer indexes of buffers in this reassembly in chronological order -
+  // including overlaps and duplicate fragments
+  u32 *cached_buffers;
+  // set to true when this reassembly is completed
+  bool is_complete;
+  // ip protocol
+  u8 ip_proto;
+  // l4 src port
+  u16 l4_src_port;
+  // l4 dst port
+  u16 l4_dst_port;
+  // lru indexes
+  u32 lru_prev;
+  u32 lru_next;
+} ip6_sv_reass_t;
+
+typedef struct
+{
+  ip6_sv_reass_t *pool;
+  u32 reass_n;
+  u32 id_counter;
+  clib_spinlock_t lock;
+  // lru indexes
+  u32 lru_first;
+  u32 lru_last;
+} ip6_sv_reass_per_thread_t;
+
+typedef struct
+{
+  // IPv6 config
+  u32 timeout_ms;
+  f64 timeout;
+  u32 expire_walk_interval_ms;
+  // maximum number of fragments in one reassembly
+  u32 max_reass_len;
+  // maximum number of reassemblies
+  u32 max_reass_n;
+
+  // IPv6 runtime
+  clib_bihash_48_8_t hash;
+
+  // per-thread data
+  ip6_sv_reass_per_thread_t *per_thread_data;
+
+  // convenience
+  vlib_main_t *vlib_main;
+  vnet_main_t *vnet_main;
+
+  // node index of ip6-drop node
+  u32 ip6_drop_idx;
+  u32 ip6_icmp_error_idx;
+  u32 ip6_sv_reass_expire_node_idx;
+
+  /** Worker handoff */
+  u32 fq_index;
+  u32 fq_feature_index;
+
+  // reference count for enabling/disabling feature
+  u32 feature_use_refcount;
+} ip6_sv_reass_main_t;
+
+extern ip6_sv_reass_main_t ip6_sv_reass_main;
+
+#ifndef CLIB_MARCH_VARIANT
+ip6_sv_reass_main_t ip6_sv_reass_main;
+#endif /* CLIB_MARCH_VARIANT */
+
+typedef enum
+{
+  IP6_SV_REASSEMBLY_NEXT_INPUT,
+  IP6_SV_REASSEMBLY_NEXT_DROP,
+  IP6_SV_REASSEMBLY_NEXT_ICMP_ERROR,
+  IP6_SV_REASSEMBLY_NEXT_HANDOFF,
+  IP6_SV_REASSEMBLY_N_NEXT,
+} ip6_sv_reass_next_t;
+
+typedef enum
+{
+  ICMP_ERROR_FL_TOO_BIG,
+  ICMP_ERROR_FL_NOT_MULT_8,
+  REASS_FRAGMENT_CACHE,
+  REASS_FINISH,
+  REASS_FRAGMENT_FORWARD,
+} ip6_sv_reass_trace_operation_e;
+
+typedef struct
+{
+  ip6_sv_reass_trace_operation_e action;
+  u32 reass_id;
+  u32 op_id;
+  u8 ip_proto;
+  u16 l4_src_port;
+  u16 l4_dst_port;
+} ip6_sv_reass_trace_t;
+
+static u8 *
+format_ip6_sv_reass_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 *);
+  ip6_sv_reass_trace_t *t = va_arg (*args, ip6_sv_reass_trace_t *);
+  s = format (s, "reass id: %u, op id: %u ", t->reass_id, t->op_id);
+  switch (t->action)
+    {
+    case ICMP_ERROR_FL_TOO_BIG:
+      s = format (s, "\n%Uicmp-error - frag_len > 65535 %U");
+      break;
+    case ICMP_ERROR_FL_NOT_MULT_8:
+      s = format (s, "\n%Uicmp-error - frag_len mod 8 != 0 %U");
+      break;
+    case REASS_FRAGMENT_CACHE:
+      s = format (s, "[cached]");
+      break;
+    case REASS_FINISH:
+      s =
+	format (s, "[finish, ip proto=%u, src_port=%u, dst_port=%u]",
+		t->ip_proto, clib_net_to_host_u16 (t->l4_src_port),
+		clib_net_to_host_u16 (t->l4_dst_port));
+      break;
+    case REASS_FRAGMENT_FORWARD:
+      s =
+	format (s, "[forward, ip proto=%u, src_port=%u, dst_port=%u]",
+		t->ip_proto, clib_net_to_host_u16 (t->l4_src_port),
+		clib_net_to_host_u16 (t->l4_dst_port));
+      break;
+    }
+  return s;
+}
+
+static void
+ip6_sv_reass_add_trace (vlib_main_t * vm, vlib_node_runtime_t * node,
+			ip6_sv_reass_main_t * rm,
+			ip6_sv_reass_t * reass, u32 bi,
+			ip6_sv_reass_trace_operation_e action,
+			u32 ip_proto, u16 l4_src_port, u16 l4_dst_port)
+{
+  vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+  ip6_sv_reass_trace_t *t = vlib_add_trace (vm, node, b, sizeof (t[0]));
+  t->reass_id = reass->id;
+  t->action = action;
+  t->op_id = reass->trace_op_counter;
+  t->ip_proto = ip_proto;
+  t->l4_src_port = l4_src_port;
+  t->l4_dst_port = l4_dst_port;
+  ++reass->trace_op_counter;
+#if 0
+  static u8 *s = NULL;
+  s = format (s, "%U", format_ip6_sv_reass_trace, NULL, NULL, t);
+  printf ("%.*s\n", vec_len (s), s);
+  fflush (stdout);
+  vec_reset_length (s);
+#endif
+}
+
+always_inline void
+ip6_sv_reass_free (vlib_main_t * vm, ip6_sv_reass_main_t * rm,
+		   ip6_sv_reass_per_thread_t * rt, ip6_sv_reass_t * reass)
+{
+  clib_bihash_kv_48_8_t kv;
+  kv.key[0] = reass->key.as_u64[0];
+  kv.key[1] = reass->key.as_u64[1];
+  kv.key[2] = reass->key.as_u64[2];
+  kv.key[3] = reass->key.as_u64[3];
+  kv.key[4] = reass->key.as_u64[4];
+  kv.key[5] = reass->key.as_u64[5];
+  clib_bihash_add_del_48_8 (&rm->hash, &kv, 0);
+  vlib_buffer_free (vm, reass->cached_buffers,
+		    vec_len (reass->cached_buffers));
+  vec_free (reass->cached_buffers);
+  reass->cached_buffers = NULL;
+  if (~0 != reass->lru_prev)
+    {
+      ip6_sv_reass_t *lru_prev =
+	pool_elt_at_index (rt->pool, reass->lru_prev);
+      lru_prev->lru_next = reass->lru_next;
+    }
+  if (~0 != reass->lru_next)
+    {
+      ip6_sv_reass_t *lru_next =
+	pool_elt_at_index (rt->pool, reass->lru_next);
+      lru_next->lru_prev = reass->lru_prev;
+    }
+  if (rt->lru_first == reass - rt->pool)
+    {
+      rt->lru_first = reass->lru_next;
+    }
+  if (rt->lru_last == reass - rt->pool)
+    {
+      rt->lru_last = reass->lru_prev;
+    }
+  pool_put (rt->pool, reass);
+  --rt->reass_n;
+}
+
+always_inline void
+ip6_sv_reass_init (ip6_sv_reass_t * reass)
+{
+  reass->cached_buffers = NULL;
+  reass->is_complete = false;
+}
+
+always_inline ip6_sv_reass_t *
+ip6_sv_reass_find_or_create (vlib_main_t * vm, vlib_node_runtime_t * node,
+			     ip6_sv_reass_main_t * rm,
+			     ip6_sv_reass_per_thread_t * rt,
+			     ip6_sv_reass_kv_t * kv, u32 * icmp_bi,
+			     u8 * do_handoff)
+{
+  ip6_sv_reass_t *reass = NULL;
+  f64 now = vlib_time_now (rm->vlib_main);
+
+  if (!clib_bihash_search_48_8
+      (&rm->hash, (clib_bihash_kv_48_8_t *) kv, (clib_bihash_kv_48_8_t *) kv))
+    {
+      if (vm->thread_index != kv->v.thread_index)
+	{
+	  *do_handoff = 1;
+	  return NULL;
+	}
+      reass = pool_elt_at_index (rt->pool, kv->v.reass_index);
+
+      if (now > reass->last_heard + rm->timeout)
+	{
+	  ip6_sv_reass_free (vm, rm, rt, reass);
+	  reass = NULL;
+	}
+    }
+
+  if (reass)
+    {
+      reass->last_heard = now;
+      return reass;
+    }
+
+  if (rt->reass_n >= rm->max_reass_n)
+    {
+      reass = pool_elt_at_index (rt->pool, rt->lru_last);
+      ip6_sv_reass_free (vm, rm, rt, reass);
+    }
+
+  pool_get (rt->pool, reass);
+  clib_memset (reass, 0, sizeof (*reass));
+  reass->id = ((u64) vm->thread_index * 1000000000) + rt->id_counter;
+  ++rt->id_counter;
+  ip6_sv_reass_init (reass);
+  ++rt->reass_n;
+
+  reass->lru_prev = reass->lru_next = ~0;
+
+  if (~0 != rt->lru_last)
+    {
+      ip6_sv_reass_t *lru_last = pool_elt_at_index (rt->pool, rt->lru_last);
+      reass->lru_prev = rt->lru_last;
+      lru_last->lru_next = rt->lru_last = reass - rt->pool;
+    }
+
+  if (~0 == rt->lru_first)
+    {
+      rt->lru_first = rt->lru_last = reass - rt->pool;
+    }
+
+  reass->key.as_u64[0] = ((clib_bihash_kv_48_8_t *) kv)->key[0];
+  reass->key.as_u64[1] = ((clib_bihash_kv_48_8_t *) kv)->key[1];
+  reass->key.as_u64[2] = ((clib_bihash_kv_48_8_t *) kv)->key[2];
+  reass->key.as_u64[3] = ((clib_bihash_kv_48_8_t *) kv)->key[3];
+  reass->key.as_u64[4] = ((clib_bihash_kv_48_8_t *) kv)->key[4];
+  reass->key.as_u64[5] = ((clib_bihash_kv_48_8_t *) kv)->key[5];
+  kv->v.reass_index = (reass - rt->pool);
+  kv->v.thread_index = vm->thread_index;
+  reass->last_heard = now;
+
+  if (clib_bihash_add_del_48_8 (&rm->hash, (clib_bihash_kv_48_8_t *) kv, 1))
+    {
+      ip6_sv_reass_free (vm, rm, rt, reass);
+      reass = NULL;
+    }
+
+  return reass;
+}
+
+always_inline ip6_sv_reass_rc_t
+ip6_sv_reass_update (vlib_main_t * vm, vlib_node_runtime_t * node,
+		     ip6_sv_reass_main_t * rm, ip6_sv_reass_per_thread_t * rt,
+		     ip6_sv_reass_t * reass, u32 bi0,
+		     ip6_frag_hdr_t * frag_hdr)
+{
+  vlib_buffer_t *fb = vlib_get_buffer (vm, bi0);
+  vnet_buffer_opaque_t *fvnb = vnet_buffer (fb);
+  fvnb->ip.reass.ip6_frag_hdr_offset =
+    (u8 *) frag_hdr - (u8 *) vlib_buffer_get_current (fb);
+  ip6_header_t *fip = vlib_buffer_get_current (fb);
+  if (fb->current_length < sizeof (*fip) ||
+      fvnb->ip.reass.ip6_frag_hdr_offset == 0 ||
+      fvnb->ip.reass.ip6_frag_hdr_offset >= fb->current_length)
+    {
+      return IP6_SV_REASS_RC_INTERNAL_ERROR;
+    }
+
+  u32 fragment_first = fvnb->ip.reass.fragment_first =
+    ip6_frag_hdr_offset_bytes (frag_hdr);
+  u32 fragment_length =
+    vlib_buffer_length_in_chain (vm, fb) -
+    (fvnb->ip.reass.ip6_frag_hdr_offset + sizeof (*frag_hdr));
+  u32 fragment_last = fvnb->ip.reass.fragment_last =
+    fragment_first + fragment_length - 1;
+  fvnb->ip.reass.range_first = fragment_first;
+  fvnb->ip.reass.range_last = fragment_last;
+  fvnb->ip.reass.next_range_bi = ~0;
+  if (0 == fragment_first)
+    {
+      ip6_ext_header_t *ext_hdr = (void *) frag_hdr;
+      while (ip6_ext_hdr (ext_hdr->next_hdr)
+	     && vlib_object_within_buffer_data (vm, fb, ext_hdr,
+						ext_hdr->n_data_u64s * 8))
+	{
+	  ext_hdr = ip6_ext_next_header (ext_hdr);
+	}
+      reass->ip_proto = ext_hdr->next_hdr;
+      reass->l4_src_port = ip6_get_port (fip, 1, fb->current_length);
+      reass->l4_dst_port = ip6_get_port (fip, 0, fb->current_length);
+      if (!reass->l4_src_port || !reass->l4_dst_port)
+	return IP6_SV_REASS_RC_UNSUPP_IP_PROTO;
+      reass->is_complete = true;
+      vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
+      if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+	{
+	  ip6_sv_reass_add_trace (vm, node, rm, reass, bi0, REASS_FINISH,
+				  reass->ip_proto, reass->l4_src_port,
+				  reass->l4_dst_port);
+	}
+    }
+  vec_add1 (reass->cached_buffers, bi0);
+  if (!reass->is_complete)
+    {
+      if (PREDICT_FALSE (fb->flags & VLIB_BUFFER_IS_TRACED))
+	{
+	  ip6_sv_reass_add_trace (vm, node, rm, reass, bi0,
+				  REASS_FRAGMENT_CACHE, reass->ip_proto,
+				  reass->l4_src_port, reass->l4_dst_port);
+	}
+      if (vec_len (reass->cached_buffers) > rm->max_reass_len)
+	{
+	  return IP6_SV_REASS_RC_TOO_MANY_FRAGMENTS;
+	}
+    }
+  return IP6_SV_REASS_RC_OK;
+}
+
+always_inline bool
+ip6_sv_reass_verify_upper_layer_present (vlib_node_runtime_t * node,
+					 vlib_buffer_t * b,
+					 ip6_frag_hdr_t * frag_hdr)
+{
+  ip6_ext_header_t *tmp = (ip6_ext_header_t *) frag_hdr;
+  while (ip6_ext_hdr (tmp->next_hdr))
+    {
+      tmp = ip6_ext_next_header (tmp);
+    }
+  if (IP_PROTOCOL_IP6_NONXT == tmp->next_hdr)
+    {
+      icmp6_error_set_vnet_buffer (b, ICMP6_parameter_problem,
+				   ICMP6_parameter_problem_first_fragment_has_incomplete_header_chain,
+				   0);
+      b->error = node->errors[IP6_ERROR_REASS_MISSING_UPPER];
+
+      return false;
+    }
+  return true;
+}
+
+always_inline bool
+ip6_sv_reass_verify_fragment_multiple_8 (vlib_main_t * vm,
+					 vlib_node_runtime_t * node,
+					 vlib_buffer_t * b,
+					 ip6_frag_hdr_t * frag_hdr)
+{
+  vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+  ip6_header_t *ip = vlib_buffer_get_current (b);
+  int more_fragments = ip6_frag_hdr_more (frag_hdr);
+  u32 fragment_length =
+    vlib_buffer_length_in_chain (vm, b) -
+    (vnb->ip.reass.ip6_frag_hdr_offset + sizeof (*frag_hdr));
+  if (more_fragments && 0 != fragment_length % 8)
+    {
+      icmp6_error_set_vnet_buffer (b, ICMP6_parameter_problem,
+				   ICMP6_parameter_problem_erroneous_header_field,
+				   (u8 *) & ip->payload_length - (u8 *) ip);
+      return false;
+    }
+  return true;
+}
+
+always_inline bool
+ip6_sv_reass_verify_packet_size_lt_64k (vlib_main_t * vm,
+					vlib_node_runtime_t * node,
+					vlib_buffer_t * b,
+					ip6_frag_hdr_t * frag_hdr)
+{
+  vnet_buffer_opaque_t *vnb = vnet_buffer (b);
+  u32 fragment_first = ip6_frag_hdr_offset_bytes (frag_hdr);
+  u32 fragment_length =
+    vlib_buffer_length_in_chain (vm, b) -
+    (vnb->ip.reass.ip6_frag_hdr_offset + sizeof (*frag_hdr));
+  if (fragment_first + fragment_length > 65535)
+    {
+      ip6_header_t *ip0 = vlib_buffer_get_current (b);
+      icmp6_error_set_vnet_buffer (b, ICMP6_parameter_problem,
+				   ICMP6_parameter_problem_erroneous_header_field,
+				   (u8 *) & frag_hdr->fragment_offset_and_more
+				   - (u8 *) ip0);
+      return false;
+    }
+  return true;
+}
+
+always_inline uword
+ip6_sv_reassembly_inline (vlib_main_t * vm,
+			  vlib_node_runtime_t * node,
+			  vlib_frame_t * frame, bool is_feature)
+{
+  u32 *from = vlib_frame_vector_args (frame);
+  u32 n_left_from, n_left_to_next, *to_next, next_index;
+  ip6_sv_reass_main_t *rm = &ip6_sv_reass_main;
+  ip6_sv_reass_per_thread_t *rt = &rm->per_thread_data[vm->thread_index];
+  clib_spinlock_lock (&rt->lock);
+
+  n_left_from = frame->n_vectors;
+  next_index = node->cached_next_index;
+
+  while (n_left_from > 0)
+    {
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+
+      while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	  u32 bi0;
+	  vlib_buffer_t *b0;
+	  u32 next0 = IP6_SV_REASSEMBLY_NEXT_DROP;
+	  u32 error0 = IP6_ERROR_NONE;
+	  u32 icmp_bi = ~0;
+
+	  bi0 = from[0];
+	  b0 = vlib_get_buffer (vm, bi0);
+
+	  ip6_header_t *ip0 = vlib_buffer_get_current (b0);
+	  ip6_frag_hdr_t *frag_hdr = NULL;
+	  ip6_ext_header_t *prev_hdr;
+	  if (ip6_ext_hdr (ip0->protocol))
+	    {
+	      frag_hdr =
+		ip6_ext_header_find (vm, b0, ip0,
+				     IP_PROTOCOL_IPV6_FRAGMENTATION,
+				     &prev_hdr);
+	    }
+	  if (!frag_hdr)
+	    {
+	      // this is a regular packet - no fragmentation
+	      vnet_buffer (b0)->ip.reass.ip_proto = ip0->protocol;
+	      vnet_buffer (b0)->ip.reass.l4_src_port =
+		ip6_get_port (ip0, 1, b0->current_length);
+	      vnet_buffer (b0)->ip.reass.l4_dst_port =
+		ip6_get_port (ip0, 0, b0->current_length);
+	      next0 = IP6_SV_REASSEMBLY_NEXT_INPUT;
+	      goto packet_enqueue;
+	    }
+	  if (0 == ip6_frag_hdr_offset (frag_hdr))
+	    {
+	      // first fragment - verify upper-layer is present
+	      if (!ip6_sv_reass_verify_upper_layer_present
+		  (node, b0, frag_hdr))
+		{
+		  next0 = IP6_SV_REASSEMBLY_NEXT_ICMP_ERROR;
+		  goto packet_enqueue;
+		}
+	    }
+	  if (!ip6_sv_reass_verify_fragment_multiple_8
+	      (vm, node, b0, frag_hdr)
+	      || !ip6_sv_reass_verify_packet_size_lt_64k (vm, node, b0,
+							  frag_hdr))
+	    {
+	      next0 = IP6_SV_REASSEMBLY_NEXT_ICMP_ERROR;
+	      goto packet_enqueue;
+	    }
+	  vnet_buffer (b0)->ip.reass.ip6_frag_hdr_offset =
+	    (u8 *) frag_hdr - (u8 *) ip0;
+
+	  ip6_sv_reass_kv_t kv;
+	  u8 do_handoff = 0;
+
+	  kv.k.as_u64[0] = ip0->src_address.as_u64[0];
+	  kv.k.as_u64[1] = ip0->src_address.as_u64[1];
+	  kv.k.as_u64[2] = ip0->dst_address.as_u64[0];
+	  kv.k.as_u64[3] = ip0->dst_address.as_u64[1];
+	  kv.k.as_u64[4] =
+	    ((u64) vec_elt (ip6_main.fib_index_by_sw_if_index,
+			    vnet_buffer (b0)->sw_if_index[VLIB_RX])) << 32 |
+	    (u64) frag_hdr->identification;
+	  kv.k.as_u64[5] = ip0->protocol;
+
+	  ip6_sv_reass_t *reass =
+	    ip6_sv_reass_find_or_create (vm, node, rm, rt, &kv, &icmp_bi,
+					 &do_handoff);
+
+	  if (PREDICT_FALSE (do_handoff))
+	    {
+	      next0 = IP6_SV_REASSEMBLY_NEXT_HANDOFF;
+	      vnet_buffer (b0)->ip.reass.owner_thread_index =
+		kv.v.thread_index;
+	    }
+
+	  if (!reass)
+	    {
+	      next0 = IP6_SV_REASSEMBLY_NEXT_DROP;
+	      error0 = IP6_ERROR_REASS_LIMIT_REACHED;
+	      goto packet_enqueue;
+	    }
+
+	  if (reass->is_complete)
+	    {
+	      vnet_buffer (b0)->ip.reass.ip_proto = reass->ip_proto;
+	      vnet_buffer (b0)->ip.reass.l4_src_port = reass->l4_src_port;
+	      vnet_buffer (b0)->ip.reass.l4_dst_port = reass->l4_dst_port;
+	      next0 = IP6_SV_REASSEMBLY_NEXT_INPUT;
+	      error0 = IP6_ERROR_NONE;
+	      if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+		{
+		  ip6_sv_reass_add_trace (vm, node, rm, reass, bi0,
+					  REASS_FRAGMENT_FORWARD,
+					  reass->ip_proto,
+					  reass->l4_src_port,
+					  reass->l4_dst_port);
+		}
+	      goto packet_enqueue;
+	    }
+
+	  switch (ip6_sv_reass_update
+		  (vm, node, rm, rt, reass, bi0, frag_hdr))
+	    {
+	    case IP6_SV_REASS_RC_OK:
+	      /* nothing to do here */
+	      break;
+	    case IP6_SV_REASS_RC_TOO_MANY_FRAGMENTS:
+	      vlib_node_increment_counter (vm, node->node_index,
+					   IP6_ERROR_REASS_FRAGMENT_CHAIN_TOO_LONG,
+					   1);
+	      ip6_sv_reass_free (vm, rm, rt, reass);
+	      goto next_packet;
+	      break;
+	    case IP6_SV_REASS_RC_UNSUPP_IP_PROTO:
+	      vlib_node_increment_counter (vm, node->node_index,
+					   IP6_ERROR_REASS_UNSUPP_IP_PROTO,
+					   1);
+	      ip6_sv_reass_free (vm, rm, rt, reass);
+	      goto next_packet;
+	      break;
+	    case IP6_SV_REASS_RC_INTERNAL_ERROR:
+	      vlib_node_increment_counter (vm, node->node_index,
+					   IP6_ERROR_REASS_INTERNAL_ERROR, 1);
+	      ip6_sv_reass_free (vm, rm, rt, reass);
+	      goto next_packet;
+	      break;
+	    }
+
+	  b0->error = node->errors[error0];
+
+	  if (reass->is_complete)
+	    {
+	      u32 idx;
+	      vec_foreach_index (idx, reass->cached_buffers)
+	      {
+		u32 bi0 = vec_elt (reass->cached_buffers, idx);
+		if (0 == n_left_to_next)
+		  {
+		    vlib_put_next_frame (vm, node, next_index,
+					 n_left_to_next);
+		    vlib_get_next_frame (vm, node, next_index, to_next,
+					 n_left_to_next);
+		  }
+		to_next[0] = bi0;
+		to_next += 1;
+		n_left_to_next -= 1;
+		b0 = vlib_get_buffer (vm, bi0);
+		if (is_feature)
+		  {
+		    vnet_feature_next (&next0, b0);
+		  }
+		vnet_buffer (b0)->ip.reass.ip_proto = reass->ip_proto;
+		vnet_buffer (b0)->ip.reass.l4_src_port = reass->l4_src_port;
+		vnet_buffer (b0)->ip.reass.l4_dst_port = reass->l4_dst_port;
+		if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
+		  {
+		    ip6_sv_reass_add_trace (vm, node, rm, reass, bi0,
+					    REASS_FRAGMENT_FORWARD,
+					    reass->ip_proto,
+					    reass->l4_src_port,
+					    reass->l4_dst_port);
+		  }
+		vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
+						 to_next, n_left_to_next, bi0,
+						 next0);
+	      }
+	      _vec_len (reass->cached_buffers) = 0;	// buffers are owned by frame now
+	    }
+	  goto next_packet;
+
+	packet_enqueue:
+	  to_next[0] = bi0;
+	  to_next += 1;
+	  n_left_to_next -= 1;
+	  if (is_feature && IP6_ERROR_NONE == error0)
+	    {
+	      b0 = vlib_get_buffer (vm, bi0);
+	      vnet_feature_next (&next0, b0);
+	    }
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+
+	  if (~0 != icmp_bi)
+	    {
+	      next0 = IP6_SV_REASSEMBLY_NEXT_ICMP_ERROR;
+	      to_next[0] = icmp_bi;
+	      to_next += 1;
+	      n_left_to_next -= 1;
+	      vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					       n_left_to_next, icmp_bi,
+					       next0);
+	    }
+
+	next_packet:
+	  from += 1;
+	  n_left_from -= 1;
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  clib_spinlock_unlock (&rt->lock);
+  return frame->n_vectors;
+}
+
+static char *ip6_sv_reassembly_error_strings[] = {
+#define _(sym, string) string,
+  foreach_ip6_error
+#undef _
+};
+
+VLIB_NODE_FN (ip6_sv_reass_node) (vlib_main_t * vm,
+				  vlib_node_runtime_t * node,
+				  vlib_frame_t * frame)
+{
+  return ip6_sv_reassembly_inline (vm, node, frame, false /* is_feature */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip6_sv_reass_node) = {
+    .name = "ip6-sv-reassembly",
+    .vector_size = sizeof (u32),
+    .format_trace = format_ip6_sv_reass_trace,
+    .n_errors = ARRAY_LEN (ip6_sv_reassembly_error_strings),
+    .error_strings = ip6_sv_reassembly_error_strings,
+    .n_next_nodes = IP6_SV_REASSEMBLY_N_NEXT,
+    .next_nodes =
+        {
+                [IP6_SV_REASSEMBLY_NEXT_INPUT] = "ip6-input",
+                [IP6_SV_REASSEMBLY_NEXT_DROP] = "ip6-drop",
+                [IP6_SV_REASSEMBLY_NEXT_ICMP_ERROR] = "ip6-icmp-error",
+                [IP6_SV_REASSEMBLY_NEXT_HANDOFF] = "ip6-sv-reassembly-handoff",
+        },
+};
+/* *INDENT-ON* */
+
+VLIB_NODE_FN (ip6_sv_reass_node_feature) (vlib_main_t * vm,
+					  vlib_node_runtime_t * node,
+					  vlib_frame_t * frame)
+{
+  return ip6_sv_reassembly_inline (vm, node, frame, true /* is_feature */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip6_sv_reass_node_feature) = {
+    .name = "ip6-sv-reassembly-feature",
+    .vector_size = sizeof (u32),
+    .format_trace = format_ip6_sv_reass_trace,
+    .n_errors = ARRAY_LEN (ip6_sv_reassembly_error_strings),
+    .error_strings = ip6_sv_reassembly_error_strings,
+    .n_next_nodes = IP6_SV_REASSEMBLY_N_NEXT,
+    .next_nodes =
+        {
+                [IP6_SV_REASSEMBLY_NEXT_INPUT] = "ip6-input",
+                [IP6_SV_REASSEMBLY_NEXT_DROP] = "ip6-drop",
+                [IP6_SV_REASSEMBLY_NEXT_ICMP_ERROR] = "ip6-icmp-error",
+                [IP6_SV_REASSEMBLY_NEXT_HANDOFF] = "ip6-sv-reass-feature-hoff",
+        },
+};
+/* *INDENT-ON* */
+
+/* *INDENT-OFF* */
+VNET_FEATURE_INIT (ip6_sv_reassembly_feature) = {
+    .arc_name = "ip6-unicast",
+    .node_name = "ip6-sv-reassembly-feature",
+    .runs_before = VNET_FEATURES ("ip6-lookup"),
+    .runs_after = 0,
+};
+/* *INDENT-ON* */
+
+#ifndef CLIB_MARCH_VARIANT
+static u32
+ip6_sv_reass_get_nbuckets ()
+{
+  ip6_sv_reass_main_t *rm = &ip6_sv_reass_main;
+  u32 nbuckets;
+  u8 i;
+
+  nbuckets = (u32) (rm->max_reass_n / IP6_SV_REASS_HT_LOAD_FACTOR);
+
+  for (i = 0; i < 31; i++)
+    if ((1 << i) >= nbuckets)
+      break;
+  nbuckets = 1 << i;
+
+  return nbuckets;
+}
+#endif /* CLIB_MARCH_VARIANT */
+
+typedef enum
+{
+  IP6_EVENT_CONFIG_CHANGED = 1,
+} ip6_sv_reass_event_t;
+
+#ifndef CLIB_MARCH_VARIANT
+typedef struct
+{
+  int failure;
+  clib_bihash_48_8_t *new_hash;
+} ip6_rehash_cb_ctx;
+
+static void
+ip6_rehash_cb (clib_bihash_kv_48_8_t * kv, void *_ctx)
+{
+  ip6_rehash_cb_ctx *ctx = _ctx;
+  if (clib_bihash_add_del_48_8 (ctx->new_hash, kv, 1))
+    {
+      ctx->failure = 1;
+    }
+}
+
+static void
+ip6_sv_reass_set_params (u32 timeout_ms, u32 max_reassemblies,
+			 u32 max_reassembly_length,
+			 u32 expire_walk_interval_ms)
+{
+  ip6_sv_reass_main.timeout_ms = timeout_ms;
+  ip6_sv_reass_main.timeout = (f64) timeout_ms / (f64) MSEC_PER_SEC;
+  ip6_sv_reass_main.max_reass_n = max_reassemblies;
+  ip6_sv_reass_main.max_reass_len = max_reassembly_length;
+  ip6_sv_reass_main.expire_walk_interval_ms = expire_walk_interval_ms;
+}
+
+vnet_api_error_t
+ip6_sv_reass_set (u32 timeout_ms, u32 max_reassemblies,
+		  u32 max_reassembly_length, u32 expire_walk_interval_ms)
+{
+  u32 old_nbuckets = ip6_sv_reass_get_nbuckets ();
+  ip6_sv_reass_set_params (timeout_ms, max_reassemblies,
+			   max_reassembly_length, expire_walk_interval_ms);
+  vlib_process_signal_event (ip6_sv_reass_main.vlib_main,
+			     ip6_sv_reass_main.ip6_sv_reass_expire_node_idx,
+			     IP6_EVENT_CONFIG_CHANGED, 0);
+  u32 new_nbuckets = ip6_sv_reass_get_nbuckets ();
+  if (ip6_sv_reass_main.max_reass_n > 0 && new_nbuckets > old_nbuckets)
+    {
+      clib_bihash_48_8_t new_hash;
+      clib_memset (&new_hash, 0, sizeof (new_hash));
+      ip6_rehash_cb_ctx ctx;
+      ctx.failure = 0;
+      ctx.new_hash = &new_hash;
+      clib_bihash_init_48_8 (&new_hash, "ip6-sv-reass", new_nbuckets,
+			     new_nbuckets * 1024);
+      clib_bihash_foreach_key_value_pair_48_8 (&ip6_sv_reass_main.hash,
+					       ip6_rehash_cb, &ctx);
+      if (ctx.failure)
+	{
+	  clib_bihash_free_48_8 (&new_hash);
+	  return -1;
+	}
+      else
+	{
+	  clib_bihash_free_48_8 (&ip6_sv_reass_main.hash);
+	  clib_memcpy_fast (&ip6_sv_reass_main.hash, &new_hash,
+			    sizeof (ip6_sv_reass_main.hash));
+	  clib_bihash_copied (&ip6_sv_reass_main.hash, &new_hash);
+	}
+    }
+  return 0;
+}
+
+vnet_api_error_t
+ip6_sv_reass_get (u32 * timeout_ms, u32 * max_reassemblies,
+		  u32 * max_reassembly_length, u32 * expire_walk_interval_ms)
+{
+  *timeout_ms = ip6_sv_reass_main.timeout_ms;
+  *max_reassemblies = ip6_sv_reass_main.max_reass_n;
+  *max_reassembly_length = ip6_sv_reass_main.max_reass_len;
+  *expire_walk_interval_ms = ip6_sv_reass_main.expire_walk_interval_ms;
+  return 0;
+}
+
+static clib_error_t *
+ip6_sv_reass_init_function (vlib_main_t * vm)
+{
+  ip6_sv_reass_main_t *rm = &ip6_sv_reass_main;
+  clib_error_t *error = 0;
+  u32 nbuckets;
+  vlib_node_t *node;
+
+  rm->vlib_main = vm;
+  rm->vnet_main = vnet_get_main ();
+
+  vec_validate (rm->per_thread_data, vlib_num_workers ());
+  ip6_sv_reass_per_thread_t *rt;
+  vec_foreach (rt, rm->per_thread_data)
+  {
+    clib_spinlock_init (&rt->lock);
+    pool_alloc (rt->pool, rm->max_reass_n);
+    rt->lru_first = rt->lru_last = ~0;
+  }
+
+  node = vlib_get_node_by_name (vm, (u8 *) "ip6-sv-reassembly-expire-walk");
+  ASSERT (node);
+  rm->ip6_sv_reass_expire_node_idx = node->index;
+
+  ip6_sv_reass_set_params (IP6_SV_REASS_TIMEOUT_DEFAULT_MS,
+			   IP6_SV_REASS_MAX_REASSEMBLIES_DEFAULT,
+			   IP6_SV_REASS_MAX_REASSEMBLY_LENGTH_DEFAULT,
+			   IP6_SV_REASS_EXPIRE_WALK_INTERVAL_DEFAULT_MS);
+
+  nbuckets = ip6_sv_reass_get_nbuckets ();
+  clib_bihash_init_48_8 (&rm->hash, "ip6-sv-reass", nbuckets,
+			 nbuckets * 1024);
+
+  node = vlib_get_node_by_name (vm, (u8 *) "ip6-drop");
+  ASSERT (node);
+  rm->ip6_drop_idx = node->index;
+  node = vlib_get_node_by_name (vm, (u8 *) "ip6-icmp-error");
+  ASSERT (node);
+  rm->ip6_icmp_error_idx = node->index;
+
+  if ((error = vlib_call_init_function (vm, ip_main_init)))
+    return error;
+  ip6_register_protocol (IP_PROTOCOL_IPV6_FRAGMENTATION,
+			 ip6_sv_reass_node.index);
+
+  rm->fq_index = vlib_frame_queue_main_init (ip6_sv_reass_node.index, 0);
+  rm->fq_feature_index =
+    vlib_frame_queue_main_init (ip6_sv_reass_node_feature.index, 0);
+
+  return error;
+}
+
+VLIB_INIT_FUNCTION (ip6_sv_reass_init_function);
+#endif /* CLIB_MARCH_VARIANT */
+
+static uword
+ip6_sv_reass_walk_expired (vlib_main_t * vm,
+			   vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+  ip6_sv_reass_main_t *rm = &ip6_sv_reass_main;
+  uword event_type, *event_data = 0;
+
+  while (true)
+    {
+      vlib_process_wait_for_event_or_clock (vm,
+					    (f64) rm->expire_walk_interval_ms
+					    / (f64) MSEC_PER_SEC);
+      event_type = vlib_process_get_events (vm, &event_data);
+
+      switch (event_type)
+	{
+	case ~0:		/* no events => timeout */
+	  /* nothing to do here */
+	  break;
+	case IP6_EVENT_CONFIG_CHANGED:
+	  break;
+	default:
+	  clib_warning ("BUG: event type 0x%wx", event_type);
+	  break;
+	}
+      f64 now = vlib_time_now (vm);
+
+      ip6_sv_reass_t *reass;
+      int *pool_indexes_to_free = NULL;
+
+      uword thread_index = 0;
+      int index;
+      const uword nthreads = vlib_num_workers () + 1;
+      u32 *vec_icmp_bi = NULL;
+      for (thread_index = 0; thread_index < nthreads; ++thread_index)
+	{
+	  ip6_sv_reass_per_thread_t *rt = &rm->per_thread_data[thread_index];
+	  clib_spinlock_lock (&rt->lock);
+
+	  vec_reset_length (pool_indexes_to_free);
+          /* *INDENT-OFF* */
+          pool_foreach_index (index, rt->pool, ({
+                                reass = pool_elt_at_index (rt->pool, index);
+                                if (now > reass->last_heard + rm->timeout)
+                                  {
+                                    vec_add1 (pool_indexes_to_free, index);
+                                  }
+                              }));
+          /* *INDENT-ON* */
+	  int *i;
+          /* *INDENT-OFF* */
+          vec_foreach (i, pool_indexes_to_free)
+          {
+            ip6_sv_reass_t *reass = pool_elt_at_index (rt->pool, i[0]);
+            ip6_sv_reass_free (vm, rm, rt, reass);
+          }
+          /* *INDENT-ON* */
+
+	  clib_spinlock_unlock (&rt->lock);
+	}
+
+      while (vec_len (vec_icmp_bi) > 0)
+	{
+	  vlib_frame_t *f =
+	    vlib_get_frame_to_node (vm, rm->ip6_icmp_error_idx);
+	  u32 *to_next = vlib_frame_vector_args (f);
+	  u32 n_left_to_next = VLIB_FRAME_SIZE - f->n_vectors;
+	  int trace_frame = 0;
+	  while (vec_len (vec_icmp_bi) > 0 && n_left_to_next > 0)
+	    {
+	      u32 bi = vec_pop (vec_icmp_bi);
+	      vlib_buffer_t *b = vlib_get_buffer (vm, bi);
+	      if (PREDICT_FALSE (b->flags & VLIB_BUFFER_IS_TRACED))
+		{
+		  trace_frame = 1;
+		}
+	      b->error = node->errors[IP6_ERROR_REASS_TIMEOUT];
+	      to_next[0] = bi;
+	      ++f->n_vectors;
+	      to_next += 1;
+	      n_left_to_next -= 1;
+	    }
+	  f->frame_flags |= (trace_frame * VLIB_FRAME_TRACE);
+	  vlib_put_frame_to_node (vm, rm->ip6_icmp_error_idx, f);
+	}
+
+      vec_free (pool_indexes_to_free);
+      vec_free (vec_icmp_bi);
+      if (event_data)
+	{
+	  _vec_len (event_data) = 0;
+	}
+    }
+
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip6_sv_reass_expire_node) = {
+    .function = ip6_sv_reass_walk_expired,
+    .format_trace = format_ip6_sv_reass_trace,
+    .type = VLIB_NODE_TYPE_PROCESS,
+    .name = "ip6-sv-reassembly-expire-walk",
+
+    .n_errors = ARRAY_LEN (ip6_sv_reassembly_error_strings),
+    .error_strings = ip6_sv_reassembly_error_strings,
+
+};
+/* *INDENT-ON* */
+
+static u8 *
+format_ip6_sv_reass_key (u8 * s, va_list * args)
+{
+  ip6_sv_reass_key_t *key = va_arg (*args, ip6_sv_reass_key_t *);
+  s = format (s, "xx_id: %u, src: %U, dst: %U, frag_id: %u, proto: %u",
+	      key->xx_id, format_ip6_address, &key->src, format_ip6_address,
+	      &key->dst, clib_net_to_host_u16 (key->frag_id), key->proto);
+  return s;
+}
+
+static u8 *
+format_ip6_sv_reass (u8 * s, va_list * args)
+{
+  vlib_main_t *vm = va_arg (*args, vlib_main_t *);
+  ip6_sv_reass_t *reass = va_arg (*args, ip6_sv_reass_t *);
+
+  s = format (s, "ID: %lu, key: %U, trace_op_counter: %u\n",
+	      reass->id, format_ip6_sv_reass_key, &reass->key,
+	      reass->trace_op_counter);
+  vlib_buffer_t *b;
+  u32 *bip;
+  u32 counter = 0;
+  vec_foreach (bip, reass->cached_buffers)
+  {
+    u32 bi = *bip;
+    do
+      {
+	b = vlib_get_buffer (vm, bi);
+	s = format (s, "  #%03u: bi: %u\n", counter, bi);
+	++counter;
+	bi = b->next_buffer;
+      }
+    while (b->flags & VLIB_BUFFER_NEXT_PRESENT);
+  }
+  return s;
+}
+
+static clib_error_t *
+show_ip6_sv_reass (vlib_main_t * vm, unformat_input_t * input,
+		   CLIB_UNUSED (vlib_cli_command_t * lmd))
+{
+  ip6_sv_reass_main_t *rm = &ip6_sv_reass_main;
+
+  vlib_cli_output (vm, "---------------------");
+  vlib_cli_output (vm, "IP6 reassembly status");
+  vlib_cli_output (vm, "---------------------");
+  bool details = false;
+  if (unformat (input, "details"))
+    {
+      details = true;
+    }
+
+  u32 sum_reass_n = 0;
+  u64 sum_buffers_n = 0;
+  ip6_sv_reass_t *reass;
+  uword thread_index;
+  const uword nthreads = vlib_num_workers () + 1;
+  for (thread_index = 0; thread_index < nthreads; ++thread_index)
+    {
+      ip6_sv_reass_per_thread_t *rt = &rm->per_thread_data[thread_index];
+      clib_spinlock_lock (&rt->lock);
+      if (details)
+	{
+          /* *INDENT-OFF* */
+          pool_foreach (reass, rt->pool, {
+            vlib_cli_output (vm, "%U", format_ip6_sv_reass, vm, reass);
+          });
+          /* *INDENT-ON* */
+	}
+      sum_reass_n += rt->reass_n;
+      clib_spinlock_unlock (&rt->lock);
+    }
+  vlib_cli_output (vm, "---------------------");
+  vlib_cli_output (vm, "Current IP6 reassemblies count: %lu\n",
+		   (long unsigned) sum_reass_n);
+  vlib_cli_output (vm, "Maximum configured concurrent IP6 reassemblies per "
+		   "worker-thread: %lu\n", (long unsigned) rm->max_reass_n);
+  vlib_cli_output (vm, "Buffers in use: %lu\n",
+		   (long unsigned) sum_buffers_n);
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (show_ip6_sv_reassembly_cmd, static) = {
+    .path = "show ip6-sv-reassembly",
+    .short_help = "show ip6-sv-reassembly [details]",
+    .function = show_ip6_sv_reass,
+};
+/* *INDENT-ON* */
+
+#ifndef CLIB_MARCH_VARIANT
+vnet_api_error_t
+ip6_sv_reass_enable_disable (u32 sw_if_index, u8 enable_disable)
+{
+  return vnet_feature_enable_disable ("ip6-unicast",
+				      "ip6-sv-reassembly-feature",
+				      sw_if_index, enable_disable, 0, 0);
+}
+#endif /* CLIB_MARCH_VARIANT */
+
+#define foreach_ip6_sv_reassembly_handoff_error                       \
+_(CONGESTION_DROP, "congestion drop")
+
+
+typedef enum
+{
+#define _(sym,str) IP6_SV_REASSEMBLY_HANDOFF_ERROR_##sym,
+  foreach_ip6_sv_reassembly_handoff_error
+#undef _
+    IP6_SV_REASSEMBLY_HANDOFF_N_ERROR,
+} ip6_sv_reassembly_handoff_error_t;
+
+static char *ip6_sv_reassembly_handoff_error_strings[] = {
+#define _(sym,string) string,
+  foreach_ip6_sv_reassembly_handoff_error
+#undef _
+};
+
+typedef struct
+{
+  u32 next_worker_index;
+} ip6_sv_reassembly_handoff_trace_t;
+
+static u8 *
+format_ip6_sv_reassembly_handoff_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 *);
+  ip6_sv_reassembly_handoff_trace_t *t =
+    va_arg (*args, ip6_sv_reassembly_handoff_trace_t *);
+
+  s =
+    format (s, "ip6-sv-reassembly-handoff: next-worker %d",
+	    t->next_worker_index);
+
+  return s;
+}
+
+always_inline uword
+ip6_sv_reassembly_handoff_inline (vlib_main_t * vm,
+				  vlib_node_runtime_t * node,
+				  vlib_frame_t * frame, bool is_feature)
+{
+  ip6_sv_reass_main_t *rm = &ip6_sv_reass_main;
+
+  vlib_buffer_t *bufs[VLIB_FRAME_SIZE], **b;
+  u32 n_enq, n_left_from, *from;
+  u16 thread_indices[VLIB_FRAME_SIZE], *ti;
+  u32 fq_index;
+
+  from = vlib_frame_vector_args (frame);
+  n_left_from = frame->n_vectors;
+  vlib_get_buffers (vm, from, bufs, n_left_from);
+
+  b = bufs;
+  ti = thread_indices;
+
+  fq_index = (is_feature) ? rm->fq_feature_index : rm->fq_index;
+
+  while (n_left_from > 0)
+    {
+      ti[0] = vnet_buffer (b[0])->ip.reass.owner_thread_index;
+
+      if (PREDICT_FALSE
+	  ((node->flags & VLIB_NODE_FLAG_TRACE)
+	   && (b[0]->flags & VLIB_BUFFER_IS_TRACED)))
+	{
+	  ip6_sv_reassembly_handoff_trace_t *t =
+	    vlib_add_trace (vm, node, b[0], sizeof (*t));
+	  t->next_worker_index = ti[0];
+	}
+
+      n_left_from -= 1;
+      ti += 1;
+      b += 1;
+    }
+  n_enq =
+    vlib_buffer_enqueue_to_thread (vm, fq_index, from, thread_indices,
+				   frame->n_vectors, 1);
+
+  if (n_enq < frame->n_vectors)
+    vlib_node_increment_counter (vm, node->node_index,
+				 IP6_SV_REASSEMBLY_HANDOFF_ERROR_CONGESTION_DROP,
+				 frame->n_vectors - n_enq);
+  return frame->n_vectors;
+}
+
+VLIB_NODE_FN (ip6_sv_reassembly_handoff_node) (vlib_main_t * vm,
+					       vlib_node_runtime_t * node,
+					       vlib_frame_t * frame)
+{
+  return ip6_sv_reassembly_handoff_inline (vm, node, frame,
+					   false /* is_feature */ );
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip6_sv_reassembly_handoff_node) = {
+  .name = "ip6-sv-reassembly-handoff",
+  .vector_size = sizeof (u32),
+  .n_errors = ARRAY_LEN(ip6_sv_reassembly_handoff_error_strings),
+  .error_strings = ip6_sv_reassembly_handoff_error_strings,
+  .format_trace = format_ip6_sv_reassembly_handoff_trace,
+
+  .n_next_nodes = 1,
+
+  .next_nodes = {
+    [0] = "error-drop",
+  },
+};
+
+
+VLIB_NODE_FN (ip6_sv_reassembly_feature_handoff_node) (vlib_main_t * vm,
+                               vlib_node_runtime_t * node, vlib_frame_t * frame)
+{
+  return ip6_sv_reassembly_handoff_inline (vm, node, frame, true /* is_feature */ );
+}
+
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (ip6_sv_reassembly_feature_handoff_node) = {
+  .name = "ip6-sv-reass-feature-hoff",
+  .vector_size = sizeof (u32),
+  .n_errors = ARRAY_LEN(ip6_sv_reassembly_handoff_error_strings),
+  .error_strings = ip6_sv_reassembly_handoff_error_strings,
+  .format_trace = format_ip6_sv_reassembly_handoff_trace,
+
+  .n_next_nodes = 1,
+
+  .next_nodes = {
+    [0] = "error-drop",
+  },
+};
+/* *INDENT-ON* */
+
+#ifndef CLIB_MARCH_VARIANT
+int
+ip6_sv_reass_enable_disable_with_refcnt (u32 sw_if_index, int is_enable)
+{
+  ip6_sv_reass_main_t *rm = &ip6_sv_reass_main;
+  if (is_enable)
+    {
+      if (!rm->feature_use_refcount)
+	{
+	  ++rm->feature_use_refcount;
+	  return vnet_feature_enable_disable ("ip6-unicast",
+					      "ip6-sv-reassembly-feature",
+					      sw_if_index, 1, 0, 0);
+	}
+      ++rm->feature_use_refcount;
+    }
+  else
+    {
+      --rm->feature_use_refcount;
+      if (!rm->feature_use_refcount)
+	return vnet_feature_enable_disable ("ip6-unicast",
+					    "ip6-sv-reassembly-feature",
+					    sw_if_index, 0, 0, 0);
+    }
+  return -1;
+}
+#endif
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/ip/reass/ip6_sv_reass.h b/src/vnet/ip/reass/ip6_sv_reass.h
new file mode 100644
index 00000000000..81ac2312bdf
--- /dev/null
+++ b/src/vnet/ip/reass/ip6_sv_reass.h
@@ -0,0 +1,56 @@
+/*
+ * Copyright (c) 2017 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.
+ */
+
+/**
+ * @file
+ * @brief IPv6 shallow virtual reassembly.
+ *
+ * This file contains the source code for IPv6 shallow virtual reassembly.
+ */
+
+#ifndef __included_ip6_sv_reass_h__
+#define __included_ip6_sv_reass_h__
+
+#include <vnet/api_errno.h>
+#include <vnet/vnet.h>
+
+/**
+ * @brief set ip6 reassembly configuration
+ */
+vnet_api_error_t ip6_sv_reass_set (u32 timeout_ms, u32 max_reassemblies,
+				   u32 max_reassembly_length,
+				   u32 expire_walk_interval_ms);
+
+/**
+ * @brief get ip6 reassembly configuration
+ */
+vnet_api_error_t ip6_sv_reass_get (u32 * timeout_ms, u32 * max_reassemblies,
+				   u32 * max_reassembly_length,
+				   u32 * expire_walk_interval_ms);
+
+vnet_api_error_t ip6_sv_reass_enable_disable (u32 sw_if_index,
+					      u8 enable_disable);
+
+int ip6_sv_reass_enable_disable_with_refcnt (u32 sw_if_index, int is_enable);
+
+#endif /* __included_ip6_sv_reass_h */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/test/test_reassembly.py b/test/test_reassembly.py
index 407b626e1d2..ee7830e1762 100644
--- a/test/test_reassembly.py
+++ b/test/test_reassembly.py
@@ -11,7 +11,7 @@ from scapy.packet import Raw
 from scapy.layers.l2 import Ether, GRE
 from scapy.layers.inet import IP, UDP, ICMP
 from scapy.layers.inet6 import HBHOptUnknown, ICMPv6ParamProblem,\
-        ICMPv6TimeExceeded, IPv6, IPv6ExtHdrFragment, IPv6ExtHdrHopByHop
+    ICMPv6TimeExceeded, IPv6, IPv6ExtHdrFragment, IPv6ExtHdrHopByHop
 from framework import VppTestCase, VppTestRunner
 from util import ppp, ppc, fragment_rfc791, fragment_rfc8200
 from vpp_gre_interface import VppGreInterface
@@ -503,6 +503,177 @@ class TestIPv4Reassembly(VppTestCase):
         self.src_if.assert_nothing_captured()
 
 
+class TestIPv4SVReassembly(VppTestCase):
+    """ IPv4 Shallow Virtual Reassembly """
+
+    @classmethod
+    def setUpClass(cls):
+        super(TestIPv4SVReassembly, cls).setUpClass()
+
+        cls.create_pg_interfaces([0, 1])
+        cls.src_if = cls.pg0
+        cls.dst_if = cls.pg1
+
+        # setup all interfaces
+        for i in cls.pg_interfaces:
+            i.admin_up()
+            i.config_ip4()
+            i.resolve_arp()
+
+    def setUp(self):
+        """ Test setup - force timeout on existing reassemblies """
+        super(TestIPv4SVReassembly, self).setUp()
+        self.vapi.ip_reassembly_enable_disable(
+            sw_if_index=self.src_if.sw_if_index, enable_ip4=True,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL)
+        self.vapi.ip_reassembly_set(
+            timeout_ms=0, max_reassemblies=1000,
+            max_reassembly_length=1000,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL,
+            expire_walk_interval_ms=10)
+        self.sleep(.25)
+        self.vapi.ip_reassembly_set(
+            timeout_ms=1000000, max_reassemblies=1000,
+            max_reassembly_length=1000,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL,
+            expire_walk_interval_ms=10000)
+
+    def tearDown(self):
+        super(TestIPv4SVReassembly, self).tearDown()
+        self.logger.debug(self.vapi.ppcli("show ip4-sv-reassembly details"))
+        self.logger.debug(self.vapi.ppcli("show buffers"))
+
+    def test_basic(self):
+        """ basic reassembly """
+        payload_len = 1000
+        payload = ""
+        counter = 0
+        while len(payload) < payload_len:
+            payload += "%u " % counter
+            counter += 1
+
+        p = (Ether(dst=self.src_if.local_mac, src=self.src_if.remote_mac) /
+             IP(id=1, src=self.src_if.remote_ip4,
+                dst=self.dst_if.remote_ip4) /
+             UDP(sport=1234, dport=5678) /
+             Raw(payload))
+        fragments = fragment_rfc791(p, payload_len/4)
+
+        # send fragment #2 - should be cached inside reassembly
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[1])
+        self.pg_start()
+        self.logger.debug(self.vapi.ppcli("show ip4-sv-reassembly details"))
+        self.logger.debug(self.vapi.ppcli("show buffers"))
+        self.logger.debug(self.vapi.ppcli("show trace"))
+        self.dst_if.assert_nothing_captured()
+
+        # send fragment #1 - reassembly is finished now and both fragments
+        # forwarded
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[0])
+        self.pg_start()
+        self.logger.debug(self.vapi.ppcli("show ip4-sv-reassembly details"))
+        self.logger.debug(self.vapi.ppcli("show buffers"))
+        self.logger.debug(self.vapi.ppcli("show trace"))
+        c = self.dst_if.get_capture(2)
+        for sent, recvd in zip([fragments[1], fragments[0]], c):
+            self.assertEqual(sent[IP].src, recvd[IP].src)
+            self.assertEqual(sent[IP].dst, recvd[IP].dst)
+            self.assertEqual(sent[Raw].payload, recvd[Raw].payload)
+
+        # send rest of fragments - should be immediately forwarded
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[2:])
+        self.pg_start()
+        c = self.dst_if.get_capture(len(fragments[2:]))
+        for sent, recvd in zip(fragments[2:], c):
+            self.assertEqual(sent[IP].src, recvd[IP].src)
+            self.assertEqual(sent[IP].dst, recvd[IP].dst)
+            self.assertEqual(sent[Raw].payload, recvd[Raw].payload)
+
+    def test_timeout(self):
+        """ reassembly timeout """
+        payload_len = 1000
+        payload = ""
+        counter = 0
+        while len(payload) < payload_len:
+            payload += "%u " % counter
+            counter += 1
+
+        p = (Ether(dst=self.src_if.local_mac, src=self.src_if.remote_mac) /
+             IP(id=1, src=self.src_if.remote_ip4,
+                dst=self.dst_if.remote_ip4) /
+             UDP(sport=1234, dport=5678) /
+             Raw(payload))
+        fragments = fragment_rfc791(p, payload_len/4)
+
+        self.vapi.ip_reassembly_set(
+            timeout_ms=100, max_reassemblies=1000,
+            max_reassembly_length=1000,
+            expire_walk_interval_ms=50,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL)
+
+        # send fragments #2 and #1 - should be forwarded
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[0:2])
+        self.pg_start()
+        self.logger.debug(self.vapi.ppcli("show ip4-sv-reassembly details"))
+        self.logger.debug(self.vapi.ppcli("show buffers"))
+        self.logger.debug(self.vapi.ppcli("show trace"))
+        c = self.dst_if.get_capture(2)
+        for sent, recvd in zip([fragments[1], fragments[0]], c):
+            self.assertEqual(sent[IP].src, recvd[IP].src)
+            self.assertEqual(sent[IP].dst, recvd[IP].dst)
+            self.assertEqual(sent[Raw].payload, recvd[Raw].payload)
+
+        # wait for cleanup
+        self.sleep(.25, "wait before sending rest of fragments")
+
+        # send rest of fragments - shouldn't be forwarded
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[2:])
+        self.pg_start()
+        self.dst_if.assert_nothing_captured()
+
+    def test_lru(self):
+        """ reassembly reuses LRU element """
+
+        self.vapi.ip_reassembly_set(
+            timeout_ms=1000000, max_reassemblies=1,
+            max_reassembly_length=1000,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL,
+            expire_walk_interval_ms=10000)
+
+        payload_len = 1000
+        payload = ""
+        counter = 0
+        while len(payload) < payload_len:
+            payload += "%u " % counter
+            counter += 1
+
+        packet_count = 10
+
+        fragments = [f
+                     for i in range(packet_count)
+                     for p in (Ether(dst=self.src_if.local_mac,
+                                     src=self.src_if.remote_mac) /
+                               IP(id=i, src=self.src_if.remote_ip4,
+                                   dst=self.dst_if.remote_ip4) /
+                               UDP(sport=1234, dport=5678) /
+                               Raw(payload))
+                     for f in fragment_rfc791(p, payload_len/4)]
+
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments)
+        self.pg_start()
+        c = self.dst_if.get_capture(len(fragments))
+        for sent, recvd in zip(fragments, c):
+            self.assertEqual(sent[IP].src, recvd[IP].src)
+            self.assertEqual(sent[IP].dst, recvd[IP].dst)
+            self.assertEqual(sent[Raw].payload, recvd[Raw].payload)
+
+
 class TestIPv4MWReassembly(VppTestCase):
     """ IPv4 Reassembly (multiple workers) """
     worker_config = "workers %d" % worker_count
@@ -1304,6 +1475,176 @@ class TestIPv6MWReassembly(VppTestCase):
             send_if.assert_nothing_captured()
 
 
+class TestIPv6SVReassembly(VppTestCase):
+    """ IPv6 Shallow Virtual Reassembly """
+
+    @classmethod
+    def setUpClass(cls):
+        super(TestIPv6SVReassembly, cls).setUpClass()
+
+        cls.create_pg_interfaces([0, 1])
+        cls.src_if = cls.pg0
+        cls.dst_if = cls.pg1
+
+        # setup all interfaces
+        for i in cls.pg_interfaces:
+            i.admin_up()
+            i.config_ip6()
+            i.resolve_ndp()
+
+    def setUp(self):
+        """ Test setup - force timeout on existing reassemblies """
+        super(TestIPv6SVReassembly, self).setUp()
+        self.vapi.ip_reassembly_enable_disable(
+            sw_if_index=self.src_if.sw_if_index, enable_ip6=True,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL)
+        self.vapi.ip_reassembly_set(
+            timeout_ms=0, max_reassemblies=1000,
+            max_reassembly_length=1000,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL,
+            expire_walk_interval_ms=10, is_ip6=1)
+        self.sleep(.25)
+        self.vapi.ip_reassembly_set(
+            timeout_ms=1000000, max_reassemblies=1000,
+            max_reassembly_length=1000,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL,
+            expire_walk_interval_ms=10000, is_ip6=1)
+
+    def tearDown(self):
+        super(TestIPv6SVReassembly, self).tearDown()
+        self.logger.debug(self.vapi.ppcli("show ip6-sv-reassembly details"))
+        self.logger.debug(self.vapi.ppcli("show buffers"))
+
+    def test_basic(self):
+        """ basic reassembly """
+        payload_len = 1000
+        payload = ""
+        counter = 0
+        while len(payload) < payload_len:
+            payload += "%u " % counter
+            counter += 1
+
+        p = (Ether(dst=self.src_if.local_mac, src=self.src_if.remote_mac) /
+             IPv6(src=self.src_if.remote_ip6, dst=self.dst_if.remote_ip6) /
+             UDP(sport=1234, dport=5678) /
+             Raw(payload))
+        fragments = fragment_rfc8200(p, 1, payload_len/4)
+
+        # send fragment #2 - should be cached inside reassembly
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[1])
+        self.pg_start()
+        self.logger.debug(self.vapi.ppcli("show ip6-sv-reassembly details"))
+        self.logger.debug(self.vapi.ppcli("show buffers"))
+        self.logger.debug(self.vapi.ppcli("show trace"))
+        self.dst_if.assert_nothing_captured()
+
+        # send fragment #1 - reassembly is finished now and both fragments
+        # forwarded
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[0])
+        self.pg_start()
+        self.logger.debug(self.vapi.ppcli("show ip6-sv-reassembly details"))
+        self.logger.debug(self.vapi.ppcli("show buffers"))
+        self.logger.debug(self.vapi.ppcli("show trace"))
+        c = self.dst_if.get_capture(2)
+        for sent, recvd in zip([fragments[1], fragments[0]], c):
+            self.assertEqual(sent[IPv6].src, recvd[IPv6].src)
+            self.assertEqual(sent[IPv6].dst, recvd[IPv6].dst)
+            self.assertEqual(sent[Raw].payload, recvd[Raw].payload)
+
+        # send rest of fragments - should be immediately forwarded
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[2:])
+        self.pg_start()
+        c = self.dst_if.get_capture(len(fragments[2:]))
+        for sent, recvd in zip(fragments[2:], c):
+            self.assertEqual(sent[IPv6].src, recvd[IPv6].src)
+            self.assertEqual(sent[IPv6].dst, recvd[IPv6].dst)
+            self.assertEqual(sent[Raw].payload, recvd[Raw].payload)
+
+    def test_timeout(self):
+        """ reassembly timeout """
+        payload_len = 1000
+        payload = ""
+        counter = 0
+        while len(payload) < payload_len:
+            payload += "%u " % counter
+            counter += 1
+
+        p = (Ether(dst=self.src_if.local_mac, src=self.src_if.remote_mac) /
+             IPv6(src=self.src_if.remote_ip6, dst=self.dst_if.remote_ip6) /
+             UDP(sport=1234, dport=5678) /
+             Raw(payload))
+        fragments = fragment_rfc8200(p, 1, payload_len/4)
+
+        self.vapi.ip_reassembly_set(
+            timeout_ms=100, max_reassemblies=1000,
+            max_reassembly_length=1000,
+            expire_walk_interval_ms=50,
+            is_ip6=1,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL)
+
+        # send fragments #2 and #1 - should be forwarded
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[0:2])
+        self.pg_start()
+        self.logger.debug(self.vapi.ppcli("show ip4-sv-reassembly details"))
+        self.logger.debug(self.vapi.ppcli("show buffers"))
+        self.logger.debug(self.vapi.ppcli("show trace"))
+        c = self.dst_if.get_capture(2)
+        for sent, recvd in zip([fragments[1], fragments[0]], c):
+            self.assertEqual(sent[IPv6].src, recvd[IPv6].src)
+            self.assertEqual(sent[IPv6].dst, recvd[IPv6].dst)
+            self.assertEqual(sent[Raw].payload, recvd[Raw].payload)
+
+        # wait for cleanup
+        self.sleep(.25, "wait before sending rest of fragments")
+
+        # send rest of fragments - shouldn't be forwarded
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments[2:])
+        self.pg_start()
+        self.dst_if.assert_nothing_captured()
+
+    def test_lru(self):
+        """ reassembly reuses LRU element """
+
+        self.vapi.ip_reassembly_set(
+            timeout_ms=1000000, max_reassemblies=1,
+            max_reassembly_length=1000,
+            type=VppEnum.vl_api_ip_reass_type_t.IP_REASS_TYPE_SHALLOW_VIRTUAL,
+            is_ip6=1, expire_walk_interval_ms=10000)
+
+        payload_len = 1000
+        payload = ""
+        counter = 0
+        while len(payload) < payload_len:
+            payload += "%u " % counter
+            counter += 1
+
+        packet_count = 10
+
+        fragments = [f
+                     for i in range(packet_count)
+                     for p in (Ether(dst=self.src_if.local_mac,
+                                     src=self.src_if.remote_mac) /
+                               IPv6(src=self.src_if.remote_ip6,
+                                    dst=self.dst_if.remote_ip6) /
+                               UDP(sport=1234, dport=5678) /
+                               Raw(payload))
+                     for f in fragment_rfc8200(p, i, payload_len/4)]
+
+        self.pg_enable_capture()
+        self.src_if.add_stream(fragments)
+        self.pg_start()
+        c = self.dst_if.get_capture(len(fragments))
+        for sent, recvd in zip(fragments, c):
+            self.assertEqual(sent[IPv6].src, recvd[IPv6].src)
+            self.assertEqual(sent[IPv6].dst, recvd[IPv6].dst)
+            self.assertEqual(sent[Raw].payload, recvd[Raw].payload)
+
+
 class TestIPv4ReassemblyLocalNode(VppTestCase):
     """ IPv4 Reassembly for packets coming to ip4-local node """