[dpdk-dev] [PATCH 2/2] kdp: add virtual PMD for kernel slow data path communication

[Ferruh Yigit <ferruh.yigit@intel.com>]

This patch provides slow data path communication to the Linux kernel.
Patch is based on librte_kni, and heavily re-uses it.

The main difference is librte_kni library converted into a PMD, to
provide ease of use for applications.

Now any application can use slow path communication without any update
in application, because of existing eal support for virtual PMD.

Also this PMD supports two methods to send packets to the Linux, first
one is custom FIFO implementation with help of KDP kernel module, second
one is Linux in-kernel tun/tap support. PMD first checks for KDP kernel
module, if fails it tries to create and use a tap interface.

With FIFO method: PMD's rx_pkt_burst() get packets from FIFO,
and tx_pkt_burst() puts packet to the FIFO.
The corresponding Linux virtual network device driver code
also gets/puts packets from FIFO as they are coming from hardware.

With tun/tap method: no external kernel module required, PMD reads from
and writes packets to the tap interface file descriptor. Tap interface
has performance penalty against FIFO implementation.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
 config/common_linuxapp                  |   1 +
 doc/guides/nics/pcap_ring.rst           | 125 ++++++++-
 doc/guides/rel_notes/release_2_3.rst    |   6 +
 drivers/net/Makefile                    |   3 +-
 drivers/net/kdp/Makefile                |  61 ++++
 drivers/net/kdp/rte_eth_kdp.c           | 481 ++++++++++++++++++++++++++++++++
 drivers/net/kdp/rte_kdp.c               | 365 ++++++++++++++++++++++++
 drivers/net/kdp/rte_kdp.h               | 126 +++++++++
 drivers/net/kdp/rte_kdp_fifo.h          |  91 ++++++
 drivers/net/kdp/rte_kdp_tap.c           |  96 +++++++
 drivers/net/kdp/rte_pmd_kdp_version.map |   4 +
 lib/librte_eal/common/include/rte_log.h |   3 +-
 mk/rte.app.mk                           |   3 +-
 13 files changed, 1359 insertions(+), 6 deletions(-)
 create mode 100644 drivers/net/kdp/Makefile
 create mode 100644 drivers/net/kdp/rte_eth_kdp.c
 create mode 100644 drivers/net/kdp/rte_kdp.c
 create mode 100644 drivers/net/kdp/rte_kdp.h
 create mode 100644 drivers/net/kdp/rte_kdp_fifo.h
 create mode 100644 drivers/net/kdp/rte_kdp_tap.c
 create mode 100644 drivers/net/kdp/rte_pmd_kdp_version.map

diff --git a/config/common_linuxapp b/config/common_linuxapp
index 73c91d8..b9dec0c 100644
--- a/config/common_linuxapp
+++ b/config/common_linuxapp
@@ -322,6 +322,7 @@ CONFIG_RTE_LIBRTE_PMD_NULL=y
 #
 # Compile KDP PMD
 #
+CONFIG_RTE_LIBRTE_PMD_KDP=y
 CONFIG_RTE_KDP_KMOD=y
 CONFIG_RTE_KDP_PREEMPT_DEFAULT=y
 
diff --git a/doc/guides/nics/pcap_ring.rst b/doc/guides/nics/pcap_ring.rst
index 46aa3ac..78b7b61 100644
--- a/doc/guides/nics/pcap_ring.rst
+++ b/doc/guides/nics/pcap_ring.rst
@@ -28,11 +28,11 @@
     (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
     OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
-Libpcap and Ring Based Poll Mode Drivers
-========================================
+Software Poll Mode Drivers
+==========================
 
 In addition to Poll Mode Drivers (PMDs) for physical and virtual hardware,
-the DPDK also includes two pure-software PMDs. These two drivers are:
+the DPDK also includes pure-software PMDs. These drivers are:
 
 *   A libpcap -based PMD (librte_pmd_pcap) that reads and writes packets using libpcap,
     - both from files on disk, as well as from physical NIC devices using standard Linux kernel drivers.
@@ -40,6 +40,10 @@ the DPDK also includes two pure-software PMDs. These two drivers are:
 *   A ring-based PMD (librte_pmd_ring) that allows a set of software FIFOs (that is, rte_ring)
     to be accessed using the PMD APIs, as though they were physical NICs.
 
+*   A slow data path PMD (librte_pmd_kdp) that allows send/get packets to/from OS network
+    stack as it is a physical NIC.
+
+
 .. note::
 
     The libpcap -based PMD is disabled by default in the build configuration files,
@@ -211,6 +215,121 @@ Multiple devices may be specified, separated by commas.
     Done.
 
 
+Kernel Data Path PMD
+~~~~~~~~~~~~~~~~~~~~
+
+Kernel Data Path (KDP) PMD is to communicate with OS network stack easily by application.
+
+.. code-block:: console
+
+        ./testpmd --vdev eth_kdp0 --vdev eth_kdp1 -- -i
+        ...
+        Configuring Port 0 (socket 0)
+        Port 0: 00:00:00:00:00:00
+        Configuring Port 1 (socket 0)
+        Port 1: 00:00:00:00:00:00
+        Checking link statuses...
+        Port 0 Link Up - speed 10000 Mbps - full-duplex
+        Port 1 Link Up - speed 10000 Mbps - full-duplex
+        Done
+
+KDP PMD supports two type of communication:
+
+* Custom FIFO implementation
+* tun/tap implementation
+
+Custom FIFO implementation gives more performance but requires KDP kernel module (rte_kdp.ko) inserted.
+
+By default FIFO communication has priority, if KDP kernel module is not inserted, tun/tap communication used.
+
+If KDP kernel module inserted, above testpmd command will create following virtual interfaces, these can be used as any interface.
+
+.. code-block:: console
+
+        # ifconfig kdp0; ifconfig kdp1
+        kdp0: flags=4098<BROADCAST,MULTICAST>  mtu 1500
+                ether 00:00:00:00:00:00  txqueuelen 1000  (Ethernet)
+                RX packets 0  bytes 0 (0.0 B)
+                RX errors 0  dropped 0  overruns 0  frame 0
+                TX packets 0  bytes 0 (0.0 B)
+                TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0
+
+        kdp1: flags=4098<BROADCAST,MULTICAST>  mtu 1500
+                ether 00:00:00:00:00:00  txqueuelen 1000  (Ethernet)
+                RX packets 0  bytes 0 (0.0 B)
+                RX errors 0  dropped 0  overruns 0  frame 0
+                TX packets 0  bytes 0 (0.0 B)
+                TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0
+
+
+With tun/tap communication method, following interfaces are created:
+
+.. code-block:: console
+
+        # ifconfig tap_kdp0; ifconfig tap_kdp1
+        tap_kdp0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1500
+                inet6 fe80::341f:afff:feb7:23db  prefixlen 64  scopeid 0x20<link>
+                ether 36:1f:af:b7:23:db  txqueuelen 500  (Ethernet)
+                RX packets 126624864  bytes 6184828655 (5.7 GiB)
+                RX errors 0  dropped 0  overruns 0  frame 0
+                TX packets 126236898  bytes 6150306636 (5.7 GiB)
+                TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0
+
+        tap_kdp1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1500
+                inet6 fe80::f030:b4ff:fe94:b720  prefixlen 64  scopeid 0x20<link>
+                ether f2:30:b4:94:b7:20  txqueuelen 500  (Ethernet)
+                RX packets 126237370  bytes 6150329717 (5.7 GiB)
+                RX errors 0  dropped 9  overruns 0  frame 0
+                TX packets 126624896  bytes 6184826874 (5.7 GiB)
+                TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0
+
+DPDK application can be used to forward packages between these interfaces:
+
+.. code-block:: console
+
+        In Linux:
+        ip l add br0 type bridge
+        ip l set tap_kdp0 master br0
+        ip l set tap_kdp1 master br0
+        ip l set br0 up
+        ip l set tap_kdp0 up
+        ip l set tap_kdp1 up
+
+
+        In testpmd:
+        testpmd> start
+          io packet forwarding - CRC stripping disabled - packets/burst=32
+          nb forwarding cores=1 - nb forwarding ports=2
+          RX queues=1 - RX desc=128 - RX free threshold=0
+          RX threshold registers: pthresh=0 hthresh=0 wthresh=0
+          TX queues=1 - TX desc=512 - TX free threshold=0
+          TX threshold registers: pthresh=0 hthresh=0 wthresh=0
+          TX RS bit threshold=0 - TXQ flags=0x0
+        testpmd> stop
+        Telling cores to stop...
+        Waiting for lcores to finish...
+
+          ---------------------- Forward statistics for port 0  ----------------------
+          RX-packets: 973900         RX-dropped: 0             RX-total: 973900
+          TX-packets: 973903         TX-dropped: 0             TX-total: 973903
+          ----------------------------------------------------------------------------
+
+          ---------------------- Forward statistics for port 1  ----------------------
+          RX-packets: 973903         RX-dropped: 0             RX-total: 973903
+          TX-packets: 973900         TX-dropped: 0             TX-total: 973900
+          ----------------------------------------------------------------------------
+
+          +++++++++++++++ Accumulated forward statistics for all ports+++++++++++++++
+          RX-packets: 1947803        RX-dropped: 0             RX-total: 1947803
+          TX-packets: 1947803        TX-dropped: 0             TX-total: 1947803
+          ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+
+        Done.
+
+
+
+
+
 Using the Poll Mode Driver from an Application
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
diff --git a/doc/guides/rel_notes/release_2_3.rst b/doc/guides/rel_notes/release_2_3.rst
index 99de186..faf6a17 100644
--- a/doc/guides/rel_notes/release_2_3.rst
+++ b/doc/guides/rel_notes/release_2_3.rst
@@ -4,6 +4,12 @@ DPDK Release 2.3
 New Features
 ------------
 
+* **Added Slow Data Path support.**
+
+  * This is based on KNI work and in long term intends to replace it.
+  * Added Kernel Data Path (KDP) kernel module.
+  * Added KDP virtual PMD.
+
 
 Resolved Issues
 ---------------
diff --git a/drivers/net/Makefile b/drivers/net/Makefile
index 6e4497e..0be06f5 100644
--- a/drivers/net/Makefile
+++ b/drivers/net/Makefile
@@ -1,6 +1,6 @@
 #   BSD LICENSE
 #
-#   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+#   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
 #   All rights reserved.
 #
 #   Redistribution and use in source and binary forms, with or without
@@ -51,6 +51,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2) += szedata2
 DIRS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD) += virtio
 DIRS-$(CONFIG_RTE_LIBRTE_VMXNET3_PMD) += vmxnet3
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_XENVIRT) += xenvirt
+DIRS-$(CONFIG_RTE_LIBRTE_PMD_KDP) += kdp
 
 include $(RTE_SDK)/mk/rte.sharelib.mk
 include $(RTE_SDK)/mk/rte.subdir.mk
diff --git a/drivers/net/kdp/Makefile b/drivers/net/kdp/Makefile
new file mode 100644
index 0000000..035056e
--- /dev/null
+++ b/drivers/net/kdp/Makefile
@@ -0,0 +1,61 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2016 Intel Corporation. All rights reserved.
+#
+#   Redistribution and use in source and binary forms, with or without
+#   modification, are permitted provided that the following conditions
+#   are met:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of Intel Corporation nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_pmd_kdp.a
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+EXPORT_MAP := rte_pmd_kdp_version.map
+
+LIBABIVER := 1
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_KDP) += rte_eth_kdp.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_KDP) += rte_kdp.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_KDP) += rte_kdp_tap.c
+
+#
+# Export include files
+#
+SYMLINK-y-include +=
+
+# this lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_KDP) += lib/librte_mbuf
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_KDP) += lib/librte_ether
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/kdp/rte_eth_kdp.c b/drivers/net/kdp/rte_eth_kdp.c
new file mode 100644
index 0000000..ac650d7
--- /dev/null
+++ b/drivers/net/kdp/rte_eth_kdp.c
@@ -0,0 +1,481 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <rte_ethdev.h>
+#include <rte_dev.h>
+#include <rte_kvargs.h>
+
+#include "rte_kdp.h"
+
+#define MAX_PACKET_SZ 2048
+
+struct kdp_queue {
+	struct pmd_internals *internals;
+	struct rte_mempool *mb_pool;
+
+	uint64_t rx_pkts;
+	uint64_t rx_bytes;
+	uint64_t rx_err_pkts;
+	uint64_t tx_pkts;
+	uint64_t tx_bytes;
+	uint64_t tx_err_pkts;
+};
+
+struct pmd_internals {
+	struct rte_kdp *kdp;
+	struct rte_kdp_tap *kdp_tap;
+
+	struct kdp_queue rx_queues[RTE_MAX_QUEUES_PER_PORT];
+	struct kdp_queue tx_queues[RTE_MAX_QUEUES_PER_PORT];
+};
+
+static struct ether_addr eth_addr = { .addr_bytes = {0} };
+static const char *drivername = "KDP PMD";
+static struct rte_eth_link pmd_link = {
+		.link_speed = 10000,
+		.link_duplex = ETH_LINK_FULL_DUPLEX,
+		.link_status = 0
+};
+
+static uint16_t
+eth_kdp_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
+{
+	struct kdp_queue *kdp_q = q;
+	struct pmd_internals *internals = kdp_q->internals;
+	uint16_t nb_pkts;
+
+	nb_pkts = rte_kdp_rx_burst(internals->kdp, bufs, nb_bufs);
+
+	kdp_q->rx_pkts += nb_pkts;
+	kdp_q->rx_err_pkts += nb_bufs - nb_pkts;
+
+	return nb_pkts;
+}
+
+static uint16_t
+eth_kdp_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
+{
+	struct kdp_queue *kdp_q = q;
+	struct pmd_internals *internals = kdp_q->internals;
+	uint16_t nb_pkts;
+
+	nb_pkts =  rte_kdp_tx_burst(internals->kdp, bufs, nb_bufs);
+
+	kdp_q->tx_pkts += nb_pkts;
+	kdp_q->tx_err_pkts += nb_bufs - nb_pkts;
+
+	return nb_pkts;
+}
+
+static uint16_t
+eth_kdp_tap_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
+{
+	struct kdp_queue *kdp_q = q;
+	struct pmd_internals *internals = kdp_q->internals;
+	struct rte_kdp_tap *kdp_tap = internals->kdp_tap;
+	struct rte_mbuf *m;
+	int ret;
+	unsigned i;
+
+	for (i = 0; i < nb_bufs; i++) {
+		m = rte_pktmbuf_alloc(kdp_q->mb_pool);
+		bufs[i] = m;
+		ret = read(kdp_tap->tap_fd, rte_pktmbuf_mtod(m, void *),
+				MAX_PACKET_SZ);
+		if (ret < 0) {
+			rte_pktmbuf_free(m);
+			break;
+		}
+
+		m->nb_segs = 1;
+		m->next = NULL;
+		m->pkt_len = (uint16_t)ret;
+		m->data_len = (uint16_t)ret;
+	}
+
+	kdp_q->rx_pkts += i;
+	kdp_q->rx_err_pkts += nb_bufs - i;
+
+	return i;
+}
+
+static uint16_t
+eth_kdp_tap_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
+{
+	struct kdp_queue *kdp_q = q;
+	struct pmd_internals *internals = kdp_q->internals;
+	struct rte_kdp_tap *kdp_tap = internals->kdp_tap;
+	struct rte_mbuf *m;
+	unsigned i;
+
+	for (i = 0; i < nb_bufs; i++) {
+		m = bufs[i];
+		write(kdp_tap->tap_fd, rte_pktmbuf_mtod(m, void*),
+				rte_pktmbuf_data_len(m));
+		rte_pktmbuf_free(m);
+	}
+
+	kdp_q->tx_pkts += i;
+	kdp_q->tx_err_pkts += nb_bufs - i;
+
+	return i;
+}
+
+static int
+kdp_start(struct rte_eth_dev *dev)
+{
+	struct pmd_internals *internals = dev->data->dev_private;
+	struct rte_kdp_conf conf;
+	uint16_t port_id = dev->data->port_id;
+	int ret = 0;
+
+	if (internals->kdp) {
+		snprintf(conf.name, RTE_KDP_NAMESIZE, "kdp%u", port_id);
+		conf.force_bind = 0;
+		conf.group_id = port_id;
+		conf.mbuf_size = MAX_PACKET_SZ;
+
+		ret = rte_kdp_start(internals->kdp,
+				internals->rx_queues[0].mb_pool,
+				&conf);
+		if (ret)
+			RTE_LOG(ERR, KDP, "Fail to create kdp for port: %d\n",
+					port_id);
+	}
+
+	return ret;
+}
+
+static int
+eth_dev_start(struct rte_eth_dev *dev)
+{
+	int ret;
+
+	ret = kdp_start(dev);
+	if (ret)
+		return -1;
+
+	dev->data->dev_link.link_status = 1;
+	return 0;
+}
+
+static void
+eth_dev_stop(struct rte_eth_dev *dev)
+{
+	struct pmd_internals *internals = dev->data->dev_private;
+
+	rte_kdp_release(internals->kdp);
+	dev->data->dev_link.link_status = 0;
+}
+
+static void
+eth_dev_close(struct rte_eth_dev *dev __rte_unused)
+{
+	rte_kdp_close();
+}
+
+static int
+eth_dev_configure(struct rte_eth_dev *dev __rte_unused)
+{
+	return 0;
+}
+
+static void
+eth_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
+{
+	struct rte_eth_dev_data *data = dev->data;
+
+	dev_info->driver_name = data->drv_name;
+	dev_info->max_mac_addrs = 1;
+	dev_info->max_rx_pktlen = (uint32_t)-1;
+	dev_info->max_rx_queues = data->nb_rx_queues;
+	dev_info->max_tx_queues = data->nb_tx_queues;
+	dev_info->min_rx_bufsize = 0;
+	dev_info->pci_dev = NULL;
+}
+
+static int
+eth_rx_queue_setup(struct rte_eth_dev *dev,
+		uint16_t rx_queue_id __rte_unused,
+		uint16_t nb_rx_desc __rte_unused,
+		unsigned int socket_id __rte_unused,
+		const struct rte_eth_rxconf *rx_conf __rte_unused,
+		struct rte_mempool *mb_pool)
+{
+	struct pmd_internals *internals = dev->data->dev_private;
+	struct kdp_queue *q;
+
+	q = &internals->rx_queues[rx_queue_id];
+	q->internals = internals;
+	q->mb_pool = mb_pool;
+
+	dev->data->rx_queues[rx_queue_id] = q;
+
+	return 0;
+}
+
+static int
+eth_tx_queue_setup(struct rte_eth_dev *dev,
+		uint16_t tx_queue_id,
+		uint16_t nb_tx_desc __rte_unused,
+		unsigned int socket_id __rte_unused,
+		const struct rte_eth_txconf *tx_conf __rte_unused)
+{
+	struct pmd_internals *internals = dev->data->dev_private;
+	struct kdp_queue *q;
+
+	q = &internals->tx_queues[tx_queue_id];
+	q->internals = internals;
+
+	dev->data->tx_queues[tx_queue_id] = q;
+
+	return 0;
+}
+
+static void
+eth_queue_release(void *q __rte_unused)
+{
+}
+
+static int
+eth_link_update(struct rte_eth_dev *dev __rte_unused,
+		int wait_to_complete __rte_unused)
+{
+	return 0;
+}
+
+static void
+eth_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
+{
+	unsigned i, num_stats;
+	unsigned long rx_packets_total = 0, rx_bytes_total = 0;
+	unsigned long tx_packets_total = 0, tx_bytes_total = 0;
+	unsigned long tx_packets_err_total = 0;
+	struct rte_eth_dev_data *data = dev->data;
+	struct kdp_queue *q;
+
+	num_stats = RTE_MIN((unsigned)RTE_ETHDEV_QUEUE_STAT_CNTRS,
+			data->nb_rx_queues);
+	for (i = 0; i < num_stats; i++) {
+		q = data->rx_queues[i];
+		stats->q_ipackets[i] = q->rx_pkts;
+		stats->q_ibytes[i] = q->rx_bytes;
+		rx_packets_total += stats->q_ipackets[i];
+		rx_bytes_total += stats->q_ibytes[i];
+	}
+
+	num_stats = RTE_MIN((unsigned)RTE_ETHDEV_QUEUE_STAT_CNTRS,
+			data->nb_tx_queues);
+	for (i = 0; i < num_stats; i++) {
+		q = data->tx_queues[i];
+		stats->q_opackets[i] = q->tx_pkts;
+		stats->q_obytes[i] = q->tx_bytes;
+		stats->q_errors[i] = q->tx_err_pkts;
+		tx_packets_total += stats->q_opackets[i];
+		tx_bytes_total += stats->q_obytes[i];
+		tx_packets_err_total += stats->q_errors[i];
+	}
+
+	stats->ipackets = rx_packets_total;
+	stats->ibytes = rx_bytes_total;
+	stats->opackets = tx_packets_total;
+	stats->obytes = tx_bytes_total;
+	stats->oerrors = tx_packets_err_total;
+}
+
+static void
+eth_stats_reset(struct rte_eth_dev *dev)
+{
+	unsigned i;
+	struct rte_eth_dev_data *data = dev->data;
+	struct kdp_queue *q;
+
+	for (i = 0; i < data->nb_rx_queues; i++) {
+		q = data->rx_queues[i];
+		q->rx_pkts = 0;
+		q->rx_bytes = 0;
+	}
+	for (i = 0; i < data->nb_tx_queues; i++) {
+		q = data->rx_queues[i];
+		q->tx_pkts = 0;
+		q->tx_bytes = 0;
+		q->tx_err_pkts = 0;
+	}
+}
+
+static const struct eth_dev_ops ops = {
+	.dev_start = eth_dev_start,
+	.dev_stop = eth_dev_stop,
+	.dev_close = eth_dev_close,
+	.dev_configure = eth_dev_configure,
+	.dev_infos_get = eth_dev_info,
+	.rx_queue_setup = eth_rx_queue_setup,
+	.tx_queue_setup = eth_tx_queue_setup,
+	.rx_queue_release = eth_queue_release,
+	.tx_queue_release = eth_queue_release,
+	.link_update = eth_link_update,
+	.stats_get = eth_stats_get,
+	.stats_reset = eth_stats_reset,
+};
+
+static struct rte_eth_dev *
+eth_dev_kdp_create(const char *name, unsigned numa_node)
+{
+	uint16_t nb_rx_queues = 1;
+	uint16_t nb_tx_queues = 1;
+	struct rte_eth_dev_data *data = NULL;
+	struct pmd_internals *internals = NULL;
+	struct rte_eth_dev *eth_dev = NULL;
+
+	if (name == NULL)
+		return NULL;
+
+	RTE_LOG(INFO, PMD, "Creating kdp ethdev on numa socket %u\n",
+			numa_node);
+
+	data = rte_zmalloc_socket(name, sizeof(*data), 0, numa_node);
+	if (data == NULL)
+		goto error;
+
+	internals = rte_zmalloc_socket(name, sizeof(*internals), 0, numa_node);
+	if (internals == NULL)
+		goto error;
+
+	/* reserve an ethdev entry */
+	eth_dev = rte_eth_dev_allocate(name, RTE_ETH_DEV_VIRTUAL);
+	if (eth_dev == NULL)
+		goto error;
+
+	data->dev_private = internals;
+	data->port_id = eth_dev->data->port_id;
+	memmove(data->name, eth_dev->data->name, sizeof(data->name));
+	data->nb_rx_queues = nb_rx_queues;
+	data->nb_tx_queues = nb_tx_queues;
+	data->dev_link = pmd_link;
+	data->mac_addrs = &eth_addr;
+
+	eth_dev->data = data;
+	eth_dev->dev_ops = &ops;
+	eth_dev->driver = NULL;
+
+	data->dev_flags = RTE_ETH_DEV_DETACHABLE;
+	data->kdrv = RTE_KDRV_NONE;
+	data->drv_name = drivername;
+	data->numa_node = numa_node;
+
+	return eth_dev;
+
+error:
+	rte_free(data);
+	rte_free(internals);
+
+	return NULL;
+}
+
+static int
+rte_pmd_kdp_devinit(const char *name, const char *params __rte_unused)
+{
+	struct rte_eth_dev *eth_dev = NULL;
+	struct pmd_internals *internals;
+	struct rte_kdp *kdp;
+	struct rte_kdp_tap *kdp_tap = NULL;
+	uint16_t port_id;
+
+	RTE_LOG(INFO, PMD, "Initializing eth_kdp for %s\n", name);
+
+	eth_dev = eth_dev_kdp_create(name, rte_socket_id());
+	if (eth_dev == NULL)
+		return -1;
+
+	internals = eth_dev->data->dev_private;
+	port_id = eth_dev->data->port_id;
+
+	kdp = rte_kdp_init(port_id);
+	if (kdp == NULL)
+		kdp_tap = rte_kdp_tap_init(port_id);
+
+	if (kdp == NULL && kdp_tap == NULL) {
+		rte_eth_dev_release_port(eth_dev);
+		rte_free(internals);
+
+		/* Not return error to prevent panic in rte_eal_init()  */
+		return 0;
+	}
+
+	internals->kdp = kdp;
+	internals->kdp_tap = kdp_tap;
+
+	if (kdp == NULL) {
+		eth_dev->rx_pkt_burst = eth_kdp_tap_rx;
+		eth_dev->tx_pkt_burst = eth_kdp_tap_tx;
+	} else {
+		eth_dev->rx_pkt_burst = eth_kdp_rx;
+		eth_dev->tx_pkt_burst = eth_kdp_tx;
+	}
+
+	return 0;
+}
+
+static int
+rte_pmd_kdp_devuninit(const char *name)
+{
+	struct rte_eth_dev *eth_dev = NULL;
+
+	if (name == NULL)
+		return -EINVAL;
+
+	RTE_LOG(INFO, PMD, "Un-Initializing eth_kdp for %s\n", name);
+
+	/* find the ethdev entry */
+	eth_dev = rte_eth_dev_allocated(name);
+	if (eth_dev == NULL)
+		return -1;
+
+	eth_dev_stop(eth_dev);
+
+	if (eth_dev->data)
+		rte_free(eth_dev->data->dev_private);
+	rte_free(eth_dev->data);
+
+	rte_eth_dev_release_port(eth_dev);
+	return 0;
+}
+
+static struct rte_driver pmd_kdp_drv = {
+	.name = "eth_kdp",
+	.type = PMD_VDEV,
+	.init = rte_pmd_kdp_devinit,
+	.uninit = rte_pmd_kdp_devuninit,
+};
+
+PMD_REGISTER_DRIVER(pmd_kdp_drv);
diff --git a/drivers/net/kdp/rte_kdp.c b/drivers/net/kdp/rte_kdp.c
new file mode 100644
index 0000000..604f697
--- /dev/null
+++ b/drivers/net/kdp/rte_kdp.c
@@ -0,0 +1,365 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef RTE_EXEC_ENV_LINUXAPP
+#error "KDP is not supported"
+#endif
+
+#include <rte_spinlock.h>
+#include <rte_ethdev.h>
+#include <rte_memzone.h>
+
+#include "rte_kdp.h"
+#include "rte_kdp_fifo.h"
+
+#define MAX_MBUF_BURST_NUM     32
+
+/* Maximum number of ring entries */
+#define KDP_FIFO_COUNT_MAX     1024
+#define KDP_FIFO_SIZE          (KDP_FIFO_COUNT_MAX * sizeof(void *) + \
+					sizeof(struct rte_kdp_fifo))
+
+static volatile int kdp_fd = -1;
+
+static const struct rte_memzone *
+kdp_memzone_reserve(const char *name, size_t len, int socket_id,
+		unsigned flags)
+{
+	const struct rte_memzone *mz = rte_memzone_lookup(name);
+
+	if (mz == NULL)
+		mz = rte_memzone_reserve(name, len, socket_id, flags);
+
+	return mz;
+}
+
+static int
+slot_init(struct rte_kdp_memzone_slot *slot)
+{
+#define OBJNAMSIZ 32
+	char obj_name[OBJNAMSIZ];
+	const struct rte_memzone *mz;
+
+	/* TX RING */
+	snprintf(obj_name, OBJNAMSIZ, "kdp_tx_%d", slot->id);
+	mz = kdp_memzone_reserve(obj_name, KDP_FIFO_SIZE, SOCKET_ID_ANY, 0);
+	if (mz == NULL)
+		goto kdp_fail;
+	slot->m_tx_q = mz;
+
+	/* RX RING */
+	snprintf(obj_name, OBJNAMSIZ, "kdp_rx_%d", slot->id);
+	mz = kdp_memzone_reserve(obj_name, KDP_FIFO_SIZE, SOCKET_ID_ANY, 0);
+	if (mz == NULL)
+		goto kdp_fail;
+	slot->m_rx_q = mz;
+
+	/* ALLOC RING */
+	snprintf(obj_name, OBJNAMSIZ, "kdp_alloc_%d", slot->id);
+	mz = kdp_memzone_reserve(obj_name, KDP_FIFO_SIZE, SOCKET_ID_ANY, 0);
+	if (mz == NULL)
+		goto kdp_fail;
+	slot->m_alloc_q = mz;
+
+	/* FREE RING */
+	snprintf(obj_name, OBJNAMSIZ, "kdp_free_%d", slot->id);
+	mz = kdp_memzone_reserve(obj_name, KDP_FIFO_SIZE, SOCKET_ID_ANY, 0);
+	if (mz == NULL)
+		goto kdp_fail;
+	slot->m_free_q = mz;
+
+	return 0;
+
+kdp_fail:
+	return -1;
+}
+
+static void
+ring_init(struct rte_kdp *kdp)
+{
+	struct rte_kdp_memzone_slot *slot = kdp->slot;
+	const struct rte_memzone *mz;
+
+	/* TX RING */
+	mz = slot->m_tx_q;
+	kdp->tx_q = mz->addr;
+	kdp_fifo_init(kdp->tx_q, KDP_FIFO_COUNT_MAX);
+
+	/* RX RING */
+	mz = slot->m_rx_q;
+	kdp->rx_q = mz->addr;
+	kdp_fifo_init(kdp->rx_q, KDP_FIFO_COUNT_MAX);
+
+	/* ALLOC RING */
+	mz = slot->m_alloc_q;
+	kdp->alloc_q = mz->addr;
+	kdp_fifo_init(kdp->alloc_q, KDP_FIFO_COUNT_MAX);
+
+	/* FREE RING */
+	mz = slot->m_free_q;
+	kdp->free_q = mz->addr;
+	kdp_fifo_init(kdp->free_q, KDP_FIFO_COUNT_MAX);
+}
+
+/* Shall be called before any allocation happens */
+struct rte_kdp *
+rte_kdp_init(uint16_t port_id)
+{
+	struct rte_kdp_memzone_slot *slot = NULL;
+	struct rte_kdp *kdp = NULL;
+	int ret;
+
+	/* Check FD and open */
+	if (kdp_fd < 0) {
+		kdp_fd = open("/dev/kdp", O_RDWR);
+		if (kdp_fd < 0) {
+			RTE_LOG(ERR, KDP, "Can not open /dev/kdp\n");
+			return NULL;
+		}
+	}
+
+	slot = rte_malloc(NULL, sizeof(struct rte_kdp_memzone_slot), 0);
+	if (slot == NULL)
+		goto kdp_fail;
+	slot->id = port_id;
+
+	kdp = rte_malloc(NULL, sizeof(struct rte_kdp), 0);
+	if (kdp == NULL)
+		goto kdp_fail;
+	kdp->slot = slot;
+
+	ret = slot_init(slot);
+	if (ret < 0)
+		goto kdp_fail;
+
+	ring_init(kdp);
+
+	return kdp;
+
+kdp_fail:
+	rte_free(slot);
+	rte_free(kdp);
+	RTE_LOG(ERR, KDP, "Unable to allocate memory\n");
+	return NULL;
+}
+
+static void
+kdp_allocate_mbufs(struct rte_kdp *kdp)
+{
+	int i, ret;
+	struct rte_mbuf *pkts[MAX_MBUF_BURST_NUM];
+
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pool) !=
+			 offsetof(struct rte_kdp_mbuf, pool));
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_addr) !=
+			 offsetof(struct rte_kdp_mbuf, buf_addr));
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, next) !=
+			 offsetof(struct rte_kdp_mbuf, next));
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, data_off) !=
+			 offsetof(struct rte_kdp_mbuf, data_off));
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, data_len) !=
+			 offsetof(struct rte_kdp_mbuf, data_len));
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pkt_len) !=
+			 offsetof(struct rte_kdp_mbuf, pkt_len));
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, ol_flags) !=
+			 offsetof(struct rte_kdp_mbuf, ol_flags));
+
+	/* Check if pktmbuf pool has been configured */
+	if (kdp->pktmbuf_pool == NULL) {
+		RTE_LOG(ERR, KDP, "No valid mempool for allocating mbufs\n");
+		return;
+	}
+
+	for (i = 0; i < MAX_MBUF_BURST_NUM; i++) {
+		pkts[i] = rte_pktmbuf_alloc(kdp->pktmbuf_pool);
+		if (unlikely(pkts[i] == NULL)) {
+			/* Out of memory */
+			RTE_LOG(ERR, KDP, "Out of memory\n");
+			break;
+		}
+	}
+
+	/* No pkt mbuf alocated */
+	if (i <= 0)
+		return;
+
+	ret = kdp_fifo_put(kdp->alloc_q, (void **)pkts, i);
+
+	/* Check if any mbufs not put into alloc_q, and then free them */
+	if (ret >= 0 && ret < i && ret < MAX_MBUF_BURST_NUM) {
+		int j;
+
+		for (j = ret; j < i; j++)
+			rte_pktmbuf_free(pkts[j]);
+	}
+}
+
+int
+rte_kdp_start(struct rte_kdp *kdp, struct rte_mempool *pktmbuf_pool,
+	      const struct rte_kdp_conf *conf)
+{
+	struct rte_kdp_memzone_slot *slot = kdp->slot;
+	struct rte_kdp_device_info dev_info;
+	char mz_name[RTE_MEMZONE_NAMESIZE];
+	const struct rte_memzone *mz;
+	int ret;
+
+	if (!kdp || !pktmbuf_pool || !conf || !conf->name[0])
+		return -1;
+
+	snprintf(kdp->name, RTE_KDP_NAMESIZE, "%s", conf->name);
+	kdp->pktmbuf_pool = pktmbuf_pool;
+	kdp->group_id = conf->group_id;
+
+	memset(&dev_info, 0, sizeof(dev_info));
+	dev_info.core_id = conf->core_id;
+	dev_info.force_bind = conf->force_bind;
+	dev_info.group_id = conf->group_id;
+	dev_info.mbuf_size = conf->mbuf_size;
+	snprintf(dev_info.name, RTE_KDP_NAMESIZE, "%s", conf->name);
+
+	dev_info.tx_phys = slot->m_tx_q->phys_addr;
+	dev_info.rx_phys = slot->m_rx_q->phys_addr;
+	dev_info.alloc_phys = slot->m_alloc_q->phys_addr;
+	dev_info.free_phys = slot->m_free_q->phys_addr;
+
+	/* MBUF mempool */
+	snprintf(mz_name, sizeof(mz_name), RTE_MEMPOOL_OBJ_NAME,
+		pktmbuf_pool->name);
+	mz = rte_memzone_lookup(mz_name);
+	if (mz == NULL)
+		goto kdp_fail;
+	dev_info.mbuf_va = mz->addr;
+	dev_info.mbuf_phys = mz->phys_addr;
+
+	ret = ioctl(kdp_fd, RTE_KDP_IOCTL_CREATE, &dev_info);
+	if (ret < 0)
+		goto kdp_fail;
+
+	kdp->in_use = 1;
+
+	/* Allocate mbufs and then put them into alloc_q */
+	kdp_allocate_mbufs(kdp);
+
+	return 0;
+
+kdp_fail:
+	return -1;
+}
+
+static void
+kdp_free_mbufs(struct rte_kdp *kdp)
+{
+	int i, ret;
+	struct rte_mbuf *pkts[MAX_MBUF_BURST_NUM];
+
+	ret = kdp_fifo_get(kdp->free_q, (void **)pkts, MAX_MBUF_BURST_NUM);
+	if (likely(ret > 0)) {
+		for (i = 0; i < ret; i++)
+			rte_pktmbuf_free(pkts[i]);
+	}
+}
+
+unsigned
+rte_kdp_tx_burst(struct rte_kdp *kdp, struct rte_mbuf **mbufs, unsigned num)
+{
+	unsigned ret = kdp_fifo_put(kdp->rx_q, (void **)mbufs, num);
+
+	/* Get mbufs from free_q and then free them */
+	kdp_free_mbufs(kdp);
+
+	return ret;
+}
+
+unsigned
+rte_kdp_rx_burst(struct rte_kdp *kdp, struct rte_mbuf **mbufs, unsigned num)
+{
+	unsigned ret = kdp_fifo_get(kdp->tx_q, (void **)mbufs, num);
+
+	/* If buffers removed, allocate mbufs and then put them into alloc_q */
+	if (ret)
+		kdp_allocate_mbufs(kdp);
+
+	return ret;
+}
+
+static void
+kdp_free_fifo(struct rte_kdp_fifo *fifo)
+{
+	int ret;
+	struct rte_mbuf *pkt;
+
+	do {
+		ret = kdp_fifo_get(fifo, (void **)&pkt, 1);
+		if (ret)
+			rte_pktmbuf_free(pkt);
+	} while (ret);
+}
+
+int
+rte_kdp_release(struct rte_kdp *kdp)
+{
+	struct rte_kdp_device_info dev_info;
+
+	if (!kdp || !kdp->in_use)
+		return -1;
+
+	snprintf(dev_info.name, sizeof(dev_info.name), "%s", kdp->name);
+	if (ioctl(kdp_fd, RTE_KDP_IOCTL_RELEASE, &dev_info) < 0) {
+		RTE_LOG(ERR, KDP, "Fail to release kdp device\n");
+		return -1;
+	}
+
+	/* mbufs in all fifo should be released, except request/response */
+	kdp_free_fifo(kdp->tx_q);
+	kdp_free_fifo(kdp->rx_q);
+	kdp_free_fifo(kdp->alloc_q);
+	kdp_free_fifo(kdp->free_q);
+
+	rte_free(kdp->slot);
+
+	/* Memset the KDP struct */
+	memset(kdp, 0, sizeof(struct rte_kdp));
+
+	return 0;
+}
+
+void
+rte_kdp_close(void)
+{
+	if (kdp_fd < 0)
+		return;
+
+	close(kdp_fd);
+	kdp_fd = -1;
+}
diff --git a/drivers/net/kdp/rte_kdp.h b/drivers/net/kdp/rte_kdp.h
new file mode 100644
index 0000000..b9db048
--- /dev/null
+++ b/drivers/net/kdp/rte_kdp.h
@@ -0,0 +1,126 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _RTE_KDP_H_
+#define _RTE_KDP_H_
+
+#include <fcntl.h>
+#include <unistd.h>
+
+#include <sys/ioctl.h>
+
+#include <rte_malloc.h>
+#include <rte_mbuf.h>
+#include <rte_memcpy.h>
+#include <rte_memory.h>
+#include <rte_mempool.h>
+
+#include <exec-env/rte_kdp_common.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * KDP memzone pool slot
+ */
+struct rte_kdp_memzone_slot {
+	uint32_t id;
+
+	/* Memzones */
+	const struct rte_memzone *m_tx_q;      /**< TX queue */
+	const struct rte_memzone *m_rx_q;      /**< RX queue */
+	const struct rte_memzone *m_alloc_q;   /**< Allocated mbufs queue */
+	const struct rte_memzone *m_free_q;    /**< To be freed mbufs queue */
+};
+
+/**
+ * KDP context
+ */
+struct rte_kdp {
+	char name[RTE_KDP_NAMESIZE];        /**< KDP interface name */
+	struct rte_mempool *pktmbuf_pool;   /**< pkt mbuf mempool */
+	struct rte_kdp_memzone_slot *slot;
+	uint16_t group_id;                  /**< Group ID of KDP devices */
+
+	struct rte_kdp_fifo *tx_q;          /**< TX queue */
+	struct rte_kdp_fifo *rx_q;          /**< RX queue */
+	struct rte_kdp_fifo *alloc_q;       /**< Allocated mbufs queue */
+	struct rte_kdp_fifo *free_q;        /**< To be freed mbufs queue */
+
+	uint8_t in_use;                     /**< kdp in use */
+};
+
+struct rte_kdp_tap {
+	char name[RTE_KDP_NAMESIZE];
+	int tap_fd;
+};
+
+/**
+ * Structure for configuring KDP device.
+ */
+struct rte_kdp_conf {
+	/*
+	 * KDP name which will be used in relevant network device.
+	 * Let the name as short as possible, as it will be part of
+	 * memzone name.
+	 */
+	char name[RTE_KDP_NAMESIZE];
+	uint32_t core_id;   /* Core ID to bind kernel thread on */
+	uint16_t group_id;
+	unsigned mbuf_size;
+
+	uint8_t force_bind; /* Flag to bind kernel thread */
+};
+
+struct rte_kdp_tap *rte_kdp_tap_init(uint16_t port_id);
+struct rte_kdp *rte_kdp_init(uint16_t port_id);
+
+int rte_kdp_start(struct rte_kdp *kdp, struct rte_mempool *pktmbuf_pool,
+	      const struct rte_kdp_conf *conf);
+
+unsigned rte_kdp_rx_burst(struct rte_kdp *kdp,
+		struct rte_mbuf **mbufs, unsigned num);
+
+unsigned rte_kdp_tx_burst(struct rte_kdp *kdp,
+		struct rte_mbuf **mbufs, unsigned num);
+
+int rte_kdp_release(struct rte_kdp *kdp);
+
+void rte_kdp_close(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _RTE_KDP_H_ */
diff --git a/drivers/net/kdp/rte_kdp_fifo.h b/drivers/net/kdp/rte_kdp_fifo.h
new file mode 100644
index 0000000..1a7e063
--- /dev/null
+++ b/drivers/net/kdp/rte_kdp_fifo.h
@@ -0,0 +1,91 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/**
+ * Initializes the kdp fifo structure
+ */
+static void
+kdp_fifo_init(struct rte_kdp_fifo *fifo, unsigned size)
+{
+	/* Ensure size is power of 2 */
+	if (size & (size - 1))
+		rte_panic("KDP fifo size must be power of 2\n");
+
+	fifo->write = 0;
+	fifo->read = 0;
+	fifo->len = size;
+	fifo->elem_size = sizeof(void *);
+}
+
+/**
+ * Adds num elements into the fifo. Return the number actually written
+ */
+static inline unsigned
+kdp_fifo_put(struct rte_kdp_fifo *fifo, void **data, unsigned num)
+{
+	unsigned i = 0;
+	unsigned fifo_write = fifo->write;
+	unsigned fifo_read = fifo->read;
+	unsigned new_write = fifo_write;
+
+	for (i = 0; i < num; i++) {
+		new_write = (new_write + 1) & (fifo->len - 1);
+
+		if (new_write == fifo_read)
+			break;
+		fifo->buffer[fifo_write] = data[i];
+		fifo_write = new_write;
+	}
+	fifo->write = fifo_write;
+	return i;
+}
+
+/**
+ * Get up to num elements from the fifo. Return the number actully read
+ */
+static inline unsigned
+kdp_fifo_get(struct rte_kdp_fifo *fifo, void **data, unsigned num)
+{
+	unsigned i = 0;
+	unsigned new_read = fifo->read;
+	unsigned fifo_write = fifo->write;
+	for (i = 0; i < num; i++) {
+		if (new_read == fifo_write)
+			break;
+
+		data[i] = fifo->buffer[new_read];
+		new_read = (new_read + 1) & (fifo->len - 1);
+	}
+	fifo->read = new_read;
+	return i;
+}
diff --git a/drivers/net/kdp/rte_kdp_tap.c b/drivers/net/kdp/rte_kdp_tap.c
new file mode 100644
index 0000000..f07ba98
--- /dev/null
+++ b/drivers/net/kdp/rte_kdp_tap.c
@@ -0,0 +1,96 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <string.h>
+
+#include <sys/socket.h>
+#include <linux/if.h>
+#include <linux/if_tun.h>
+
+#include "rte_kdp.h"
+
+static int
+tap_create(char *name)
+{
+	struct ifreq ifr;
+	int fd, ret;
+
+	fd = open("/dev/net/tun", O_RDWR);
+	if (fd < 0)
+		return fd;
+
+	memset(&ifr, 0, sizeof(ifr));
+
+	/* TAP device without packet information */
+	ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
+
+	if (name && *name)
+		snprintf(ifr.ifr_name, IFNAMSIZ, "%s", name);
+
+	ret = ioctl(fd, TUNSETIFF, (void *)&ifr);
+	if (ret < 0) {
+		close(fd);
+		return ret;
+	}
+
+	if (name)
+		snprintf(name, IFNAMSIZ, "%s", ifr.ifr_name);
+
+	return fd;
+}
+
+struct rte_kdp_tap *
+rte_kdp_tap_init(uint16_t port_id)
+{
+	struct rte_kdp_tap *kdp_tap = NULL;
+	int flags;
+
+	kdp_tap = rte_malloc(NULL, sizeof(struct rte_kdp_tap), 0);
+	if (kdp_tap == NULL)
+		goto error;
+
+	snprintf(kdp_tap->name, IFNAMSIZ, "tap_kdp%u", port_id);
+	kdp_tap->tap_fd = tap_create(kdp_tap->name);
+	if (kdp_tap->tap_fd < 0)
+		goto error;
+
+	flags = fcntl(kdp_tap->tap_fd, F_GETFL, 0);
+	fcntl(kdp_tap->tap_fd, F_SETFL, flags | O_NONBLOCK);
+
+	return kdp_tap;
+
+error:
+	rte_free(kdp_tap);
+	return NULL;
+}
+
diff --git a/drivers/net/kdp/rte_pmd_kdp_version.map b/drivers/net/kdp/rte_pmd_kdp_version.map
new file mode 100644
index 0000000..0812bb1
--- /dev/null
+++ b/drivers/net/kdp/rte_pmd_kdp_version.map
@@ -0,0 +1,4 @@
+DPDK_2.3 {
+
+	local: *;
+};
diff --git a/lib/librte_eal/common/include/rte_log.h b/lib/librte_eal/common/include/rte_log.h
index 2e47e7f..5a0048b 100644
--- a/lib/librte_eal/common/include/rte_log.h
+++ b/lib/librte_eal/common/include/rte_log.h
@@ -1,7 +1,7 @@
 /*-
  *   BSD LICENSE
  *
- *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
+ *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
  *   All rights reserved.
  *
  *   Redistribution and use in source and binary forms, with or without
@@ -79,6 +79,7 @@ extern struct rte_logs rte_logs;
 #define RTE_LOGTYPE_PIPELINE 0x00008000 /**< Log related to pipeline. */
 #define RTE_LOGTYPE_MBUF    0x00010000 /**< Log related to mbuf. */
 #define RTE_LOGTYPE_CRYPTODEV 0x00020000 /**< Log related to cryptodev. */
+#define RTE_LOGTYPE_KDP     0x00080000 /**< Log related to KDP. */
 
 /* these log types can be used in an application */
 #define RTE_LOGTYPE_USER1   0x01000000 /**< User-defined log type 1. */
diff --git a/mk/rte.app.mk b/mk/rte.app.mk
index 8ecab41..eb18972 100644
--- a/mk/rte.app.mk
+++ b/mk/rte.app.mk
@@ -1,6 +1,6 @@
 #   BSD LICENSE
 #
-#   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+#   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
 #   Copyright(c) 2014-2015 6WIND S.A.
 #   All rights reserved.
 #
@@ -154,6 +154,7 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_PCAP)       += -lrte_pmd_pcap
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_AF_PACKET)  += -lrte_pmd_af_packet
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_NULL)       += -lrte_pmd_null
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_QAT)        += -lrte_pmd_qat
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_KDP)        += -lrte_pmd_kdp
 
 # AESNI MULTI BUFFER is dependent on the IPSec_MB library
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_AESNI_MB)   += -lrte_pmd_aesni_mb
-- 
2.5.0