examples/bond: add example application for link bonding mode 6
authorMichal Jastrzebski <michalx.k.jastrzebski@intel.com>
Fri, 20 Feb 2015 16:09:21 +0000 (17:09 +0100)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Fri, 20 Feb 2015 22:07:02 +0000 (23:07 +0100)
This patch contains an example for link bonding mode 6.
It interact with user by a command prompt. Available commands are:
Start - starts ARP_thread which respond to ARP_requests and sends
ARP_updates (this
Is enabled by default after startup),
Stop  -stops ARP_thread,
Send count ip - send count ARP requests for IP,
Show - prints basic bond information, like IPv4 statistics from clients
Help,
Quit.
The best way to test mode 6 is to use this example together with
previous patch:
[PATCH 3/4] bond: add debug info for mode 6 link bonding.
Connect clients thru switch to bonding machine and send:
arping -c 1 bond_ip or
generate IPv4 traffic to bond_ip (IPv4 traffic from different clients
should be then balanced on slaves in round robin manner).

Signed-off-by: Michal Jastrzebski <michalx.k.jastrzebski@intel.com>
Signed-off-by: Maciej Gajdzica <maciejx.t.gajdzica@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
MAINTAINERS
examples/Makefile
examples/bond/Makefile [new file with mode: 0644]
examples/bond/main.c [new file with mode: 0644]
examples/bond/main.h [new file with mode: 0644]

index 7ac6d59..baf8e36 100644 (file)
@@ -186,6 +186,7 @@ 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*
+F: examples/bond/
 
 Linux KNI
 M: Helin Zhang <helin.zhang@intel.com>
index 095bad2..380c5d6 100644 (file)
@@ -37,6 +37,7 @@ RTE_TARGET ?= x86_64-native-linuxapp-gcc
 
 include $(RTE_SDK)/mk/rte.vars.mk
 
+DIRS-$(CONFIG_RTE_LIBRTE_PMD_BOND) += bond
 DIRS-y += cmdline
 DIRS-$(CONFIG_RTE_LIBRTE_DISTRIBUTOR) += distributor
 ifneq ($(ICP_ROOT),)
diff --git a/examples/bond/Makefile b/examples/bond/Makefile
new file mode 100644 (file)
index 0000000..626d79d
--- /dev/null
@@ -0,0 +1,57 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+#   All rights reserved.
+#
+#   Redistribution and use in source and binary forms, with or without
+#   modification, are permitted provided that the following conditions
+#   are met:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of Intel Corporation nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+ifeq ($(RTE_SDK),)
+$(error "Please define RTE_SDK environment variable")
+endif
+
+# Default target, can be overridden by command line or environment
+RTE_TARGET ?= x86_64-native-linuxapp-gcc
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# binary name
+APP = bond_app
+
+# all source are stored in SRCS-y
+SRCS-y := main.c
+
+CFLAGS += $(WERROR_FLAGS)
+
+# workaround for a gcc bug with noreturn attribute
+# http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12603
+ifeq ($(CONFIG_RTE_TOOLCHAIN_GCC),y)
+CFLAGS_main.o += -Wno-return-type
+endif
+
+CFLAGS += -O3
+
+include $(RTE_SDK)/mk/rte.extapp.mk
diff --git a/examples/bond/main.c b/examples/bond/main.c
new file mode 100644 (file)
index 0000000..e4457f2
--- /dev/null
@@ -0,0 +1,796 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 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 <stdint.h>
+#include <sys/queue.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <assert.h>
+#include <errno.h>
+#include <signal.h>
+#include <stdarg.h>
+#include <inttypes.h>
+#include <getopt.h>
+#include <termios.h>
+#include <unistd.h>
+#include <pthread.h>
+
+#include <rte_common.h>
+#include <rte_log.h>
+#include <rte_memory.h>
+#include <rte_memcpy.h>
+#include <rte_memzone.h>
+#include <rte_tailq.h>
+#include <rte_eal.h>
+#include <rte_per_lcore.h>
+#include <rte_launch.h>
+#include <rte_atomic.h>
+#include <rte_cycles.h>
+#include <rte_prefetch.h>
+#include <rte_lcore.h>
+#include <rte_per_lcore.h>
+#include <rte_branch_prediction.h>
+#include <rte_interrupts.h>
+#include <rte_pci.h>
+#include <rte_random.h>
+#include <rte_debug.h>
+#include <rte_ether.h>
+#include <rte_ethdev.h>
+#include <rte_ring.h>
+#include <rte_log.h>
+#include <rte_mempool.h>
+#include <rte_mbuf.h>
+#include <rte_memcpy.h>
+#include <rte_ip.h>
+#include <rte_tcp.h>
+#include <rte_arp.h>
+#include <rte_spinlock.h>
+
+#include <cmdline_rdline.h>
+#include <cmdline_parse.h>
+#include <cmdline_parse_num.h>
+#include <cmdline_parse_string.h>
+#include <cmdline_parse_ipaddr.h>
+#include <cmdline_parse_etheraddr.h>
+#include <cmdline_socket.h>
+#include <cmdline.h>
+
+#include "main.h"
+
+#include <rte_devargs.h>
+
+
+#include "rte_byteorder.h"
+#include "rte_cpuflags.h"
+#include "rte_eth_bond.h"
+
+#define RTE_LOGTYPE_DCB RTE_LOGTYPE_USER1
+
+#define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
+#define NB_MBUF   (1024*8)
+
+#define MAX_PKT_BURST 32
+#define BURST_TX_DRAIN_US 100      /* TX drain every ~100us */
+#define BURST_RX_INTERVAL_NS (10) /* RX poll interval ~100ns */
+
+/*
+ * RX and TX Prefetch, Host, and Write-back threshold values should be
+ * carefully set for optimal performance. Consult the network
+ * controller's datasheet and supporting DPDK documentation for guidance
+ * on how these parameters should be set.
+ */
+#define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
+#define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
+#define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
+#define RX_FTHRESH (MAX_PKT_BURST * 2)/**< Default values of RX free threshold reg. */
+
+/*
+ * These default values are optimized for use with the Intel(R) 82599 10 GbE
+ * Controller and the DPDK ixgbe PMD. Consider using other values for other
+ * network controllers and/or network drivers.
+ */
+#define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
+#define TX_HTHRESH 0  /**< Default values of TX host threshold reg. */
+#define TX_WTHRESH 0  /**< Default values of TX write-back threshold reg. */
+
+/*
+ * Configurable number of RX/TX ring descriptors
+ */
+#define RTE_RX_DESC_DEFAULT 128
+#define RTE_TX_DESC_DEFAULT 512
+
+#define BOND_IP_1      7
+#define BOND_IP_2      0
+#define BOND_IP_3      0
+#define BOND_IP_4      10
+
+/* not defined under linux */
+#ifndef NIPQUAD
+#define NIPQUAD_FMT "%u.%u.%u.%u"
+#endif
+
+#define MAX_PORTS      4
+#define PRINT_MAC(addr)                printf("%02"PRIx8":%02"PRIx8":%02"PRIx8 \
+               ":%02"PRIx8":%02"PRIx8":%02"PRIx8,      \
+               addr.addr_bytes[0], addr.addr_bytes[1], addr.addr_bytes[2], \
+               addr.addr_bytes[3], addr.addr_bytes[4], addr.addr_bytes[5])
+
+uint8_t slaves[RTE_MAX_ETHPORTS];
+uint8_t slaves_count;
+
+static uint8_t BOND_PORT = 0xff;
+
+static struct rte_mempool *mbuf_pool;
+
+static struct rte_eth_conf port_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 */
+       },
+       .rx_adv_conf = {
+               .rss_conf = {
+                       .rss_key = NULL,
+                       .rss_hf = ETH_RSS_IP,
+               },
+       },
+       .txmode = {
+               .mq_mode = ETH_MQ_TX_NONE,
+       },
+};
+
+static void
+slave_port_init(uint8_t portid, struct rte_mempool *mbuf_pool)
+{
+       int retval;
+
+       if (portid >= rte_eth_dev_count())
+               rte_exit(EXIT_FAILURE, "Invalid port\n");
+
+       retval = rte_eth_dev_configure(portid, 1, 1, &port_conf);
+       if (retval != 0)
+               rte_exit(EXIT_FAILURE, "port %u: configuration failed (res=%d)\n",
+                               portid, retval);
+
+       /* RX setup */
+       retval = rte_eth_rx_queue_setup(portid, 0, RTE_RX_DESC_DEFAULT,
+                                       rte_eth_dev_socket_id(portid), NULL,
+                                       mbuf_pool);
+       if (retval < 0)
+               rte_exit(retval, " port %u: RX queue 0 setup failed (res=%d)",
+                               portid, retval);
+
+       /* TX setup */
+       retval = rte_eth_tx_queue_setup(portid, 0, RTE_TX_DESC_DEFAULT,
+                               rte_eth_dev_socket_id(portid), NULL);
+
+       if (retval < 0)
+               rte_exit(retval, "port %u: TX queue 0 setup failed (res=%d)",
+                               portid, retval);
+
+       retval  = rte_eth_dev_start(portid);
+       if (retval < 0)
+               rte_exit(retval,
+                               "Start port %d failed (res=%d)",
+                               portid, retval);
+
+       struct ether_addr addr;
+
+       rte_eth_macaddr_get(portid, &addr);
+       printf("Port %u MAC: ", (unsigned)portid);
+       PRINT_MAC(addr);
+       printf("\n");
+}
+
+static void
+bond_port_init(struct rte_mempool *mbuf_pool)
+{
+       int retval;
+       uint8_t i;
+
+       retval = rte_eth_bond_create("bond0", BONDING_MODE_ALB,
+                       0 /*SOCKET_ID_ANY*/);
+       if (retval < 0)
+               rte_exit(EXIT_FAILURE,
+                               "Faled to create bond port\n");
+
+       BOND_PORT = (uint8_t)retval;
+
+       retval = rte_eth_dev_configure(BOND_PORT, 1, 1, &port_conf);
+       if (retval != 0)
+               rte_exit(EXIT_FAILURE, "port %u: configuration failed (res=%d)\n",
+                               BOND_PORT, retval);
+
+       /* RX setup */
+       retval = rte_eth_rx_queue_setup(BOND_PORT, 0, RTE_RX_DESC_DEFAULT,
+                                       rte_eth_dev_socket_id(BOND_PORT), NULL,
+                                       mbuf_pool);
+       if (retval < 0)
+               rte_exit(retval, " port %u: RX queue 0 setup failed (res=%d)",
+                               BOND_PORT, retval);
+
+       /* TX setup */
+       retval = rte_eth_tx_queue_setup(BOND_PORT, 0, RTE_TX_DESC_DEFAULT,
+                               rte_eth_dev_socket_id(BOND_PORT), NULL);
+
+       if (retval < 0)
+               rte_exit(retval, "port %u: TX queue 0 setup failed (res=%d)",
+                               BOND_PORT, retval);
+
+       for (i = 0; i < slaves_count; i++) {
+               if (rte_eth_bond_slave_add(BOND_PORT, slaves[i]) == -1)
+                       rte_exit(-1, "Oooops! adding slave (%u) to bond (%u) failed!\n",
+                                       slaves[i], BOND_PORT);
+
+       }
+
+       retval  = rte_eth_dev_start(BOND_PORT);
+       if (retval < 0)
+               rte_exit(retval, "Start port %d failed (res=%d)", BOND_PORT, retval);
+
+       rte_eth_promiscuous_enable(BOND_PORT);
+
+       struct ether_addr addr;
+
+       rte_eth_macaddr_get(BOND_PORT, &addr);
+       printf("Port %u MAC: ", (unsigned)BOND_PORT);
+               PRINT_MAC(addr);
+               printf("\n");
+}
+
+static inline size_t
+get_vlan_offset(struct ether_hdr *eth_hdr, uint16_t *proto)
+{
+       size_t vlan_offset = 0;
+
+       if (rte_cpu_to_be_16(ETHER_TYPE_VLAN) == *proto) {
+               struct vlan_hdr *vlan_hdr = (struct vlan_hdr *)(eth_hdr + 1);
+
+               vlan_offset = sizeof(struct vlan_hdr);
+               *proto = vlan_hdr->eth_proto;
+
+               if (rte_cpu_to_be_16(ETHER_TYPE_VLAN) == *proto) {
+                       vlan_hdr = vlan_hdr + 1;
+
+                       *proto = vlan_hdr->eth_proto;
+                       vlan_offset += sizeof(struct vlan_hdr);
+               }
+       }
+       return vlan_offset;
+}
+
+struct global_flag_stru_t {
+       int LcoreMainIsRunning;
+       int LcoreMainCore;
+       uint32_t port_packets[4];
+       rte_spinlock_t lock;
+};
+struct global_flag_stru_t global_flag_stru;
+struct global_flag_stru_t *global_flag_stru_p = &global_flag_stru;
+
+/*
+ * Main thread that does the work, reading from INPUT_PORT
+ * and writing to OUTPUT_PORT
+ */
+static int lcore_main(__attribute__((unused)) void *arg1)
+{
+       struct rte_mbuf *pkts[MAX_PKT_BURST] __rte_cache_aligned;
+       struct ether_addr d_addr;
+
+       struct ether_hdr *eth_hdr;
+       struct arp_hdr *arp_hdr;
+       struct ipv4_hdr *ipv4_hdr;
+       uint16_t ether_type, offset;
+
+       uint16_t rx_cnt;
+       uint32_t bond_ip;
+       int i = 0;
+       uint8_t is_free;
+
+       bond_ip = BOND_IP_1 | (BOND_IP_2 << 8) |
+                               (BOND_IP_3 << 16) | (BOND_IP_4 << 24);
+
+       rte_spinlock_trylock(&global_flag_stru_p->lock);
+
+       while (global_flag_stru_p->LcoreMainIsRunning) {
+               rte_spinlock_unlock(&global_flag_stru_p->lock);
+               rx_cnt = rte_eth_rx_burst(BOND_PORT, 0, pkts, MAX_PKT_BURST);
+               is_free = 0;
+
+               /* If didn't receive any packets, wait and go to next iteration */
+               if (rx_cnt == 0) {
+                       rte_delay_us(50);
+                       continue;
+               }
+
+               /* Search incoming data for ARP packets and prepare response */
+               for (i = 0; i < rx_cnt; i++) {
+                       if (rte_spinlock_trylock(&global_flag_stru_p->lock) == 1) {
+                               global_flag_stru_p->port_packets[0]++;
+                               rte_spinlock_unlock(&global_flag_stru_p->lock);
+                       }
+                       eth_hdr = rte_pktmbuf_mtod(pkts[i], struct ether_hdr *);
+                       ether_type = eth_hdr->ether_type;
+                       if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_VLAN))
+                               printf("VLAN taged frame, offset:");
+                       offset = get_vlan_offset(eth_hdr, &ether_type);
+                       if (offset > 0)
+                               printf("%d\n", offset);
+                       if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_ARP)) {
+                               if (rte_spinlock_trylock(&global_flag_stru_p->lock) == 1)     {
+                                       global_flag_stru_p->port_packets[1]++;
+                                       rte_spinlock_unlock(&global_flag_stru_p->lock);
+                               }
+                               arp_hdr = (struct arp_hdr *)((char *)(eth_hdr + 1) + offset);
+                               if (arp_hdr->arp_data.arp_tip == bond_ip) {
+                                       if (arp_hdr->arp_op == rte_cpu_to_be_16(ARP_OP_REQUEST)) {
+                                               arp_hdr->arp_op = rte_cpu_to_be_16(ARP_OP_REPLY);
+                                               /* Switch src and dst data and set bonding MAC */
+                                               ether_addr_copy(&eth_hdr->s_addr, &eth_hdr->d_addr);
+                                               rte_eth_macaddr_get(BOND_PORT, &eth_hdr->s_addr);
+                                               ether_addr_copy(&arp_hdr->arp_data.arp_sha, &arp_hdr->arp_data.arp_tha);
+                                               arp_hdr->arp_data.arp_tip = arp_hdr->arp_data.arp_sip;
+                                               rte_eth_macaddr_get(BOND_PORT, &d_addr);
+                                               ether_addr_copy(&d_addr, &arp_hdr->arp_data.arp_sha);
+                                               arp_hdr->arp_data.arp_sip = bond_ip;
+                                               rte_eth_tx_burst(BOND_PORT, 0, &pkts[i], 1);
+                                               is_free = 1;
+                                       } else {
+                                               rte_eth_tx_burst(BOND_PORT, 0, NULL, 0);
+                                       }
+                               }
+                       } else if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv4)) {
+                               if (rte_spinlock_trylock(&global_flag_stru_p->lock) == 1)     {
+                                       global_flag_stru_p->port_packets[2]++;
+                                       rte_spinlock_unlock(&global_flag_stru_p->lock);
+                                }
+                               ipv4_hdr = (struct ipv4_hdr *)((char *)(eth_hdr + 1) + offset);
+                               if (ipv4_hdr->dst_addr == bond_ip) {
+                                       ether_addr_copy(&eth_hdr->s_addr, &eth_hdr->d_addr);
+                                       rte_eth_macaddr_get(BOND_PORT, &eth_hdr->s_addr);
+                                       ipv4_hdr->dst_addr = ipv4_hdr->src_addr;
+                                       ipv4_hdr->src_addr = bond_ip;
+                                       rte_eth_tx_burst(BOND_PORT, 0, &pkts[i], 1);
+                               }
+
+                       }
+
+                       /* Free processed packets */
+                       if (is_free == 0)
+                               rte_pktmbuf_free(pkts[i]);
+               }
+               rte_spinlock_trylock(&global_flag_stru_p->lock);
+       }
+       rte_spinlock_unlock(&global_flag_stru_p->lock);
+       printf("BYE lcore_main\n");
+       return 0;
+}
+
+struct cmd_obj_send_result {
+       cmdline_fixed_string_t action;
+       cmdline_ipaddr_t ip;
+};
+static inline void get_string(struct cmd_obj_send_result *res, char *buf, uint8_t size)
+{
+       snprintf(buf, size, NIPQUAD_FMT,
+               ((unsigned)((unsigned char *)&(res->ip.addr.ipv4))[0]),
+               ((unsigned)((unsigned char *)&(res->ip.addr.ipv4))[1]),
+               ((unsigned)((unsigned char *)&(res->ip.addr.ipv4))[2]),
+               ((unsigned)((unsigned char *)&(res->ip.addr.ipv4))[3])
+               );
+}
+static void cmd_obj_send_parsed(void *parsed_result,
+               __attribute__((unused)) struct cmdline *cl,
+                              __attribute__((unused)) void *data)
+{
+
+       struct cmd_obj_send_result *res = parsed_result;
+       char ip_str[INET6_ADDRSTRLEN];
+
+       struct rte_mbuf *created_pkt;
+       struct ether_hdr *eth_hdr;
+       struct arp_hdr *arp_hdr;
+
+       uint32_t bond_ip;
+       size_t pkt_size;
+
+       if (res->ip.family == AF_INET)
+               get_string(res, ip_str, INET_ADDRSTRLEN);
+       else
+               cmdline_printf(cl, "Wrong IP format. Only IPv4 is supported\n");
+
+       bond_ip = BOND_IP_1 | (BOND_IP_2 << 8) |
+                               (BOND_IP_3 << 16) | (BOND_IP_4 << 24);
+
+       created_pkt = rte_pktmbuf_alloc(mbuf_pool);
+       pkt_size = sizeof(struct ether_hdr) + sizeof(struct arp_hdr);
+       created_pkt->data_len = pkt_size;
+       created_pkt->pkt_len = pkt_size;
+
+       eth_hdr = rte_pktmbuf_mtod(created_pkt, struct ether_hdr *);
+       rte_eth_macaddr_get(BOND_PORT, &eth_hdr->s_addr);
+       memset(&eth_hdr->d_addr, 0xFF, ETHER_ADDR_LEN);
+       eth_hdr->ether_type = rte_cpu_to_be_16(ETHER_TYPE_ARP);
+
+       arp_hdr = (struct arp_hdr *)((char *)eth_hdr + sizeof(struct ether_hdr));
+       arp_hdr->arp_hrd = rte_cpu_to_be_16(ARP_HRD_ETHER);
+       arp_hdr->arp_pro = rte_cpu_to_be_16(ETHER_TYPE_IPv4);
+       arp_hdr->arp_hln = ETHER_ADDR_LEN;
+       arp_hdr->arp_pln = sizeof(uint32_t);
+       arp_hdr->arp_op = rte_cpu_to_be_16(ARP_OP_REQUEST);
+
+       rte_eth_macaddr_get(BOND_PORT, &arp_hdr->arp_data.arp_sha);
+       arp_hdr->arp_data.arp_sip = bond_ip;
+       memset(&arp_hdr->arp_data.arp_tha, 0, ETHER_ADDR_LEN);
+       arp_hdr->arp_data.arp_tip =
+                         ((unsigned char *)&res->ip.addr.ipv4)[0]        |
+                        (((unsigned char *)&res->ip.addr.ipv4)[1] << 8)  |
+                        (((unsigned char *)&res->ip.addr.ipv4)[2] << 16) |
+                        (((unsigned char *)&res->ip.addr.ipv4)[3] << 24);
+       rte_eth_tx_burst(BOND_PORT, 0, &created_pkt, 1);
+
+       rte_delay_ms(100);
+       cmdline_printf(cl, "\n");
+}
+
+cmdline_parse_token_string_t cmd_obj_action_send =
+       TOKEN_STRING_INITIALIZER(struct cmd_obj_send_result, action, "send");
+cmdline_parse_token_ipaddr_t cmd_obj_ip =
+       TOKEN_IPV4_INITIALIZER(struct cmd_obj_send_result, ip);
+
+cmdline_parse_inst_t cmd_obj_send = {
+       .f = cmd_obj_send_parsed,  /* function to call */
+       .data = NULL,      /* 2nd arg of func */
+       .help_str = "send client_ip",
+       .tokens = {        /* token list, NULL terminated */
+               (void *)&cmd_obj_action_send,
+               (void *)&cmd_obj_ip,
+               NULL,
+       },
+};
+
+struct cmd_start_result {
+       cmdline_fixed_string_t start;
+};
+
+static void cmd_start_parsed(__attribute__((unused)) void *parsed_result,
+                              struct cmdline *cl,
+                              __attribute__((unused)) void *data)
+{
+       int slave_core_id = rte_lcore_id();
+
+       rte_spinlock_trylock(&global_flag_stru_p->lock);
+       if (global_flag_stru_p->LcoreMainIsRunning == 0)        {
+               if (lcore_config[global_flag_stru_p->LcoreMainCore].state != WAIT)      {
+                       rte_spinlock_unlock(&global_flag_stru_p->lock);
+                       return;
+               }
+               rte_spinlock_unlock(&global_flag_stru_p->lock);
+       } else {
+               cmdline_printf(cl, "lcore_main already running on core:%d\n",
+                               global_flag_stru_p->LcoreMainCore);
+               rte_spinlock_unlock(&global_flag_stru_p->lock);
+               return;
+       }
+
+       /* start lcore main on core != master_core - ARP response thread */
+       slave_core_id = rte_get_next_lcore(rte_lcore_id(), 1, 0);
+       if ((slave_core_id >= RTE_MAX_LCORE) || (slave_core_id == 0))
+               return;
+
+       rte_spinlock_trylock(&global_flag_stru_p->lock);
+       global_flag_stru_p->LcoreMainIsRunning = 1;
+       rte_spinlock_unlock(&global_flag_stru_p->lock);
+       cmdline_printf(cl,
+                       "Starting lcore_main on core %d:%d "
+                       "Our IP:%d.%d.%d.%d\n",
+                       slave_core_id,
+                       rte_eal_remote_launch(lcore_main, NULL, slave_core_id),
+                       BOND_IP_1,
+                       BOND_IP_2,
+                       BOND_IP_3,
+                       BOND_IP_4
+               );
+}
+
+cmdline_parse_token_string_t cmd_start_start =
+       TOKEN_STRING_INITIALIZER(struct cmd_start_result, start, "start");
+
+cmdline_parse_inst_t cmd_start = {
+       .f = cmd_start_parsed,  /* function to call */
+       .data = NULL,      /* 2nd arg of func */
+       .help_str = "starts listening if not started at startup",
+       .tokens = {        /* token list, NULL terminated */
+               (void *)&cmd_start_start,
+               NULL,
+       },
+};
+
+struct cmd_help_result {
+       cmdline_fixed_string_t help;
+};
+
+static void cmd_help_parsed(__attribute__((unused)) void *parsed_result,
+                           struct cmdline *cl,
+                           __attribute__((unused)) void *data)
+{
+       cmdline_printf(cl,
+                       "ALB - link bonding mode 6 example\n"
+                       "send IP        - sends one ARPrequest thru bonding for IP.\n"
+                       "start          - starts listening ARPs.\n"
+                       "stop           - stops lcore_main.\n"
+                       "show           - shows some bond info: ex. active slaves etc.\n"
+                       "help           - prints help.\n"
+                       "quit           - terminate all threads and quit.\n"
+                      );
+}
+
+cmdline_parse_token_string_t cmd_help_help =
+       TOKEN_STRING_INITIALIZER(struct cmd_help_result, help, "help");
+
+cmdline_parse_inst_t cmd_help = {
+       .f = cmd_help_parsed,  /* function to call */
+       .data = NULL,      /* 2nd arg of func */
+       .help_str = "show help",
+       .tokens = {        /* token list, NULL terminated */
+               (void *)&cmd_help_help,
+               NULL,
+       },
+};
+
+struct cmd_stop_result {
+       cmdline_fixed_string_t stop;
+};
+
+static void cmd_stop_parsed(__attribute__((unused)) void *parsed_result,
+                           struct cmdline *cl,
+                           __attribute__((unused)) void *data)
+{
+       rte_spinlock_trylock(&global_flag_stru_p->lock);
+       if (global_flag_stru_p->LcoreMainIsRunning == 0)        {
+               cmdline_printf(cl,
+                                       "lcore_main not running on core:%d\n",
+                                       global_flag_stru_p->LcoreMainCore);
+               rte_spinlock_unlock(&global_flag_stru_p->lock);
+               return;
+       }
+       global_flag_stru_p->LcoreMainIsRunning = 0;
+       rte_eal_wait_lcore(global_flag_stru_p->LcoreMainCore);
+       cmdline_printf(cl,
+                       "lcore_main stopped on core:%d\n",
+                       global_flag_stru_p->LcoreMainCore);
+       rte_spinlock_unlock(&global_flag_stru_p->lock);
+}
+
+cmdline_parse_token_string_t cmd_stop_stop =
+       TOKEN_STRING_INITIALIZER(struct cmd_stop_result, stop, "stop");
+
+cmdline_parse_inst_t cmd_stop = {
+       .f = cmd_stop_parsed,  /* function to call */
+       .data = NULL,      /* 2nd arg of func */
+       .help_str = "this command do not handle any arguments",
+       .tokens = {        /* token list, NULL terminated */
+               (void *)&cmd_stop_stop,
+               NULL,
+       },
+};
+
+struct cmd_quit_result {
+       cmdline_fixed_string_t quit;
+};
+
+static void cmd_quit_parsed(__attribute__((unused)) void *parsed_result,
+                           struct cmdline *cl,
+                           __attribute__((unused)) void *data)
+{
+       rte_spinlock_trylock(&global_flag_stru_p->lock);
+       if (global_flag_stru_p->LcoreMainIsRunning == 0)        {
+               cmdline_printf(cl,
+                                       "lcore_main not running on core:%d\n",
+                                       global_flag_stru_p->LcoreMainCore);
+               rte_spinlock_unlock(&global_flag_stru_p->lock);
+               cmdline_quit(cl);
+               return;
+       }
+       global_flag_stru_p->LcoreMainIsRunning = 0;
+       rte_eal_wait_lcore(global_flag_stru_p->LcoreMainCore);
+       cmdline_printf(cl,
+                       "lcore_main stopped on core:%d\n",
+                       global_flag_stru_p->LcoreMainCore);
+       rte_spinlock_unlock(&global_flag_stru_p->lock);
+       cmdline_quit(cl);
+}
+
+cmdline_parse_token_string_t cmd_quit_quit =
+       TOKEN_STRING_INITIALIZER(struct cmd_quit_result, quit, "quit");
+
+cmdline_parse_inst_t cmd_quit = {
+       .f = cmd_quit_parsed,  /* function to call */
+       .data = NULL,      /* 2nd arg of func */
+       .help_str = "this command do not handle any arguments",
+       .tokens = {        /* token list, NULL terminated */
+               (void *)&cmd_quit_quit,
+               NULL,
+       },
+};
+
+extern void print_client_stats(void);
+struct cmd_show_result {
+       cmdline_fixed_string_t show;
+};
+
+static void cmd_show_parsed(__attribute__((unused)) void *parsed_result,
+                           struct cmdline *cl,
+                           __attribute__((unused)) void *data)
+{
+       uint8_t slaves[16] = {0};
+       uint8_t len = 16;
+       struct ether_addr addr;
+       uint8_t i = 0;
+
+       while (i < slaves_count)        {
+               rte_eth_macaddr_get(i, &addr);
+               PRINT_MAC(addr);
+               printf("\n");
+               i++;
+       }
+
+       rte_spinlock_trylock(&global_flag_stru_p->lock);
+       cmdline_printf(cl,
+                       "Active_slaves:%d "
+                       "packets received:Tot:%d Arp:%d IPv4:%d\n",
+                       rte_eth_bond_active_slaves_get(BOND_PORT, slaves, len),
+                       global_flag_stru_p->port_packets[0],
+                       global_flag_stru_p->port_packets[1],
+                       global_flag_stru_p->port_packets[2]);
+       rte_spinlock_unlock(&global_flag_stru_p->lock);
+
+#if defined(RTE_LIBRTE_BOND_DEBUG_ALB_L1) || defined(RTE_LIBRTE_BOND_DEBUG_ALB)
+       print_client_stats();
+#endif
+}
+
+cmdline_parse_token_string_t cmd_show_show =
+       TOKEN_STRING_INITIALIZER(struct cmd_show_result, show, "show");
+
+cmdline_parse_inst_t cmd_show = {
+       .f = cmd_show_parsed,  /* function to call */
+       .data = NULL,      /* 2nd arg of func */
+       .help_str = "this command do not handle any arguments",
+       .tokens = {        /* token list, NULL terminated */
+               (void *)&cmd_show_show,
+               NULL,
+       },
+};
+
+/****** CONTEXT (list of instruction) */
+
+cmdline_parse_ctx_t main_ctx[] = {
+       (cmdline_parse_inst_t *)&cmd_start,
+       (cmdline_parse_inst_t *)&cmd_obj_send,
+       (cmdline_parse_inst_t *)&cmd_stop,
+       (cmdline_parse_inst_t *)&cmd_show,
+       (cmdline_parse_inst_t *)&cmd_quit,
+       (cmdline_parse_inst_t *)&cmd_help,
+       NULL,
+};
+
+/* prompt function, called from main on MASTER lcore */
+static void *prompt(__attribute__((unused)) void *arg1)
+{
+       struct cmdline *cl;
+
+       cl = cmdline_stdin_new(main_ctx, "bond6>");
+       if (cl != NULL) {
+               cmdline_interact(cl);
+               cmdline_stdin_exit(cl);
+       }
+}
+
+/* Main function, does initialisation and calls the per-lcore functions */
+int
+main(int argc, char *argv[])
+{
+       int ret;
+       uint8_t nb_ports, i;
+
+       /* init EAL */
+       ret = rte_eal_init(argc, argv);
+       rte_eal_devargs_dump(stdout);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
+       argc -= ret;
+       argv += ret;
+
+       nb_ports = rte_eth_dev_count();
+       if (nb_ports == 0)
+               rte_exit(EXIT_FAILURE, "Give at least one port\n");
+       else if (nb_ports > MAX_PORTS)
+               rte_exit(EXIT_FAILURE, "You can have max 4 ports\n");
+
+       mbuf_pool = rte_mempool_create("MBUF_POOL", NB_MBUF,
+                                      MBUF_SIZE, 32,
+                                      sizeof(struct rte_pktmbuf_pool_private),
+                                      rte_pktmbuf_pool_init, NULL,
+                                      rte_pktmbuf_init, NULL,
+                                      rte_socket_id(), 0);
+       if (mbuf_pool == NULL)
+               rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
+
+       /* initialize all ports */
+       slaves_count = nb_ports;
+       for (i = 0; i < nb_ports; i++) {
+               slave_port_init(i, mbuf_pool);
+               slaves[i] = i;
+       }
+
+       bond_port_init(mbuf_pool);
+
+       rte_spinlock_init(&global_flag_stru_p->lock);
+       int slave_core_id = rte_lcore_id();
+
+       /* check state of lcores */
+       RTE_LCORE_FOREACH_SLAVE(slave_core_id) {
+       if (lcore_config[slave_core_id].state != WAIT)
+               return -EBUSY;
+       }
+       /* start lcore main on core != master_core - ARP response thread */
+       slave_core_id = rte_get_next_lcore(rte_lcore_id(), 1, 0);
+       if ((slave_core_id >= RTE_MAX_LCORE) || (slave_core_id == 0))
+               return -EPERM;
+
+       global_flag_stru_p->LcoreMainIsRunning = 1;
+       global_flag_stru_p->LcoreMainCore = slave_core_id;
+       printf("Starting lcore_main on core %d:%d Our IP:%d.%d.%d.%d\n",
+                       slave_core_id,
+                       rte_eal_remote_launch((lcore_function_t *)lcore_main,
+                                       NULL,
+                                       slave_core_id),
+                       BOND_IP_1,
+                       BOND_IP_2,
+                       BOND_IP_3,
+                       BOND_IP_4
+               );
+
+       /* Start prompt for user interact */
+       prompt(NULL);
+
+       rte_delay_ms(100);
+       return 0;
+}
diff --git a/examples/bond/main.h b/examples/bond/main.h
new file mode 100644 (file)
index 0000000..ea331e5
--- /dev/null
@@ -0,0 +1,39 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _MAIN_H_
+#define _MAIN_H_
+
+int main(int argc, char *argv[]);
+
+#endif /* ifndef _MAIN_H_ */