examples/l2fwd_fork: new app
authorIntel <intel.com>
Wed, 18 Sep 2013 10:00:00 +0000 (12:00 +0200)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Wed, 9 Oct 2013 14:16:16 +0000 (16:16 +0200)
Signed-off-by: Intel
examples/multi_process/l2fwd_fork/Makefile [new file with mode: 0644]
examples/multi_process/l2fwd_fork/flib.c [new file with mode: 0644]
examples/multi_process/l2fwd_fork/flib.h [new file with mode: 0644]
examples/multi_process/l2fwd_fork/main.c [new file with mode: 0644]
examples/multi_process/l2fwd_fork/main.h [new file with mode: 0644]

diff --git a/examples/multi_process/l2fwd_fork/Makefile b/examples/multi_process/l2fwd_fork/Makefile
new file mode 100644 (file)
index 0000000..c7168f4
--- /dev/null
@@ -0,0 +1,50 @@
+#   BSD LICENSE
+# 
+#   Copyright(c) 2010-2013 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 overriden by command line or environment
+RTE_TARGET ?= x86_64-default-linuxapp-gcc
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# binary name
+APP = l2fwd_fork
+
+# all source are stored in SRCS-y
+SRCS-y := main.c flib.c
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+include $(RTE_SDK)/mk/rte.extapp.mk
diff --git a/examples/multi_process/l2fwd_fork/flib.c b/examples/multi_process/l2fwd_fork/flib.c
new file mode 100644 (file)
index 0000000..563035c
--- /dev/null
@@ -0,0 +1,314 @@
+/*-
+ *   BSD LICENSE
+ * 
+ *   Copyright(c) 2010-2013 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 <unistd.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
+#include <inttypes.h>
+#include <sys/types.h>
+#include <sys/queue.h>
+#include <sys/wait.h>
+#include <sys/prctl.h>
+#include <netinet/in.h>
+#include <setjmp.h>
+#include <stdarg.h>
+#include <ctype.h>
+#include <errno.h>
+#include <getopt.h>
+#include <dirent.h>
+#include <signal.h>
+
+#include <rte_common.h>
+#include <rte_log.h>
+#include <rte_malloc.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_mempool.h>
+#include <rte_mbuf.h>
+#include <rte_string_fns.h>
+
+#include "flib.h"
+
+#define SIG_PARENT_EXIT SIGUSR1
+
+struct lcore_stat {
+       pid_t pid;           /**< pthread identifier */
+       lcore_function_t *f; /**< function to call */
+       void *arg;           /**< argument of function */
+       slave_exit_notify *cb_fn;
+} __rte_cache_aligned;
+
+
+static struct lcore_stat *core_cfg;
+static uint16_t *lcore_cfg = NULL;
+
+/* signal handler to be notified after parent leaves */
+static void
+sighand_parent_exit(int sig)
+{
+       printf("lcore = %u : Find parent leaves, sig=%d\n", rte_lcore_id(),
+                       sig);
+       printf("Child leaving\n");
+       exit(0);
+
+       return;
+}
+
+/**
+ * Real function entrance ran in slave process
+ **/
+static int
+slave_proc_func(void)
+{
+       struct rte_config *config;
+       unsigned slave_id = rte_lcore_id();
+       struct lcore_stat *cfg = &core_cfg[slave_id];
+
+       if (prctl(PR_SET_PDEATHSIG, SIG_PARENT_EXIT, 0, 0, 0, 0) != 0)
+               printf("Warning: Slave can't register for being notified in"
+               "case master process exited\n");
+       else {
+               struct sigaction act;
+               memset(&act, 0 , sizeof(act));
+               act.sa_handler = sighand_parent_exit;
+               if (sigaction(SIG_PARENT_EXIT, &act, NULL) != 0)
+                       printf("Fail to register signal handler:%d\n", SIG_PARENT_EXIT);
+       }
+
+       /* Set slave process to SECONDARY to avoid operation like dev_start/stop etc */
+       config = rte_eal_get_configuration();
+       if (NULL == config)
+               printf("Warning:Can't get rte_config\n");
+       else
+               config->process_type = RTE_PROC_SECONDARY;
+
+       printf("Core %u is ready (pid=%d)\n", slave_id, (int)cfg->pid);
+
+       exit(cfg->f(cfg->arg));
+}
+
+/** 
+ * function entrance ran in master thread, which will spawn slave process and wait until
+ * specific slave exited. 
+ **/
+static int
+lcore_func(void *arg __attribute__((unused)))
+{
+       unsigned slave_id = rte_lcore_id();
+       struct lcore_stat *cfg = &core_cfg[slave_id];
+       int pid, stat;
+
+       if (rte_get_master_lcore() == slave_id)
+               return cfg->f(cfg->arg);
+
+       /* fork a slave process */
+       pid = fork();
+
+       if (pid == -1) {
+               printf("Failed to fork\n");
+               return -1;
+       } else if (pid == 0) /* child */
+               return slave_proc_func();
+       else { /* parent */
+               cfg->pid = pid;
+
+               waitpid(pid, &stat, 0);
+
+               cfg->pid = 0;
+               cfg->f = NULL;
+               cfg->arg = NULL;
+               /* Notify slave's exit if applicable */
+               if(cfg->cb_fn)
+                       cfg->cb_fn(slave_id, stat);
+               return stat;
+       }
+}
+
+static int 
+lcore_id_init(void)
+{
+       int i;
+       /* Setup lcore ID allocation map */
+       lcore_cfg = rte_zmalloc("LCORE_ID_MAP",
+                                               sizeof(uint16_t) * RTE_MAX_LCORE,
+                                               CACHE_LINE_SIZE);
+
+       if(lcore_cfg == NULL)
+               rte_panic("Failed to malloc\n");
+
+       for (i = 0; i < RTE_MAX_LCORE; i++) {
+               if (rte_lcore_is_enabled(i))
+                       lcore_cfg[i] = 1;
+       }
+       return 0;
+}
+
+int
+flib_assign_lcore_id(void)
+{
+       unsigned i;
+       int ret;
+
+       /**
+        * thread assigned a lcore id previously, or a  slave thread. But still have 
+        * a bug here: If the core mask includes core 0, and that core call this
+        * function, it still can get a new lcore id. 
+        **/
+       if (rte_lcore_id() != 0)
+               return -1;
+
+       do {
+               /* Find a lcore id not used yet, avoid to use lcore ID 0 */
+               for (i = 1; i < RTE_MAX_LCORE; i++) {
+                       if (lcore_cfg[i] == 0)
+                               break;
+               }
+               if (i == RTE_MAX_LCORE)
+                       return -1;
+
+               /* Assign new lcore id to this thread */
+
+               ret = rte_atomic16_cmpset(&lcore_cfg[i], 0, 1);
+       } while (unlikely(ret == 0));
+
+       RTE_PER_LCORE(_lcore_id) = i;
+       return i;
+}
+
+void
+flib_free_lcore_id(unsigned lcore_id)
+{
+       /* id is not valid or belongs to pinned core id */
+       if (lcore_id >= RTE_MAX_LCORE || lcore_id == 0 || 
+               rte_lcore_is_enabled(lcore_id))
+               return;
+
+       lcore_cfg[lcore_id] = 0;
+}
+
+int
+flib_register_slave_exit_notify(unsigned slave_id, 
+       slave_exit_notify *cb)
+{
+       if (cb == NULL)
+               return -EFAULT;
+
+       if (!rte_lcore_is_enabled(slave_id))
+               return -ENOENT;
+
+       core_cfg[slave_id].cb_fn = cb;
+
+       return 0;
+}
+
+enum slave_stat 
+flib_query_slave_status(unsigned slave_id)
+{
+       if (!rte_lcore_is_enabled(slave_id))
+               return ST_FREEZE;
+       /* pid only be set when slave process spawned */
+       if (core_cfg[slave_id].pid != 0)
+               return ST_RUN;
+       else
+               return ST_IDLE;
+}
+
+int
+flib_remote_launch(lcore_function_t *f,
+                                       void *arg, unsigned slave_id)
+{
+       if (f == NULL)
+               return -1;
+
+       if (!rte_lcore_is_enabled(slave_id))
+               return -1;
+
+       /* Wait until specific lcore state change to WAIT */
+       rte_eal_wait_lcore(slave_id);
+
+       core_cfg[slave_id].f = f;
+       core_cfg[slave_id].arg = arg;
+
+       return rte_eal_remote_launch(lcore_func, NULL, slave_id);
+}
+
+int
+flib_mp_remote_launch(lcore_function_t *f, void *arg,
+                       enum rte_rmt_call_master_t call_master)
+{
+       int i;
+
+       RTE_LCORE_FOREACH_SLAVE(i) {
+               core_cfg[i].arg = arg;
+               core_cfg[i].f = f;
+       }
+
+       return rte_eal_mp_remote_launch(lcore_func, NULL, call_master);
+}
+
+int 
+flib_init(void)
+{
+       if ((core_cfg = rte_zmalloc("core_cfg",
+               sizeof(struct lcore_stat) * RTE_MAX_LCORE,
+               CACHE_LINE_SIZE)) == NULL ) {
+               printf("rte_zmalloc failed\n");
+               return -1;
+       }
+
+       if (lcore_id_init() != 0) {
+               printf("lcore_id_init failed\n");
+               return -1;
+       }
+
+       return 0;
+}
diff --git a/examples/multi_process/l2fwd_fork/flib.h b/examples/multi_process/l2fwd_fork/flib.h
new file mode 100644 (file)
index 0000000..0c34d59
--- /dev/null
@@ -0,0 +1,149 @@
+/*-
+ *   BSD LICENSE
+ * 
+ *   Copyright(c) 2010-2013 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 __FLIB_H
+#define __FLIB_H
+
+/* callback function pointer when specific slave leaves */
+typedef void (slave_exit_notify)(unsigned slaveid, int stat);
+
+enum slave_stat{
+       ST_FREEZE = 1,
+       ST_IDLE,
+       ST_RUN,
+       ST_ZOMBIE,      /* Not implemented yet */
+};
+
+/**
+ * Initialize the fork lib.
+ *
+ * @return
+ *    - 0 : fork lib initialized successfully
+ *    - -1 : fork lib initialized failed
+ */
+int flib_init(void);
+
+/**
+ * Check that every SLAVE lcores are in WAIT state, then call
+ * flib_remote_launch() for all of them. If call_master is true
+ * (set to CALL_MASTER), also call the function on the master lcore.
+ * 
+ * @param f:
+ *     function pointer need to run
+ * @param arg:
+ *     argument for f to carry
+ * @param call_master
+ *     - SKIP_MASTER : only launch function on slave lcores
+ *     - CALL_MASTER : launch function on master and slave lcores
+ * @return
+ *    - 0 : function  execute successfully
+ *    - -1 :  function  execute  failed
+ */
+int flib_mp_remote_launch(lcore_function_t *f, 
+               void *arg, enum rte_rmt_call_master_t call_master);
+
+/**
+ * Send a message to a slave lcore identified by slave_id to call a
+ * function f with argument arg.
+ * 
+ * @param f:
+ *     function pointer need to run
+ * @param arg:
+ *     argument for f to carry
+ * @param slave_id
+ *     slave lcore id to run on
+ * @return
+ *    - 0 : function  execute successfully
+ *    - -1 :  function  execute  failed
+ */
+int flib_remote_launch(lcore_function_t *f, 
+                                       void *arg, unsigned slave_id);
+
+/**
+ * Query the running stat for specific slave, wont' work in with master id 
+ * 
+ * @param slave_id:
+ *     lcore id which should not be master id
+ * @return
+ *    - ST_FREEZE : lcore is not in enabled core mask
+ *      - ST_IDLE     : lcore is idle
+ *    -  ST_RUN     : lcore is running something
+ */
+enum slave_stat 
+flib_query_slave_status(unsigned slave_id);
+
+/**
+ * Register a callback function to be notified in case specific slave exit.  
+ * 
+ * @param slave_id:
+ *     lcore id which should not be master id
+ * @param cb:
+ *     callback pointer to register
+ * @return
+ *    - 0            :  function  execute successfully
+ *    - -EFAULT  :  argument error
+ *    - -ENOENT :  slave_id not correct
+ */
+int flib_register_slave_exit_notify(unsigned slave_id, 
+       slave_exit_notify *cb);
+
+/**
+ * Assign a lcore ID to non-slave thread.  Non-slave thread refers to thread that
+ * not created by function rte_eal_remote_launch or rte_eal_mp_remote_launch.
+ * These threads can either bind lcore or float among differnt lcores.
+ * This lcore ID will be unique in multi-thread or multi-process DPDK running
+ * environment, then it can benefit from using the cache mechanism provided in
+ * mempool library.
+ * After calling successfully, use rte_lcore_id() to get the assigned lcore ID, but
+ * other lcore funtions can't guarantee to work correctly.
+ *
+ * @return
+ *   -    -1  : can't assign a lcore id with 3 possibilities.
+ *                 - it's not non-slave thread.
+ *                 - it had assign a lcore id previously
+ *                 - the lcore id is running out.
+ *   -  > 0 :  the assigned lcore id.
+ */
+int flib_assign_lcore_id(void);
+
+/**
+ * Free the lcore_id that assigned in flib_assign_lcore_id().
+ * call it in case non-slave thread is leaving or left.
+ *
+ * @param lcore_id
+ * The identifier of the lcore, which MUST be between 1 and
+ *   RTE_MAX_LCORE-1.
+ */
+void flib_free_lcore_id(unsigned lcore_id);
+
+#endif /* __FLIB_H  */
diff --git a/examples/multi_process/l2fwd_fork/main.c b/examples/multi_process/l2fwd_fork/main.c
new file mode 100644 (file)
index 0000000..1437914
--- /dev/null
@@ -0,0 +1,1351 @@
+/*-
+ *   BSD LICENSE
+ * 
+ *   Copyright(c) 2010-2013 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.
+ */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
+#include <stdint.h>
+#include <sched.h>
+#include <inttypes.h>
+#include <sys/types.h>
+#include <sys/queue.h>
+#include <netinet/in.h>
+#include <setjmp.h>
+#include <stdarg.h>
+#include <ctype.h>
+#include <errno.h>
+#include <getopt.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_spinlock.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_mempool.h>
+#include <rte_mbuf.h>
+#include <rte_malloc.h>
+
+#include "main.h"
+#include "flib.h"
+
+#define RTE_LOGTYPE_L2FWD RTE_LOGTYPE_USER1
+#define MBUF_NAME      "mbuf_pool_%d"
+#define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
+#define NB_MBUF   8192
+#define RING_MASTER_NAME       "l2fwd_ring_m2s_"
+#define RING_SLAVE_NAME                "l2fwd_ring_s2m_"
+#define MAX_NAME_LEN   32
+/* RECREATE flag indicate needs initialize resource and launch slave_core again */
+#define SLAVE_RECREATE_FLAG 0x1
+/* RESTART flag indicate needs restart port and send START command again */
+#define SLAVE_RESTART_FLAG 0x2
+#define INVALID_MAPPING_ID     ((unsigned)LCORE_ID_ANY)
+/* Maximum message buffer per slave */
+#define NB_CORE_MSGBUF 32
+enum l2fwd_cmd{
+       CMD_START,
+       CMD_STOP,
+};
+
+/*
+ * 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. */
+
+/*
+ * 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. */
+
+#define MAX_PKT_BURST 32
+#define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
+
+/*
+ * Configurable number of RX/TX ring descriptors
+ */
+#define RTE_TEST_RX_DESC_DEFAULT 128
+#define RTE_TEST_TX_DESC_DEFAULT 512
+static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
+static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
+
+/* ethernet addresses of ports */
+static struct ether_addr l2fwd_ports_eth_addr[RTE_MAX_ETHPORTS];
+
+/* mask of enabled ports */
+static uint32_t l2fwd_enabled_port_mask = 0;
+
+/* list of enabled ports */
+static uint32_t l2fwd_dst_ports[RTE_MAX_ETHPORTS];
+
+static unsigned int l2fwd_rx_queue_per_lcore = 1;
+
+struct mbuf_table {
+       unsigned len;
+       struct rte_mbuf *m_table[MAX_PKT_BURST];
+};
+
+#define MAX_RX_QUEUE_PER_LCORE 16
+#define MAX_TX_QUEUE_PER_PORT 16
+struct lcore_queue_conf {
+       unsigned n_rx_port;
+       unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
+       struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
+
+} __rte_cache_aligned;
+struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
+
+struct lcore_resource_struct {
+       int enabled;    /* Only set in case this lcore involved into packet forwarding */
+       int flags;          /* Set only slave need to restart or recreate */
+       unsigned lcore_id;  /*  lcore ID */
+       unsigned pair_id;       /* dependency lcore ID on port */
+       char ring_name[2][MAX_NAME_LEN];        
+       /* ring[0] for master send cmd, slave read */
+       /* ring[1] for slave send ack, master read */
+       struct rte_ring *ring[2]; 
+       int port_num;                                   /* Total port numbers */
+       uint8_t port[RTE_MAX_ETHPORTS]; /* Port id for that lcore to receive packets */         
+}__attribute__((packed)) __rte_cache_aligned;
+
+static struct lcore_resource_struct lcore_resource[RTE_MAX_LCORE]; 
+static struct rte_mempool *message_pool;
+static rte_spinlock_t res_lock = RTE_SPINLOCK_INITIALIZER;
+/* use floating processes */
+static int float_proc = 0;
+/* Save original cpu affinity */
+struct cpu_aff_arg{
+       cpu_set_t set;
+       size_t size;
+}cpu_aff;
+
+static const struct rte_eth_conf port_conf = {
+       .rxmode = {
+               .split_hdr_size = 0,
+               .header_split   = 0, /**< Header Split disabled */
+               .hw_ip_checksum = 0, /**< IP checksum offload disabled */
+               .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,
+       },
+};
+
+static const struct rte_eth_rxconf rx_conf = {
+       .rx_thresh = {
+               .pthresh = RX_PTHRESH,
+               .hthresh = RX_HTHRESH,
+               .wthresh = RX_WTHRESH,
+       },
+};
+
+static const struct rte_eth_txconf tx_conf = {
+       .tx_thresh = {
+               .pthresh = TX_PTHRESH,
+               .hthresh = TX_HTHRESH,
+               .wthresh = TX_WTHRESH,
+       },
+       .tx_free_thresh = 0, /* Use PMD default values */
+       .tx_rs_thresh = 0, /* Use PMD default values */
+       .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS | ETH_TXQ_FLAGS_NOOFFLOADS,
+};
+
+static struct rte_mempool * l2fwd_pktmbuf_pool[RTE_MAX_ETHPORTS];
+
+/* Per-port statistics struct */
+struct l2fwd_port_statistics {
+       uint64_t tx;
+       uint64_t rx;
+       uint64_t dropped;
+} __rte_cache_aligned;
+struct l2fwd_port_statistics *port_statistics;
+/**
+ * pointer to lcore ID mapping array, used to return lcore id in case slave
+ * process exited unexpectedly, use only floating process option applied
+ **/
+unsigned *mapping_id;
+
+/* A tsc-based timer responsible for triggering statistics printout */
+#define TIMER_MILLISECOND 2000000ULL /* around 1ms at 2 Ghz */
+#define MAX_TIMER_PERIOD 86400 /* 1 day max */
+static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* default period is 10 seconds */
+
+static int l2fwd_launch_one_lcore(void *dummy);
+
+/* Print out statistics on packets dropped */
+static void
+print_stats(void)
+{
+       uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
+       unsigned portid;
+
+       total_packets_dropped = 0;
+       total_packets_tx = 0;
+       total_packets_rx = 0;
+
+       const char clr[] = { 27, '[', '2', 'J', '\0' };
+       const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
+
+               /* Clear screen and move to top left */
+       printf("%s%s", clr, topLeft);
+
+       printf("\nPort statistics ====================================");
+
+       for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
+               /* skip disabled ports */
+               if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
+                       continue;
+               printf("\nStatistics for port %u ------------------------------"
+                          "\nPackets sent: %24"PRIu64
+                          "\nPackets received: %20"PRIu64
+                          "\nPackets dropped: %21"PRIu64,
+                          portid,
+                          port_statistics[portid].tx,
+                          port_statistics[portid].rx,
+                          port_statistics[portid].dropped);
+
+               total_packets_dropped += port_statistics[portid].dropped;
+               total_packets_tx += port_statistics[portid].tx;
+               total_packets_rx += port_statistics[portid].rx;
+       }
+       printf("\nAggregate statistics ==============================="
+                  "\nTotal packets sent: %18"PRIu64
+                  "\nTotal packets received: %14"PRIu64
+                  "\nTotal packets dropped: %15"PRIu64,
+                  total_packets_tx,
+                  total_packets_rx,
+                  total_packets_dropped);
+       printf("\n====================================================\n");
+}
+
+static int
+clear_cpu_affinity(void)
+{
+       int s;
+
+       s = sched_setaffinity(0, cpu_aff.size, &cpu_aff.set);
+       if (s != 0) {
+               printf("sched_setaffinity failed:%s\n", strerror(errno));
+               return -1;
+       }
+
+       return 0;
+}
+
+static int
+get_cpu_affinity(void)
+{
+       int s;
+
+       cpu_aff.size = sizeof(cpu_set_t);
+       CPU_ZERO(&cpu_aff.set);
+
+       s = sched_getaffinity(0, cpu_aff.size, &cpu_aff.set);
+       if (s != 0) {
+               printf("sched_getaffinity failed:%s\n", strerror(errno));
+               return -1;
+       }
+
+       return 0;
+}
+
+/**
+ * This fnciton demonstrates the approach to create ring in first instance 
+ * or re-attach an existed ring in later instance. 
+ **/
+static struct rte_ring *
+create_ring(const char *name, unsigned count,
+                                       int socket_id,unsigned flags)
+{
+       struct rte_ring *ring;
+
+       if (name == NULL)
+               return NULL;
+       
+       /* If already create, just attached it */
+       if (likely((ring = rte_ring_lookup(name)) != NULL))
+               return ring;
+
+       /* First call it, create one */
+       return rte_ring_create(name, count, socket_id, flags);
+}
+
+/* Malloc with rte_malloc on structures that shared by master and slave */
+static int 
+l2fwd_malloc_shared_struct(void)
+{
+       port_statistics = rte_zmalloc("port_stat",
+                                               sizeof(struct l2fwd_port_statistics) * RTE_MAX_ETHPORTS,
+                                               0);
+       if (port_statistics == NULL)
+               return -1;
+
+       /* allocate  mapping_id array */
+       if (float_proc) {
+               int i;
+               mapping_id = rte_malloc("mapping_id", sizeof(unsigned) * RTE_MAX_LCORE,
+                                                               0);
+
+               if (mapping_id == NULL)
+                       return -1;
+
+               for (i = 0 ;i < RTE_MAX_LCORE; i++)
+                       mapping_id[i] = INVALID_MAPPING_ID;
+       }
+       return 0;
+}
+
+/* Create ring which used for communicate among master and slave */
+static int 
+create_ms_ring(unsigned slaveid)
+{
+       unsigned flag = RING_F_SP_ENQ | RING_F_SC_DEQ;
+       struct lcore_resource_struct *res = &lcore_resource[slaveid];
+       unsigned socketid = rte_socket_id();
+       
+       /* Always assume create ring on master socket_id */
+       /* Default only create a ring size 32 */
+       snprintf(res->ring_name[0], MAX_NAME_LEN, "%s%u",
+                       RING_MASTER_NAME, slaveid);
+       if ((res->ring[0] = create_ring(res->ring_name[0], NB_CORE_MSGBUF, 
+                               socketid, flag)) == NULL) {
+               printf("Create m2s ring %s failed\n", res->ring_name[0]);
+               return -1;
+       }
+       
+       snprintf(res->ring_name[1], MAX_NAME_LEN, "%s%u",
+                       RING_SLAVE_NAME, slaveid);
+       if ((res->ring[1] = create_ring(res->ring_name[1], NB_CORE_MSGBUF, 
+               socketid, flag)) == NULL) {
+               printf("Create s2m ring %s failed\n", res->ring_name[1]);
+               return -1;
+       }       
+
+       return 0;
+}
+
+/* send command to pair in paired master and slave ring */
+static inline int 
+sendcmd(unsigned slaveid, enum l2fwd_cmd cmd, int is_master)
+{
+       struct lcore_resource_struct *res = &lcore_resource[slaveid];
+       void *msg;
+       int fd = !is_master;
+
+       /* Only check master, it must be enabled and running if it is slave */
+       if (is_master && !res->enabled)
+               return -1;
+       
+       if (res->ring[fd] == NULL)
+               return -1;      
+       
+       if (rte_mempool_get(message_pool, &msg) < 0) {
+               printf("Error to get message buffer\n");
+               return -1;
+       }
+
+       *(enum l2fwd_cmd *)msg = cmd;
+
+       if (rte_ring_enqueue(res->ring[fd], msg) != 0) {
+               printf("Enqueue error\n");
+               rte_mempool_put(message_pool, msg);
+               return -1;
+       }
+               
+       return 0;
+}
+
+/* Get command from pair in paired master and slave ring */
+static inline int
+getcmd(unsigned slaveid, enum l2fwd_cmd *cmd, int is_master)
+{
+       struct lcore_resource_struct *res = &lcore_resource[slaveid];
+       void *msg;
+       int fd = !!is_master;
+       int ret;
+       /* Only check master, it must be enabled and running if it is slave */
+       if (is_master && (!res->enabled))
+               return -1;
+
+       if (res->ring[fd] == NULL)
+               return -1;
+
+       ret = rte_ring_dequeue(res->ring[fd], &msg);
+
+       if (ret == 0) {
+               *cmd = *(enum l2fwd_cmd *)msg;
+               rte_mempool_put(message_pool, msg);
+       }
+       return ret;
+}
+
+/* Master send command to slave and wait until ack received or error met */
+static int 
+master_sendcmd_with_ack(unsigned slaveid, enum l2fwd_cmd cmd)
+{
+       enum l2fwd_cmd ack_cmd;
+       int ret = -1;
+
+       if (sendcmd(slaveid, cmd, 1) != 0)
+               rte_exit(EXIT_FAILURE, "Failed to send message\n");
+
+       /* Get ack */
+       while (1) {
+               ret = getcmd(slaveid, &ack_cmd, 1);
+               if (ret == 0 && cmd == ack_cmd)
+                       break;
+
+               /* If slave not running yet, return an error */
+               if (flib_query_slave_status(slaveid) != ST_RUN) {
+                       ret = -ENOENT;
+                       break;
+               }
+       }
+
+       return ret;
+}
+
+/* restart all port that assigned to that slave lcore */
+static int 
+reset_slave_all_ports(unsigned slaveid)
+{
+       struct lcore_resource_struct *slave = &lcore_resource[slaveid];
+       int i, ret = 0;
+
+       /* stop/start port */
+       for (i = 0; i < slave->port_num; i++) {
+               char buf_name[RTE_MEMPOOL_NAMESIZE];
+               struct rte_mempool *pool;
+               printf("Stop port :%d\n", slave->port[i]);
+               rte_eth_dev_stop(slave->port[i]);
+               snprintf(buf_name, RTE_MEMPOOL_NAMESIZE, MBUF_NAME, slave->port[i]);
+               pool = rte_mempool_lookup(buf_name);
+               if (pool) 
+                       printf("Port %d mempool free object is %u(%u)\n", slave->port[i], 
+                               rte_mempool_count(pool), (unsigned)NB_MBUF);
+               else
+                       printf("Can't find mempool %s\n", buf_name);
+               
+               printf("Start port :%d\n", slave->port[i]);
+               ret = rte_eth_dev_start(slave->port[i]);
+               if (ret != 0)
+                       break;
+       }
+       return ret;
+}
+
+static int 
+reset_shared_structures(unsigned slaveid)
+{
+       int ret;
+       /* Only port are shared resource here */
+       ret = reset_slave_all_ports(slaveid);
+
+       return ret;
+}
+
+/** 
+ * Call this function to re-create resource that needed for slave process that 
+ * exited in last instance
+ **/
+static int 
+init_slave_res(unsigned slaveid)
+{
+       struct lcore_resource_struct *slave = &lcore_resource[slaveid];
+       enum l2fwd_cmd cmd;
+       
+       if (!slave->enabled) {
+               printf("Something wrong with lcore=%u enabled=%d\n",slaveid, 
+                       slave->enabled);
+               return -1;
+       }
+
+       /* Initialize ring */
+       if (create_ms_ring(slaveid) != 0)
+               rte_exit(EXIT_FAILURE, "failed to create ring for slave %u\n",
+                               slaveid);
+
+       /* drain un-read buffer if have */
+       while (getcmd(slaveid, &cmd, 1) == 0);
+       while (getcmd(slaveid, &cmd, 0) == 0);
+
+       return 0;
+}
+
+static int 
+recreate_one_slave(unsigned slaveid)
+{
+       int ret = 0;
+       /* Re-initialize resource for stalled slave */
+       if ((ret = init_slave_res(slaveid)) != 0) {
+               printf("Init slave=%u failed\n", slaveid);
+               return ret;
+       }       
+       
+       if ((ret = flib_remote_launch(l2fwd_launch_one_lcore, NULL, slaveid))
+               != 0)
+               printf("Launch slave %u failed\n", slaveid);
+       
+       return ret;
+}
+
+/**
+ * remapping resource belong to slave_id to new lcore that gets from flib_assign_lcore_id(),
+ * used only floating process option applied.
+ *
+ * @param slaveid
+ *   original lcore_id that apply for remapping
+ */
+static void
+remapping_slave_resource(unsigned slaveid, unsigned map_id)
+{
+
+       /* remapping lcore_resource */
+       memcpy(&lcore_resource[map_id], &lcore_resource[slaveid],
+                       sizeof(struct lcore_resource_struct));
+
+       /* remapping lcore_queue_conf */
+       memcpy(&lcore_queue_conf[map_id], &lcore_queue_conf[slaveid],
+                       sizeof(struct lcore_queue_conf));
+}
+
+static int
+reset_pair(unsigned slaveid, unsigned pairid)
+{
+       int ret;
+       if ((ret = reset_shared_structures(slaveid)) != 0)
+               goto back;
+
+       if((ret = reset_shared_structures(pairid)) != 0)
+               goto back;
+
+       if (float_proc) {
+               unsigned map_id = mapping_id[slaveid];
+
+               if (map_id != INVALID_MAPPING_ID) {
+                       printf("%u return mapping id %u\n", slaveid, map_id);
+                       flib_free_lcore_id(map_id);
+                       mapping_id[slaveid] = INVALID_MAPPING_ID;
+               }
+
+               map_id = mapping_id[pairid];
+               if (map_id != INVALID_MAPPING_ID) {
+                       printf("%u return mapping id %u\n", pairid, map_id);
+                       flib_free_lcore_id(map_id);
+                       mapping_id[pairid] = INVALID_MAPPING_ID;
+               }
+       }
+
+       if((ret = recreate_one_slave(slaveid)) != 0)
+               goto back;
+
+       ret = recreate_one_slave(pairid);
+
+back:
+       return ret;
+}
+
+static void 
+slave_exit_cb(unsigned slaveid, __attribute__((unused))int stat)
+{
+       struct lcore_resource_struct *slave = &lcore_resource[slaveid];
+
+       printf("Get slave %u leave info\n", slaveid);
+       if (!slave->enabled) {
+               printf("Lcore=%u not registered for it's exit\n", slaveid);
+               return;
+       }
+       rte_spinlock_lock(&res_lock);
+
+       /* Change the state and wait master to start them */
+       slave->flags = SLAVE_RECREATE_FLAG;
+
+       rte_spinlock_unlock(&res_lock);
+}
+
+/* Send the packet on an output interface */
+static int
+l2fwd_send_burst(struct lcore_queue_conf *qconf, unsigned n, uint8_t port)
+{
+       struct rte_mbuf **m_table;
+       unsigned ret;
+       unsigned queueid =0;
+
+       m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
+
+       ret = rte_eth_tx_burst(port, (uint16_t) queueid, m_table, (uint16_t) n);
+       port_statistics[port].tx += ret;
+       if (unlikely(ret < n)) {
+               port_statistics[port].dropped += (n - ret);
+               do {
+                       rte_pktmbuf_free(m_table[ret]);
+               } while (++ret < n);
+       }
+
+       return 0;
+}
+
+/* Send the packet on an output interface */
+static int
+l2fwd_send_packet(struct rte_mbuf *m, uint8_t port)
+{
+       unsigned lcore_id, len;
+       struct lcore_queue_conf *qconf;
+
+       lcore_id = rte_lcore_id();
+
+       qconf = &lcore_queue_conf[lcore_id];
+       len = qconf->tx_mbufs[port].len;
+       qconf->tx_mbufs[port].m_table[len] = m;
+       len++;
+
+       /* enough pkts to be sent */
+       if (unlikely(len == MAX_PKT_BURST)) {
+               l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
+               len = 0;
+       }
+
+       qconf->tx_mbufs[port].len = len;
+       return 0;
+}
+
+static void
+l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
+{
+       struct ether_hdr *eth;
+       void *tmp;
+       unsigned dst_port;
+
+       dst_port = l2fwd_dst_ports[portid];
+       eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
+
+       /* 02:00:00:00:00:xx */
+       tmp = &eth->d_addr.addr_bytes[0];
+       *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
+
+       /* src addr */
+       ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], &eth->s_addr);
+
+       l2fwd_send_packet(m, (uint8_t) dst_port);
+}
+
+/* main processing loop */
+static void
+l2fwd_main_loop(void)
+{
+       struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+       struct rte_mbuf *m;
+       unsigned lcore_id;
+       uint64_t prev_tsc, diff_tsc, cur_tsc;
+       unsigned i, j, portid, nb_rx;
+       struct lcore_queue_conf *qconf;
+       const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
+       
+       prev_tsc = 0;
+
+       lcore_id = rte_lcore_id();
+
+       qconf = &lcore_queue_conf[lcore_id];
+
+       if (qconf->n_rx_port == 0) {
+               RTE_LOG(INFO, L2FWD, "lcore %u has nothing to do\n", lcore_id);
+               return;
+       }
+
+       RTE_LOG(INFO, L2FWD, "entering main loop on lcore %u\n", lcore_id);
+
+       for (i = 0; i < qconf->n_rx_port; i++) {
+               portid = qconf->rx_port_list[i];
+               RTE_LOG(INFO, L2FWD, " -- lcoreid=%u portid=%u\n", lcore_id,
+                       portid);
+       }
+
+       while (1) {
+               enum l2fwd_cmd cmd;
+               cur_tsc = rte_rdtsc();
+
+               if (unlikely(getcmd(lcore_id, &cmd, 0) == 0)) {
+                       sendcmd(lcore_id, cmd, 0);
+
+                       /* If get stop command, stop forwarding and exit */
+                       if (cmd == CMD_STOP) {
+                               return;
+                       } 
+               }
+               
+               /*
+                * TX burst queue drain
+                */
+               diff_tsc = cur_tsc - prev_tsc;
+               if (unlikely(diff_tsc > drain_tsc)) {
+
+                       for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
+                               if (qconf->tx_mbufs[portid].len == 0)
+                                       continue;
+                               l2fwd_send_burst(&lcore_queue_conf[lcore_id],
+                                                qconf->tx_mbufs[portid].len,
+                                                (uint8_t) portid);
+                               qconf->tx_mbufs[portid].len = 0;
+                       }
+               }
+
+               /*
+                * Read packet from RX queues
+                */
+               for (i = 0; i < qconf->n_rx_port; i++) {
+
+                       portid = qconf->rx_port_list[i];
+                       nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
+                                                pkts_burst, MAX_PKT_BURST);
+
+                       port_statistics[portid].rx += nb_rx;
+
+                       for (j = 0; j < nb_rx; j++) {
+                               m = pkts_burst[j];
+                               rte_prefetch0(rte_pktmbuf_mtod(m, void *));
+                               l2fwd_simple_forward(m, portid);
+                       }
+               }
+       }
+}
+
+static int
+l2fwd_launch_one_lcore(__attribute__((unused)) void *dummy)
+{
+       unsigned lcore_id = rte_lcore_id();
+
+       if (float_proc) {
+               unsigned flcore_id;
+
+               /* Change it to floating process, also change it's lcore_id */
+               clear_cpu_affinity();
+               RTE_PER_LCORE(_lcore_id) = 0;
+               /* Get a lcore_id */
+               if (flib_assign_lcore_id() < 0 ) {
+                       printf("flib_assign_lcore_id failed\n");
+                       return -1;
+               }
+               flcore_id = rte_lcore_id();
+               /* Set mapping id, so master can return it after slave exited */
+               mapping_id[lcore_id] = flcore_id;
+               printf("Org lcore_id = %u, cur lcore_id = %u\n",
+                               lcore_id, flcore_id);
+               remapping_slave_resource(lcore_id, flcore_id);
+       }
+
+       l2fwd_main_loop();
+
+       /* return lcore_id before return */
+       if (float_proc) {
+               flib_free_lcore_id(rte_lcore_id());
+               mapping_id[lcore_id] = INVALID_MAPPING_ID;
+       }
+       return 0;
+}
+
+/* display usage */
+static void
+l2fwd_usage(const char *prgname)
+{
+       printf("%s [EAL options] -- -p PORTMASK -s COREMASK [-q NQ] -f\n"
+              "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
+              "  -q NQ: number of queue (=ports) per lcore (default is 1)\n"
+              "  -f use floating process which won't bind to any core to run\n"
+                  "  -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
+              prgname);
+}
+
+static int
+l2fwd_parse_portmask(const char *portmask)
+{
+       char *end = NULL;
+       unsigned long pm;
+
+       /* parse hexadecimal string */
+       pm = strtoul(portmask, &end, 16);
+       if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
+               return -1;
+
+       if (pm == 0)
+               return -1;
+
+       return pm;
+}
+
+static unsigned int
+l2fwd_parse_nqueue(const char *q_arg)
+{
+       char *end = NULL;
+       unsigned long n;
+
+       /* parse hexadecimal string */
+       n = strtoul(q_arg, &end, 10);
+       if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
+               return 0;
+       if (n == 0)
+               return 0;
+       if (n >= MAX_RX_QUEUE_PER_LCORE)
+               return 0;
+
+       return n;
+}
+
+static int
+l2fwd_parse_timer_period(const char *q_arg)
+{
+       char *end = NULL;
+       int n;
+
+       /* parse number string */
+       n = strtol(q_arg, &end, 10);
+       if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
+               return -1;
+       if (n >= MAX_TIMER_PERIOD)
+               return -1;
+
+       return n;
+}
+
+/* Parse the argument given in the command line of the application */
+static int
+l2fwd_parse_args(int argc, char **argv)
+{
+       int opt, ret;
+       char **argvopt;
+       int option_index;
+       char *prgname = argv[0];
+       static struct option lgopts[] = {
+               {NULL, 0, 0, 0}
+       };
+       int has_pmask = 0;
+       
+       argvopt = argv;
+
+       while ((opt = getopt_long(argc, argvopt, "p:q:T:f",
+                                 lgopts, &option_index)) != EOF) {
+
+               switch (opt) {
+               /* portmask */
+               case 'p':
+                       l2fwd_enabled_port_mask = l2fwd_parse_portmask(optarg);
+                       if (l2fwd_enabled_port_mask == 0) {
+                               printf("invalid portmask\n");
+                               l2fwd_usage(prgname);
+                               return -1;
+                       }
+                       has_pmask = 1;
+                       break;
+
+               /* nqueue */
+               case 'q':
+                       l2fwd_rx_queue_per_lcore = l2fwd_parse_nqueue(optarg);
+                       if (l2fwd_rx_queue_per_lcore == 0) {
+                               printf("invalid queue number\n");
+                               l2fwd_usage(prgname);
+                               return -1;
+                       }
+                       break;
+
+               /* timer period */
+               case 'T':
+                       timer_period = l2fwd_parse_timer_period(optarg) * 1000 * TIMER_MILLISECOND;
+                       if (timer_period < 0) {
+                               printf("invalid timer period\n");
+                               l2fwd_usage(prgname);
+                               return -1;
+                       }
+                       break;
+
+               /* use floating process */
+               case 'f':
+                       float_proc = 1;
+                       break;
+                       
+               /* long options */
+               case 0:
+                       l2fwd_usage(prgname);
+                       return -1;
+
+               default:
+                       l2fwd_usage(prgname);
+                       return -1;
+               }
+       }
+
+       if (optind >= 0)
+               argv[optind-1] = prgname;
+
+       if (!has_pmask) {
+               l2fwd_usage(prgname);
+               return -1;
+       }
+       ret = optind-1;
+       optind = 0; /* reset getopt lib */
+       return ret;
+}
+
+/* Check the link status of all ports in up to 9s, and print them finally */
+static void
+check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
+{
+#define CHECK_INTERVAL 100 /* 100ms */
+#define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
+       uint8_t portid, count, all_ports_up, print_flag = 0;
+       struct rte_eth_link link;
+
+       printf("\nChecking link status");
+       fflush(stdout);
+       for (count = 0; count <= MAX_CHECK_TIME; count++) {
+               all_ports_up = 1;
+               for (portid = 0; portid < port_num; portid++) {
+                       if ((port_mask & (1 << portid)) == 0)
+                               continue;
+                       memset(&link, 0, sizeof(link));
+                       rte_eth_link_get_nowait(portid, &link);
+                       /* print link status if flag set */
+                       if (print_flag == 1) {
+                               if (link.link_status)
+                                       printf("Port %d Link Up - speed %u "
+                                               "Mbps - %s\n", (uint8_t)portid,
+                                               (unsigned)link.link_speed,
+                               (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
+                                       ("full-duplex") : ("half-duplex\n"));
+                               else
+                                       printf("Port %d Link Down\n",
+                                               (uint8_t)portid);
+                               continue;
+                       }
+                       /* clear all_ports_up flag if any link down */
+                       if (link.link_status == 0) {
+                               all_ports_up = 0;
+                               break;
+                       }
+               }
+               /* after finally printing all link status, get out */
+               if (print_flag == 1)
+                       break;
+
+               if (all_ports_up == 0) {
+                       printf(".");
+                       fflush(stdout);
+                       rte_delay_ms(CHECK_INTERVAL);
+               }
+
+               /* set the print_flag if all ports up or timeout */
+               if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
+                       print_flag = 1;
+                       printf("done\n");
+               }
+       }
+}
+
+int
+MAIN(int argc, char **argv)
+{
+       struct lcore_queue_conf *qconf;
+       struct rte_eth_dev_info dev_info;
+       int ret;
+       uint8_t nb_ports;
+       uint8_t nb_ports_available;
+       uint8_t portid, last_port;
+       unsigned rx_lcore_id;
+       unsigned nb_ports_in_mask = 0;
+       unsigned i;
+       int flags = 0;
+       uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
+
+       /* Save cpu_affinity first, restore it in case it's floating process option */
+       if (get_cpu_affinity() != 0)
+               rte_exit(EXIT_FAILURE, "get_cpu_affinity error\n");
+
+       /* Also tries to set cpu affinity to detect whether  it will fail in child process */
+       if(clear_cpu_affinity() != 0)
+               rte_exit(EXIT_FAILURE, "clear_cpu_affinity error\n");
+
+       /* init EAL */
+       ret = rte_eal_init(argc, argv);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
+       argc -= ret;
+       argv += ret;
+
+       /* parse application arguments (after the EAL ones) */
+       ret = l2fwd_parse_args(argc, argv);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
+
+       /*flib init */
+       if (flib_init() != 0)
+               rte_exit(EXIT_FAILURE, "flib init error");
+
+       /**
+         * Allocated structures that slave lcore would change. For those that slaves are 
+         * read only, needn't use malloc to share and global or static variables is ok since
+         * slave inherit all the knowledge that master initialized.
+         **/
+       if (l2fwd_malloc_shared_struct() != 0)
+               rte_exit(EXIT_FAILURE, "malloc mem failed\n");
+
+       /* Initialize lcore_resource structures */
+       memset(lcore_resource, 0, sizeof(lcore_resource));
+       for (i = 0; i < RTE_MAX_LCORE; i++)
+               lcore_resource[i].lcore_id = i;
+       
+       /* init driver(s) */
+       if (rte_pmd_init_all() < 0)
+               rte_exit(EXIT_FAILURE, "Cannot init pmd\n");
+
+       if (rte_eal_pci_probe() < 0)
+               rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
+
+       nb_ports = rte_eth_dev_count();
+       if (nb_ports == 0)
+               rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
+
+       if (nb_ports > RTE_MAX_ETHPORTS)
+               nb_ports = RTE_MAX_ETHPORTS;
+
+       /* create the mbuf pool */
+       for (portid = 0; portid < nb_ports; portid++) {
+               /* skip ports that are not enabled */
+               if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
+                       continue;         
+               char buf_name[RTE_MEMPOOL_NAMESIZE];
+               flags = MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET;
+               snprintf(buf_name, RTE_MEMPOOL_NAMESIZE, MBUF_NAME, portid);
+               l2fwd_pktmbuf_pool[portid] =
+                       rte_mempool_create(buf_name, NB_MBUF,
+                                          MBUF_SIZE, 32,
+                                          sizeof(struct rte_pktmbuf_pool_private),
+                                          rte_pktmbuf_pool_init, NULL,
+                                          rte_pktmbuf_init, NULL,
+                                          rte_socket_id(), flags);
+               if (l2fwd_pktmbuf_pool[portid] == NULL)
+                       rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
+
+               printf("Create mbuf %s\n", buf_name);
+       }
+
+       /* reset l2fwd_dst_ports */
+       for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
+               l2fwd_dst_ports[portid] = 0;
+       last_port = 0;
+
+       /*
+        * Each logical core is assigned a dedicated TX queue on each port.
+        */
+       for (portid = 0; portid < nb_ports; portid++) {
+               /* skip ports that are not enabled */
+               if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
+                       continue;
+
+               if (nb_ports_in_mask % 2) {
+                       l2fwd_dst_ports[portid] = last_port;
+                       l2fwd_dst_ports[last_port] = portid;
+               }
+               else
+                       last_port = portid;
+
+               nb_ports_in_mask++;
+
+               rte_eth_dev_info_get(portid, &dev_info);
+       }
+       if (nb_ports_in_mask % 2) {
+               printf("Notice: odd number of ports in portmask.\n");
+               l2fwd_dst_ports[last_port] = last_port;
+       }
+
+       rx_lcore_id = 0;
+       qconf = NULL;
+
+       /* Initialize the port/queue configuration of each logical core */
+       for (portid = 0; portid < nb_ports; portid++) {
+               struct lcore_resource_struct *res;
+               /* skip ports that are not enabled */
+               if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
+                       continue;
+
+               /* get the lcore_id for this port */
+               /* skip master lcore */
+               while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
+                          rte_get_master_lcore() == rx_lcore_id || 
+                      lcore_queue_conf[rx_lcore_id].n_rx_port ==
+                      l2fwd_rx_queue_per_lcore) {
+
+                       rx_lcore_id++;
+                       if (rx_lcore_id >= RTE_MAX_LCORE)
+                               rte_exit(EXIT_FAILURE, "Not enough cores\n");
+               }
+
+               if (qconf != &lcore_queue_conf[rx_lcore_id])
+                       /* Assigned a new logical core in the loop above. */
+                       qconf = &lcore_queue_conf[rx_lcore_id];
+
+               qconf->rx_port_list[qconf->n_rx_port] = portid;
+               qconf->n_rx_port++;
+
+               /* Save the port resource info into lcore_resource strucutres */
+               res = &lcore_resource[rx_lcore_id];
+               res->enabled = 1;
+               res->port[res->port_num++] = portid;
+               
+               printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
+       }
+
+       nb_ports_available = nb_ports;
+
+       /* Initialise each port */
+       for (portid = 0; portid < nb_ports; portid++) {
+               /* skip ports that are not enabled */
+               if ((l2fwd_enabled_port_mask & (1 << portid)) == 0) {
+                       printf("Skipping disabled port %u\n", (unsigned) portid);
+                       nb_ports_available--;
+                       continue;
+               }
+               /* init port */
+               printf("Initializing port %u... ", (unsigned) portid);
+               fflush(stdout);
+               ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
+               if (ret < 0)
+                       rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
+                                 ret, (unsigned) portid);
+
+               rte_eth_macaddr_get(portid,&l2fwd_ports_eth_addr[portid]);
+
+               /* init one RX queue */
+               fflush(stdout);
+               ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
+                                            rte_eth_dev_socket_id(portid), &rx_conf,
+                                            l2fwd_pktmbuf_pool[portid]);
+               if (ret < 0)
+                       rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
+                                 ret, (unsigned) portid);
+
+               /* init one TX queue on each port */
+               fflush(stdout);
+               ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
+                               rte_eth_dev_socket_id(portid), &tx_conf);
+               if (ret < 0)
+                       rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
+                               ret, (unsigned) portid);
+
+               /* Start device */
+               ret = rte_eth_dev_start(portid);
+               if (ret < 0)
+                       rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
+                                 ret, (unsigned) portid);
+
+               printf("done: \n");
+
+               rte_eth_promiscuous_enable(portid);
+
+               printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
+                               (unsigned) portid,
+                               l2fwd_ports_eth_addr[portid].addr_bytes[0],
+                               l2fwd_ports_eth_addr[portid].addr_bytes[1],
+                               l2fwd_ports_eth_addr[portid].addr_bytes[2],
+                               l2fwd_ports_eth_addr[portid].addr_bytes[3],
+                               l2fwd_ports_eth_addr[portid].addr_bytes[4],
+                               l2fwd_ports_eth_addr[portid].addr_bytes[5]);
+
+               /* initialize port stats */
+               //memset(&port_statistics, 0, sizeof(port_statistics));
+       }
+
+       if (!nb_ports_available) {
+               rte_exit(EXIT_FAILURE,
+                       "All available ports are disabled. Please set portmask.\n");
+       }
+
+       check_all_ports_link_status(nb_ports, l2fwd_enabled_port_mask);
+
+       /* Record pair lcore */
+       /**
+        * Since l2fwd example would create pair between different neighbour port, that's 
+        * port 0 receive and forward to port 1, the same to port 1, these 2 ports will have
+        * dependency. If one port stopped working (killed, for example), the port need to
+        * be stopped/started again. During the time, another port need to wait until stop/start
+        * procedure completed. So, record the pair relationship for those lcores working
+        * on ports. 
+        **/
+       for (portid = 0; portid < nb_ports; portid++) {
+               uint32_t pair_port;
+               unsigned lcore = 0, pair_lcore = 0;
+               unsigned j, find_lcore, find_pair_lcore;
+               /* skip ports that are not enabled */
+               if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
+                       continue;
+
+               /* Find pair ports' lcores */
+               find_lcore = find_pair_lcore = 0;
+               pair_port = l2fwd_dst_ports[portid];
+               for (i = 0; i < RTE_MAX_LCORE; i++) {
+                       if (!rte_lcore_is_enabled(i))
+                               continue;
+                       for (j = 0; j < lcore_queue_conf[i].n_rx_port;j++) {
+                               if (lcore_queue_conf[i].rx_port_list[j] == portid) {
+                                       lcore = i;      
+                                       find_lcore = 1;
+                                       break;
+                               }
+                               if (lcore_queue_conf[i].rx_port_list[j] == pair_port) {
+                                       pair_lcore = i; 
+                                       find_pair_lcore = 1;
+                                       break;
+                               }
+                       }
+                       if (find_lcore && find_pair_lcore)
+                               break;
+               }
+               if (!find_lcore || !find_pair_lcore)
+                       rte_exit(EXIT_FAILURE, "Not find port=%d pair\n", portid);
+
+               printf("lcore %u and %u paired\n", lcore, pair_lcore);
+               lcore_resource[lcore].pair_id = pair_lcore;
+               lcore_resource[pair_lcore].pair_id = lcore;
+       }
+
+       /* Create message buffer for all master and slave */
+       message_pool = rte_mempool_create("ms_msg_pool", 
+                          NB_CORE_MSGBUF * RTE_MAX_LCORE,
+                          sizeof(enum l2fwd_cmd), NB_CORE_MSGBUF / 2,
+                          0,
+                          rte_pktmbuf_pool_init, NULL,
+                          rte_pktmbuf_init, NULL,
+                          rte_socket_id(), 0); 
+
+       if (message_pool == NULL)
+               rte_exit(EXIT_FAILURE, "Create msg mempool failed\n");
+
+       /* Create ring for each master and slave pair, also register cb when slave leaves */
+       for (i = 0; i < RTE_MAX_LCORE; i++) {
+               /** 
+                * Only create ring and register slave_exit cb in case that core involved into 
+                * packet forwarding
+                **/
+               if (lcore_resource[i].enabled) {
+                       /* Create ring for master and slave communication */
+                       ret = create_ms_ring(i);
+                       if (ret != 0)
+                               rte_exit(EXIT_FAILURE, "Create ring for lcore=%u failed",
+                               i);
+
+                       if (flib_register_slave_exit_notify(i,
+                               slave_exit_cb) != 0)
+                               rte_exit(EXIT_FAILURE, 
+                                               "Register master_trace_slave_exit failed");
+               }
+       }
+
+       /* launch per-lcore init on every lcore except master */
+       flib_mp_remote_launch(l2fwd_launch_one_lcore, NULL, SKIP_MASTER);
+
+       /* print statistics 10 second */
+       prev_tsc = cur_tsc = rte_rdtsc();
+       timer_tsc = 0;
+       while (1) {
+               sleep(1);
+               cur_tsc = rte_rdtsc();
+               diff_tsc = cur_tsc - prev_tsc;
+               /* if timer is enabled */
+               if (timer_period > 0) {
+               
+                       /* advance the timer */
+                       timer_tsc += diff_tsc;
+               
+                       /* if timer has reached its timeout */
+                       if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
+               
+                               print_stats();
+                               /* reset the timer */
+                               timer_tsc = 0;
+                       }
+               }
+               
+               prev_tsc = cur_tsc;
+
+               /* Check any slave need restart or recreate */
+               rte_spinlock_lock(&res_lock);
+               for (i = 0; i < RTE_MAX_LCORE; i++) {
+                       struct lcore_resource_struct *res  = &lcore_resource[i];
+                       struct lcore_resource_struct *pair = &lcore_resource[res->pair_id];
+
+                       /* If find slave exited, try to reset pair */
+                       if (res->enabled && res->flags && pair->enabled) {
+                               if (!pair->flags) {
+                                       master_sendcmd_with_ack(pair->lcore_id, CMD_STOP);
+                                       rte_spinlock_unlock(&res_lock);
+                                       sleep(1);
+                                       rte_spinlock_lock(&res_lock);
+                                       if (pair->flags)
+                                               continue;
+                               }
+                               if (reset_pair(res->lcore_id, pair->lcore_id) != 0)
+                                       rte_exit(EXIT_FAILURE, "failed to reset slave");
+                               res->flags  = 0;
+                               pair->flags = 0;
+                       }
+               }
+               rte_spinlock_unlock(&res_lock);
+       }
+
+}
diff --git a/examples/multi_process/l2fwd_fork/main.h b/examples/multi_process/l2fwd_fork/main.h
new file mode 100644 (file)
index 0000000..93dca37
--- /dev/null
@@ -0,0 +1,45 @@
+/*-
+ *   BSD LICENSE
+ * 
+ *   Copyright(c) 2010-2013 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_
+
+#ifdef RTE_EXEC_ENV_BAREMETAL
+#define MAIN _main
+#else
+#define MAIN main
+#endif
+
+int MAIN(int argc, char **argv);
+
+#endif /* _MAIN_H_ */