5 files changed, 1317 insertions, 0 deletions

diff --git a/utils/extras/rtinject/tap_inject.c b/utils/extras/rtinject/tap_inject.c
new file mode 100644
index 000000000..f41ae86c8
--- /dev/null
+++ b/utils/extras/rtinject/tap_inject.c
@@ -0,0 +1,380 @@
+/*
+ * Copyright 2016 Intel Corporation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "tap_inject.h"
+
+#include <vnet/mfib/mfib_table.h>
+#include <vnet/ip/ip.h>
+#include <vnet/ip/lookup.h>
+#include <vnet/fib/fib.h>
+
+static tap_inject_main_t tap_inject_main;
+extern dpo_type_t tap_inject_dpo_type;
+
+tap_inject_main_t *
+tap_inject_get_main (void)
+{
+  return &tap_inject_main;
+}
+
+void
+tap_inject_insert_tap (u32 sw_if_index, u32 tap_fd, u32 tap_if_index)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  vec_validate_init_empty (im->sw_if_index_to_tap_fd, sw_if_index, ~0);
+  vec_validate_init_empty (im->sw_if_index_to_tap_if_index, sw_if_index, ~0);
+
+  vec_validate_init_empty (im->tap_fd_to_sw_if_index, tap_fd, ~0);
+
+  im->sw_if_index_to_tap_fd[sw_if_index] = tap_fd;
+  im->sw_if_index_to_tap_if_index[sw_if_index] = tap_if_index;
+
+  im->tap_fd_to_sw_if_index[tap_fd] = sw_if_index;
+
+  hash_set (im->tap_if_index_to_sw_if_index, tap_if_index, sw_if_index);
+}
+
+void
+tap_inject_delete_tap (u32 sw_if_index)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+  u32 tap_fd = im->sw_if_index_to_tap_fd[sw_if_index];
+  u32 tap_if_index = im->sw_if_index_to_tap_if_index[sw_if_index];
+
+  im->sw_if_index_to_tap_if_index[sw_if_index] = ~0;
+  im->sw_if_index_to_tap_fd[sw_if_index] = ~0;
+  im->tap_fd_to_sw_if_index[tap_fd] = ~0;
+
+  hash_unset (im->tap_if_index_to_sw_if_index, tap_if_index);
+}
+
+u32
+tap_inject_lookup_tap_fd (u32 sw_if_index)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  vec_validate_init_empty (im->sw_if_index_to_tap_fd, sw_if_index, ~0);
+  return im->sw_if_index_to_tap_fd[sw_if_index];
+}
+
+u32
+tap_inject_lookup_sw_if_index_from_tap_fd (u32 tap_fd)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  vec_validate_init_empty (im->tap_fd_to_sw_if_index, tap_fd, ~0);
+  return im->tap_fd_to_sw_if_index[tap_fd];
+}
+
+u32
+tap_inject_lookup_sw_if_index_from_tap_if_index (u32 tap_if_index)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+  uword * sw_if_index;
+
+  sw_if_index = hash_get (im->tap_if_index_to_sw_if_index, tap_if_index);
+  return sw_if_index ? *(u32 *)sw_if_index : ~0;
+}
+
+/* *INDENT-OFF* */
+VLIB_PLUGIN_REGISTER () = {
+    // .version = VPP_BUILD_VER, FIXME
+    .description = "router",
+};
+/* *INDENT-ON* */
+
+
+static void
+tap_inject_disable (void)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  im->flags &= ~TAP_INJECT_F_ENABLED;
+
+  clib_warning ("tap-inject is not actually disabled.");
+}
+
+static clib_error_t *
+tap_inject_enable (void)
+{
+  vlib_main_t * vm = vlib_get_main ();
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  if (tap_inject_is_enabled ())
+    return 0;
+
+  tap_inject_enable_netlink ();
+
+  /* Only enable netlink? */
+  if (im->flags & TAP_INJECT_F_CONFIG_NETLINK)
+    {
+      im->flags |= TAP_INJECT_F_ENABLED;
+      return 0;
+    }
+
+  /* Register ARP and ICMP6 as neighbor nodes. */
+  ethernet_register_input_type (vm, ETHERNET_TYPE_ARP, im->neighbor_node_index);
+  ip6_register_protocol (IP_PROTOCOL_ICMP6, im->neighbor_node_index);
+
+  /* Register remaining protocols. */
+  ip4_register_protocol (IP_PROTOCOL_ICMP, im->tx_node_index);
+
+  ip4_register_protocol (IP_PROTOCOL_OSPF, im->tx_node_index);
+  ip4_register_protocol (IP_PROTOCOL_TCP, im->tx_node_index);
+  ip4_register_protocol (IP_PROTOCOL_UDP, im->tx_node_index);
+
+  ip6_register_protocol (IP_PROTOCOL_OSPF, im->tx_node_index);
+  ip6_register_protocol (IP_PROTOCOL_TCP, im->tx_node_index);
+  ip6_register_protocol (IP_PROTOCOL_UDP, im->tx_node_index);
+
+  {
+    dpo_id_t dpo = DPO_INVALID;
+
+    const mfib_prefix_t pfx_224_0_0_0 = {
+        .fp_len = 24,
+        .fp_proto = FIB_PROTOCOL_IP4,
+        .fp_grp_addr = {
+            .ip4.as_u32 = clib_host_to_net_u32(0xe0000000),
+        },
+        .fp_src_addr = {
+            .ip4.as_u32 = 0,
+        },
+    };
+
+    dpo_set(&dpo, tap_inject_dpo_type, DPO_PROTO_IP4, ~0);
+
+    index_t repi = replicate_create(1, DPO_PROTO_IP4);
+    replicate_set_bucket(repi, 0, &dpo);
+
+    mfib_table_entry_special_add(0,
+                                 &pfx_224_0_0_0,
+                                 MFIB_SOURCE_API,
+                                 MFIB_ENTRY_FLAG_ACCEPT_ALL_ITF,
+                                 repi);
+
+    dpo_reset(&dpo);
+  }
+
+  im->flags |= TAP_INJECT_F_ENABLED;
+
+  return 0;
+}
+
+static uword
+tap_inject_iface_isr (vlib_main_t * vm, vlib_node_runtime_t * node,
+                      vlib_frame_t * f)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+  vnet_hw_interface_t * hw;
+  u32 * hw_if_index;
+  clib_error_t * err = 0;
+
+  vec_foreach (hw_if_index, im->interfaces_to_enable)
+    {
+      hw = vnet_get_hw_interface (vnet_get_main (), *hw_if_index);
+
+      if (hw->hw_class_index == ethernet_hw_interface_class.index)
+        {
+          err = tap_inject_tap_connect (hw);
+          if (err)
+            break;
+        }
+    }
+
+  vec_foreach (hw_if_index, im->interfaces_to_disable)
+    tap_inject_tap_disconnect (*hw_if_index);
+
+  vec_free (im->interfaces_to_enable);
+  vec_free (im->interfaces_to_disable);
+
+  return err ? -1 : 0;
+}
+
+VLIB_REGISTER_NODE (tap_inject_iface_isr_node, static) = {
+  .function = tap_inject_iface_isr,
+  .name = "tap-inject-iface-isr",
+  .type = VLIB_NODE_TYPE_INPUT,
+  .state = VLIB_NODE_STATE_INTERRUPT,
+  .vector_size = sizeof (u32),
+};
+
+
+static clib_error_t *
+tap_inject_interface_add_del (struct vnet_main_t * vnet_main, u32 hw_if_index,
+                              u32 add)
+{
+  vlib_main_t * vm = vlib_get_main ();
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  if (!tap_inject_is_config_enabled ())
+    return 0;
+
+  tap_inject_enable ();
+
+  if (add)
+    vec_add1 (im->interfaces_to_enable, hw_if_index);
+  else
+    vec_add1 (im->interfaces_to_disable, hw_if_index);
+
+  vlib_node_set_interrupt_pending (vm, tap_inject_iface_isr_node.index);
+
+  return 0;
+}
+
+VNET_HW_INTERFACE_ADD_DEL_FUNCTION (tap_inject_interface_add_del);
+
+
+static clib_error_t *
+tap_inject_enable_disable_all_interfaces (int enable)
+{
+  vnet_main_t * vnet_main = vnet_get_main ();
+  tap_inject_main_t * im = tap_inject_get_main ();
+  vnet_hw_interface_t * interfaces;
+  vnet_hw_interface_t * hw;
+  u32 ** indices;
+
+  if (enable)
+    tap_inject_enable ();
+  else
+    tap_inject_disable ();
+
+  /* Collect all the interface indices. */
+  interfaces = vnet_main->interface_main.hw_interfaces;
+  indices = enable ? &im->interfaces_to_enable : &im->interfaces_to_disable;
+  pool_foreach (hw, interfaces, vec_add1 (*indices, hw - interfaces));
+
+  if (tap_inject_iface_isr (vlib_get_main (), 0, 0))
+    return clib_error_return (0, "tap-inject interface add del isr failed");
+
+  return 0;
+}
+
+static clib_error_t *
+tap_inject_cli (vlib_main_t * vm, unformat_input_t * input,
+                 vlib_cli_command_t * cmd)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  if (cmd->function_arg)
+    {
+      clib_error_t * err;
+
+      if (tap_inject_is_config_disabled ())
+        return clib_error_return (0,
+            "tap-inject is disabled in config, thus cannot be enabled.");
+
+      /* Enable */
+      err = tap_inject_enable_disable_all_interfaces (1);
+      if (err)
+        {
+          tap_inject_enable_disable_all_interfaces (0);
+          return err;
+        }
+
+      im->flags |= TAP_INJECT_F_CONFIG_ENABLE;
+    }
+  else
+    {
+      /* Disable */
+      tap_inject_enable_disable_all_interfaces (0);
+      im->flags &= ~TAP_INJECT_F_CONFIG_ENABLE;
+    }
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (tap_inject_enable_cmd, static) = {
+  .path = "enable tap-inject",
+  .short_help = "enable tap-inject",
+  .function = tap_inject_cli,
+  .function_arg = 1,
+};
+
+VLIB_CLI_COMMAND (tap_inject_disable_cmd, static) = {
+  .path = "disable tap-inject",
+  .short_help = "disable tap-inject",
+  .function = tap_inject_cli,
+  .function_arg = 0,
+};
+
+
+static clib_error_t *
+show_tap_inject (vlib_main_t * vm, unformat_input_t * input,
+                 vlib_cli_command_t * cmd)
+{
+  vnet_main_t * vnet_main = vnet_get_main ();
+  tap_inject_main_t * im = tap_inject_get_main ();
+  u32 k, v;
+
+  if (tap_inject_is_config_disabled ())
+    {
+      vlib_cli_output (vm, "tap-inject is disabled in config.\n");
+      return 0;
+    }
+
+  if (!tap_inject_is_enabled ())
+    {
+      vlib_cli_output (vm, "tap-inject is not enabled.\n");
+      return 0;
+    }
+
+  hash_foreach (k, v, im->tap_if_index_to_sw_if_index, {
+    vlib_cli_output (vm, "%U -> %U",
+            format_vnet_sw_interface_name, vnet_main,
+            vnet_get_sw_interface (vnet_main, v),
+            format_tap_inject_tap_name, k);
+  });
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (show_tap_inject_cmd, static) = {
+  .path = "show tap-inject",
+  .short_help = "show tap-inject",
+  .function = show_tap_inject,
+};
+
+
+static clib_error_t *
+tap_inject_config (vlib_main_t * vm, unformat_input_t * input)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (input, "enable"))
+        im->flags |= TAP_INJECT_F_CONFIG_ENABLE;
+
+      else if (unformat (input, "disable"))
+        im->flags |= TAP_INJECT_F_CONFIG_DISABLE;
+
+      else if (unformat (input, "netlink-only"))
+        im->flags |= TAP_INJECT_F_CONFIG_NETLINK;
+
+      else
+        return clib_error_return (0, "syntax error `%U'",
+                                  format_unformat_error, input);
+    }
+
+  if (tap_inject_is_config_enabled () && tap_inject_is_config_disabled ())
+    return clib_error_return (0,
+              "tap-inject cannot be both enabled and disabled.");
+
+  return 0;
+}
+
+VLIB_CONFIG_FUNCTION (tap_inject_config, "tap-inject");
diff --git a/utils/extras/rtinject/tap_inject.h b/utils/extras/rtinject/tap_inject.h
new file mode 100644
index 000000000..ec5121a09
--- /dev/null
+++ b/utils/extras/rtinject/tap_inject.h
@@ -0,0 +1,108 @@
+/*
+ * Copyright 2016 Intel Corporation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef _TAP_INJECT_H
+#define _TAP_INJECT_H
+
+#include <vnet/plugin/plugin.h>
+#include <vnet/ip/ip.h>
+
+#ifndef ETHER_ADDR_LEN
+#define ETHER_ADDR_LEN 6
+#endif
+
+typedef struct {
+  /*
+   * tap-inject can be enabled or disabled in config file or during runtime.
+   * When disabled in config, it is not possible to enable during runtime.
+   *
+   * When the netlink-only option is used, netlink configuration is monitored
+   * and mirrored to the data plane but no traffic is passed between the host
+   * and the data plane.
+   */
+#define TAP_INJECT_F_CONFIG_ENABLE  (1U << 0)
+#define TAP_INJECT_F_CONFIG_DISABLE (1U << 1)
+#define TAP_INJECT_F_CONFIG_NETLINK (1U << 2)
+#define TAP_INJECT_F_ENABLED        (1U << 3)
+
+  u32 flags;
+
+  u32 * sw_if_index_to_tap_fd;
+  u32 * sw_if_index_to_tap_if_index;
+  u32 * tap_fd_to_sw_if_index;
+  u32 * tap_if_index_to_sw_if_index;
+
+  u32 * interfaces_to_enable;
+  u32 * interfaces_to_disable;
+
+  u32 * rx_file_descriptors;
+
+  u32 rx_node_index;
+  u32 tx_node_index;
+  u32 neighbor_node_index;
+
+  u32 * rx_buffers;
+
+} tap_inject_main_t;
+
+
+tap_inject_main_t * tap_inject_get_main (void);
+
+void tap_inject_insert_tap (u32 sw_if_index, u32 tap_fd, u32 tap_if_index);
+void tap_inject_delete_tap (u32 sw_if_index);
+
+u32 tap_inject_lookup_tap_fd (u32 sw_if_index);
+u32 tap_inject_lookup_sw_if_index_from_tap_fd (u32 tap_fd);
+u32 tap_inject_lookup_sw_if_index_from_tap_if_index (u32 tap_if_index);
+
+static inline int
+tap_inject_is_enabled (void)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  return !!(im->flags & TAP_INJECT_F_ENABLED);
+}
+
+static inline int
+tap_inject_is_config_enabled (void)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  return !!(im->flags & TAP_INJECT_F_CONFIG_ENABLE);
+}
+
+static inline int
+tap_inject_is_config_disabled (void)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  return !!(im->flags & TAP_INJECT_F_CONFIG_DISABLE);
+}
+
+
+/* Netlink */
+
+void tap_inject_enable_netlink (void);
+
+
+/* Tap */
+
+clib_error_t * tap_inject_tap_connect (vnet_hw_interface_t * hw);
+clib_error_t * tap_inject_tap_disconnect (u32 sw_if_index);
+
+u8 * format_tap_inject_tap_name (u8 * s, va_list * args);
+
+#endif /* _TAP_INJECT_H */
diff --git a/utils/extras/rtinject/tap_inject_netlink.c b/utils/extras/rtinject/tap_inject_netlink.c
new file mode 100644
index 000000000..c0e0ce995
--- /dev/null
+++ b/utils/extras/rtinject/tap_inject_netlink.c
@@ -0,0 +1,285 @@
+/*
+ * Copyright 2016 Intel Corporation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "../devices/rtnetlink/netns.h"
+#include <vlibmemory/api.h>
+#include <vnet/ip/ip6_neighbor.h>
+#include <vnet/ip/lookup.h>
+#include <vnet/fib/fib.h>
+#include <vnet/ethernet/arp.h>
+#include <arpa/inet.h>
+#include <linux/mpls.h>
+#include <vnet/mpls/packet.h>
+
+#include "tap_inject.h"
+
+static void
+add_del_addr (ns_addr_t * a, int is_del)
+{
+  vlib_main_t * vm = vlib_get_main ();
+  u32 sw_if_index;
+
+  sw_if_index = tap_inject_lookup_sw_if_index_from_tap_if_index (
+                                                                 a->ifaddr.ifa_index);
+
+  if (sw_if_index == ~0)
+    return;
+
+  if (a->ifaddr.ifa_family == AF_INET)
+    {
+      ip4_add_del_interface_address (vm, sw_if_index,
+                                     (ip4_address_t *) a->local, a->ifaddr.ifa_prefixlen, is_del);
+    }
+  else if (a->ifaddr.ifa_family == AF_INET6)
+    {
+      ip6_add_del_interface_address (vm, sw_if_index,
+                                     (ip6_address_t *) a->addr, a->ifaddr.ifa_prefixlen, is_del);
+    }
+}
+
+
+struct set_flags_args {
+  u32 index;
+  u8 flags;
+};
+
+static void
+set_flags_cb (struct set_flags_args * a)
+{
+  vnet_sw_interface_set_flags (vnet_get_main (), a->index, a->flags);
+}
+
+static void
+add_del_link (ns_link_t * l, int is_del)
+{
+  struct set_flags_args args = { ~0, 0 };
+  vnet_sw_interface_t * sw;
+  u8 flags = 0;
+  u32 sw_if_index;
+
+  sw_if_index = tap_inject_lookup_sw_if_index_from_tap_if_index (
+                                                                 l->ifi.ifi_index);
+
+  if (sw_if_index == ~0)
+    return;
+
+  sw = vnet_get_sw_interface (vnet_get_main (), sw_if_index);
+
+  flags = sw->flags;
+
+  if (l->ifi.ifi_flags & IFF_UP)
+    flags |= VNET_SW_INTERFACE_FLAG_ADMIN_UP;
+  else
+    flags &= ~VNET_SW_INTERFACE_FLAG_ADMIN_UP;
+
+  args.index = sw_if_index;
+  args.flags = flags;
+
+  vl_api_rpc_call_main_thread (set_flags_cb, (u8 *)&args, sizeof (args));
+}
+
+
+static void
+add_del_neigh (ns_neigh_t * n, int is_del)
+{
+  vnet_main_t * vnet_main = vnet_get_main ();
+  vlib_main_t * vm = vlib_get_main ();
+  u32 sw_if_index;
+
+  sw_if_index = tap_inject_lookup_sw_if_index_from_tap_if_index (
+                                                                 n->nd.ndm_ifindex);
+
+  if (sw_if_index == ~0)
+    return;
+
+  if (n->nd.ndm_family == AF_INET)
+    {
+      ethernet_arp_ip4_over_ethernet_address_t a;
+
+      memset (&a, 0, sizeof (a));
+
+      clib_memcpy (&a.mac, n->lladdr, ETHER_ADDR_LEN);
+      clib_memcpy (&a.ip4, n->dst, sizeof (a.ip4));
+
+
+      if (n->nd.ndm_state & NUD_REACHABLE)
+        {
+          vnet_arp_set_ip4_over_ethernet (vnet_main, sw_if_index,
+                                          &a,
+                                          IP_NEIGHBOR_FLAG_NO_FIB_ENTRY);
+
+        }
+      else if (n->nd.ndm_state & NUD_FAILED)
+        {
+          vnet_arp_unset_ip4_over_ethernet (vnet_main, sw_if_index, &a);
+        }
+    }
+  else if (n->nd.ndm_family == AF_INET6)
+    {
+      if (n->nd.ndm_state & NUD_REACHABLE)
+        {
+
+          mac_address_t * mac1;
+	  mac1=malloc(sizeof(mac_address_t));
+          memcpy (mac1, n->lladdr, ETHER_ADDR_LEN);
+          vnet_set_ip6_ethernet_neighbor (vm, sw_if_index,
+                                          (ip6_address_t *) n->dst, (mac_address_t *) mac1,
+                                           IP_NEIGHBOR_FLAG_NO_FIB_ENTRY);
+        }
+      else
+        vnet_unset_ip6_ethernet_neighbor (vm, sw_if_index,
+                                          (ip6_address_t *) n->dst);
+    }
+}
+
+
+#define TAP_INJECT_HOST_ROUTE_TABLE_MAIN 254
+
+static void
+get_mpls_label_stack(struct mpls_label *addr, u32* l)
+{
+  u32 entry = ntohl(addr[0].entry);
+  u32 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
+
+  for(int i = 1; label != 0; i++) {
+    *l++ = label;
+    if(entry & MPLS_LS_S_MASK)
+      return;
+    entry = ntohl(addr[i].entry);
+    label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
+  }
+}
+
+static void
+add_del_route (ns_route_t * r, int is_del)
+{
+  u32 sw_if_index;
+
+  sw_if_index = tap_inject_lookup_sw_if_index_from_tap_if_index (r->oif);
+
+  if (sw_if_index == ~0)
+    return;
+
+  if (r->rtm.rtm_family == AF_INET)
+    {
+      u32 stack[MPLS_STACK_DEPTH] = {0};
+
+      fib_prefix_t prefix;
+      ip46_address_t nh;
+
+      memset (&prefix, 0, sizeof (prefix));
+      prefix.fp_len = r->rtm.rtm_dst_len;
+      prefix.fp_proto = FIB_PROTOCOL_IP4;
+      clib_memcpy (&prefix.fp_addr.ip4, r->dst, sizeof (prefix.fp_addr.ip4));
+      get_mpls_label_stack(r->encap, stack);
+      memset (&nh, 0, sizeof (nh));
+      clib_memcpy (&nh.ip4, r->gateway, sizeof (nh.ip4));
+      if(*stack == 0)
+        fib_table_entry_path_add (0, &prefix, FIB_SOURCE_API,
+                                  FIB_ENTRY_FLAG_NONE, prefix.fp_proto,
+                                  &nh, sw_if_index, 0,
+                                  0 /* weight */, NULL,
+                                  FIB_ROUTE_PATH_FLAG_NONE);
+      else {
+        fib_route_path_t *rpaths = NULL, rpath;
+        memset(&rpath, 0, sizeof(rpath));
+        rpath.frp_weight = 1;
+        rpath.frp_proto = DPO_PROTO_IP4;
+        clib_memcpy(&rpath.frp_addr.ip4, r->gateway, sizeof(rpath.frp_addr.ip4));
+        rpath.frp_sw_if_index = sw_if_index;
+        for(int i = 0; i < MPLS_STACK_DEPTH && stack[i] != 0; i++) {
+          fib_mpls_label_t fib_label = {stack[i],0,0,0};
+          vec_add1(rpath.frp_label_stack, fib_label);
+        }
+        vec_add1(rpaths, rpath);
+        fib_table_entry_path_add2(0,
+                                  &prefix,
+                                  FIB_SOURCE_API,
+                                  FIB_ENTRY_FLAG_NONE,
+                                  rpaths);
+      }
+    }
+  else if (r->rtm.rtm_family == AF_INET6)
+    {
+      fib_prefix_t prefix;
+      ip46_address_t nh;
+      memset (&prefix, 0, sizeof (prefix));
+      prefix.fp_len = r->rtm.rtm_dst_len;
+      prefix.fp_proto = FIB_PROTOCOL_IP6;
+      clib_memcpy (&prefix.fp_addr.ip6, r->dst, sizeof (prefix.fp_addr.ip6));
+      memset (&nh, 0, sizeof (nh));
+      clib_memcpy (&nh.ip6, r->gateway, sizeof (nh.ip6));
+      fib_table_entry_path_add (0, &prefix, FIB_SOURCE_API,
+                                FIB_ENTRY_FLAG_NONE, prefix.fp_proto,
+                                &nh, sw_if_index, 0,
+                                0 /* weight */, NULL,
+                                FIB_ROUTE_PATH_FLAG_NONE);
+    }
+/*     else if (r->rtm.rtm_family == AF_MPLS)
+    {
+      u32 dst_label;
+      get_mpls_label_stack((struct mpls_label*) r->dst, &dst_label);
+      struct rtvia *via = (struct rtvia*) r->via;
+      fib_prefix_t prefix;
+      fib_route_path_t *rpaths = NULL, rpath;
+      memset (&prefix, 0, sizeof (prefix));
+      prefix.fp_len = 21;
+      prefix.fp_label = dst_label;
+      prefix.fp_proto = FIB_PROTOCOL_MPLS;
+      prefix.fp_payload_proto = DPO_PROTO_IP4;
+      memset(&rpath, 0, sizeof(rpath));
+      clib_memcpy (&rpath.frp_addr.ip4, via->rtvia_addr, sizeof (rpath.frp_addr.ip4));
+      rpath.frp_weight = 1;
+      rpath.frp_proto = DPO_PROTO_IP4;
+      rpath.frp_fib_index = 0;
+      rpath.frp_sw_if_index = sw_if_index;
+      vec_add1(rpaths, rpath);
+      fib_table_entry_path_add2(0,
+                                &prefix,
+                                FIB_SOURCE_API,
+                                FIB_ENTRY_FLAG_NONE,
+                                rpaths);
+    }*/
+}
+
+
+static void
+netns_notify_cb (void * obj, netns_type_t type, u32 flags, uword opaque)
+{
+  if (type == NETNS_TYPE_ADDR)
+    add_del_addr ((ns_addr_t *)obj, flags & NETNS_F_DEL);
+
+  else if (type == NETNS_TYPE_LINK)
+    add_del_link ((ns_link_t *)obj, flags & NETNS_F_DEL);
+
+  else if (type == NETNS_TYPE_NEIGH)
+    add_del_neigh ((ns_neigh_t *)obj, flags & NETNS_F_DEL);
+
+  else if (type == NETNS_TYPE_ROUTE)
+    add_del_route ((ns_route_t *)obj, flags & NETNS_F_DEL);
+}
+
+void
+tap_inject_enable_netlink (void)
+{
+  char nsname = 0;
+  netns_sub_t sub = {
+    .notify = netns_notify_cb,
+    .opaque = 0,
+  };
+
+  netns_open (&nsname, &sub);
+}
diff --git a/utils/extras/rtinject/tap_inject_node.c b/utils/extras/rtinject/tap_inject_node.c
new file mode 100644
index 000000000..73c296451
--- /dev/null
+++ b/utils/extras/rtinject/tap_inject_node.c
@@ -0,0 +1,374 @@
+/*
+ * Copyright 2016 Intel Corporation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "tap_inject.h"
+#include <sys/uio.h>
+#include <netinet/in.h>
+#include <vnet/ethernet/arp_packet.h>
+
+vlib_node_registration_t tap_inject_rx_node;
+vlib_node_registration_t tap_inject_tx_node;
+vlib_node_registration_t tap_inject_neighbor_node;
+
+enum {
+  NEXT_NEIGHBOR_ARP,
+  NEXT_NEIGHBOR_ICMP6,
+};
+
+/**
+ * @brief Dynamically added tap_inject DPO type
+ */
+dpo_type_t tap_inject_dpo_type;
+
+static inline void
+tap_inject_tap_send_buffer (int fd, vlib_buffer_t * b)
+{
+  struct iovec iov;
+  ssize_t n_bytes;
+
+  iov.iov_base = vlib_buffer_get_current (b);
+  iov.iov_len = b->current_length;
+
+  n_bytes = writev (fd, &iov, 1);
+
+  if (n_bytes < 0)
+    clib_warning ("writev failed");
+  else if (n_bytes < b->current_length || b->flags & VLIB_BUFFER_NEXT_PRESENT)
+    clib_warning ("buffer truncated");
+}
+
+static uword
+tap_inject_tx (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+  vlib_buffer_t * b;
+  u32 * pkts;
+  u32 fd;
+  u32 i;
+
+  pkts = vlib_frame_vector_args (f);
+
+  for (i = 0; i < f->n_vectors; ++i)
+    {
+      b = vlib_get_buffer (vm, pkts[i]);
+
+      fd = tap_inject_lookup_tap_fd (vnet_buffer (b)->sw_if_index[VLIB_RX]);
+      if (fd == ~0)
+        continue;
+
+      /* Re-wind the buffer to the start of the Ethernet header. */
+      vlib_buffer_advance (b, -b->current_data);
+
+      tap_inject_tap_send_buffer (fd, b);
+    }
+
+  vlib_buffer_free (vm, pkts, f->n_vectors);
+  return f->n_vectors;
+}
+
+VLIB_REGISTER_NODE (tap_inject_tx_node) = {
+  .function = tap_inject_tx,
+  .name = "tap-inject-tx",
+  .vector_size = sizeof (u32),
+  .type = VLIB_NODE_TYPE_INTERNAL,
+};
+
+
+static uword
+tap_inject_neighbor (vlib_main_t * vm,
+                     vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+  vlib_buffer_t * b;
+  u32 * pkts;
+  u32 fd;
+  u32 i;
+  u32 bi;
+  u32 next_index = node->cached_next_index;
+  u32 next = ~0;
+  u32 n_left;
+  u32 * to_next;
+
+  pkts = vlib_frame_vector_args (f);
+
+  for (i = 0; i < f->n_vectors; ++i)
+    {
+      bi = pkts[i];
+      b = vlib_get_buffer (vm, bi);
+
+      fd = tap_inject_lookup_tap_fd (vnet_buffer (b)->sw_if_index[VLIB_RX]);
+      if (fd == ~0)
+        {
+          vlib_buffer_free (vm, &bi, 1);
+          continue;
+        }
+
+      /* Re-wind the buffer to the start of the Ethernet header. */
+      vlib_buffer_advance (b, -b->current_data);
+
+      tap_inject_tap_send_buffer (fd, b);
+
+      /* Send the buffer to a neighbor node too? */
+      {
+        ethernet_header_t * eth = vlib_buffer_get_current (b);
+        u16 ether_type = htons (eth->type);
+
+        if (ether_type == ETHERNET_TYPE_ARP)
+          {
+            ethernet_arp_header_t * arp = (void *)(eth + 1);
+
+            if (arp->opcode == ntohs (ETHERNET_ARP_OPCODE_reply))
+              next = NEXT_NEIGHBOR_ARP;
+          }
+        else if (ether_type == ETHERNET_TYPE_IP6)
+          {
+            ip6_header_t * ip = (void *)(eth + 1);
+            icmp46_header_t * icmp = (void *)(ip + 1);
+
+            if (ip->protocol == IP_PROTOCOL_ICMP6 &&
+                icmp->type == ICMP6_neighbor_advertisement)
+              next = NEXT_NEIGHBOR_ICMP6;
+          }
+      }
+
+      if (next == ~0)
+        {
+          vlib_buffer_free (vm, &bi, 1);
+          continue;
+        }
+
+      /* ARP and ICMP6 expect to start processing after the Ethernet header. */
+      vlib_buffer_advance (b, sizeof (ethernet_header_t));
+
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left);
+
+      *(to_next++) = bi;
+      --n_left;
+
+      vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+                                       n_left, bi, next);
+      vlib_put_next_frame (vm, node, next_index, n_left);
+    }
+
+  return f->n_vectors;
+}
+
+VLIB_REGISTER_NODE (tap_inject_neighbor_node) = {
+  .function = tap_inject_neighbor,
+  .name = "tap-inject-neighbor",
+  .vector_size = sizeof (u32),
+  .type = VLIB_NODE_TYPE_INTERNAL,
+  .n_next_nodes = 2,
+  .next_nodes = {
+    [NEXT_NEIGHBOR_ARP] = "arp-input",
+    [NEXT_NEIGHBOR_ICMP6] = "icmp6-neighbor-solicitation",
+  },
+};
+
+
+#define MTU 1500
+#define MTU_BUFFERS ((MTU + vlib_buffer_get_default_data_size(vm) - 1) / vlib_buffer_get_default_data_size(vm))
+#define NUM_BUFFERS_TO_ALLOC 32
+
+static inline uword
+tap_rx (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * f, int fd)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+  u32 sw_if_index;
+  struct iovec iov[MTU_BUFFERS];
+  u32 bi[MTU_BUFFERS];
+  vlib_buffer_t * b;
+  ssize_t n_bytes;
+  ssize_t n_bytes_left;
+  u32 i, j;
+
+  sw_if_index = tap_inject_lookup_sw_if_index_from_tap_fd (fd);
+  if (sw_if_index == ~0)
+    return 0;
+
+  /* Allocate buffers in bulk when there are less than enough to rx an MTU. */
+  if (vec_len (im->rx_buffers) < MTU_BUFFERS)
+    {
+      u32 len = vec_len (im->rx_buffers);
+
+
+      u8 index = vlib_buffer_pool_get_default_for_numa (vm,0);
+      len = vlib_buffer_alloc_from_pool(vm,
+                    &im->rx_buffers[len], NUM_BUFFERS_TO_ALLOC,
+                    index);
+
+      _vec_len (im->rx_buffers) += len;
+
+      if (vec_len (im->rx_buffers) < MTU_BUFFERS)
+        {
+          clib_warning ("failed to allocate buffers");
+          return 0;
+        }
+    }
+
+  /* Fill buffers from the end of the list to make it easier to resize. */
+  for (i = 0, j = vec_len (im->rx_buffers) - 1; i < MTU_BUFFERS; ++i, --j)
+    {
+      vlib_buffer_t * b;
+
+      bi[i] = im->rx_buffers[j];
+
+      b = vlib_get_buffer (vm, bi[i]);
+
+      iov[i].iov_base = b->data;
+      iov[i].iov_len = VLIB_BUFFER_DEFAULT_DATA_SIZE;
+    }
+
+  n_bytes = readv (fd, iov, MTU_BUFFERS);
+  if (n_bytes < 0)
+    {
+      clib_warning ("readv failed");
+      return 0;
+    }
+
+  b = vlib_get_buffer (vm, bi[0]);
+
+  vnet_buffer (b)->sw_if_index[VLIB_RX] = sw_if_index;
+  vnet_buffer (b)->sw_if_index[VLIB_TX] = sw_if_index;
+
+  n_bytes_left = n_bytes - VLIB_BUFFER_DEFAULT_DATA_SIZE;
+
+  if (n_bytes_left > 0)
+    {
+      b->total_length_not_including_first_buffer = n_bytes_left;
+      b->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID;
+    }
+
+  b->current_length = n_bytes;
+
+  /* If necessary, configure any remaining buffers in the chain. */
+  for (i = 1; n_bytes_left > 0; ++i, n_bytes_left -= VLIB_BUFFER_DEFAULT_DATA_SIZE)
+    {
+      b = vlib_get_buffer (vm, bi[i - 1]);
+      b->current_length = VLIB_BUFFER_DEFAULT_DATA_SIZE;
+      b->flags |= VLIB_BUFFER_NEXT_PRESENT;
+      b->next_buffer = bi[i];
+
+      b = vlib_get_buffer (vm, bi[i]);
+      b->current_length = n_bytes_left;
+    }
+
+  _vec_len (im->rx_buffers) -= i;
+
+  /* Get the packet to the output node. */
+  {
+    vnet_hw_interface_t * hw;
+    vlib_frame_t * new_frame;
+    u32 * to_next;
+
+    hw = vnet_get_hw_interface (vnet_get_main (), sw_if_index);
+
+    new_frame = vlib_get_frame_to_node (vm, hw->output_node_index);
+    to_next = vlib_frame_vector_args (new_frame);
+    to_next[0] = bi[0];
+    new_frame->n_vectors = 1;
+
+    vlib_put_frame_to_node (vm, hw->output_node_index, new_frame);
+  }
+
+  return 1;
+}
+
+static uword
+tap_inject_rx (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * f)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+  u32 * fd;
+  uword count = 0;
+
+  vec_foreach (fd, im->rx_file_descriptors)
+    {
+      if (tap_rx (vm, node, f, *fd) != 1)
+        {
+          clib_warning ("rx failed");
+          count = 0;
+          break;
+        }
+      ++count;
+    }
+
+  vec_free (im->rx_file_descriptors);
+
+  return count;
+}
+
+VLIB_REGISTER_NODE (tap_inject_rx_node) = {
+  .function = tap_inject_rx,
+  .name = "tap-inject-rx",
+  .type = VLIB_NODE_TYPE_INPUT,
+  .state = VLIB_NODE_STATE_INTERRUPT,
+  .vector_size = sizeof (u32),
+};
+
+/**
+ * @brief no-op lock function.
+ */
+static void
+tap_inject_dpo_lock (dpo_id_t * dpo)
+{
+}
+
+/**
+ * @brief no-op unlock function.
+ */
+static void
+tap_inject_dpo_unlock (dpo_id_t * dpo)
+{
+}
+
+u8 *
+format_tap_inject_dpo (u8 * s, va_list * args)
+{
+  return (format (s, "tap-inject:[%d]", 0));
+}
+
+const static dpo_vft_t tap_inject_vft = {
+  .dv_lock = tap_inject_dpo_lock,
+  .dv_unlock = tap_inject_dpo_unlock,
+  .dv_format = format_tap_inject_dpo,
+};
+
+const static char *const tap_inject_tx_nodes[] = {
+  "tap-inject-tx",
+  NULL,
+};
+
+const static char *const *const tap_inject_nodes[DPO_PROTO_NUM] = {
+  [DPO_PROTO_IP4] = tap_inject_tx_nodes,
+  [DPO_PROTO_IP6] = tap_inject_tx_nodes,
+};
+
+static clib_error_t *
+tap_inject_init (vlib_main_t * vm)
+{
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  im->rx_node_index = tap_inject_rx_node.index;
+  im->tx_node_index = tap_inject_tx_node.index;
+  im->neighbor_node_index = tap_inject_neighbor_node.index;
+
+  tap_inject_dpo_type = dpo_register_new_type (&tap_inject_vft, tap_inject_nodes);
+
+  vec_alloc (im->rx_buffers, NUM_BUFFERS_TO_ALLOC);
+  vec_reset_length (im->rx_buffers);
+
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (tap_inject_init);
diff --git a/utils/extras/rtinject/tap_inject_tap.c b/utils/extras/rtinject/tap_inject_tap.c
new file mode 100644
index 000000000..a3ec9ffef
--- /dev/null
+++ b/utils/extras/rtinject/tap_inject_tap.c
@@ -0,0 +1,170 @@
+/*
+ * Copyright 2016 Intel Corporation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "tap_inject.h"
+
+#include <fcntl.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <linux/if.h>
+#include <linux/if_arp.h>
+#include <linux/if_ether.h>
+#include <linux/if_tun.h>
+#include <netinet/in.h>
+#include <vnet/unix/tuntap.h>
+
+#include <vlib/unix/unix.h>
+
+
+static clib_error_t *
+tap_inject_tap_read (clib_file_t * f)
+{
+  vlib_main_t * vm = vlib_get_main ();
+  tap_inject_main_t * im = tap_inject_get_main ();
+
+  vec_add1 (im->rx_file_descriptors, f->file_descriptor);
+
+  vlib_node_set_interrupt_pending (vm, im->rx_node_index);
+
+  return 0;
+}
+
+#define TAP_INJECT_TAP_BASE_NAME "vpp"
+
+clib_error_t *
+tap_inject_tap_connect (vnet_hw_interface_t * hw)
+{
+  vnet_main_t * vnet_main = vnet_get_main ();
+  vnet_sw_interface_t * sw = vnet_get_sw_interface (vnet_main, hw->hw_if_index);
+  static const int one = 1;
+  int fd;
+  struct ifreq ifr;
+  clib_file_t template;
+  u32 tap_fd;
+  u8 * name;
+
+  memset (&ifr, 0, sizeof (ifr));
+  memset (&template, 0, sizeof (template));
+
+  ASSERT (hw->hw_if_index == sw->sw_if_index);
+
+  /* Create the tap. */
+  tap_fd = open ("/dev/net/tun", O_RDWR);
+
+  if ((int)tap_fd < 0)
+    return clib_error_return (0, "failed to open tun device");
+
+  name = format (0, TAP_INJECT_TAP_BASE_NAME "%u%c", hw->hw_instance, 0);
+
+  strncpy (ifr.ifr_name, (char *) name, sizeof (ifr.ifr_name) - 1);
+  ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
+
+  if (ioctl (tap_fd, TUNSETIFF, (void *)&ifr) < 0)
+    {
+      close (tap_fd);
+      return clib_error_return (0, "failed to create tap");
+    }
+
+  if (ioctl (tap_fd, FIONBIO, &one) < 0)
+    {
+      close (tap_fd);
+      return clib_error_return (0, "failed to set tap to non-blocking io");
+    }
+
+  /* Open a socket to configure the device. */
+  fd = socket (PF_PACKET, SOCK_RAW, htons (ETH_P_ALL));
+
+  if (fd < 0)
+    {
+      close (tap_fd);
+      return clib_error_return (0, "failed to configure tap");
+    }
+
+  if (hw->hw_address)
+    clib_memcpy (ifr.ifr_hwaddr.sa_data, hw->hw_address, ETHER_ADDR_LEN);
+
+  ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
+
+  /* Set the hardware address. */
+  if (ioctl (fd, SIOCSIFHWADDR, &ifr) < 0)
+    {
+      close (tap_fd);
+      close (fd);
+      return clib_error_return (0, "failed to set tap hardware address");
+    }
+
+  /* Get the tap if index. */
+  if (ioctl (fd, SIOCGIFINDEX, &ifr) < 0)
+    {
+      close (tap_fd);
+      close (fd);
+      return clib_error_return (0, "failed to procure tap if index");
+    }
+
+  close (fd);
+
+  /* Get notified when the tap needs to be read. */
+  template.read_function = tap_inject_tap_read;
+  template.file_descriptor = tap_fd;
+
+  clib_file_add (&file_main, &template);
+
+  tap_inject_insert_tap (sw->sw_if_index, tap_fd, ifr.ifr_ifindex);
+
+  return 0;
+}
+
+clib_error_t *
+tap_inject_tap_disconnect (u32 sw_if_index)
+{
+  u32 tap_fd;
+
+  tap_fd = tap_inject_lookup_tap_fd (sw_if_index);
+  if (tap_fd == ~0)
+    return clib_error_return (0, "failed to disconnect tap");
+
+  tap_inject_delete_tap (sw_if_index);
+
+  close (tap_fd);
+  return 0;
+}
+
+
+u8 *
+format_tap_inject_tap_name (u8 * s, va_list * args)
+{
+  int fd;
+  struct ifreq ifr;
+
+  fd = socket (PF_PACKET, SOCK_RAW, htons (ETH_P_ALL));
+
+  if (fd < 0)
+    return 0;
+
+  memset (&ifr, 0, sizeof (ifr));
+
+  ifr.ifr_ifindex = va_arg (*args, u32);
+
+  if (ioctl (fd, SIOCGIFNAME, &ifr) < 0)
+    {
+      close (fd);
+      return 0;
+    }
+
+  close (fd);
+
+  return format (s, "%s", ifr.ifr_name);
+}