szedata2: add new poll mode driver
authorMatej Vido <matejvido@gmail.com>
Tue, 10 Nov 2015 14:18:13 +0000 (15:18 +0100)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Fri, 20 Nov 2015 16:17:04 +0000 (17:17 +0100)
Add virtual PMD which communicates with COMBO cards through sze2
layer using libsze2 library.

Since link_speed is uint16_t, there can not be used number for 100G
speed, therefore link_speed is set to ETH_LINK_SPEED_10G until the
type of link_speed is solved.

Signed-off-by: Matej Vido <matejvido@gmail.com>
13 files changed:
MAINTAINERS
config/common_bsdapp
config/common_linuxapp
doc/guides/nics/index.rst
doc/guides/nics/szedata2.rst [new file with mode: 0644]
doc/guides/prog_guide/source_org.rst
doc/guides/rel_notes/release_2_2.rst
drivers/net/Makefile
drivers/net/szedata2/Makefile [new file with mode: 0644]
drivers/net/szedata2/rte_eth_szedata2.c [new file with mode: 0644]
drivers/net/szedata2/rte_eth_szedata2.h [new file with mode: 0644]
drivers/net/szedata2/rte_pmd_szedata2_version.map [new file with mode: 0644]
mk/rte.app.mk

index 8ae688d..840faeb 100644 (file)
@@ -246,6 +246,11 @@ M: John Daley <johndale@cisco.com>
 M: Sujith Sankar <ssujith@cisco.com>
 F: drivers/net/enic/
 
+Combo szedata2
+M: Matej Vido <matejvido@gmail.com>
+F: drivers/net/szedata2/
+F: doc/guides/nics/szedata2.rst
+
 Intel e1000
 M: Wenzhuo Lu <wenzhuo.lu@intel.com>
 F: drivers/net/e1000/
index 7df0763..ae7374a 100644 (file)
@@ -254,6 +254,11 @@ CONFIG_RTE_LIBRTE_CXGBE_DEBUG_RX=n
 CONFIG_RTE_LIBRTE_ENIC_PMD=y
 CONFIG_RTE_LIBRTE_ENIC_DEBUG=n
 
+#
+# Compile software PMD backed by SZEDATA2 device
+#
+CONFIG_RTE_LIBRTE_PMD_SZEDATA2=n
+
 #
 # Compile burst-oriented VIRTIO PMD driver
 #
index 52173d5..0954eef 100644 (file)
@@ -252,6 +252,11 @@ CONFIG_RTE_LIBRTE_CXGBE_DEBUG_RX=n
 CONFIG_RTE_LIBRTE_ENIC_PMD=y
 CONFIG_RTE_LIBRTE_ENIC_DEBUG=n
 
+#
+# Compile software PMD backed by SZEDATA2 device
+#
+CONFIG_RTE_LIBRTE_PMD_SZEDATA2=n
+
 #
 # Compile burst-oriented VIRTIO PMD driver
 #
index 2d4936d..0a0b724 100644 (file)
@@ -46,6 +46,7 @@ Network Interface Controller Drivers
     intel_vf
     mlx4
     mlx5
+    szedata2
     virtio
     vmxnet3
     pcap_ring
diff --git a/doc/guides/nics/szedata2.rst b/doc/guides/nics/szedata2.rst
new file mode 100644 (file)
index 0000000..b952b46
--- /dev/null
@@ -0,0 +1,127 @@
+..  BSD LICENSE
+    Copyright 2015 CESNET
+    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 CESNET 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.
+
+SZEDATA2 poll mode driver library
+=================================
+
+The SZEDATA2 poll mode driver library implements support for cards from COMBO
+family (**COMBO-80G**, **COMBO-100G**).
+The SZEDATA2 PMD is virtual PMD which uses interface provided by libsze2
+library to communicate with COMBO cards over sze2 layer.
+
+More information about family of
+`COMBO cards <https://www.liberouter.org/technologies/cards/>`_
+and used technology
+(`NetCOPE platform <https://www.liberouter.org/technologies/netcope/>`_) can be
+found on the `Liberouter website <https://www.liberouter.org/>`_.
+
+.. note::
+
+   This driver has external dependencies.
+   Therefore it is disabled in default configuration files.
+   It can be enabled by setting ``CONFIG_RTE_LIBRTE_PMD_SZEDATA2=y``
+   and recompiling.
+
+Prerequisities
+--------------
+
+This PMD requires kernel modules which are responsible for initialization and
+allocation of resources needed for sze2 layer function.
+Communication between PMD and kernel modules is mediated by libsze2 library.
+These kernel modules and library are not part of DPDK and must be installed
+separately:
+
+*  **libsze2 library**
+
+   The library provides API for initialization of sze2 transfers, receiving and
+   transmitting data segments.
+
+*  **Kernel modules**
+
+   * combov3
+   * szedata2_cv3
+
+   Kernel modules manage initialization of hardware, allocation and
+   sharing of resources for user space applications:
+
+Information about getting the dependencies can be found `here
+<https://www.liberouter.org/technologies/netcope/access-to-libsze2-library/>`_.
+
+
+Using the SZEDATA2 PMD
+----------------------
+
+SZEDATA2 PMD can be created by passing ``--vdev=`` option to EAL in the
+following format:
+
+.. code-block:: console
+
+   --vdev 'DEVICE,dev_path=PATH,rx_ifaces=RX_MASK,tx_ifaces=TX_MASK'
+
+``DEVICE`` and options ``dev_path``, ``rx_ifaces``, ``tx_ifaces`` are mandatory
+and must be separated by commas.
+
+*  ``DEVICE``: contains prefix ``eth_szedata2`` followed by numbers or letters,
+   must be unique for each virtual device
+
+*  ``dev_path``: Defines path to szedata2 device.
+   Value is valid path to szedata2 device. Example:
+
+   .. code-block:: console
+
+      dev_path=/dev/szedataII0
+
+*  ``rx_ifaces``: Defines which receive channels will be used.
+   For each channel is created one queue. Value is mask for selecting which
+   receive channels are required. Example:
+
+   .. code-block:: console
+
+      rx_ifaces=0x3
+
+*  ``tx_ifaces``: Defines which transmit channels will be used.
+   For each channel is created one queue. Value is mask for selecting which
+   transmit channels are required. Example:
+
+   .. code-block:: console
+
+      tx_ifaces=0x3
+
+Example of usage
+----------------
+
+Read packets from 0. and 1. receive channel and write them to 0. and 1.
+transmit channel:
+
+.. code-block:: console
+
+   $RTE_TARGET/app/testpmd -c 0xf -n 2 \
+   --vdev 'eth_szedata20,dev_path=/dev/szedataII0,rx_ifaces=0x3,tx_ifaces=0x3' \
+   -- --port-topology=chained --rxq=2 --txq=2 --nb-cores=2
index ae11b3b..0c06d47 100644 (file)
@@ -106,6 +106,7 @@ The drivers directory has a *net* subdirectory which contains::
     +-- null               # NULL poll mode driver for testing
     +-- pcap               # PCAP poll mode driver
     +-- ring               # Ring poll mode driver
+    +-- szedata2           # SZEDATA2 poll mode driver
     +-- virtio             # Virtio poll mode driver
     +-- vmxnet3            # VMXNET3 poll mode driver
     +-- xenvirt            # Xen virtio poll mode driver
index 2d91bdc..4fea433 100644 (file)
@@ -98,6 +98,11 @@ New Features
   Like mlx4, this PMD is only available for Linux and is disabled by default
   due to external dependencies (libibverbs and libmlx5).
 
+* **Added virtual szedata2 driver for COMBO cards.**
+
+  Added virtual PMD for COMBO-100G and COMBO-80G cards.
+  PMD is disabled in default configuration.
+
 * **Enhanced support for virtio driver.**
 
   * Virtio ring layout optimization (fixed avail ring)
index 6da1ce2..cddcd57 100644 (file)
@@ -46,6 +46,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_MPIPE_PMD) += mpipe
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_NULL) += null
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_PCAP) += pcap
 DIRS-$(CONFIG_RTE_LIBRTE_PMD_RING) += ring
+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
diff --git a/drivers/net/szedata2/Makefile b/drivers/net/szedata2/Makefile
new file mode 100644 (file)
index 0000000..c3c42e5
--- /dev/null
@@ -0,0 +1,62 @@
+#   BSD LICENSE
+#
+#   Copyright (c) 2015 CESNET
+#   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 CESNET 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_szedata2.a
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+EXPORT_MAP := rte_pmd_szedata2_version.map
+
+LIBABIVER := 1
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2) += rte_eth_szedata2.c
+
+#
+# Export include files
+#
+SYMLINK-y-include +=
+
+# this lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2) += lib/librte_mbuf
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2) += lib/librte_ether
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2) += lib/librte_malloc
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2) += lib/librte_kvargs
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/szedata2/rte_eth_szedata2.c b/drivers/net/szedata2/rte_eth_szedata2.c
new file mode 100644 (file)
index 0000000..5f3a19c
--- /dev/null
@@ -0,0 +1,827 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright (c) 2015 CESNET
+ *   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 CESNET 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 <stdint.h>
+#include <unistd.h>
+#include <stdbool.h>
+#include <err.h>
+
+#include <libsze2.h>
+
+#include <rte_mbuf.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+#include <rte_memcpy.h>
+#include <rte_kvargs.h>
+#include <rte_dev.h>
+
+#include "rte_eth_szedata2.h"
+
+#define RTE_ETH_SZEDATA2_DEV_PATH_ARG "dev_path"
+#define RTE_ETH_SZEDATA2_RX_IFACES_ARG "rx_ifaces"
+#define RTE_ETH_SZEDATA2_TX_IFACES_ARG "tx_ifaces"
+
+#define RTE_ETH_SZEDATA2_MAX_RX_QUEUES 32
+#define RTE_ETH_SZEDATA2_MAX_TX_QUEUES 32
+#define RTE_ETH_SZEDATA2_TX_LOCK_SIZE (32 * 1024 * 1024)
+
+/**
+ * size of szedata2_packet header with alignment
+ */
+#define RTE_SZE2_PACKET_HEADER_SIZE_ALIGNED 8
+
+struct szedata2_rx_queue {
+       struct szedata *sze;
+       uint8_t rx_channel;
+       uint8_t in_port;
+       struct rte_mempool *mb_pool;
+       volatile uint64_t rx_pkts;
+       volatile uint64_t rx_bytes;
+       volatile uint64_t err_pkts;
+};
+
+struct szedata2_tx_queue {
+       struct szedata *sze;
+       uint8_t tx_channel;
+       volatile uint64_t tx_pkts;
+       volatile uint64_t err_pkts;
+       volatile uint64_t tx_bytes;
+};
+
+struct rxtx_szedata2 {
+       uint32_t num_of_rx;
+       uint32_t num_of_tx;
+       uint32_t sze_rx_mask_req;
+       uint32_t sze_tx_mask_req;
+       char *sze_dev;
+};
+
+struct pmd_internals {
+       struct szedata2_rx_queue rx_queue[RTE_ETH_SZEDATA2_MAX_RX_QUEUES];
+       struct szedata2_tx_queue tx_queue[RTE_ETH_SZEDATA2_MAX_TX_QUEUES];
+       unsigned nb_rx_queues;
+       unsigned nb_tx_queues;
+       uint32_t num_of_rx;
+       uint32_t num_of_tx;
+       uint32_t sze_rx_req;
+       uint32_t sze_tx_req;
+       int if_index;
+       char *sze_dev;
+};
+
+static const char *valid_arguments[] = {
+       RTE_ETH_SZEDATA2_DEV_PATH_ARG,
+       RTE_ETH_SZEDATA2_RX_IFACES_ARG,
+       RTE_ETH_SZEDATA2_TX_IFACES_ARG,
+       NULL
+};
+
+static struct ether_addr eth_addr = {
+       .addr_bytes = { 0x00, 0x11, 0x17, 0x00, 0x00, 0x00 }
+};
+static const char *drivername = "SZEdata2 PMD";
+static struct rte_eth_link pmd_link = {
+               .link_speed = ETH_LINK_SPEED_10G,
+               .link_duplex = ETH_LINK_FULL_DUPLEX,
+               .link_status = 0
+};
+
+
+static uint32_t
+count_ones(uint32_t num)
+{
+       num = num - ((num >> 1) & 0x55555555); /* reuse input as temporary */
+       num = (num & 0x33333333) + ((num >> 2) & 0x33333333);        /* temp */
+       return (((num + (num >> 4)) & 0xF0F0F0F) * 0x1010101) >> 24; /* count */
+}
+
+static int
+init_rx_channels(struct rte_eth_dev *dev, int v)
+{
+       struct pmd_internals *internals = dev->data->dev_private;
+       int ret;
+       uint32_t i;
+       uint32_t count = internals->num_of_rx;
+       uint32_t num_sub = 0;
+       uint32_t x;
+       uint32_t rx;
+       uint32_t tx;
+
+       rx = internals->sze_rx_req;
+       tx = 0;
+
+       for (i = 0; i < count; i++) {
+               /*
+                * Open, subscribe rx,tx channels and start device
+                */
+               if (v)
+                       RTE_LOG(INFO, PMD, "Opening SZE device %u. time\n", i);
+
+               internals->rx_queue[num_sub].sze =
+                       szedata_open(internals->sze_dev);
+               if (internals->rx_queue[num_sub].sze == NULL)
+                       return -1;
+
+               /* separate least significant non-zero bit */
+               x = rx & ((~rx) + 1);
+
+               if (v)
+                       RTE_LOG(INFO, PMD, "Subscribing rx channel: 0x%x "
+                               "tx channel: 0x%x\n", x, tx);
+
+               ret = szedata_subscribe3(internals->rx_queue[num_sub].sze,
+                               &x, &tx);
+               if (ret) {
+                       szedata_close(internals->rx_queue[num_sub].sze);
+                       internals->rx_queue[num_sub].sze = NULL;
+                       return -1;
+               }
+
+               if (v)
+                       RTE_LOG(INFO, PMD, "Subscribed rx channel: 0x%x "
+                               "tx channel: 0x%x\n", x, tx);
+
+               if (x) {
+                       if (v)
+                               RTE_LOG(INFO, PMD, "Starting SZE device for "
+                                       "rx queue: %u\n", num_sub);
+
+                       ret = szedata_start(internals->rx_queue[num_sub].sze);
+                       if (ret) {
+                               szedata_close(internals->rx_queue[num_sub].sze);
+                               internals->rx_queue[num_sub].sze = NULL;
+                               return -1;
+                       }
+
+                       /*
+                        * set to 1 all bits lower than bit set to 1
+                        * and that bit to 0
+                        */
+                       x -= 1;
+                       internals->rx_queue[num_sub].rx_channel =
+                               count_ones(x);
+
+                       if (v)
+                               RTE_LOG(INFO, PMD, "Subscribed rx channel "
+                                       "no: %u\n",
+                                       internals->rx_queue[num_sub].rx_channel
+                                       );
+
+                       num_sub++;
+                       internals->nb_rx_queues = num_sub;
+               } else {
+                       if (v)
+                               RTE_LOG(INFO, PMD,
+                                       "Could not subscribe any rx channel. "
+                                       "Closing SZE device\n");
+
+                       szedata_close(internals->rx_queue[num_sub].sze);
+                       internals->rx_queue[num_sub].sze = NULL;
+               }
+
+               /* set least significant non-zero bit to zero */
+               rx = rx & (rx - 1);
+       }
+
+       dev->data->nb_rx_queues = (uint16_t)num_sub;
+
+       if (v)
+               RTE_LOG(INFO, PMD, "Successfully opened rx channels: %u\n",
+                       num_sub);
+
+       return 0;
+}
+
+static int
+init_tx_channels(struct rte_eth_dev *dev, int v)
+{
+       struct pmd_internals *internals = dev->data->dev_private;
+       int ret;
+       uint32_t i;
+       uint32_t count = internals->num_of_tx;
+       uint32_t num_sub = 0;
+       uint32_t x;
+       uint32_t rx;
+       uint32_t tx;
+
+       rx = 0;
+       tx = internals->sze_tx_req;
+
+       for (i = 0; i < count; i++) {
+               /*
+                * Open, subscribe rx,tx channels and start device
+                */
+               if (v)
+                       RTE_LOG(INFO, PMD, "Opening SZE device %u. time\n",
+                               i + internals->num_of_rx);
+
+               internals->tx_queue[num_sub].sze =
+                       szedata_open(internals->sze_dev);
+               if (internals->tx_queue[num_sub].sze == NULL)
+                       return -1;
+
+               /* separate least significant non-zero bit */
+               x = tx & ((~tx) + 1);
+
+               if (v)
+                       RTE_LOG(INFO, PMD, "Subscribing rx channel: 0x%x "
+                               "tx channel: 0x%x\n", rx, x);
+
+               ret = szedata_subscribe3(internals->tx_queue[num_sub].sze,
+                               &rx, &x);
+               if (ret) {
+                       szedata_close(internals->tx_queue[num_sub].sze);
+                       internals->tx_queue[num_sub].sze = NULL;
+                       return -1;
+               }
+
+               if (v)
+                       RTE_LOG(INFO, PMD, "Subscribed rx channel: 0x%x "
+                               "tx channel: 0x%x\n", rx, x);
+
+               if (x) {
+                       if (v)
+                               RTE_LOG(INFO, PMD, "Starting SZE device for "
+                                       "tx queue: %u\n", num_sub);
+
+                       ret = szedata_start(internals->tx_queue[num_sub].sze);
+                       if (ret) {
+                               szedata_close(internals->tx_queue[num_sub].sze);
+                               internals->tx_queue[num_sub].sze = NULL;
+                               return -1;
+                       }
+
+                       /*
+                        * set to 1 all bits lower than bit set to 1
+                        * and that bit to 0
+                        */
+                       x -= 1;
+                       internals->tx_queue[num_sub].tx_channel =
+                               count_ones(x);
+
+                       if (v)
+                               RTE_LOG(INFO, PMD, "Subscribed tx channel "
+                                       "no: %u\n",
+                                       internals->tx_queue[num_sub].tx_channel
+                                       );
+
+                       num_sub++;
+                       internals->nb_tx_queues = num_sub;
+               } else {
+                       if (v)
+                               RTE_LOG(INFO, PMD,
+                                       "Could not subscribe any tx channel. "
+                                       "Closing SZE device\n");
+
+                       szedata_close(internals->tx_queue[num_sub].sze);
+                       internals->tx_queue[num_sub].sze = NULL;
+               }
+
+               /* set least significant non-zero bit to zero */
+               tx = tx & (tx - 1);
+       }
+
+       dev->data->nb_tx_queues = (uint16_t)num_sub;
+
+       if (v)
+               RTE_LOG(INFO, PMD, "Successfully opened tx channels: %u\n",
+                       num_sub);
+
+       return 0;
+}
+
+static void
+close_rx_channels(struct rte_eth_dev *dev)
+{
+       struct pmd_internals *internals = dev->data->dev_private;
+       uint32_t i;
+       uint32_t num_sub = internals->nb_rx_queues;
+
+       for (i = 0; i < num_sub; i++) {
+               if (internals->rx_queue[i].sze != NULL) {
+                       szedata_close(internals->rx_queue[i].sze);
+                       internals->rx_queue[i].sze = NULL;
+               }
+       }
+       /* set number of rx queues to zero */
+       internals->nb_rx_queues = 0;
+       dev->data->nb_rx_queues = (uint16_t)0;
+}
+
+static void
+close_tx_channels(struct rte_eth_dev *dev)
+{
+       struct pmd_internals *internals = dev->data->dev_private;
+       uint32_t i;
+       uint32_t num_sub = internals->nb_tx_queues;
+
+       for (i = 0; i < num_sub; i++) {
+               if (internals->tx_queue[i].sze != NULL) {
+                       szedata_close(internals->tx_queue[i].sze);
+                       internals->tx_queue[i].sze = NULL;
+               }
+       }
+       /* set number of rx queues to zero */
+       internals->nb_tx_queues = 0;
+       dev->data->nb_tx_queues = (uint16_t)0;
+}
+
+static int
+eth_dev_start(struct rte_eth_dev *dev)
+{
+       struct pmd_internals *internals = dev->data->dev_private;
+       int ret;
+
+       if (internals->nb_rx_queues == 0) {
+               ret = init_rx_channels(dev, 0);
+               if (ret != 0) {
+                       close_rx_channels(dev);
+                       return -1;
+               }
+       }
+
+       if (internals->nb_tx_queues == 0) {
+               ret = init_tx_channels(dev, 0);
+               if (ret != 0) {
+                       close_tx_channels(dev);
+                       close_rx_channels(dev);
+                       return -1;
+               }
+       }
+
+       dev->data->dev_link.link_status = 1;
+       return 0;
+}
+
+static void
+eth_dev_stop(struct rte_eth_dev *dev)
+{
+       unsigned i;
+       struct pmd_internals *internals = dev->data->dev_private;
+
+       for (i = 0; i < internals->nb_rx_queues; i++) {
+               if (internals->rx_queue[i].sze != NULL) {
+                       szedata_close(internals->rx_queue[i].sze);
+                       internals->rx_queue[i].sze = NULL;
+               }
+       }
+
+       for (i = 0; i < internals->nb_tx_queues; i++) {
+               if (internals->tx_queue[i].sze != NULL) {
+                       szedata_close(internals->tx_queue[i].sze);
+                       internals->tx_queue[i].sze = NULL;
+               }
+       }
+
+       internals->nb_rx_queues = 0;
+       internals->nb_tx_queues = 0;
+
+       dev->data->nb_rx_queues = (uint16_t)0;
+       dev->data->nb_tx_queues = (uint16_t)0;
+
+       dev->data->dev_link.link_status = 0;
+}
+
+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 pmd_internals *internals = dev->data->dev_private;
+       dev_info->driver_name = drivername;
+       dev_info->if_index = internals->if_index;
+       dev_info->max_mac_addrs = 1;
+       dev_info->max_rx_pktlen = (uint32_t)-1;
+       dev_info->max_rx_queues = (uint16_t)internals->nb_rx_queues;
+       dev_info->max_tx_queues = (uint16_t)internals->nb_tx_queues;
+       dev_info->min_rx_bufsize = 0;
+       dev_info->pci_dev = NULL;
+}
+
+static void
+eth_stats_get(struct rte_eth_dev *dev,
+               struct rte_eth_stats *stats)
+{
+       unsigned i;
+       uint64_t rx_total = 0;
+       uint64_t tx_total = 0;
+       uint64_t tx_err_total = 0;
+       uint64_t rx_total_bytes = 0;
+       uint64_t tx_total_bytes = 0;
+       const struct pmd_internals *internal = dev->data->dev_private;
+
+       for (i = 0; i < RTE_ETHDEV_QUEUE_STAT_CNTRS &&
+                       i < internal->nb_rx_queues; i++) {
+               stats->q_ipackets[i] = internal->rx_queue[i].rx_pkts;
+               stats->q_ibytes[i] = internal->rx_queue[i].rx_bytes;
+               rx_total += stats->q_ipackets[i];
+               rx_total_bytes += stats->q_ibytes[i];
+       }
+
+       for (i = 0; i < RTE_ETHDEV_QUEUE_STAT_CNTRS &&
+                       i < internal->nb_tx_queues; i++) {
+               stats->q_opackets[i] = internal->tx_queue[i].tx_pkts;
+               stats->q_errors[i] = internal->tx_queue[i].err_pkts;
+               stats->q_obytes[i] = internal->tx_queue[i].tx_bytes;
+               tx_total += stats->q_opackets[i];
+               tx_err_total += stats->q_errors[i];
+               tx_total_bytes += stats->q_obytes[i];
+       }
+
+       stats->ipackets = rx_total;
+       stats->opackets = tx_total;
+       stats->ibytes = rx_total_bytes;
+       stats->obytes = tx_total_bytes;
+       stats->oerrors = tx_err_total;
+}
+
+static void
+eth_stats_reset(struct rte_eth_dev *dev)
+{
+       unsigned i;
+       struct pmd_internals *internal = dev->data->dev_private;
+       for (i = 0; i < internal->nb_rx_queues; i++) {
+               internal->rx_queue[i].rx_pkts = 0;
+               internal->rx_queue[i].rx_bytes = 0;
+       }
+       for (i = 0; i < internal->nb_tx_queues; i++) {
+               internal->tx_queue[i].tx_pkts = 0;
+               internal->tx_queue[i].err_pkts = 0;
+               internal->tx_queue[i].tx_bytes = 0;
+       }
+}
+
+static void
+eth_dev_close(struct rte_eth_dev *dev)
+{
+       unsigned i;
+       struct pmd_internals *internals = dev->data->dev_private;
+
+       for (i = 0; i < internals->nb_rx_queues; i++) {
+               if (internals->rx_queue[i].sze != NULL) {
+                       szedata_close(internals->rx_queue[i].sze);
+                       internals->rx_queue[i].sze = NULL;
+               }
+       }
+
+       for (i = 0; i < internals->nb_tx_queues; i++) {
+               if (internals->tx_queue[i].sze != NULL) {
+                       szedata_close(internals->tx_queue[i].sze);
+                       internals->tx_queue[i].sze = NULL;
+               }
+       }
+
+       internals->nb_rx_queues = 0;
+       internals->nb_tx_queues = 0;
+
+       dev->data->nb_rx_queues = (uint16_t)0;
+       dev->data->nb_tx_queues = (uint16_t)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 int
+eth_rx_queue_setup(struct rte_eth_dev *dev,
+               uint16_t rx_queue_id,
+               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 szedata2_rx_queue *szedata2_q =
+               &internals->rx_queue[rx_queue_id];
+       szedata2_q->mb_pool = mb_pool;
+       dev->data->rx_queues[rx_queue_id] = szedata2_q;
+       szedata2_q->in_port = dev->data->port_id;
+       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;
+       dev->data->tx_queues[tx_queue_id] = &internals->tx_queue[tx_queue_id];
+       return 0;
+}
+
+static void
+eth_mac_addr_set(struct rte_eth_dev *dev __rte_unused,
+               struct ether_addr *mac_addr __rte_unused)
+{
+}
+
+static 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,
+               .mac_addr_set       = eth_mac_addr_set,
+};
+
+static int
+parse_mask(const char *mask_str, uint32_t *mask_num)
+{
+       char *endptr;
+       long int value;
+
+       value = strtol(mask_str, &endptr, 0);
+       if (*endptr != '\0' || value > UINT32_MAX || value < 0)
+               return -1;
+
+       *mask_num = (uint32_t)value;
+       return 0;
+}
+
+static int
+add_rx_mask(const char *key __rte_unused, const char *value, void *extra_args)
+{
+       struct rxtx_szedata2 *szedata2 = extra_args;
+       uint32_t mask;
+
+       if (parse_mask(value, &mask) != 0)
+               return -1;
+
+       szedata2->sze_rx_mask_req |= mask;
+       return 0;
+}
+
+static int
+add_tx_mask(const char *key __rte_unused, const char *value, void *extra_args)
+{
+       struct rxtx_szedata2 *szedata2 = extra_args;
+       uint32_t mask;
+
+       if (parse_mask(value, &mask) != 0)
+               return -1;
+
+       szedata2->sze_tx_mask_req |= mask;
+       return 0;
+}
+
+static int
+rte_pmd_init_internals(const char *name, const unsigned nb_rx_queues,
+               const unsigned nb_tx_queues,
+               const unsigned numa_node,
+               struct pmd_internals **internals,
+               struct rte_eth_dev **eth_dev)
+{
+       struct rte_eth_dev_data *data = NULL;
+
+       RTE_LOG(INFO, PMD,
+                       "Creating szedata2-backed ethdev on numa socket %u\n",
+                       numa_node);
+
+       /*
+        * now do all data allocation - for eth_dev structure
+        * and internal (private) data
+        */
+       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;
+
+       /*
+        * now put it all together
+        * - store queue data in internals,
+        * - store numa_node info in pci_driver
+        * - point eth_dev_data to internals
+        * - and point eth_dev structure to new eth_dev_data structure
+        *
+        * NOTE: we'll replace the data element, of originally allocated eth_dev
+        * so the rings are local per-process
+        */
+
+       (*internals)->nb_rx_queues = nb_rx_queues;
+       (*internals)->nb_tx_queues = nb_tx_queues;
+
+       (*internals)->if_index = 0;
+
+       data->dev_private = *internals;
+       data->port_id = (*eth_dev)->data->port_id;
+       snprintf(data->name, sizeof(data->name), "%s", (*eth_dev)->data->name);
+       data->nb_rx_queues = (uint16_t)nb_rx_queues;
+       data->nb_tx_queues = (uint16_t)nb_tx_queues;
+       data->dev_link = pmd_link;
+       data->mac_addrs = &eth_addr;
+
+       (*eth_dev)->data = data;
+       (*eth_dev)->dev_ops = &ops;
+       (*eth_dev)->data->dev_flags = RTE_ETH_DEV_DETACHABLE;
+       (*eth_dev)->driver = NULL;
+       (*eth_dev)->data->kdrv = RTE_KDRV_NONE;
+       (*eth_dev)->data->drv_name = drivername;
+       (*eth_dev)->data->numa_node = numa_node;
+
+       return 0;
+
+error:
+       rte_free(data);
+       rte_free(*internals);
+       return -1;
+}
+
+static int
+rte_eth_from_szedata2(const char *name,
+               struct rxtx_szedata2 *szedata2,
+               const unsigned numa_node)
+{
+       struct pmd_internals *internals = NULL;
+       struct rte_eth_dev *eth_dev = NULL;
+       int ret;
+
+       if (rte_pmd_init_internals(name, 0, 0, numa_node,
+                       &internals, &eth_dev) < 0)
+               return -1;
+
+       internals->sze_dev = szedata2->sze_dev;
+       internals->sze_rx_req = szedata2->sze_rx_mask_req;
+       internals->sze_tx_req = szedata2->sze_tx_mask_req;
+       internals->num_of_rx = szedata2->num_of_rx;
+       internals->num_of_tx = szedata2->num_of_tx;
+
+       RTE_LOG(INFO, PMD, "Number of rx channels to open: %u mask: 0x%x\n",
+                       internals->num_of_rx, internals->sze_rx_req);
+       RTE_LOG(INFO, PMD, "Number of tx channels to open: %u mask: 0x%x\n",
+                       internals->num_of_tx, internals->sze_tx_req);
+
+       ret = init_rx_channels(eth_dev, 1);
+       if (ret != 0) {
+               close_rx_channels(eth_dev);
+               return -1;
+       }
+
+       ret = init_tx_channels(eth_dev, 1);
+       if (ret != 0) {
+               close_tx_channels(eth_dev);
+               close_rx_channels(eth_dev);
+               return -1;
+       }
+
+       eth_dev->rx_pkt_burst = NULL;
+       eth_dev->tx_pkt_burst = NULL;
+
+       return 0;
+}
+
+
+static int
+rte_pmd_szedata2_devinit(const char *name, const char *params)
+{
+       unsigned numa_node;
+       int ret;
+       struct rte_kvargs *kvlist;
+       unsigned k_idx;
+       struct rte_kvargs_pair *pair = NULL;
+       struct rxtx_szedata2 szedata2 = { 0, 0, 0, 0, NULL };
+       bool dev_path_missing = true;
+
+       RTE_LOG(INFO, PMD, "Initializing pmd_szedata2 for %s\n", name);
+
+       numa_node = rte_socket_id();
+
+       kvlist = rte_kvargs_parse(params, valid_arguments);
+       if (kvlist == NULL)
+               return -1;
+
+       /*
+        * Get szedata2 device path and rx,tx channels from passed arguments.
+        */
+
+       if (rte_kvargs_count(kvlist, RTE_ETH_SZEDATA2_DEV_PATH_ARG) != 1)
+               goto err;
+
+       if (rte_kvargs_count(kvlist, RTE_ETH_SZEDATA2_RX_IFACES_ARG) < 1)
+               goto err;
+
+       if (rte_kvargs_count(kvlist, RTE_ETH_SZEDATA2_TX_IFACES_ARG) < 1)
+               goto err;
+
+       for (k_idx = 0; k_idx < kvlist->count; k_idx++) {
+               pair = &kvlist->pairs[k_idx];
+               if (strstr(pair->key, RTE_ETH_SZEDATA2_DEV_PATH_ARG) != NULL) {
+                       szedata2.sze_dev = pair->value;
+                       dev_path_missing = false;
+                       break;
+               }
+       }
+
+       if (dev_path_missing)
+               goto err;
+
+       ret = rte_kvargs_process(kvlist, RTE_ETH_SZEDATA2_RX_IFACES_ARG,
+                       &add_rx_mask, &szedata2);
+       if (ret < 0)
+               goto err;
+
+       ret = rte_kvargs_process(kvlist, RTE_ETH_SZEDATA2_TX_IFACES_ARG,
+                       &add_tx_mask, &szedata2);
+       if (ret < 0)
+               goto err;
+
+       szedata2.num_of_rx = count_ones(szedata2.sze_rx_mask_req);
+       szedata2.num_of_tx = count_ones(szedata2.sze_tx_mask_req);
+
+       RTE_LOG(INFO, PMD, "SZE device found at path %s\n", szedata2.sze_dev);
+
+       return rte_eth_from_szedata2(name, &szedata2, numa_node);
+err:
+       rte_kvargs_free(kvlist);
+       return -1;
+}
+
+static int
+rte_pmd_szedata2_devuninit(const char *name)
+{
+       struct rte_eth_dev *dev = NULL;
+
+       RTE_LOG(INFO, PMD, "Uninitializing pmd_szedata2 for %s "
+                       "on numa socket %u\n", name, rte_socket_id());
+
+       if (name == NULL)
+               return -1;
+
+       dev = rte_eth_dev_allocated(name);
+       if (dev == NULL)
+               return -1;
+
+       rte_free(dev->data->dev_private);
+       rte_free(dev->data);
+       rte_eth_dev_release_port(dev);
+       return 0;
+}
+
+static struct rte_driver pmd_szedata2_drv = {
+       .name = "eth_szedata2",
+       .type = PMD_VDEV,
+       .init = rte_pmd_szedata2_devinit,
+       .uninit = rte_pmd_szedata2_devuninit,
+};
+
+PMD_REGISTER_DRIVER(pmd_szedata2_drv);
diff --git a/drivers/net/szedata2/rte_eth_szedata2.h b/drivers/net/szedata2/rte_eth_szedata2.h
new file mode 100644 (file)
index 0000000..4d09a98
--- /dev/null
@@ -0,0 +1,102 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright (c) 2015 CESNET
+ *   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 CESNET 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_PMD_SZEDATA2_H_
+#define RTE_PMD_SZEDATA2_H_
+
+/* szedata2_packet header length == 4 bytes == 2B segment size + 2B hw size */
+#define RTE_SZE2_PACKET_HEADER_SIZE 4
+
+#define RTE_SZE2_MMIO_MAX 10
+
+/*!
+ * Round 'what' to the nearest larger (or equal) multiple of '8'
+ * (szedata2 packet is aligned to 8 bytes)
+ */
+#define RTE_SZE2_ALIGN8(what) (((what) + ((8) - 1)) & (~((8) - 1)))
+
+/*! main handle structure */
+struct szedata {
+       int fd;
+       struct sze2_instance_info *info;
+       uint32_t *write_size;
+       void *space[RTE_SZE2_MMIO_MAX];
+       struct szedata_lock lock[2][2];
+
+       __u32 *rx_asize, *tx_asize;
+
+       /* szedata_read_next variables - to keep context (ct) */
+
+       /*
+        * rx
+        */
+       /** initial sze lock ptr */
+       const struct szedata_lock   *ct_rx_lck_orig;
+       /** current sze lock ptr (initial or next) */
+       const struct szedata_lock   *ct_rx_lck;
+       /** remaining bytes (not read) within current lock */
+       unsigned int                ct_rx_rem_bytes;
+       /** current pointer to locked memory */
+       unsigned char               *ct_rx_cur_ptr;
+       /**
+        * allocated buffer to store RX packet if it was split
+        * into 2 buffers
+        */
+       unsigned char               *ct_rx_buffer;
+       /** registered function to provide filtering based on hwdata */
+       int (*ct_rx_filter)(u_int16_t hwdata_len, u_char *hwdata);
+
+       /*
+        * tx
+        */
+       /**
+        * buffer for tx - packet is prepared here
+        * (in future for burst write)
+        */
+       unsigned char               *ct_tx_buffer;
+       /** initial sze TX lock ptrs - number according to TX interfaces */
+       const struct szedata_lock   **ct_tx_lck_orig;
+       /** current sze TX lock ptrs - number according to TX interfaces */
+       const struct szedata_lock   **ct_tx_lck;
+       /** already written bytes in both locks */
+       unsigned int                *ct_tx_written_bytes;
+       /** remaining bytes (not written) within current lock */
+       unsigned int                *ct_tx_rem_bytes;
+       /** current pointers to locked memory */
+       unsigned char               **ct_tx_cur_ptr;
+       /** NUMA node closest to PCIe device, or -1 */
+       int                         numa_node;
+};
+
+
+#endif
diff --git a/drivers/net/szedata2/rte_pmd_szedata2_version.map b/drivers/net/szedata2/rte_pmd_szedata2_version.map
new file mode 100644 (file)
index 0000000..ad607bb
--- /dev/null
@@ -0,0 +1,3 @@
+DPDK_2.2 {
+       local: *;
+};
index 724efa7..148653e 100644 (file)
@@ -92,6 +92,8 @@ endif # ! CONFIG_RTE_BUILD_COMBINE_LIBS
 
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_PCAP)       += -lpcap
 
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2)   += -lsze2
+
 ifeq ($(CONFIG_RTE_LIBRTE_VHOST_NUMA),y)
 _LDLIBS-$(CONFIG_RTE_LIBRTE_VHOST)          += -lnuma
 endif
@@ -142,6 +144,7 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_IXGBE_PMD)      += -lrte_pmd_ixgbe
 _LDLIBS-$(CONFIG_RTE_LIBRTE_E1000_PMD)      += -lrte_pmd_e1000
 _LDLIBS-$(CONFIG_RTE_LIBRTE_MLX4_PMD)       += -lrte_pmd_mlx4
 _LDLIBS-$(CONFIG_RTE_LIBRTE_MLX5_PMD)       += -lrte_pmd_mlx5
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_SZEDATA2)   += -lrte_pmd_szedata2
 _LDLIBS-$(CONFIG_RTE_LIBRTE_MPIPE_PMD)      += -lrte_pmd_mpipe -lgxio
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_RING)       += -lrte_pmd_ring
 _LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_PCAP)       += -lrte_pmd_pcap