Add vpp native linux kernel AF_PACKET interface support

This is 1st drop of VPP native driver for linux AF_PACKET.

New CLI:

  create host-interface name <host-if-name>  [hw-addr <mac-address>]

References:
 - Documentation/networking/packet_mmap.txt in the Linux kernel tree
 - man 7 packet

Known issues:
 - attaching to linux bridge doesn't work
 - it is not expected to work in multicore setup

Change-Id: I1cb1c3d305f349759e90e76e25696718b73bd73d
Signed-off-by: Damjan Marion <damarion@cisco.com>

9 files changed, 920 insertions, 1 deletions

diff --git a/vlib/vlib/unix/input.c b/vlib/vlib/unix/input.c
index f9d7fece..62b32544 100644
--- a/vlib/vlib/unix/input.c
+++ b/vlib/vlib/unix/input.c
@@ -72,6 +72,8 @@ linux_epoll_file_update (unix_file_t * f,
   e.events = EPOLLIN;
   if (f->flags & UNIX_FILE_DATA_AVAILABLE_TO_WRITE)
     e.events |= EPOLLOUT;
+    if (f->flags & UNIX_FILE_EVENT_EDGE_TRIGGERED)
+      e.events |= EPOLLET;
   e.data.u32 = f - um->file_pool;
 
   if (epoll_ctl (em->epoll_fd,
diff --git a/vlib/vlib/unix/unix.h b/vlib/vlib/unix/unix.h
index 0802a93b..e3e6aa14 100644
--- a/vlib/vlib/unix/unix.h
+++ b/vlib/vlib/unix/unix.h
@@ -51,6 +51,7 @@ typedef struct unix_file {
 
   u32 flags;
 #define UNIX_FILE_DATA_AVAILABLE_TO_WRITE (1 << 0)
+#define UNIX_FILE_EVENT_EDGE_TRIGGERED   (1 << 1)
 
   /* Data available for function's use. */
   uword private_data;
diff --git a/vnet/Makefile.am b/vnet/Makefile.am
index cf1fb1e0..ff8a31da 100644
--- a/vnet/Makefile.am
+++ b/vnet/Makefile.am
@@ -607,6 +607,19 @@ nobase_include_HEADERS +=			\
   vnet/devices/ssvm/ssvm_eth.h
 
 ########################################
+# Linux packet interface
+########################################
+
+libvnet_la_SOURCES +=				\
+  vnet/devices/af_packet/af_packet.c		\
+  vnet/devices/af_packet/device.c		\
+  vnet/devices/af_packet/node.c			\
+  vnet/devices/af_packet/cli.c
+
+nobase_include_HEADERS +=			\
+  vnet/devices/af_packet/af_packet.h
+
+########################################
 # Unix kernel related
 ########################################
 
diff --git a/vnet/vnet/api_errno.h b/vnet/vnet/api_errno.h
index b4b55354..69535d3f 100644
--- a/vnet/vnet/api_errno.h
+++ b/vnet/vnet/api_errno.h
@@ -68,7 +68,8 @@ _(UNEXPECTED_INTF_STATE, -74, "Unexpected interface state")             \
 _(TUNNEL_EXIST, -75, "Tunnel already exists")                           \
 _(INVALID_DECAP_NEXT, -76, "Invalid decap-next")			\
 _(RESPONSE_NOT_READY, -77, "Response not ready")			\
-_(NOT_CONNECTED, -78, "Not connected to the data plane")
+_(NOT_CONNECTED, -78, "Not connected to the data plane")                \
+_(IF_ALREADY_EXISTS, -79, "Interface already exists")
 
 typedef enum {
 #define _(a,b,c) VNET_API_ERROR_##a = (b),
diff --git a/vnet/vnet/devices/af_packet/af_packet.c b/vnet/vnet/devices/af_packet/af_packet.c
new file mode 100644
index 00000000..93c8e0c6
--- /dev/null
+++ b/vnet/vnet/devices/af_packet/af_packet.c
@@ -0,0 +1,279 @@
+/*
+ *------------------------------------------------------------------
+ * af_packet.c - linux kernel packet interface
+ *
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#include <linux/if_ether.h>
+#include <linux/if_packet.h>
+
+#include <vlib/vlib.h>
+#include <vlib/unix/unix.h>
+#include <vnet/ip/ip.h>
+#include <vnet/ethernet/ethernet.h>
+
+#include <vnet/devices/af_packet/af_packet.h>
+
+#define AF_PACKET_DEBUG_SOCKET		0
+
+#define AF_PACKET_TX_FRAMES_PER_BLOCK	1024
+#define AF_PACKET_TX_FRAME_SIZE	 	(2048 * 5)
+#define AF_PACKET_TX_BLOCK_NR		1
+#define AF_PACKET_TX_FRAME_NR		(AF_PACKET_TX_BLOCK_NR * \
+					 AF_PACKET_TX_FRAMES_PER_BLOCK)
+#define AF_PACKET_TX_BLOCK_SIZE	 	(AF_PACKET_TX_FRAME_SIZE * \
+					 AF_PACKET_TX_FRAMES_PER_BLOCK)
+
+#define AF_PACKET_RX_FRAMES_PER_BLOCK	1024
+#define AF_PACKET_RX_FRAME_SIZE	 	(2048 * 5)
+#define AF_PACKET_RX_BLOCK_NR		1
+#define AF_PACKET_RX_FRAME_NR		(AF_PACKET_RX_BLOCK_NR * \
+					 AF_PACKET_RX_FRAMES_PER_BLOCK)
+#define AF_PACKET_RX_BLOCK_SIZE		(AF_PACKET_RX_FRAME_SIZE * \
+					 AF_PACKET_RX_FRAMES_PER_BLOCK)
+
+#if AF_PACKET_DEBUG_SOCKET == 1
+#define DBG_SOCK(args...) clib_warning(args);
+#else
+#define DBG_SOCK(args...)
+#endif
+
+/*defined in net/if.h but clashes with dpdk headers */
+unsigned int if_nametoindex(const char *ifname);
+
+typedef struct tpacket_req tpacket_req_t;
+
+static u32
+af_packet_eth_flag_change (vnet_main_t * vnm, vnet_hw_interface_t * hi, u32 flags)
+{
+  /* nothing for now */
+  return 0;
+}
+
+static clib_error_t * af_packet_fd_read_ready (unix_file_t * uf)
+{
+  vlib_main_t * vm = vlib_get_main();
+  af_packet_main_t * apm = &af_packet_main;
+  u32 idx = uf->private_data;
+
+  apm->pending_input_bitmap = clib_bitmap_set (apm->pending_input_bitmap, idx, 1);
+
+  /* Schedule the rx node */
+  vlib_node_set_interrupt_pending (vm, af_packet_input_node.index);
+
+  return 0;
+}
+
+static int
+create_packet_v2_sock(u8 * name, tpacket_req_t * rx_req, tpacket_req_t * tx_req,
+		      int *fd, u8 ** ring)
+{
+  int ret, err;
+  struct sockaddr_ll sll;
+  uint host_if_index;
+  int ver = TPACKET_V2;
+  socklen_t req_sz = sizeof(struct tpacket_req);
+  u32 ring_sz = rx_req->tp_block_size * rx_req->tp_block_nr +
+		tx_req->tp_block_size * tx_req->tp_block_nr;
+
+  host_if_index = if_nametoindex((const char *) name);
+
+  if (!host_if_index)
+    {
+      DBG_SOCK("Wrong host interface name");
+      ret = VNET_API_ERROR_INVALID_INTERFACE;
+      goto error;
+    }
+
+  if ((*fd = socket(AF_PACKET, SOCK_RAW, htons(ETH_P_ALL))) < 0)
+    {
+      DBG_SOCK("Failed to create socket");
+      ret = VNET_API_ERROR_SYSCALL_ERROR_1;
+      goto error;
+    }
+
+  if ((err = setsockopt(*fd, SOL_PACKET, PACKET_VERSION, &ver, sizeof(ver))) < 0)
+    {
+      DBG_SOCK("Failed to set rx packet interface version");
+      ret = VNET_API_ERROR_SYSCALL_ERROR_1;
+      goto error;
+    }
+
+  int opt = 1;
+  if ((err = setsockopt(*fd, SOL_PACKET, PACKET_LOSS, &opt, sizeof(opt))) < 0)
+    {
+      DBG_SOCK("Failed to set rx packet interface version");
+      ret = VNET_API_ERROR_SYSCALL_ERROR_1;
+      goto error;
+    }
+
+  if ((err = setsockopt(*fd, SOL_PACKET, PACKET_RX_RING, rx_req, req_sz)) < 0)
+    {
+      DBG_SOCK("Failed to set packet rx ring options");
+      ret = VNET_API_ERROR_SYSCALL_ERROR_1;
+      goto error;
+    }
+
+  if ((err = setsockopt(*fd, SOL_PACKET, PACKET_TX_RING, tx_req, req_sz)) < 0)
+    {
+      DBG_SOCK("Failed to set packet rx ring options");
+      ret = VNET_API_ERROR_SYSCALL_ERROR_1;
+      goto error;
+    }
+
+  *ring = mmap(NULL, ring_sz, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_LOCKED, *fd, 0);
+  if (*ring == MAP_FAILED)
+    {
+      DBG_SOCK("mmap failure");
+      ret = VNET_API_ERROR_SYSCALL_ERROR_1;
+      goto error;
+    }
+
+  memset(&sll, 0, sizeof(sll));
+  sll.sll_family = PF_PACKET;
+  sll.sll_protocol = htons(ETH_P_ALL);
+  sll.sll_ifindex = host_if_index;
+
+  if ((err = bind(*fd, (struct sockaddr *) &sll, sizeof(sll))) < 0)
+    {
+      DBG_SOCK("Failed to bind rx packet socket (error %d)", err);
+      ret = VNET_API_ERROR_SYSCALL_ERROR_1;
+      goto error;
+    }
+
+  return 0;
+error:
+  close(*fd);
+  return ret;
+}
+
+int
+af_packet_create_if(vlib_main_t * vm, u8 * host_if_name, u8 * hw_addr_set)
+{
+  af_packet_main_t * apm = &af_packet_main;
+  int ret, fd = -1;
+  struct tpacket_req * rx_req = 0;
+  struct tpacket_req * tx_req = 0;
+  u8 * ring = 0;
+  af_packet_if_t * apif = 0;
+  u8 hw_addr[6];
+  clib_error_t * error;
+  vnet_sw_interface_t * sw;
+  vnet_main_t *vnm = vnet_get_main();
+  uword * p;
+  uword if_index;
+
+  p = mhash_get (&apm->if_index_by_host_if_name, host_if_name);
+  if (p)
+    {
+      return VNET_API_ERROR_SUBIF_ALREADY_EXISTS;
+    }
+
+  vec_validate(rx_req, 0);
+  rx_req->tp_block_size = AF_PACKET_RX_BLOCK_SIZE;
+  rx_req->tp_frame_size = AF_PACKET_RX_FRAME_SIZE;
+  rx_req->tp_block_nr = AF_PACKET_RX_BLOCK_NR;
+  rx_req->tp_frame_nr = AF_PACKET_RX_FRAME_NR;
+
+  vec_validate(tx_req, 0);
+  tx_req->tp_block_size = AF_PACKET_TX_BLOCK_SIZE;
+  tx_req->tp_frame_size = AF_PACKET_TX_FRAME_SIZE;
+  tx_req->tp_block_nr = AF_PACKET_TX_BLOCK_NR;
+  tx_req->tp_frame_nr = AF_PACKET_TX_FRAME_NR;
+
+  ret = create_packet_v2_sock(host_if_name, rx_req, tx_req, &fd, &ring);
+
+  if (ret != 0)
+    goto error;
+
+  /* So far everything looks good, let's create interface */
+  vec_add2 (apm->interfaces, apif, 1);
+  if_index = apif - apm->interfaces;
+
+  apif->fd = fd;
+  apif->rx_ring = ring;
+  apif->tx_ring = ring + rx_req->tp_block_size * rx_req->tp_block_nr;
+  apif->rx_req = rx_req;
+  apif->tx_req = tx_req;
+  apif->host_if_name = host_if_name;
+
+  {
+    unix_file_t template = {0};
+    template.read_function = af_packet_fd_read_ready;
+    template.file_descriptor = fd;
+    template.private_data = if_index;
+    template.flags = UNIX_FILE_EVENT_EDGE_TRIGGERED;
+    apif->unix_file_index = unix_file_add (&unix_main, &template);
+  }
+
+  /*use configured or generate random MAC address */
+  if (hw_addr_set)
+    memcpy(hw_addr, hw_addr_set, 6);
+  else
+    {
+      f64 now = vlib_time_now(vm);
+      u32 rnd;
+      rnd = (u32) (now * 1e6);
+      rnd = random_u32 (&rnd);
+
+      memcpy (hw_addr+2, &rnd, sizeof(rnd));
+      hw_addr[0] = 2;
+      hw_addr[1] = 0xfe;
+    }
+
+  error = ethernet_register_interface(vnm, af_packet_device_class.index,
+				      if_index, hw_addr, &apif->hw_if_index,
+				      af_packet_eth_flag_change);
+
+  if (error)
+    {
+      memset(apif, 0, sizeof(*apif));
+      _vec_len(apm->interfaces) -= 1;
+      clib_error_report (error);
+      ret = VNET_API_ERROR_SYSCALL_ERROR_1;
+      goto error;
+    }
+
+  sw = vnet_get_hw_sw_interface (vnm, apif->hw_if_index);
+  apif->sw_if_index = sw->sw_if_index;
+
+  vnet_hw_interface_set_flags (vnm, apif->hw_if_index,
+			       VNET_HW_INTERFACE_FLAG_LINK_UP);
+
+  mhash_set_mem (&apm->if_index_by_host_if_name, host_if_name, &if_index, 0);
+
+  return 0;
+
+error:
+  vec_free(host_if_name);
+  vec_free(rx_req);
+  vec_free(tx_req);
+  return ret;
+}
+
+static clib_error_t *
+af_packet_init (vlib_main_t * vm)
+{
+  af_packet_main_t * apm = &af_packet_main;
+
+  memset (apm, 0, sizeof (af_packet_main_t));
+
+  mhash_init_vec_string (&apm->if_index_by_host_if_name, sizeof (uword));
+
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (af_packet_init);
diff --git a/vnet/vnet/devices/af_packet/af_packet.h b/vnet/vnet/devices/af_packet/af_packet.h
new file mode 100644
index 00000000..dd9fa148
--- /dev/null
+++ b/vnet/vnet/devices/af_packet/af_packet.h
@@ -0,0 +1,57 @@
+/*
+ *------------------------------------------------------------------
+ * af_packet.h - linux kernel packet interface header file
+ *
+ * Copyright (c) 2016 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.
+ *------------------------------------------------------------------
+ */
+
+typedef struct {
+  CLIB_CACHE_LINE_ALIGN_MARK(cacheline0);
+  u8 * host_if_name;
+  int fd;
+  struct tpacket_req * rx_req;
+  struct tpacket_req * tx_req;
+  u8 * rx_ring;
+  u8 * tx_ring;
+  u32 hw_if_index;
+  u32 sw_if_index;
+  u32 unix_file_index;
+
+  u32 next_rx_frame;
+  u32 next_tx_frame;
+
+  u32 per_interface_next_index;
+  u8 is_admin_up;
+} af_packet_if_t;
+
+typedef struct {
+  CLIB_CACHE_LINE_ALIGN_MARK(cacheline0);
+  af_packet_if_t * interfaces;
+
+  /* bitmap of pending rx interfaces */
+  uword * pending_input_bitmap;
+
+  /* rx buffer cache */
+  u32 * rx_buffers;
+
+  /* hash of host interface names */
+  mhash_t if_index_by_host_if_name;
+} af_packet_main_t;
+
+af_packet_main_t af_packet_main;
+extern vnet_device_class_t af_packet_device_class;
+extern vlib_node_registration_t af_packet_input_node;
+
+int af_packet_create_if(vlib_main_t * vm, u8 * host_if_name, u8 * hw_addr_set);
diff --git a/vnet/vnet/devices/af_packet/cli.c b/vnet/vnet/devices/af_packet/cli.c
new file mode 100644
index 00000000..d8d829f6
--- /dev/null
+++ b/vnet/vnet/devices/af_packet/cli.c
@@ -0,0 +1,89 @@
+/*
+ *------------------------------------------------------------------
+ * af_packet.c - linux kernel packet interface
+ *
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#include <fcntl.h>		/* for open */
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <sys/uio.h>		/* for iovec */
+#include <netinet/in.h>
+
+#include <vlib/vlib.h>
+#include <vlib/unix/unix.h>
+#include <vnet/ip/ip.h>
+#include <vnet/ethernet/ethernet.h>
+
+#include <vnet/devices/af_packet/af_packet.h>
+
+static clib_error_t *
+af_packet_create_command_fn (vlib_main_t * vm, unformat_input_t * input,
+			     vlib_cli_command_t * cmd)
+{
+  unformat_input_t _line_input, * line_input = &_line_input;
+  u8 * host_if_name = NULL;
+  u8 hwaddr [6];
+  u8 * hw_addr_ptr = 0;
+  int r;
+
+  /* Get a line of input. */
+  if (! unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat (line_input, "name %s", &host_if_name))
+	;
+      else if (unformat (line_input, "hw-addr %U", unformat_ethernet_address, hwaddr))
+	hw_addr_ptr = hwaddr;
+      else
+	return clib_error_return (0, "unknown input `%U'", format_unformat_error, input);
+    }
+  unformat_free (line_input);
+
+  if (host_if_name == NULL)
+      return clib_error_return (0, "missing host interface name");
+
+  r = af_packet_create_if(vm, host_if_name, hw_addr_ptr);
+
+  if (r == VNET_API_ERROR_SYSCALL_ERROR_1)
+    return clib_error_return(0, "%s (errno %d)", strerror (errno), errno);
+
+  if (r == VNET_API_ERROR_INVALID_INTERFACE)
+    return clib_error_return(0, "Invalid interface name");
+
+  if (r == VNET_API_ERROR_SUBIF_ALREADY_EXISTS)
+    return clib_error_return(0, "Interface elready exists");
+
+  return 0;
+}
+
+VLIB_CLI_COMMAND (af_packet_create_command, static) = {
+  .path = "create host-interface",
+  .short_help = "create host-interface name <interface name> [hw-addr <mac>]",
+  .function = af_packet_create_command_fn,
+};
+
+clib_error_t *
+af_packet_cli_init (vlib_main_t * vm)
+{
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (af_packet_cli_init);
diff --git a/vnet/vnet/devices/af_packet/device.c b/vnet/vnet/devices/af_packet/device.c
new file mode 100644
index 00000000..c00c6895
--- /dev/null
+++ b/vnet/vnet/devices/af_packet/device.c
@@ -0,0 +1,198 @@
+/*
+ *------------------------------------------------------------------
+ * af_packet.c - linux kernel packet interface
+ *
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#include <linux/if_packet.h>
+
+#include <vlib/vlib.h>
+#include <vlib/unix/unix.h>
+#include <vnet/ip/ip.h>
+#include <vnet/ethernet/ethernet.h>
+
+#include <vnet/devices/af_packet/af_packet.h>
+
+#define foreach_af_packet_tx_func_error	       \
+_(FRAME_NOT_READY, "tx frame not ready")
+
+typedef enum {
+#define _(f,s) AF_PACKET_TX_ERROR_##f,
+  foreach_af_packet_tx_func_error
+#undef _
+  AF_PACKET_TX_N_ERROR,
+} af_packet_tx_func_error_t;
+
+static char * af_packet_tx_func_error_strings[] = {
+#define _(n,s) s,
+    foreach_af_packet_tx_func_error
+#undef _
+};
+
+
+static u8 * format_af_packet_device_name (u8 * s, va_list * args)
+{
+  u32 i = va_arg (*args, u32);
+  af_packet_main_t * apm = &af_packet_main;
+  af_packet_if_t * apif = vec_elt_at_index (apm->interfaces, i);
+
+  s = format (s, "host-%s", apif->host_if_name);
+  return s;
+}
+
+static u8 * format_af_packet_device (u8 * s, va_list * args)
+{
+  s = format (s, "Linux PACKET socket interface");
+  return s;
+}
+
+static u8 * format_af_packet_tx_trace (u8 * s, va_list * args)
+{
+  s = format (s, "Unimplemented...");
+  return s;
+}
+
+static uword
+af_packet_interface_tx (vlib_main_t * vm,
+		       vlib_node_runtime_t * node,
+		       vlib_frame_t * frame)
+{
+  af_packet_main_t * apm = &af_packet_main;
+  u32 * buffers = vlib_frame_args (frame);
+  u32 n_left = frame->n_vectors;
+  u32 n_sent = 0;
+  vnet_interface_output_runtime_t * rd = (void *) node->runtime_data;
+  af_packet_if_t * apif = vec_elt_at_index (apm->interfaces, rd->dev_instance);
+  int block = 0;
+  u32 block_size = apif->tx_req->tp_block_size;
+  u32 frame_size = apif->tx_req->tp_frame_size;
+  u32 frame_num = apif->tx_req->tp_frame_nr;
+  u8 * block_start = apif->tx_ring + block * block_size;
+  u32 tx_frame = apif->next_tx_frame;
+  struct tpacket2_hdr * tph;
+  u32 frame_not_ready = 0;
+
+  while(n_left > 0)
+    {
+      u32 len;
+      u32 offset = 0;
+      vlib_buffer_t * b0;
+      n_left--;
+      u32 bi = buffers[0];
+      buffers++;
+
+      tph = (struct tpacket2_hdr *) (block_start + tx_frame * frame_size);
+
+      if(tph->tp_status & (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING))
+	{
+	  frame_not_ready++;
+	  goto next;
+	}
+
+      do
+	{
+	  b0 = vlib_get_buffer (vm, bi);
+	  len = b0->current_length;
+	  memcpy((u8 *) tph + TPACKET_ALIGN(sizeof(struct tpacket2_hdr)) + offset,
+		 vlib_buffer_get_current(b0), len);
+	  offset += len;
+	}
+      while ((bi = b0->next_buffer));
+
+      tph->tp_len = tph->tp_snaplen = offset;
+      tph->tp_status = TP_STATUS_SEND_REQUEST;
+      n_sent++;
+next:
+      tx_frame = (tx_frame + 1) % frame_num;
+    }
+
+  CLIB_MEMORY_BARRIER();
+
+  if (n_sent)
+    {
+      apif->next_tx_frame = tx_frame;
+      if (sendto(apif->fd, NULL, 0, MSG_DONTWAIT, NULL, 0) == -1)
+	clib_unix_error("tx sendto failure");
+    }
+
+  if (frame_not_ready)
+    vlib_error_count (vm, node->node_index, AF_PACKET_TX_ERROR_FRAME_NOT_READY,
+		      frame_not_ready);
+
+  vlib_buffer_free (vm, vlib_frame_args (frame), frame->n_vectors);
+  return frame->n_vectors;
+}
+
+static void
+af_packet_set_interface_next_node (vnet_main_t *vnm, u32 hw_if_index,
+				  u32 node_index)
+{
+  af_packet_main_t * apm = &af_packet_main;
+  vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
+  af_packet_if_t * apif = pool_elt_at_index (apm->interfaces, hw->dev_instance);
+
+  /* Shut off redirection */
+  if (node_index == ~0)
+    {
+      apif->per_interface_next_index = node_index;
+      return;
+    }
+
+  apif->per_interface_next_index =
+    vlib_node_add_next (vlib_get_main(), af_packet_input_node.index, node_index);
+}
+
+static void af_packet_clear_hw_interface_counters (u32 instance)
+{
+  /* Nothing for now */
+}
+
+static clib_error_t *
+af_packet_interface_admin_up_down (vnet_main_t * vnm, u32 hw_if_index, u32 flags)
+{
+  af_packet_main_t * apm = &af_packet_main;
+  vnet_hw_interface_t *hw = vnet_get_hw_interface (vnm, hw_if_index);
+  af_packet_if_t * apif = pool_elt_at_index (apm->interfaces, hw->dev_instance);
+
+  apif->is_admin_up = (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) != 0;
+
+  return 0;
+}
+
+static clib_error_t *
+af_packet_subif_add_del_function (vnet_main_t * vnm,
+				 u32 hw_if_index,
+				 struct vnet_sw_interface_t * st,
+				 int is_add)
+{
+  /* Nothing for now */
+  return 0;
+}
+
+VNET_DEVICE_CLASS (af_packet_device_class) = {
+  .name = "af-packet",
+  .tx_function = af_packet_interface_tx,
+  .format_device_name = format_af_packet_device_name,
+  .format_device = format_af_packet_device,
+  .format_tx_trace = format_af_packet_tx_trace,
+  .tx_function_n_errors = AF_PACKET_TX_N_ERROR,
+  .tx_function_error_strings = af_packet_tx_func_error_strings,
+  .rx_redirect_to_node = af_packet_set_interface_next_node,
+  .clear_counters = af_packet_clear_hw_interface_counters,
+  .admin_up_down_function = af_packet_interface_admin_up_down,
+  .subif_add_del_function = af_packet_subif_add_del_function,
+  .no_flatten_output_chains = 1,
+};
\ No newline at end of file
diff --git a/vnet/vnet/devices/af_packet/node.c b/vnet/vnet/devices/af_packet/node.c
new file mode 100644
index 00000000..eadb5923
--- /dev/null
+++ b/vnet/vnet/devices/af_packet/node.c
@@ -0,0 +1,279 @@
+/*
+ *------------------------------------------------------------------
+ * af_packet.c - linux kernel packet interface
+ *
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *------------------------------------------------------------------
+ */
+
+#include <linux/if_packet.h>
+
+#include <vlib/vlib.h>
+#include <vlib/unix/unix.h>
+#include <vnet/ip/ip.h>
+#include <vnet/ethernet/ethernet.h>
+
+#include <vnet/devices/af_packet/af_packet.h>
+
+#define foreach_af_packet_input_error
+
+typedef enum {
+#define _(f,s) AF_PACKET_INPUT_ERROR_##f,
+  foreach_af_packet_input_error
+#undef _
+  AF_PACKET_INPUT_N_ERROR,
+} af_packet_input_error_t;
+
+static char * af_packet_input_error_strings[] = {
+#define _(n,s) s,
+    foreach_af_packet_input_error
+#undef _
+};
+
+enum {
+  AF_PACKET_INPUT_NEXT_DROP,
+  AF_PACKET_INPUT_NEXT_ETHERNET_INPUT,
+  AF_PACKET_INPUT_N_NEXT,
+};
+
+typedef struct {
+  u32 next_index;
+  u32 hw_if_index;
+  int block;
+  struct tpacket2_hdr tph;
+} af_packet_input_trace_t;
+
+static u8 * format_af_packet_input_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 *);
+  af_packet_input_trace_t * t = va_arg (*args, af_packet_input_trace_t *);
+  uword indent = format_get_indent (s);
+
+  s = format (s, "af_packet: hw_if_index %d next-index %d",
+	      t->hw_if_index, t->next_index);
+
+  s = format (s, "\n%Utpacket2_hdr:\n%Ustatus 0x%x len %u snaplen %u mac %u net %u"
+	      "\n%Usec 0x%x nsec 0x%x vlan_tci %u"
+#ifdef TP_STATUS_VLAN_TPID_VALID
+	      " vlan_tpid %u"
+#endif
+	      ,
+	      format_white_space, indent + 2,
+	      format_white_space, indent + 4,
+	      t->tph.tp_status,
+	      t->tph.tp_len,
+	      t->tph.tp_snaplen,
+	      t->tph.tp_mac,
+	      t->tph.tp_net,
+	      format_white_space, indent + 4,
+	      t->tph.tp_sec,
+	      t->tph.tp_nsec,
+	      t->tph.tp_vlan_tci,
+#ifdef TP_STATUS_VLAN_TPID_VALID
+	      t->tph.tp_vlan_tpid,
+#endif
+	      t->tph.tp_net);
+  return s;
+}
+
+always_inline void
+buffer_add_to_chain(vlib_main_t *vm, u32 bi, u32 first_bi, u32 prev_bi)
+{
+  vlib_buffer_t * b = vlib_get_buffer (vm, bi);
+  vlib_buffer_t * first_b = vlib_get_buffer (vm, first_bi);
+  vlib_buffer_t * prev_b = vlib_get_buffer (vm, prev_bi);
+#if DPDK > 0
+  struct rte_mbuf * mbuf = ((struct rte_mbuf *) b) - 1;
+  struct rte_mbuf * first_mbuf = ((struct rte_mbuf *) first_b) - 1;
+  struct rte_mbuf * prev_mbuf = ((struct rte_mbuf *) prev_b) - 1;
+#endif
+
+  /* update first buffer */
+  first_b->total_length_not_including_first_buffer +=  b->current_length;
+
+  /* update previous buffer */
+  prev_b->next_buffer = bi;
+  prev_b->flags |= VLIB_BUFFER_NEXT_PRESENT;
+
+  /* update current buffer */
+  b->next_buffer = 0;
+
+#if DPDK > 0
+  first_mbuf->nb_segs++;
+  prev_mbuf->next = mbuf;
+  mbuf->data_len = b->current_length;
+  mbuf->data_off = RTE_PKTMBUF_HEADROOM + b->current_data;
+  mbuf->next = 0;
+#endif
+}
+
+always_inline uword
+af_packet_device_input_fn  (vlib_main_t * vm, vlib_node_runtime_t * node,
+			    vlib_frame_t * frame, u32 device_idx)
+{
+  af_packet_main_t * apm = &af_packet_main;
+  af_packet_if_t * apif = vec_elt_at_index(apm->interfaces, device_idx);
+  struct tpacket2_hdr *tph;
+  u32 next_index = AF_PACKET_INPUT_NEXT_ETHERNET_INPUT;
+  u32 block = 0;
+  u32 rx_frame;
+  u32 n_free_bufs;
+  u32 n_rx_packets = 0;
+  u32 n_rx_bytes = 0;
+  u32 * to_next = 0;
+  u32 block_size = apif->rx_req->tp_block_size;
+  u32 frame_size = apif->rx_req->tp_frame_size;
+  u32 frame_num = apif->rx_req->tp_frame_nr;
+  u8 * block_start = apif->rx_ring + block * block_size;
+  uword n_trace = vlib_get_trace_count (vm, node);
+  u32 n_buffer_bytes = vlib_buffer_free_list_buffer_size (vm,
+    VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX);
+  u32 min_bufs = apif->rx_req->tp_frame_size / n_buffer_bytes;
+
+  n_free_bufs = vec_len (apm->rx_buffers);
+  if (PREDICT_FALSE(n_free_bufs < VLIB_FRAME_SIZE))
+    {
+      vec_validate(apm->rx_buffers, VLIB_FRAME_SIZE + n_free_bufs - 1);
+      n_free_bufs += vlib_buffer_alloc(vm, &apm->rx_buffers[n_free_bufs], VLIB_FRAME_SIZE);
+      _vec_len (apm->rx_buffers) = n_free_bufs;
+    }
+
+  rx_frame = apif->next_rx_frame;
+  tph = (struct tpacket2_hdr *) (block_start + rx_frame * frame_size);
+  while ((tph->tp_status & TP_STATUS_USER) && (n_free_bufs > min_bufs))
+    {
+      vlib_buffer_t * b0, * first_b0 = 0;
+      u32 next0 = AF_PACKET_INPUT_NEXT_ETHERNET_INPUT;
+
+      u32 n_left_to_next;
+      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+      while ((tph->tp_status & TP_STATUS_USER) && (n_free_bufs > min_bufs) &&
+	     n_left_to_next)
+	{
+	  u32 data_len = tph->tp_snaplen;
+	  u32 offset = 0;
+	  u32 bi0, first_bi0 = 0, prev_bi0;
+
+	  while (data_len)
+	    {
+	      /* grab free buffer */
+	      u32 last_empty_buffer = vec_len (apm->rx_buffers) - 1;
+	      prev_bi0 = bi0;
+	      bi0 = apm->rx_buffers[last_empty_buffer];
+	      b0 = vlib_get_buffer (vm, bi0);
+	      _vec_len (apm->rx_buffers) = last_empty_buffer;
+	      n_free_bufs--;
+
+	      /* copy data */
+	      u32 bytes_to_copy = data_len > n_buffer_bytes ? n_buffer_bytes : data_len;
+	      b0->current_data = 0;
+	      memcpy (vlib_buffer_get_current (b0), (u8 *) tph + tph->tp_mac + offset, bytes_to_copy);
+
+	      /* fill buffer header */
+	      b0->clone_count = 0;
+	      b0->current_length = bytes_to_copy;
+
+	      if (offset == 0)
+		{
+		  b0->total_length_not_including_first_buffer = 0;
+		  b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID;
+		  vnet_buffer(b0)->sw_if_index[VLIB_RX] = apif->sw_if_index;
+		  vnet_buffer(b0)->sw_if_index[VLIB_TX] = (u32)~0;
+		  first_bi0 = bi0;
+		  first_b0 = vlib_get_buffer(vm, first_bi0);
+		}
+	      else
+		buffer_add_to_chain(vm, bi0, first_bi0, prev_bi0);
+
+	      offset += bytes_to_copy;
+	      data_len -= bytes_to_copy;
+	    }
+	  n_rx_packets++;
+	  n_rx_bytes += tph->tp_snaplen;
+	  to_next[0] = first_bi0;
+	  to_next += 1;
+	  n_left_to_next--;
+
+	  /* trace */
+	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT(first_b0);
+	  if (PREDICT_FALSE(n_trace > 0))
+	    {
+	      af_packet_input_trace_t *tr;
+	      vlib_trace_buffer (vm, node, next0, first_b0, /* follow_chain */ 0);
+	      vlib_set_trace_count (vm, node, --n_trace);
+	      tr = vlib_add_trace (vm, node, first_b0, sizeof (*tr));
+	      tr->next_index = next0;
+	      tr->hw_if_index = apif->hw_if_index;
+	      memcpy(&tr->tph, tph, sizeof(struct tpacket2_hdr));
+	    }
+	  /* enque and take next packet */
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, first_bi0, next0);
+
+	   /* next packet */
+	  tph->tp_status = TP_STATUS_KERNEL;
+	  rx_frame = (rx_frame + 1) % frame_num;
+	  tph = (struct tpacket2_hdr *) (block_start + rx_frame * frame_size);
+	}
+
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  apif->next_rx_frame = rx_frame;
+
+  vlib_increment_combined_counter
+    (vnet_get_main()->interface_main.combined_sw_if_counters
+     + VNET_INTERFACE_COUNTER_RX,
+     os_get_cpu_number(),
+     apif->hw_if_index,
+     n_rx_packets, n_rx_bytes);
+
+  return n_rx_packets;
+}
+
+static uword
+af_packet_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
+		    vlib_frame_t * frame)
+{
+  int i;
+  u32 n_rx_packets = 0;
+
+  af_packet_main_t * apm = &af_packet_main;
+
+  clib_bitmap_foreach (i, apm->pending_input_bitmap,
+    ({
+      clib_bitmap_set (apm->pending_input_bitmap, i, 1);
+      n_rx_packets += af_packet_device_input_fn(vm, node, frame, i);
+    }));
+
+  return n_rx_packets;
+}
+
+
+VLIB_REGISTER_NODE (af_packet_input_node) = {
+  .function = af_packet_input_fn,
+  .name = "af-packet-input",
+  .format_trace = format_af_packet_input_trace,
+  .type = VLIB_NODE_TYPE_INPUT,
+  .state = VLIB_NODE_STATE_INTERRUPT,
+  .n_errors = AF_PACKET_INPUT_N_ERROR,
+  .error_strings = af_packet_input_error_strings,
+
+  .n_next_nodes = AF_PACKET_INPUT_N_NEXT,
+  .next_nodes = {
+    [AF_PACKET_INPUT_NEXT_DROP] = "error-drop",
+    [AF_PACKET_INPUT_NEXT_ETHERNET_INPUT] = "ethernet-input",
+  },
+};