app/test: unit tests for bonding mode 4
authorTomasz Kulasek <tomaszx.kulasek@intel.com>
Fri, 13 Feb 2015 10:38:14 +0000 (11:38 +0100)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Wed, 18 Feb 2015 17:58:55 +0000 (18:58 +0100)
This patch adds unit tests for mode 4. It is split into separate
file to avoid problems with other modes that does not need to
look into packets payload.
This patch includes also a modification of maximum number of ports
used in their tests for bonding modes 0-3 from 16 to 6.

Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
MAINTAINERS
app/test/Makefile
app/test/test_link_bonding.c
app/test/test_link_bonding_mode4.c [new file with mode: 0644]

index d7d672c..7ac6d59 100644 (file)
@@ -185,7 +185,7 @@ Link bonding
 M: Declan Doherty <declan.doherty@intel.com>
 F: lib/librte_pmd_bond/
 F: doc/guides/prog_guide/link_bonding_poll_mode_drv_lib.rst
-F: app/test/test_link_bonding.c
+F: app/test/test_link_bonding*
 
 Linux KNI
 M: Helin Zhang <helin.zhang@intel.com>
index 24b27d7..a4e4bae 100644 (file)
@@ -131,6 +131,7 @@ SRCS-y += virtual_pmd.c
 SRCS-y += packet_burst_generator.c
 SRCS-$(CONFIG_RTE_LIBRTE_ACL) += test_acl.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_BOND) += test_link_bonding.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_BOND) += test_link_bonding_mode4.c
 SRCS-$(CONFIG_RTE_LIBRTE_PMD_RING) += test_pmd_ring.c
 SRCS-$(CONFIG_RTE_LIBRTE_KVARGS) += test_kvargs.c
 
index 579ebbf..c5bd78d 100644 (file)
@@ -57,7 +57,7 @@
 
 #include "test.h"
 
-#define TEST_MAX_NUMBER_OF_PORTS (16)
+#define TEST_MAX_NUMBER_OF_PORTS (6)
 
 #define RX_RING_SIZE 128
 #define RX_FREE_THRESH 32
diff --git a/app/test/test_link_bonding_mode4.c b/app/test/test_link_bonding_mode4.c
new file mode 100644 (file)
index 0000000..c35129f
--- /dev/null
@@ -0,0 +1,1413 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2014 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 <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <inttypes.h>
+#include <errno.h>
+#include <rte_cycles.h>
+#include <sys/queue.h>
+
+#include <rte_byteorder.h>
+#include <rte_common.h>
+#include <rte_debug.h>
+#include <rte_ethdev.h>
+#include <rte_log.h>
+#include <rte_lcore.h>
+#include <rte_memory.h>
+
+#include <rte_string_fns.h>
+
+#include <rte_eth_ring.h>
+#include <rte_errno.h>
+#include <rte_eth_bond.h>
+#include <rte_eth_bond_8023ad.h>
+
+#include "packet_burst_generator.h"
+
+#include "test.h"
+
+#define SLAVE_COUNT (4)
+
+#define RX_RING_SIZE 128
+#define TX_RING_SIZE 512
+
+#define MBUF_PAYLOAD_SIZE          (2048)
+#define MBUF_SIZE (MBUF_PAYLOAD_SIZE + sizeof(struct rte_mbuf) + \
+       RTE_PKTMBUF_HEADROOM)
+#define MBUF_CACHE_SIZE         (250)
+#define BURST_SIZE              (32)
+
+#define DEFAULT_MBUF_DATA_SIZE  (2048)
+#define TEST_RX_DESC_MAX        (2048)
+#define TEST_TX_DESC_MAX        (2048)
+#define MAX_PKT_BURST           (32)
+#define DEF_PKT_BURST           (16)
+
+#define BONDED_DEV_NAME         ("unit_test_mode4_bond_dev")
+
+#define SLAVE_DEV_NAME_FMT      ("unit_test_mode4_slave_%d")
+#define SLAVE_RX_QUEUE_FMT      ("unit_test_mode4_slave_%d_rx")
+#define SLAVE_TX_QUEUE_FMT      ("unit_test_mode4_slave_%d_tx")
+
+#define INVALID_SOCKET_ID       (-1)
+#define INVALID_PORT_ID         (0xFF)
+#define INVALID_BONDING_MODE    (-1)
+
+static const struct ether_addr slave_mac_default = {
+       { 0x00, 0xFF, 0x00, 0xFF, 0x00, 0x00 }
+};
+
+static const struct ether_addr parnter_mac_default = {
+       { 0x22, 0xBB, 0xFF, 0xBB, 0x00, 0x00 }
+};
+
+static const struct ether_addr parnter_system = {
+       { 0x33, 0xFF, 0xBB, 0xFF, 0x00, 0x00 }
+};
+
+static const struct ether_addr slow_protocol_mac_addr = {
+       { 0x01, 0x80, 0xC2, 0x00, 0x00, 0x02 }
+};
+
+struct slave_conf {
+       struct rte_ring *rx_queue;
+       struct rte_ring *tx_queue;
+       uint8_t port_id;
+       uint8_t bonded : 1;
+
+       uint8_t lacp_parnter_state;
+};
+
+struct ether_vlan_hdr {
+       struct ether_hdr pkt_eth_hdr;
+       struct vlan_hdr vlan_hdr;
+};
+
+struct link_bonding_unittest_params {
+       uint8_t bonded_port_id;
+       struct slave_conf slave_ports[SLAVE_COUNT];
+
+       struct rte_mempool *mbuf_pool;
+};
+
+#define TEST_DEFAULT_SLAVE_COUNT     RTE_DIM(test_params.slave_ports)
+#define TEST_RX_SLAVE_COUT           TEST_DEFAULT_SLAVE_COUNT
+#define TEST_TX_SLAVE_COUNT          TEST_DEFAULT_SLAVE_COUNT
+#define TEST_MARKER_SLAVE_COUT       TEST_DEFAULT_SLAVE_COUNT
+#define TEST_EXPIRED_SLAVE_COUNT     TEST_DEFAULT_SLAVE_COUNT
+#define TEST_PROMISC_SLAVE_COUNT     TEST_DEFAULT_SLAVE_COUNT
+
+static struct link_bonding_unittest_params test_params  = {
+       .bonded_port_id = INVALID_PORT_ID,
+       .slave_ports = { [0 ... SLAVE_COUNT - 1] = { .port_id = INVALID_PORT_ID} },
+
+       .mbuf_pool = NULL,
+};
+
+static struct rte_eth_conf default_pmd_conf = {
+       .rxmode = {
+               .mq_mode = ETH_MQ_RX_NONE,
+               .max_rx_pkt_len = ETHER_MAX_LEN,
+               .split_hdr_size = 0,
+               .header_split   = 0, /**< Header Split disabled */
+               .hw_ip_checksum = 0, /**< IP checksum offload enabled */
+               .hw_vlan_filter = 0, /**< VLAN filtering disabled */
+               .jumbo_frame    = 0, /**< Jumbo Frame Support disabled */
+               .hw_strip_crc   = 0, /**< CRC stripped by hardware */
+       },
+       .txmode = {
+               .mq_mode = ETH_MQ_TX_NONE,
+       },
+       .lpbk_mode = 0,
+};
+
+#define FOR_EACH(_i, _item, _array, _size) \
+       for (_i = 0, _item = &_array[0]; _i < _size && (_item = &_array[_i]); _i++)
+
+/* Macro for iterating over every port that can be used as a slave
+ * in this test.
+ * _i variable used as an index in test_params->slave_ports
+ * _slave pointer to &test_params->slave_ports[_idx]
+ */
+#define FOR_EACH_PORT(_i, _port) \
+       FOR_EACH(_i, _port, test_params.slave_ports, \
+               RTE_DIM(test_params.slave_ports))
+
+/* Macro for iterating over every port that can be used as a slave
+ * in this test and satisfy given condition.
+ *
+ * _i variable used as an index in test_params->slave_ports
+ * _slave pointer to &test_params->slave_ports[_idx]
+ * _condition condition that need to be checked
+ */
+#define FOR_EACH_PORT_IF(_i, _port, _condition) FOR_EACH_PORT((_i), (_port)) \
+       if (!!(_condition))
+
+/* Macro for iterating over every port that is currently a slave of a bonded
+ * device.
+ * _i variable used as an index in test_params->slave_ports
+ * _slave pointer to &test_params->slave_ports[_idx]
+ * */
+#define FOR_EACH_SLAVE(_i, _slave) \
+       FOR_EACH_PORT_IF(_i, _slave, (_slave)->bonded != 0)
+
+/*
+ * Returns packets from slaves TX queue.
+ * slave slave port
+ * buffer for packets
+ * size size of buffer
+ * return number of packets or negative error number
+ */
+static int
+slave_get_pkts(struct slave_conf *slave, struct rte_mbuf **buf, uint16_t size)
+{
+       return rte_ring_dequeue_burst(slave->tx_queue, (void **)buf, size);
+}
+
+/*
+ * Injects given packets into slaves RX queue.
+ * slave slave port
+ * buffer for packets
+ * size number of packets to be injected
+ * return number of queued packets or negative error number
+ */
+static int
+slave_put_pkts(struct slave_conf *slave, struct rte_mbuf **buf, uint16_t size)
+{
+       return rte_ring_enqueue_burst(slave->rx_queue, (void **)buf, size);
+}
+
+static uint16_t
+bond_rx(struct rte_mbuf **buf, uint16_t size)
+{
+       return rte_eth_rx_burst(test_params.bonded_port_id, 0, buf, size);
+}
+
+static uint16_t
+bond_tx(struct rte_mbuf **buf, uint16_t size)
+{
+       return rte_eth_tx_burst(test_params.bonded_port_id, 0, buf, size);
+}
+
+static void
+free_pkts(struct rte_mbuf **pkts, uint16_t count)
+{
+       uint16_t i;
+
+       for (i = 0; i < count; i++) {
+               if (pkts[i] != NULL)
+                       rte_pktmbuf_free(pkts[i]);
+       }
+}
+
+static int
+configure_ethdev(uint8_t port_id, uint8_t start)
+{
+       TEST_ASSERT(rte_eth_dev_configure(port_id, 1, 1, &default_pmd_conf) == 0,
+               "Failed to configure device %u", port_id);
+
+       TEST_ASSERT(rte_eth_rx_queue_setup(port_id, 0, RX_RING_SIZE,
+               rte_eth_dev_socket_id(port_id), NULL, test_params.mbuf_pool) == 0,
+               "Failed to setup rx queue.");
+
+       TEST_ASSERT(rte_eth_tx_queue_setup(port_id, 0, TX_RING_SIZE,
+               rte_eth_dev_socket_id(port_id), NULL) == 0,
+               "Failed to setup tx queue.");
+
+       if (start) {
+               TEST_ASSERT(rte_eth_dev_start(port_id) == 0,
+               "Failed to start device (%d).", port_id);
+       }
+       return 0;
+}
+
+static int
+add_slave(struct slave_conf *slave, uint8_t start)
+{
+       struct ether_addr addr, addr_check;
+
+       /* Some sanity check */
+       RTE_VERIFY(test_params.slave_ports <= slave &&
+               slave - test_params.slave_ports < (int)RTE_DIM(test_params.slave_ports));
+       RTE_VERIFY(slave->bonded == 0);
+       RTE_VERIFY(slave->port_id != INVALID_PORT_ID);
+
+       ether_addr_copy(&slave_mac_default, &addr);
+       addr.addr_bytes[ETHER_ADDR_LEN - 1] = slave->port_id;
+
+       rte_eth_dev_mac_addr_remove(slave->port_id, &addr);
+
+       TEST_ASSERT_SUCCESS(rte_eth_dev_mac_addr_add(slave->port_id, &addr, 0),
+               "Failed to set slave MAC address");
+
+       TEST_ASSERT_SUCCESS(rte_eth_bond_slave_add(test_params.bonded_port_id,
+               slave->port_id),
+                       "Failed to add slave (idx=%u, id=%u) to bonding (id=%u)",
+                       (uint8_t)(slave - test_params.slave_ports), slave->port_id,
+                       test_params.bonded_port_id);
+
+       slave->bonded = 1;
+       if (start) {
+               TEST_ASSERT_SUCCESS(rte_eth_dev_start(slave->port_id),
+                       "Failed to start slave %u", slave->port_id);
+       }
+
+       rte_eth_macaddr_get(slave->port_id, &addr_check);
+       TEST_ASSERT_EQUAL(is_same_ether_addr(&addr, &addr_check), 1,
+                       "Slave MAC address is not as expected");
+
+       RTE_VERIFY(slave->lacp_parnter_state == 0);
+       return 0;
+}
+
+static int
+remove_slave(struct slave_conf *slave)
+{
+       ptrdiff_t slave_idx = slave - test_params.slave_ports;
+
+       RTE_VERIFY(test_params.slave_ports <= slave &&
+               slave_idx < (ptrdiff_t)RTE_DIM(test_params.slave_ports));
+
+       RTE_VERIFY(slave->bonded == 1);
+       RTE_VERIFY(slave->port_id != INVALID_PORT_ID);
+
+       TEST_ASSERT_EQUAL(rte_ring_count(slave->rx_queue), 0,
+               "Slave %u tx queue not empty while removing from bonding.",
+               slave->port_id);
+
+       TEST_ASSERT_EQUAL(rte_ring_count(slave->rx_queue), 0,
+               "Slave %u tx queue not empty while removing from bonding.",
+               slave->port_id);
+
+       TEST_ASSERT_EQUAL(rte_eth_bond_slave_remove(test_params.bonded_port_id,
+                       slave->port_id), 0,
+                       "Failed to remove slave (idx=%u, id=%u) from bonding (id=%u)",
+                       (uint8_t)slave_idx, slave->port_id,
+                       test_params.bonded_port_id);
+
+       slave->bonded = 0;
+       slave->lacp_parnter_state = 0;
+       return 0;
+}
+
+static int
+initialize_bonded_device_with_slaves(uint8_t slave_count, uint8_t start)
+{
+       uint8_t i;
+
+       RTE_VERIFY(test_params.bonded_port_id != INVALID_PORT_ID);
+
+       for (i = 0; i < slave_count; i++) {
+               TEST_ASSERT_SUCCESS(add_slave(&test_params.slave_ports[i], 1),
+                       "Failed to add port %u to bonded device.\n",
+                       test_params.slave_ports[i].port_id);
+       }
+
+       /* Reset mode 4 configuration */
+       rte_eth_bond_8023ad_setup(test_params.bonded_port_id, NULL);
+       rte_eth_promiscuous_disable(test_params.bonded_port_id);
+
+       if (start)
+               TEST_ASSERT_SUCCESS(rte_eth_dev_start(test_params.bonded_port_id),
+                       "Failed to start bonded device");
+
+       return TEST_SUCCESS;
+}
+
+static int
+remove_slaves_and_stop_bonded_device(void)
+{
+       struct slave_conf *slave;
+       int retval;
+       uint8_t slaves[RTE_MAX_ETHPORTS];
+       uint8_t i;
+
+       rte_eth_dev_stop(test_params.bonded_port_id);
+
+       FOR_EACH_SLAVE(i, slave)
+               remove_slave(slave);
+
+       retval = rte_eth_bond_slaves_get(test_params.bonded_port_id, slaves,
+               RTE_DIM(slaves));
+
+       TEST_ASSERT_EQUAL(retval, 0,
+               "Expected bonded device %u have 0 slaves but returned %d.",
+                       test_params.bonded_port_id, retval);
+
+       FOR_EACH_PORT(i, slave) {
+               rte_eth_dev_stop(slave->port_id);
+
+               TEST_ASSERT(slave->bonded == 0,
+                       "Port id=%u is still marked as enslaved.", slave->port_id);
+       }
+
+       return TEST_SUCCESS;
+}
+
+static int
+test_setup(void)
+{
+       int retval, nb_mbuf_per_pool;
+       char name[RTE_ETH_NAME_MAX_LEN];
+       struct slave_conf *port;
+       const uint8_t socket_id = rte_socket_id();
+       uint8_t i;
+
+       if (test_params.mbuf_pool == NULL) {
+               nb_mbuf_per_pool = TEST_RX_DESC_MAX + DEF_PKT_BURST +
+                                       TEST_TX_DESC_MAX + MAX_PKT_BURST;
+               test_params.mbuf_pool = rte_mempool_create("TEST_MODE4",
+                               nb_mbuf_per_pool, MBUF_SIZE, MBUF_CACHE_SIZE,
+                               sizeof(struct rte_pktmbuf_pool_private),
+                               rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
+                               socket_id, 0);
+
+               TEST_ASSERT(test_params.mbuf_pool != NULL,
+                       "rte_mempool_create failed\n");
+       }
+
+       /* Create / initialize ring eth devs. */
+       FOR_EACH_PORT(i, port) {
+               port = &test_params.slave_ports[i];
+
+               if (port->rx_queue == NULL) {
+                       retval = snprintf(name, RTE_DIM(name), SLAVE_RX_QUEUE_FMT, i);
+                       TEST_ASSERT(retval <= (int)RTE_DIM(name) - 1, "Name too long");
+                       port->rx_queue = rte_ring_create(name, RX_RING_SIZE, socket_id, 0);
+                       TEST_ASSERT(port->rx_queue != NULL,
+                               "Failed to allocate rx ring '%s': %s", name,
+                               rte_strerror(rte_errno));
+               }
+
+               if (port->tx_queue == NULL) {
+                       retval = snprintf(name, RTE_DIM(name), SLAVE_TX_QUEUE_FMT, i);
+                       TEST_ASSERT(retval <= (int)RTE_DIM(name) - 1, "Name too long");
+                       port->tx_queue = rte_ring_create(name, TX_RING_SIZE, socket_id, 0);
+                       TEST_ASSERT_NOT_NULL(port->tx_queue,
+                               "Failed to allocate tx ring '%s': %s", name,
+                               rte_strerror(rte_errno));
+               }
+
+               if (port->port_id == INVALID_PORT_ID) {
+                       retval = snprintf(name, RTE_DIM(name), SLAVE_DEV_NAME_FMT, i);
+                       TEST_ASSERT(retval < (int)RTE_DIM(name) - 1, "Name too long");
+                       retval = rte_eth_from_rings(name, &port->rx_queue, 1,
+                                       &port->tx_queue, 1, socket_id);
+                       TEST_ASSERT(retval >= 0,
+                               "Failed to create ring ethdev '%s'\n", name);
+
+                       port->port_id = rte_eth_dev_count() - 1;
+               }
+
+               retval = configure_ethdev(port->port_id, 1);
+               TEST_ASSERT_SUCCESS(retval, "Failed to configure virtual ethdev %s\n",
+                       name);
+       }
+
+       if (test_params.bonded_port_id == INVALID_PORT_ID) {
+               retval = rte_eth_bond_create(BONDED_DEV_NAME, BONDING_MODE_8023AD,
+                               socket_id);
+
+               TEST_ASSERT(retval >= 0, "Failed to create bonded ethdev %s",
+                               BONDED_DEV_NAME);
+
+               test_params.bonded_port_id = retval;
+               TEST_ASSERT_SUCCESS(configure_ethdev(test_params.bonded_port_id, 0),
+                               "Failed to configure bonded ethdev %s", BONDED_DEV_NAME);
+       } else if (rte_eth_bond_mode_get(test_params.bonded_port_id) !=
+                       BONDING_MODE_8023AD) {
+               TEST_ASSERT(rte_eth_bond_mode_set(test_params.bonded_port_id,
+                       BONDING_MODE_8023AD) == 0,
+                       "Failed to set ethdev %d to mode %d",
+                       test_params.bonded_port_id, BONDING_MODE_8023AD);
+       }
+
+       return 0;
+}
+
+static int
+testsuite_teardown(void)
+{
+       struct slave_conf *port;
+       uint8_t i;
+
+       /* Only stop ports.
+        * Any cleanup/reset state is done when particular test is
+        * started. */
+
+       rte_eth_dev_stop(test_params.bonded_port_id);
+
+       FOR_EACH_PORT(i, port)
+               rte_eth_dev_stop(port->port_id);
+
+       return 0;
+}
+
+/*
+ * Check if given LACP packet. If it is, make make replay packet to force
+ * COLLECTING state.
+ * return 0 when pkt is LACP frame, 1 if it is not slow frame, 2 if it is slow
+ * frame but not LACP
+ */
+static int
+make_lacp_reply(struct slave_conf *slave, struct rte_mbuf *pkt)
+{
+       struct ether_hdr *hdr;
+       struct slow_protocol_frame *slow_hdr;
+       struct lacpdu *lacp;
+
+       /* look for LACP */
+       hdr = rte_pktmbuf_mtod(pkt, struct ether_hdr *);
+       if (hdr->ether_type != rte_cpu_to_be_16(ETHER_TYPE_SLOW))
+               return 1;
+
+       slow_hdr = rte_pktmbuf_mtod(pkt, struct slow_protocol_frame *);
+       /* ignore packets of other types */
+       if (slow_hdr->slow_protocol.subtype != SLOW_SUBTYPE_LACP)
+               return 2;
+
+       slow_hdr = rte_pktmbuf_mtod(pkt, struct slow_protocol_frame *);
+
+       /* Change source address to partner address */
+       ether_addr_copy(&parnter_mac_default, &slow_hdr->eth_hdr.s_addr);
+       slow_hdr->eth_hdr.s_addr.addr_bytes[ETHER_ADDR_LEN - 1] = slave->port_id;
+
+       lacp = (struct lacpdu *) &slow_hdr->slow_protocol;
+       /* Save last received state */
+       slave->lacp_parnter_state = lacp->actor.state;
+       /* Change it into LACP replay by matching parameters. */
+       memcpy(&lacp->partner.port_params, &lacp->actor.port_params,
+               sizeof(struct port_params));
+
+       lacp->partner.state = lacp->actor.state;
+
+       ether_addr_copy(&parnter_system, &lacp->actor.port_params.system);
+       lacp->actor.state = STATE_LACP_ACTIVE |
+                                               STATE_SYNCHRONIZATION |
+                                               STATE_AGGREGATION |
+                                               STATE_COLLECTING |
+                                               STATE_DISTRIBUTING;
+
+       return 0;
+}
+
+/*
+ * Reads packets from given slave, search for LACP packet and reply them.
+ *
+ * Receives burst of packets from slave. Looks for LACP packet. Drops
+ * all other packets. Prepares response LACP and sends it back.
+ *
+ * return number of LACP received and replied, -1 on error.
+ */
+static int
+bond_handshake_reply(struct slave_conf *slave)
+{
+       int retval;
+       struct rte_mbuf *rx_buf[MAX_PKT_BURST];
+       struct rte_mbuf *lacp_tx_buf[MAX_PKT_BURST];
+       uint16_t lacp_tx_buf_cnt = 0, i;
+
+       retval = slave_get_pkts(slave, rx_buf, RTE_DIM(rx_buf));
+       TEST_ASSERT(retval >= 0, "Getting slave %u packets failed.",
+                       slave->port_id);
+
+       for (i = 0; i < (uint16_t)retval; i++) {
+               if (make_lacp_reply(slave, rx_buf[i]) == 0) {
+                       /* reply with actor's LACP */
+                       lacp_tx_buf[lacp_tx_buf_cnt++] = rx_buf[i];
+               } else
+                       rte_pktmbuf_free(rx_buf[i]);
+       }
+
+       if (lacp_tx_buf_cnt == 0)
+               return 0;
+
+       retval = slave_put_pkts(slave, lacp_tx_buf, lacp_tx_buf_cnt);
+       if (retval <= lacp_tx_buf_cnt) {
+               /* retval might be negative */
+               for (i = RTE_MAX(0, retval); retval < lacp_tx_buf_cnt; retval++)
+                       rte_pktmbuf_free(lacp_tx_buf[i]);
+       }
+
+       TEST_ASSERT_EQUAL(retval, lacp_tx_buf_cnt,
+               "Failed to equeue lacp packets into slave %u tx queue.",
+               slave->port_id);
+
+       return lacp_tx_buf_cnt;
+}
+
+/*
+ * Function check if given slave tx queue contains packets that make mode 4
+ * handshake complete. It will drain slave queue.
+ * return 0 if handshake not completed, 1 if handshake was complete,
+ */
+static int
+bond_handshake_done(struct slave_conf *slave)
+{
+       const uint8_t expected_state = STATE_LACP_ACTIVE | STATE_SYNCHRONIZATION |
+                       STATE_AGGREGATION | STATE_COLLECTING | STATE_DISTRIBUTING;
+
+       return slave->lacp_parnter_state == expected_state;
+}
+
+static unsigned
+bond_get_update_timeout_ms(void)
+{
+       struct rte_eth_bond_8023ad_conf conf;
+
+       rte_eth_bond_8023ad_conf_get(test_params.bonded_port_id, &conf);
+       return conf.update_timeout_ms;
+}
+
+/*
+ * Exchanges LACP packets with partner to achieve dynamic port configuration.
+ * return TEST_SUCCESS if initial handshake succeed, TEST_FAILED otherwise.
+ */
+static int
+bond_handshake(void)
+{
+       struct slave_conf *slave;
+       struct rte_mbuf *buf[MAX_PKT_BURST];
+       uint16_t nb_pkts;
+       uint8_t all_slaves_done, i, j;
+       uint8_t status[RTE_DIM(test_params.slave_ports)] = { 0 };
+       const unsigned delay = bond_get_update_timeout_ms();
+
+       /* Exchange LACP frames */
+       all_slaves_done = 0;
+       for (i = 0; i < 30 && all_slaves_done == 0; ++i) {
+               rte_delay_ms(delay);
+
+               all_slaves_done = 1;
+               FOR_EACH_SLAVE(j, slave) {
+                       /* If response already send, skip slave */
+                       if (status[j] != 0)
+                               continue;
+
+                       if (bond_handshake_reply(slave) < 0) {
+                               all_slaves_done = 0;
+                               break;
+                       }
+
+                       status[j] = bond_handshake_done(slave);
+                       if (status[j] == 0)
+                               all_slaves_done = 0;
+               }
+
+               nb_pkts = bond_tx(NULL, 0);
+               TEST_ASSERT_EQUAL(nb_pkts, 0, "Packets transmitted unexpectedly");
+
+               nb_pkts = bond_rx(buf, RTE_DIM(buf));
+               free_pkts(buf, nb_pkts);
+               TEST_ASSERT_EQUAL(nb_pkts, 0, "Packets received unexpectedly");
+       }
+       /* If response didn't send - report failure */
+       TEST_ASSERT_EQUAL(all_slaves_done, 1, "Bond handshake failed\n");
+
+       /* If flags doesn't match - report failure */
+       return all_slaves_done = 1 ? TEST_SUCCESS : TEST_FAILED;
+}
+
+#define TEST_LACP_SLAVE_COUT RTE_DIM(test_params.slave_ports)
+static int
+test_mode4_lacp(void)
+{
+       int retval;
+
+       retval = initialize_bonded_device_with_slaves(TEST_LACP_SLAVE_COUT, 1);
+       TEST_ASSERT_SUCCESS(retval, "Failed to initialize bonded device");
+
+       /* Test LACP handshake function */
+       retval = bond_handshake();
+       TEST_ASSERT_SUCCESS(retval, "Initial handshake failed");
+
+       retval = remove_slaves_and_stop_bonded_device();
+       TEST_ASSERT_SUCCESS(retval, "Test cleanup failed.");
+
+       return TEST_SUCCESS;
+}
+
+static int
+generate_packets(struct ether_addr *src_mac,
+       struct ether_addr *dst_mac, uint16_t count, struct rte_mbuf **buf)
+{
+       uint16_t pktlen = PACKET_BURST_GEN_PKT_LEN;
+       uint8_t vlan_enable = 0;
+       uint16_t vlan_id = 0;
+       uint8_t ip4_type = 1; /* 0 - ipv6 */
+
+       uint16_t src_port = 10, dst_port = 20;
+
+       uint32_t ip_src[4] = { [0 ... 2] = 0xDEADBEEF, [3] = IPv4(192, 168, 0, 1) };
+       uint32_t ip_dst[4] = { [0 ... 2] = 0xFEEDFACE, [3] = IPv4(192, 168, 0, 2) };
+
+       struct ether_hdr pkt_eth_hdr;
+       struct udp_hdr pkt_udp_hdr;
+       union {
+               struct ipv4_hdr v4;
+               struct ipv6_hdr v6;
+       } pkt_ip_hdr;
+
+       int retval;
+
+       initialize_eth_header(&pkt_eth_hdr, src_mac, dst_mac, ip4_type,
+                       vlan_enable, vlan_id);
+
+       if (ip4_type)
+               initialize_ipv4_header(&pkt_ip_hdr.v4, ip_src[3], ip_dst[3], pktlen);
+       else
+               initialize_ipv6_header(&pkt_ip_hdr.v6, (uint8_t *)ip_src,
+                       (uint8_t *)&ip_dst, pktlen);
+
+       initialize_udp_header(&pkt_udp_hdr, src_port, dst_port, 16);
+
+       retval = generate_packet_burst(test_params.mbuf_pool, buf,
+                       &pkt_eth_hdr, vlan_enable, &pkt_ip_hdr, 1, &pkt_udp_hdr,
+                       count, pktlen, 1);
+
+       if (retval > 0 && retval != count)
+               free_pkts(&buf[count - retval], retval);
+
+       TEST_ASSERT_EQUAL(retval, count, "Failed to generate %u packets",
+               count);
+
+       return count;
+}
+
+static int
+generate_and_put_packets(struct slave_conf *slave, struct ether_addr *src_mac,
+               struct ether_addr *dst_mac, uint16_t count)
+{
+       struct rte_mbuf *pkts[MAX_PKT_BURST];
+       int retval;
+
+       retval = generate_packets(src_mac, dst_mac, count, pkts);
+       if (retval != (int)count)
+               return retval;
+
+       retval = slave_put_pkts(slave, pkts, count);
+       if (retval > 0 && retval != count)
+               free_pkts(&pkts[retval], count - retval);
+
+       TEST_ASSERT_EQUAL(retval, count,
+               "Failed to enqueue packets into slave %u RX queue", slave->port_id);
+
+       return TEST_SUCCESS;
+}
+
+static int
+test_mode4_rx(void)
+{
+       struct slave_conf *slave;
+       uint16_t i, j;
+
+       uint16_t expected_pkts_cnt;
+       struct rte_mbuf *pkts[MAX_PKT_BURST];
+       int retval;
+       unsigned delay;
+
+       struct ether_hdr *hdr;
+
+       struct ether_addr src_mac = { { 0x00, 0xFF, 0x00, 0xFF, 0x00, 0x00 } };
+       struct ether_addr dst_mac;
+       struct ether_addr bonded_mac;
+
+       retval = initialize_bonded_device_with_slaves(TEST_PROMISC_SLAVE_COUNT, 1);
+       TEST_ASSERT_SUCCESS(retval, "Failed to initialize bonded device");
+
+       retval = bond_handshake();
+       TEST_ASSERT_SUCCESS(retval, "Initial handshake failed");
+
+       rte_eth_macaddr_get(test_params.bonded_port_id, &bonded_mac);
+       ether_addr_copy(&bonded_mac, &dst_mac);
+
+       /* Assert that dst address is not bonding address */
+       dst_mac.addr_bytes[0]++;
+
+       /* First try with promiscuous mode enabled.
+        * Add 2 packets to each slave. First with bonding MAC address, second with
+        * different. Check if we received all of them. */
+       rte_eth_promiscuous_enable(test_params.bonded_port_id);
+
+       expected_pkts_cnt = 0;
+       FOR_EACH_SLAVE(i, slave) {
+               retval = generate_and_put_packets(slave, &src_mac, &bonded_mac, 1);
+               TEST_ASSERT_SUCCESS(retval, "Failed to enqueue packets to slave %u",
+                       slave->port_id);
+
+               retval = generate_and_put_packets(slave, &src_mac, &dst_mac, 1);
+               TEST_ASSERT_SUCCESS(retval, "Failed to enqueue packets to slave %u",
+                       slave->port_id);
+
+               /* Expect 2 packets per slave */
+               expected_pkts_cnt += 2;
+       }
+
+       retval = rte_eth_rx_burst(test_params.bonded_port_id, 0, pkts,
+               RTE_DIM(pkts));
+
+       if (retval == expected_pkts_cnt) {
+               int cnt[2] = { 0, 0 };
+
+               for (i = 0; i < expected_pkts_cnt; i++) {
+                       hdr = rte_pktmbuf_mtod(pkts[i], struct ether_hdr *);
+                       cnt[is_same_ether_addr(&hdr->d_addr, &bonded_mac)]++;
+               }
+
+               free_pkts(pkts, expected_pkts_cnt);
+
+               /* For division by 2 expected_pkts_cnt must be even */
+               RTE_VERIFY((expected_pkts_cnt & 1) == 0);
+               TEST_ASSERT(cnt[0] == expected_pkts_cnt / 2 &&
+                       cnt[1] == expected_pkts_cnt / 2,
+                       "Expected %u packets with the same MAC and %u with different but "
+                       "got %u with the same and %u with diffrent MAC",
+                       expected_pkts_cnt / 2, expected_pkts_cnt / 2, cnt[1], cnt[0]);
+       } else if (retval > 0)
+               free_pkts(pkts, retval);
+
+       TEST_ASSERT_EQUAL(retval, expected_pkts_cnt,
+               "Expected %u packets but received only %d", expected_pkts_cnt, retval);
+
+       /* Now, disable promiscuous mode. When promiscuous mode is disabled we
+        * expect to receive only packets that are directed to bonding port. */
+       rte_eth_promiscuous_disable(test_params.bonded_port_id);
+
+       expected_pkts_cnt = 0;
+       FOR_EACH_SLAVE(i, slave) {
+               retval = generate_and_put_packets(slave, &src_mac, &bonded_mac, 1);
+               TEST_ASSERT_SUCCESS(retval, "Failed to enqueue packets to slave %u",
+                       slave->port_id);
+
+               retval = generate_and_put_packets(slave, &src_mac, &dst_mac, 1);
+               TEST_ASSERT_SUCCESS(retval, "Failed to enqueue packets to slave %u",
+                       slave->port_id);
+
+               /* Expect only one packet per slave */
+               expected_pkts_cnt += 1;
+       }
+
+       retval = rte_eth_rx_burst(test_params.bonded_port_id, 0, pkts,
+               RTE_DIM(pkts));
+
+       if (retval == expected_pkts_cnt) {
+               int eq_cnt = 0;
+
+               for (i = 0; i < expected_pkts_cnt; i++) {
+                       hdr = rte_pktmbuf_mtod(pkts[i], struct ether_hdr *);
+                       eq_cnt += is_same_ether_addr(&hdr->d_addr, &bonded_mac);
+               }
+
+               free_pkts(pkts, expected_pkts_cnt);
+               TEST_ASSERT_EQUAL(eq_cnt, expected_pkts_cnt, "Packet address mismatch");
+       } else if (retval > 0)
+               free_pkts(pkts, retval);
+
+       TEST_ASSERT_EQUAL(retval, expected_pkts_cnt,
+               "Expected %u packets but received only %d", expected_pkts_cnt, retval);
+
+       /* Link down test: simulate link down for first slave. */
+       delay = bond_get_update_timeout_ms();
+
+       uint8_t slave_down_id = INVALID_PORT_ID;
+
+       /* Find first slave and make link down on it*/
+       FOR_EACH_SLAVE(i, slave) {
+               rte_eth_dev_set_link_down(slave->port_id);
+               slave_down_id = slave->port_id;
+               break;
+       }
+
+       RTE_VERIFY(slave_down_id != INVALID_PORT_ID);
+
+       /* Give some time to rearrange bonding */
+       for (i = 0; i < 3; i++) {
+               rte_delay_ms(delay);
+               bond_handshake();
+       }
+
+       TEST_ASSERT_SUCCESS(bond_handshake(), "Handshake after link down failed");
+
+       /* Put packet to each slave */
+       FOR_EACH_SLAVE(i, slave) {
+               void *pkt = NULL;
+
+               dst_mac.addr_bytes[ETHER_ADDR_LEN - 1] = slave->port_id;
+               retval = generate_and_put_packets(slave, &src_mac, &dst_mac, 1);
+               TEST_ASSERT_SUCCESS(retval, "Failed to generate test packet burst.");
+
+               src_mac.addr_bytes[ETHER_ADDR_LEN - 1] = slave->port_id;
+               retval = generate_and_put_packets(slave, &src_mac, &bonded_mac, 1);
+               TEST_ASSERT_SUCCESS(retval, "Failed to generate test packet burst.");
+
+               retval = bond_rx(pkts, RTE_DIM(pkts));
+
+               /* Clean anything */
+               if (retval > 0)
+                       free_pkts(pkts, retval);
+
+               while (rte_ring_dequeue(slave->rx_queue, (void **)&pkt) == 0)
+                       rte_pktmbuf_free(pkt);
+
+               if (slave_down_id == slave->port_id)
+                       TEST_ASSERT_EQUAL(retval, 0, "Packets received unexpectedly.");
+               else
+                       TEST_ASSERT_NOT_EQUAL(retval, 0,
+                               "Expected to receive some packets on slave %u.",
+                               slave->port_id);
+               rte_eth_dev_start(slave->port_id);
+
+               for (j = 0; j < 5; j++) {
+                       TEST_ASSERT(bond_handshake_reply(slave) >= 0,
+                               "Handshake after link up");
+
+                       if (bond_handshake_done(slave) == 1)
+                               break;
+               }
+
+               TEST_ASSERT(j < 5, "Failed to agregate slave after link up");
+       }
+
+       remove_slaves_and_stop_bonded_device();
+       return TEST_SUCCESS;
+}
+
+static int
+test_mode4_tx_burst(void)
+{
+       struct slave_conf *slave;
+       uint16_t i, j;
+
+       uint16_t exp_pkts_cnt, pkts_cnt = 0;
+       struct rte_mbuf *pkts[MAX_PKT_BURST];
+       int retval;
+       unsigned delay;
+
+       struct ether_addr dst_mac = { { 0x00, 0xFF, 0x00, 0xFF, 0x00, 0x00 } };
+       struct ether_addr bonded_mac;
+
+       retval = initialize_bonded_device_with_slaves(TEST_TX_SLAVE_COUNT, 1);
+       TEST_ASSERT_SUCCESS(retval, "Failed to initialize bonded device");
+
+       retval = bond_handshake();
+       TEST_ASSERT_SUCCESS(retval, "Initial handshake failed");
+
+       rte_eth_macaddr_get(test_params.bonded_port_id, &bonded_mac);
+
+       /* Prepare burst */
+       for (pkts_cnt = 0; pkts_cnt < RTE_DIM(pkts); pkts_cnt++) {
+               dst_mac.addr_bytes[ETHER_ADDR_LEN - 1] = pkts_cnt;
+               retval = generate_packets(&bonded_mac, &dst_mac, 1, &pkts[pkts_cnt]);
+
+               if (retval != 1)
+                       free_pkts(pkts, pkts_cnt);
+
+               TEST_ASSERT_EQUAL(retval, 1, "Failed to generate packet %u", pkts_cnt);
+       }
+       exp_pkts_cnt = pkts_cnt;
+
+       /* Transmit packets on bonded device */
+       retval = bond_tx(pkts, pkts_cnt);
+       if (retval > 0 && retval < pkts_cnt)
+               free_pkts(&pkts[retval], pkts_cnt - retval);
+
+       TEST_ASSERT_EQUAL(retval, pkts_cnt, "TX on bonded device failed");
+
+       /* Check if packets were transmitted properly. Every slave should have
+        * at least one packet, and sum must match. Under normal operation
+        * there should be no LACP nor MARKER frames. */
+       pkts_cnt = 0;
+       FOR_EACH_SLAVE(i, slave) {
+               uint16_t normal_cnt, slow_cnt;
+
+               retval = slave_get_pkts(slave, pkts, RTE_DIM(pkts));
+               normal_cnt = 0;
+               slow_cnt = 0;
+
+               for (j = 0; j < retval; j++) {
+                       if (make_lacp_reply(slave, pkts[j]) == 1)
+                               normal_cnt++;
+                       else
+                               slow_cnt++;
+               }
+
+               free_pkts(pkts, normal_cnt + slow_cnt);
+               TEST_ASSERT_EQUAL(slow_cnt, 0,
+                       "slave %u unexpectedly transmitted %d SLOW packets", slave->port_id,
+                       slow_cnt);
+
+               TEST_ASSERT_NOT_EQUAL(normal_cnt, 0,
+                       "slave %u did not transmitted any packets", slave->port_id);
+
+               pkts_cnt += normal_cnt;
+       }
+
+       TEST_ASSERT_EQUAL(exp_pkts_cnt, pkts_cnt,
+               "Expected %u packets but transmitted only %d", exp_pkts_cnt, pkts_cnt);
+
+       /* Link down test:
+        * simulate link down for first slave. */
+       delay = bond_get_update_timeout_ms();
+
+       uint8_t slave_down_id = INVALID_PORT_ID;
+
+       FOR_EACH_SLAVE(i, slave) {
+               rte_eth_dev_set_link_down(slave->port_id);
+               slave_down_id = slave->port_id;
+               break;
+       }
+
+       RTE_VERIFY(slave_down_id != INVALID_PORT_ID);
+
+       /* Give some time to rearrange bonding. */
+       for (i = 0; i < 3; i++) {
+               bond_handshake();
+               rte_delay_ms(delay);
+       }
+
+       TEST_ASSERT_SUCCESS(bond_handshake(), "Handshake after link down failed");
+
+       /* Prepare burst. */
+       for (pkts_cnt = 0; pkts_cnt < RTE_DIM(pkts); pkts_cnt++) {
+               dst_mac.addr_bytes[ETHER_ADDR_LEN - 1] = pkts_cnt;
+               retval = generate_packets(&bonded_mac, &dst_mac, 1, &pkts[pkts_cnt]);
+
+               if (retval != 1)
+                       free_pkts(pkts, pkts_cnt);
+
+               TEST_ASSERT_EQUAL(retval, 1, "Failed to generate test packet %u",
+                       pkts_cnt);
+       }
+       exp_pkts_cnt = pkts_cnt;
+
+       /* Transmit packets on bonded device. */
+       retval = bond_tx(pkts, pkts_cnt);
+       if (retval > 0 && retval < pkts_cnt)
+               free_pkts(&pkts[retval], pkts_cnt - retval);
+
+       TEST_ASSERT_EQUAL(retval, pkts_cnt, "TX on bonded device failed");
+
+       /* Check if packets was transmitted properly. Every slave should have
+        * at least one packet, and sum must match. Under normal operation
+        * there should be no LACP nor MARKER frames. */
+       pkts_cnt = 0;
+       FOR_EACH_SLAVE(i, slave) {
+               uint16_t normal_cnt, slow_cnt;
+
+               retval = slave_get_pkts(slave, pkts, RTE_DIM(pkts));
+               normal_cnt = 0;
+               slow_cnt = 0;
+
+               for (j = 0; j < retval; j++) {
+                       if (make_lacp_reply(slave, pkts[j]) == 1)
+                               normal_cnt++;
+                       else
+                               slow_cnt++;
+               }
+
+               free_pkts(pkts, normal_cnt + slow_cnt);
+
+               if (slave_down_id == slave->port_id) {
+                       TEST_ASSERT_EQUAL(normal_cnt + slow_cnt, 0,
+                               "slave %u enexpectedly transmitted %u packets",
+                               normal_cnt + slow_cnt, slave->port_id);
+               } else {
+                       TEST_ASSERT_EQUAL(slow_cnt, 0,
+                               "slave %u unexpectedly transmitted %d SLOW packets",
+                               slave->port_id, slow_cnt);
+
+                       TEST_ASSERT_NOT_EQUAL(normal_cnt, 0,
+                               "slave %u did not transmitted any packets", slave->port_id);
+               }
+
+               pkts_cnt += normal_cnt;
+       }
+
+       TEST_ASSERT_EQUAL(exp_pkts_cnt, pkts_cnt,
+               "Expected %u packets but transmitted only %d", exp_pkts_cnt, pkts_cnt);
+
+       return remove_slaves_and_stop_bonded_device();
+}
+
+static void
+init_marker(struct rte_mbuf *pkt, struct slave_conf *slave)
+{
+       struct marker_header *marker_hdr = rte_pktmbuf_mtod(pkt,
+                       struct marker_header *);
+
+       /* Copy multicast destination address */
+       ether_addr_copy(&slow_protocol_mac_addr, &marker_hdr->eth_hdr.d_addr);
+
+       /* Init source address */
+       ether_addr_copy(&parnter_mac_default, &marker_hdr->eth_hdr.s_addr);
+       marker_hdr->eth_hdr.s_addr.addr_bytes[ETHER_ADDR_LEN-1] = slave->port_id;
+
+       marker_hdr->eth_hdr.ether_type = rte_cpu_to_be_16(ETHER_TYPE_SLOW);
+
+       marker_hdr->marker.subtype = SLOW_SUBTYPE_MARKER;
+       marker_hdr->marker.version_number = 1;
+       marker_hdr->marker.tlv_type_marker = MARKER_TLV_TYPE_INFO;
+       marker_hdr->marker.info_length =
+                       offsetof(struct marker, reserved_90) -
+                       offsetof(struct marker, requester_port);
+       RTE_VERIFY(marker_hdr->marker.info_length == 16);
+       marker_hdr->marker.requester_port = slave->port_id + 1;
+       marker_hdr->marker.tlv_type_terminator = TLV_TYPE_TERMINATOR_INFORMATION;
+       marker_hdr->marker.terminator_length = 0;
+}
+
+static int
+test_mode4_marker(void)
+{
+       struct slave_conf *slave;
+       struct rte_mbuf *pkts[MAX_PKT_BURST];
+       struct rte_mbuf *marker_pkt;
+       struct marker_header *marker_hdr;
+
+       unsigned delay;
+       int retval;
+       uint16_t nb_pkts;
+       uint8_t i, j;
+       const uint16_t ethtype_slow_be = rte_be_to_cpu_16(ETHER_TYPE_SLOW);
+
+       retval = initialize_bonded_device_with_slaves(TEST_MARKER_SLAVE_COUT, 1);
+       TEST_ASSERT_SUCCESS(retval, "Failed to initialize bonded device");
+
+       /* Test LACP handshake function */
+       retval = bond_handshake();
+       TEST_ASSERT_SUCCESS(retval, "Initial handshake failed");
+
+       delay = bond_get_update_timeout_ms();
+       FOR_EACH_SLAVE(i, slave) {
+               marker_pkt = rte_pktmbuf_alloc(test_params.mbuf_pool);
+               TEST_ASSERT_NOT_NULL(marker_pkt, "Failed to allocate marker packet");
+               init_marker(marker_pkt, slave);
+
+               retval = slave_put_pkts(slave, &marker_pkt, 1);
+               if (retval != 1)
+                       rte_pktmbuf_free(marker_pkt);
+
+               TEST_ASSERT_EQUAL(retval, 1,
+                       "Failed to send marker packet to slave %u", slave->port_id);
+
+               for (j = 0; j < 20; ++j) {
+                       rte_delay_ms(delay);
+                       retval = rte_eth_rx_burst(test_params.bonded_port_id, 0, pkts,
+                               RTE_DIM(pkts));
+
+                       if (retval > 0)
+                               free_pkts(pkts, retval);
+
+                       TEST_ASSERT_EQUAL(retval, 0, "Received packets unexpectedly");
+
+                       retval = rte_eth_tx_burst(test_params.bonded_port_id, 0, NULL, 0);
+                       TEST_ASSERT_EQUAL(retval, 0,
+                               "Requested TX of 0 packets but %d transmitted", retval);
+
+                       /* Check if LACP packet was send by state machines
+                          First and only packet must be a maker response */
+                       retval = slave_get_pkts(slave, pkts, MAX_PKT_BURST);
+                       if (retval == 0)
+                               continue;
+                       if (retval > 1)
+                               free_pkts(pkts, retval);
+
+                       TEST_ASSERT_EQUAL(retval, 1, "failed to get slave packets");
+                       nb_pkts = retval;
+
+                       marker_hdr = rte_pktmbuf_mtod(pkts[0], struct marker_header *);
+                       /* Check if it's slow packet*/
+                       if (marker_hdr->eth_hdr.ether_type != ethtype_slow_be)
+                               retval = -1;
+                       /* Check if it's marker packet */
+                       else if (marker_hdr->marker.subtype != SLOW_SUBTYPE_MARKER)
+                               retval = -2;
+                       else if (marker_hdr->marker.tlv_type_marker != MARKER_TLV_TYPE_RESP)
+                               retval = -3;
+
+                       free_pkts(pkts, nb_pkts);
+
+                       TEST_ASSERT_NOT_EQUAL(retval, -1, "Unexpected protocol type");
+                       TEST_ASSERT_NOT_EQUAL(retval, -2, "Unexpected sub protocol type");
+                       TEST_ASSERT_NOT_EQUAL(retval, -3, "Unexpected marker type");
+                       break;
+               }
+
+               TEST_ASSERT(j < 20, "Marker response not found");
+       }
+
+       retval = remove_slaves_and_stop_bonded_device();
+       TEST_ASSERT_SUCCESS(retval,     "Test cleanup failed.");
+
+       return TEST_SUCCESS;
+}
+
+static int
+test_mode4_expired(void)
+{
+       struct slave_conf *slave, *exp_slave = NULL;
+       struct rte_mbuf *pkts[MAX_PKT_BURST];
+       struct rte_mbuf *pkt = NULL;
+       int retval;
+       uint32_t old_delay;
+
+       uint8_t i;
+       uint16_t j;
+
+       struct rte_eth_bond_8023ad_conf conf;
+
+       retval = initialize_bonded_device_with_slaves(TEST_EXPIRED_SLAVE_COUNT, 1);
+       /* Set custom timeouts to make test last shorter. */
+       rte_eth_bond_8023ad_conf_get(test_params.bonded_port_id, &conf);
+       conf.fast_periodic_ms = 100;
+       conf.slow_periodic_ms = 600;
+       conf.short_timeout_ms = 300;
+       conf.long_timeout_ms = 900;
+       conf.aggregate_wait_timeout_ms = 200;
+       conf.tx_period_ms = 100;
+       old_delay = conf.update_timeout_ms;
+       conf.update_timeout_ms = 10;
+       rte_eth_bond_8023ad_setup(test_params.bonded_port_id, &conf);
+
+       /* Wait for new settings to be applied. */
+       for (i = 0; i < old_delay/conf.update_timeout_ms * 2; i++) {
+               FOR_EACH_SLAVE(j, slave)
+                       bond_handshake_reply(slave);
+
+               rte_delay_ms(conf.update_timeout_ms);
+       }
+
+       retval = bond_handshake();
+       TEST_ASSERT_SUCCESS(retval, "Initial handshake failed");
+
+       /* Find first slave */
+       FOR_EACH_SLAVE(i, slave) {
+               exp_slave = slave;
+               break;
+       }
+
+       RTE_VERIFY(exp_slave != NULL);
+
+       /* When one of partners do not send or respond to LACP frame in
+        * conf.long_timeout_ms time, internal state machines should detect this
+        * and transit to expired state. */
+       for (j = 0; j < conf.long_timeout_ms/conf.update_timeout_ms + 2; j++) {
+               rte_delay_ms(conf.update_timeout_ms);
+
+               retval = bond_tx(NULL, 0);
+               TEST_ASSERT_EQUAL(retval, 0, "Unexpectedly received %d packets",
+                       retval);
+
+               FOR_EACH_SLAVE(i, slave) {
+                       retval = bond_handshake_reply(slave);
+                       TEST_ASSERT(retval >= 0, "Handshake failed");
+
+                       /* Remove replay for slave that supose to be expired. */
+                       if (slave == exp_slave) {
+                               while (rte_ring_count(slave->rx_queue) > 0) {
+                                       rte_ring_dequeue(slave->rx_queue, (void **)&pkt);
+                                       rte_pktmbuf_free(pkt);
+                               }
+                       }
+               }
+
+               retval = bond_rx(pkts, RTE_DIM(pkts));
+               if (retval > 0)
+                       free_pkts(pkts, retval);
+
+               TEST_ASSERT_EQUAL(retval, 0, "Unexpectedly received %d packets",
+                       retval);
+       }
+
+       /* After test only expected slave should be in EXPIRED state */
+       FOR_EACH_SLAVE(i, slave) {
+               if (slave == exp_slave)
+                       TEST_ASSERT(slave->lacp_parnter_state & STATE_EXPIRED,
+                               "Slave %u should be in expired.", slave->port_id);
+               else
+                       TEST_ASSERT_EQUAL(bond_handshake_done(slave), 1,
+                               "Slave %u should be operational.", slave->port_id);
+       }
+
+       retval = remove_slaves_and_stop_bonded_device();
+       TEST_ASSERT_SUCCESS(retval, "Test cleanup failed.");
+
+       return TEST_SUCCESS;
+}
+
+static int
+check_environment(void)
+{
+       struct slave_conf *port;
+       uint8_t i, env_state;
+       uint8_t slaves[RTE_DIM(test_params.slave_ports)];
+       int slaves_count;
+
+       env_state = 0;
+       FOR_EACH_PORT(i, port) {
+               if (rte_ring_count(port->rx_queue) != 0)
+                       env_state |= 0x01;
+
+               if (rte_ring_count(port->tx_queue) != 0)
+                       env_state |= 0x02;
+
+               if (port->bonded != 0)
+                       env_state |= 0x04;
+
+               if (port->lacp_parnter_state != 0)
+                       env_state |= 0x08;
+
+               if (env_state != 0)
+                       break;
+       }
+
+       slaves_count = rte_eth_bond_slaves_get(test_params.bonded_port_id,
+                       slaves, RTE_DIM(slaves));
+
+       if (slaves_count != 0)
+               env_state |= 0x10;
+
+       TEST_ASSERT_EQUAL(env_state, 0,
+               "Environment not clean (port %u):%s%s%s%s%s",
+               port->port_id,
+               env_state & 0x01 ? " slave rx queue not clean" : "",
+               env_state & 0x02 ? " slave tx queue not clean" : "",
+               env_state & 0x04 ? " port marked as enslaved" : "",
+               env_state & 0x80 ? " slave state is not reset" : "",
+               env_state & 0x10 ? " slave count not equal 0" : ".");
+
+
+       return TEST_SUCCESS;
+}
+
+static int
+test_mode4_executor(int (*test_func)(void))
+{
+       struct slave_conf *port;
+       int test_result;
+       uint8_t i;
+       void *pkt;
+
+       /* Check if environment is clean. Fail to launch a test if there was
+        * a critical error before that prevented to reset environment. */
+       TEST_ASSERT_SUCCESS(check_environment(),
+               "Refusing to launch test in dirty environment.");
+
+       RTE_VERIFY(test_func != NULL);
+       test_result = (*test_func)();
+
+       /* If test succeed check if environment wast left in good condition. */
+       if (test_result == TEST_SUCCESS)
+               test_result = check_environment();
+
+       /* Reset environment in case test failed to do that. */
+       if (test_result != TEST_SUCCESS) {
+               TEST_ASSERT_SUCCESS(remove_slaves_and_stop_bonded_device(),
+                       "Failed to stop bonded device");
+
+               FOR_EACH_PORT(i, port) {
+                       while (rte_ring_count(port->rx_queue) != 0) {
+                               if (rte_ring_dequeue(port->rx_queue, &pkt) == 0)
+                                       rte_pktmbuf_free(pkt);
+                       }
+
+                       while (rte_ring_count(port->tx_queue) != 0) {
+                               if (rte_ring_dequeue(port->tx_queue, &pkt) == 0)
+                                       rte_pktmbuf_free(pkt);
+                       }
+               }
+       }
+
+       return test_result;
+}
+
+static int
+test_mode4_lacp_wrapper(void)
+{
+       return test_mode4_executor(&test_mode4_lacp);
+}
+
+static int
+test_mode4_marker_wrapper(void)
+{
+       return test_mode4_executor(&test_mode4_marker);
+}
+
+static int
+test_mode4_rx_wrapper(void)
+{
+       return test_mode4_executor(&test_mode4_rx);
+}
+
+static int
+test_mode4_tx_burst_wrapper(void)
+{
+       return test_mode4_executor(&test_mode4_tx_burst);
+}
+
+static int
+test_mode4_expired_wrapper(void)
+{
+       return test_mode4_executor(&test_mode4_expired);
+}
+
+static struct unit_test_suite link_bonding_mode4_test_suite  = {
+       .suite_name = "Link Bonding mode 4 Unit Test Suite",
+       .setup = test_setup,
+       .teardown = testsuite_teardown,
+       .unit_test_cases = {
+               TEST_CASE_NAMED("test_mode4_lacp", test_mode4_lacp_wrapper),
+               TEST_CASE_NAMED("test_mode4_rx", test_mode4_rx_wrapper),
+               TEST_CASE_NAMED("test_mode4_tx_burst", test_mode4_tx_burst_wrapper),
+               TEST_CASE_NAMED("test_mode4_marker", test_mode4_marker_wrapper),
+               TEST_CASE_NAMED("test_mode4_expired", test_mode4_expired_wrapper),
+               { NULL, NULL, NULL, NULL, NULL } /**< NULL terminate unit test array */
+       }
+};
+
+static int
+test_link_bonding_mode4(void)
+{
+       return unit_test_suite_runner(&link_bonding_mode4_test_suite);
+}
+
+static struct test_command link_bonding_cmd = {
+       .command = "link_bonding_mode4_autotest",
+       .callback = test_link_bonding_mode4,
+};
+
+REGISTER_TEST_COMMAND(link_bonding_cmd);