Handle older OSes without DLT_PRISM_HEADER/DLT_AIRONET_HEADER/DLT_PPI.

[libpcap] / pcap-dpdk.c

diff --git a/pcap-dpdk.c b/pcap-dpdk.c
index 3d680a9fb5e162eb9c93b1b12dbbc22f0dbb54a0..4175a11d748a50f09b3618d597a3bef9626b1aed 100644 (file)
--- a/pcap-dpdk.c
+++ b/pcap-dpdk.c
@@ -1,4 +1,16 @@
 /*
 /*

+ * Copyright (C) 2018 jingle YANG. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *   1. Redistributions of source code must retain the above copyright
+ *      notice, this list of conditions and the following disclaimer.
+ *   2. 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.
+ *

  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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
  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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


@@ -16,22 +28,24 @@
 Date: Dec 16, 2018
 
 Description:
 Date: Dec 16, 2018
 
 Description:

-1. Pcap-dpdk provides libpcap the ability to use DPDK with the device name as dpdk:[portid], such as dpdk:0.
+1. Pcap-dpdk provides libpcap the ability to use DPDK with the device name as dpdk:{portid}, such as dpdk:0.

 2. DPDK is a set of libraries and drivers for fast packet processing. (https://www.dpdk.org/) 
 2. DPDK is a set of libraries and drivers for fast packet processing. (https://www.dpdk.org/) 

+3. The testprogs/capturetest provides 6.4Gbps/800,000 pps on Intel 10-Gigabit X540-AT2 with DPDK 18.11.

 
 Limitations:
 
 Limitations:

-1. By default enable_dpdk is no, unless you set inlcudes and lib dir
-by --with-dpdk-includes= --with-dpdk-libraries=
+1. DPDK support will be on if DPDK is available. Please set DIR for --with-dpdk[=DIR] with ./configure or -DDPDK_DIR[=DIR] with cmake if DPDK is installed manually.

 2. Only support link libdpdk.so dynamicly, because the libdpdk.a will not work correctly.
 3. Only support read operation, and packet injection has not been supported yet.
 2. Only support link libdpdk.so dynamicly, because the libdpdk.a will not work correctly.
 3. Only support read operation, and packet injection has not been supported yet.

-4. I have tested on DPDK v18.11.
+

 Usage:
 Usage:

-1. compile DPDK as shared library and install.(https://github.com/DPDK/dpdk.git)
+1. Compile DPDK as shared library and install.(https://github.com/DPDK/dpdk.git)

 
 You shall modify the file $RTE_SDK/$RTE_TARGET/.config and set:
 CONFIG_RTE_BUILD_SHARED_LIB=y
 
 You shall modify the file $RTE_SDK/$RTE_TARGET/.config and set:
 CONFIG_RTE_BUILD_SHARED_LIB=y

+By the following command:
+sed -i 's/CONFIG_RTE_BUILD_SHARED_LIB=n/CONFIG_RTE_BUILD_SHARED_LIB=y/' $RTE_SDK/$RTE_TARGET/.config

 
 

-2. launch l2fwd that is one of DPDK examples correctly, and get device information.
+2. Launch l2fwd that is one of DPDK examples correctly, and get device information.

 
 You shall learn how to bind nic with DPDK-compatible driver by $RTE_SDK/usertools/dpdk-devbind.py, such as igb_uio.
 And enable hugepages by dpdk-setup.sh
 
 You shall learn how to bind nic with DPDK-compatible driver by $RTE_SDK/usertools/dpdk-devbind.py, such as igb_uio.
 And enable hugepages by dpdk-setup.sh


@@ -39,17 +53,22 @@ And enable hugepages by dpdk-setup.sh
 Then launch the l2fwd with dynamic dirver support. For example:
 $RTE_SDK/examples/l2fwd/$RTE_TARGET/l2fwd -dlibrte_pmd_e1000.so -dlibrte_pmd_ixgbe.so -dlibrte_mempool_ring.so -- -p 0x1
 
 Then launch the l2fwd with dynamic dirver support. For example:
 $RTE_SDK/examples/l2fwd/$RTE_TARGET/l2fwd -dlibrte_pmd_e1000.so -dlibrte_pmd_ixgbe.so -dlibrte_mempool_ring.so -- -p 0x1
 

-3. compile libpcap with dpdk options.
+3. Compile libpcap with dpdk options.
+
+If DPDK has not been found automatically, you shall export DPDK envionment variable which are used for compiling DPDK. And then pass $RTE_SDK/$RTE_TARGET to --with-dpdk or -DDPDK_DIR
+
+export RTE_SDK={your DPDK base directory}
+export RTE_TARGET={your target name}

 
 

-you shall run the following command to generate a new configure
+3.1 With configure

 
 

-make clean
-autoreconf
+./configure --with-dpdk=$RTE_SDK/$RTE_TARGET && make -s all && make -s testprogs && make install

 
 

-Then, run configure with dpdk options.
-For Ubuntu, they are --with-dpdk-includes=/usr/local/include/dpdk/ --with-dpdk-libraries=/usr/local/lib
+3.2 With cmake

 
 

-4. link your own program with libpcap, and use DPDK with the device name as dpdk:{portid}, such as dpdk:0.
+mkdir -p build && cd build && cmake -DDPDK_DIR=$RTE_SDK/$RTE_TARGET ../ && make -s all && make -s testprogs && make install 
+
+4. Link your own program with libpcap, and use DPDK with the device name as dpdk:{portid}, such as dpdk:0.

 And you shall set DPDK configure options by environment variable DPDK_CFG
 For example, the testprogs/capturetest could be lanched by: 
 
 And you shall set DPDK configure options by environment variable DPDK_CFG
 For example, the testprogs/capturetest could be lanched by: 
 


@@ -99,11 +118,13 @@ env DPDK_CFG="--log-level=debug -l0 -dlibrte_pmd_e1000.so -dlibrte_pmd_ixgbe.so
 static int is_dpdk_pre_inited=0;
 #define DPDK_LIB_NAME "libpcap_dpdk"
 #define DPDK_DESC "Data Plane Development Kit (DPDK) Interface"
 static int is_dpdk_pre_inited=0;
 #define DPDK_LIB_NAME "libpcap_dpdk"
 #define DPDK_DESC "Data Plane Development Kit (DPDK) Interface"

+#define DPDK_ERR_PERM_MSG "permission denied, DPDK needs root permission"

 #define DPDK_ARGC_MAX 64 
 #define DPDK_CFG_MAX_LEN 1024
 #define DPDK_DEV_NAME_MAX 32
 #define DPDK_DEV_DESC_MAX 512 
 #define DPDK_CFG_ENV_NAME "DPDK_CFG"
 #define DPDK_ARGC_MAX 64 
 #define DPDK_CFG_MAX_LEN 1024
 #define DPDK_DEV_NAME_MAX 32
 #define DPDK_DEV_DESC_MAX 512 
 #define DPDK_CFG_ENV_NAME "DPDK_CFG"

+#define DPDK_DEF_MIN_SLEEP_MS 1

 static char dpdk_cfg_buf[DPDK_CFG_MAX_LEN];
 #define DPDK_MAC_ADDR_SIZE 32
 #define DPDK_DEF_MAC_ADDR "00:00:00:00:00:00"
 static char dpdk_cfg_buf[DPDK_CFG_MAX_LEN];
 #define DPDK_MAC_ADDR_SIZE 32
 #define DPDK_DEF_MAC_ADDR "00:00:00:00:00:00"


@@ -136,15 +157,10 @@ struct dpdk_ts_helper{
 struct pcap_dpdk{
        pcap_t * orig;
        uint16_t portid; // portid of DPDK
 struct pcap_dpdk{
        pcap_t * orig;
        uint16_t portid; // portid of DPDK

-       pcap_handler cb; //callback and argument
-       u_char *cb_arg;
-       int max_cnt;

        int must_clear_promisc;
        int must_clear_promisc;

-       int filter_in_userland;
-       uint64_t rx_pkts;

        uint64_t bpf_drop;
        uint64_t bpf_drop;

-       struct ether_addr eth_addr;
-       char mac_addr[DPDK_MAC_ADDR_SIZE];
+       int nonblock;
+       struct timeval required_select_timeout;

        struct timeval prev_ts;
        struct rte_eth_stats prev_stats;
        struct timeval curr_ts;
        struct timeval prev_ts;
        struct rte_eth_stats prev_stats;
        struct timeval curr_ts;


@@ -153,9 +169,10 @@ struct pcap_dpdk{
        uint64_t bps;
        struct rte_mempool * pktmbuf_pool;
        struct dpdk_ts_helper ts_helper;
        uint64_t bps;
        struct rte_mempool * pktmbuf_pool;
        struct dpdk_ts_helper ts_helper;

+       struct ether_addr eth_addr;
+       char mac_addr[DPDK_MAC_ADDR_SIZE];

        char pci_addr[DPDK_PCI_ADDR_SIZE];
        unsigned char pcap_tmp_buf[RTE_ETH_PCAP_SNAPLEN];
        char pci_addr[DPDK_PCI_ADDR_SIZE];
        unsigned char pcap_tmp_buf[RTE_ETH_PCAP_SNAPLEN];

-       volatile sig_atomic_t break_loop;

 };
 
 static struct rte_eth_conf port_conf = {
 };
 
 static struct rte_eth_conf port_conf = {


@@ -187,10 +204,10 @@ static inline void calculate_timestamp(struct dpdk_ts_helper *helper,struct time
        timeradd(&(helper->start_time), &cur_time, ts);
 }
 
        timeradd(&(helper->start_time), &cur_time, ts);
 }
 

-static unsigned int dpdk_gather_data(unsigned char *data, struct rte_mbuf *mbuf)
+static uint32_t dpdk_gather_data(unsigned char *data, int len, struct rte_mbuf *mbuf)

 {
 {

-       unsigned int total_len = 0;
-       while (mbuf && (total_len+mbuf->data_len) < RTE_ETH_PCAP_SNAPLEN ){
+       uint32_t total_len = 0;
+       while (mbuf && (total_len+mbuf->data_len) < len ){

                rte_memcpy(data+total_len, rte_pktmbuf_mtod(mbuf,void *),mbuf->data_len);
                total_len+=mbuf->data_len;
                mbuf=mbuf->next;
                rte_memcpy(data+total_len, rte_pktmbuf_mtod(mbuf,void *),mbuf->data_len);
                total_len+=mbuf->data_len;
                mbuf=mbuf->next;


@@ -198,57 +215,95 @@ static unsigned int dpdk_gather_data(unsigned char *data, struct rte_mbuf *mbuf)
        return total_len;
 }
 
        return total_len;
 }
 

-static void pcap_dpdk_breakloop(pcap_t *p)
-{
-       pcap_breakloop_common(p);
+
+static int dpdk_read_with_timeout(pcap_t *p, uint16_t portid, uint16_t queueid,struct rte_mbuf **pkts_burst, const uint16_t burst_cnt){

        struct pcap_dpdk *pd = (struct pcap_dpdk*)(p->priv);
        struct pcap_dpdk *pd = (struct pcap_dpdk*)(p->priv);

-       pd->break_loop = p->break_loop; 
+       int nb_rx = 0;  
+       int timeout_ms = p->opt.timeout;
+       int sleep_ms = 0;
+       if (pd->nonblock){
+               // In non-blocking mode, just read once, no mater how many packets are captured.
+               nb_rx = (int)rte_eth_rx_burst(pd->portid, 0, pkts_burst, burst_cnt);
+       }else{
+               // In blocking mode, read many times until packets are captured or timeout or break_loop is setted.     
+               // if timeout_ms == 0, it may be blocked forever.
+               while (timeout_ms == 0 || sleep_ms < timeout_ms){
+                       nb_rx = (int)rte_eth_rx_burst(pd->portid, 0, pkts_burst, burst_cnt);
+                       if (nb_rx){ // got packets within timeout_ms
+                               break;  
+                       }else{ // no packet arrives at this round.
+                               if (p->break_loop){
+                                       break;
+                               }
+                               // sleep for a very short while.
+                               // block sleep is the only choice, since usleep() will impact performance dramatically.
+                               rte_delay_us_block(DPDK_DEF_MIN_SLEEP_MS*1000);
+                               sleep_ms += DPDK_DEF_MIN_SLEEP_MS;
+                       }
+               }
+       }
+       return nb_rx;

 }
 }

-static void dpdk_dispatch_inter(void *dpdk_user)
+
+static int pcap_dpdk_dispatch(pcap_t *p, int max_cnt, pcap_handler cb, u_char *cb_arg)

 {
 {

-       if (dpdk_user == NULL){
-               return;
-       }
-       pcap_t *p = dpdk_user;

        struct pcap_dpdk *pd = (struct pcap_dpdk*)(p->priv);
        struct pcap_dpdk *pd = (struct pcap_dpdk*)(p->priv);

-       int max_cnt = pd->max_cnt;
-       pcap_handler cb = pd->cb;
-       u_char *cb_arg = pd->cb_arg;
-       int nb_rx=0;
+       int burst_cnt = 0;
+       int nb_rx = 0;

        struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
        struct rte_mbuf *m;
        struct pcap_pkthdr pcap_header;
        uint16_t portid = pd->portid;
        struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
        struct rte_mbuf *m;
        struct pcap_pkthdr pcap_header;
        uint16_t portid = pd->portid;

-       unsigned lcore_id = rte_lcore_id();
-       unsigned master_lcore_id = rte_get_master_lcore();
-       uint16_t data_len = 0;
+       // In DPDK, pkt_len is sum of lengths for all segments. And data_len is for one segment
+       uint32_t pkt_len = 0;
+       int caplen = 0;

        u_char *bp = NULL;
        int i=0;
        unsigned int gather_len =0;
        int pkt_cnt = 0;
        int is_accepted=0;
        u_char *bp = NULL;
        int i=0;
        unsigned int gather_len =0;
        int pkt_cnt = 0;
        int is_accepted=0;

-               
-       if(lcore_id == master_lcore_id){
-               RTE_LOG(DEBUG, USER1, "dpdk: lcoreid=%u runs for portid=%u\n", lcore_id, portid);
+       u_char *large_buffer=NULL;
+       int timeout_ms = p->opt.timeout;
+
+       if ( !PACKET_COUNT_IS_UNLIMITED(max_cnt) && max_cnt < MAX_PKT_BURST){
+               burst_cnt = max_cnt;

        }else{
        }else{

-               RTE_LOG(DEBUG, USER1, "dpdk: lcore %u has nothing to do\n", lcore_id);
-       }
-       //only use master lcore
-       if (lcore_id != master_lcore_id){
-               return;
+               burst_cnt = MAX_PKT_BURST;

        }
        }

-       while( max_cnt==-1 || pkt_cnt < max_cnt){
-               if (pd->break_loop){
-                       break;
+
+       while( PACKET_COUNT_IS_UNLIMITED(max_cnt) || pkt_cnt < max_cnt){
+               if (p->break_loop){
+                       p->break_loop = 0;
+                       return PCAP_ERROR_BREAK;
+               }
+               // read once in non-blocking mode, or try many times waiting for timeout_ms.
+               // if timeout_ms == 0, it will be blocked until one packet arrives or break_loop is setted.
+               nb_rx = dpdk_read_with_timeout(p, portid, 0, pkts_burst, burst_cnt);
+               if (nb_rx == 0){ 
+                       if (pd->nonblock){
+                               RTE_LOG(DEBUG, USER1, "dpdk: no packets available in non-blocking mode.\n");
+                       }else{
+                               if (p->break_loop){
+                                       RTE_LOG(DEBUG, USER1, "dpdk: no packets available and break_loop is setted in blocking mode.\n");
+                                       p->break_loop = 0;
+                                       return PCAP_ERROR_BREAK;
+
+                               }
+                               RTE_LOG(DEBUG, USER1, "dpdk: no packets available for timeout %d ms in blocking mode.\n", timeout_ms);
+                       }
+                       // break if dpdk reads 0 packet, no matter in blocking(timeout) or non-blocking mode.
+                       break;  

                }
                }

-               nb_rx = (int)rte_eth_rx_burst(portid, 0, pkts_burst, MAX_PKT_BURST);

                pkt_cnt += nb_rx;
                for ( i = 0; i < nb_rx; i++) {
                        m = pkts_burst[i];
                        calculate_timestamp(&(pd->ts_helper),&(pcap_header.ts));
                pkt_cnt += nb_rx;
                for ( i = 0; i < nb_rx; i++) {
                        m = pkts_burst[i];
                        calculate_timestamp(&(pd->ts_helper),&(pcap_header.ts));

-                       data_len = rte_pktmbuf_data_len(m);
-                       pcap_header.caplen = data_len; 
-                       pcap_header.len = data_len; 
+                       pkt_len = rte_pktmbuf_pkt_len(m);
+                       // caplen = min(pkt_len, p->snapshot);
+                       // caplen will not be changed, no matter how long the rte_pktmbuf
+                       caplen = pkt_len < p->snapshot ? pkt_len: p->snapshot; 
+                       pcap_header.caplen = caplen;
+                       pcap_header.len = pkt_len; 

                        // volatile prefetch
                        rte_prefetch0(rte_pktmbuf_mtod(m, void *));
                        bp = NULL;
                        // volatile prefetch
                        rte_prefetch0(rte_pktmbuf_mtod(m, void *));
                        bp = NULL;


@@ -256,25 +311,21 @@ static void dpdk_dispatch_inter(void *dpdk_user)
                        {
                                bp = rte_pktmbuf_mtod(m, u_char *);
                        }else{
                        {
                                bp = rte_pktmbuf_mtod(m, u_char *);
                        }else{

-                               if (m->pkt_len <= ETHER_MAX_JUMBO_FRAME_LEN)
+                               // use fast buffer pcap_tmp_buf if pkt_len is small, no need to call malloc and free
+                               if ( pkt_len <= ETHER_MAX_JUMBO_FRAME_LEN)

                                {
                                {

-                                       gather_len = dpdk_gather_data(pd->pcap_tmp_buf, m);
+                                       gather_len = dpdk_gather_data(pd->pcap_tmp_buf, RTE_ETH_PCAP_SNAPLEN, m);

                                        bp = pd->pcap_tmp_buf;
                                        bp = pd->pcap_tmp_buf;

-                                       pcap_header.caplen = gather_len;
-                                       pcap_header.len = gather_len;
+                               }else{ 
+                                       // need call free later
+                                       large_buffer = (u_char *)malloc(caplen*sizeof(u_char));
+                                       gather_len = dpdk_gather_data(large_buffer, caplen, m); 
+                                       bp = large_buffer;

                                }
                                }

+                               

                        }
                        if (bp){
                        }
                        if (bp){

-                               //default accpet all
-                               is_accepted=1;
-                               if (pd->filter_in_userland && p->fcode.bf_insns!=NULL)
-                               {
-                                       if (!pcap_filter(p->fcode.bf_insns, bp, pcap_header.len, pcap_header.caplen)){
-                                               //rejected
-                                               is_accepted=0;
-                                       }
-                               }
-                               if (is_accepted){
+                               if (p->fcode.bf_insns==NULL || pcap_filter(p->fcode.bf_insns, bp, pcap_header.len, pcap_header.caplen)){

                                        cb(cb_arg, &pcap_header, bp);
                                }else{
                                        pd->bpf_drop++;
                                        cb(cb_arg, &pcap_header, bp);
                                }else{
                                        pd->bpf_drop++;


@@ -282,40 +333,20 @@ static void dpdk_dispatch_inter(void *dpdk_user)
                        }
                        //free all pktmbuf
                        rte_pktmbuf_free(m);
                        }
                        //free all pktmbuf
                        rte_pktmbuf_free(m);

+                       if (large_buffer){
+                               free(large_buffer);
+                               large_buffer=NULL;
+                       }

                }
        }       
                }
        }       

-       pd->rx_pkts = pkt_cnt;
-}
-static int launch_one_lcore(void *dpdk_user)
-{
-       dpdk_dispatch_inter(dpdk_user);
-       return 0;
-}
-static int pcap_dpdk_dispatch(pcap_t *p, int max_cnt, pcap_handler cb, u_char *pcap_user)
-{
-       unsigned lcore_id = 0;  
-       struct pcap_dpdk *pd = (struct pcap_dpdk*)(p->priv);
-       pd->rx_pkts=0;
-       pd->cb = cb;
-       pd->cb_arg = pcap_user;
-       pd->max_cnt = max_cnt;
-       pd->orig = p;
-       void *dpdk_user = p;    
-       // launch_one_lcore func will be called on every lcore include master core.
-       rte_eal_mp_remote_launch(launch_one_lcore, dpdk_user, CALL_MASTER);
-       RTE_LCORE_FOREACH_SLAVE(lcore_id) {
-               if (rte_eal_wait_lcore(lcore_id) < 0) {
-                       break;
-               }
-       }
-       return pd->rx_pkts;     
+       return pkt_cnt;

 }
 
 static int pcap_dpdk_inject(pcap_t *p, const void *buf _U_, int size _U_)
 {
        //not implemented yet
        pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
 }
 
 static int pcap_dpdk_inject(pcap_t *p, const void *buf _U_, int size _U_)
 {
        //not implemented yet
        pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,

-               errno, "dpdk error: Inject function has not be implemented yet");
+               errno, "dpdk error: Inject function has not been implemented yet");

        return PCAP_ERROR;
 }
 
        return PCAP_ERROR;
 }
 


@@ -332,22 +363,9 @@ static void pcap_dpdk_close(pcap_t *p)
        }
        rte_eth_dev_stop(pd->portid);
        rte_eth_dev_close(pd->portid);
        }
        rte_eth_dev_stop(pd->portid);
        rte_eth_dev_close(pd->portid);

-       // free pcap_dpdk?

        pcap_cleanup_live_common(p);
 } 
 
        pcap_cleanup_live_common(p);
 } 
 

-static int pcap_dpdk_setfilter(pcap_t *p, struct bpf_program *fp)
-{
-       //init bpf for dpdk, only support userspace bfp 
-       struct pcap_dpdk * pd = p->priv;
-       int ret=0;
-       ret = install_bpf_program(p, fp); 
-       if (ret==0){
-               pd->filter_in_userland = 1;
-       }
-       return ret;
-}
-

 static void nic_stats_display(struct pcap_dpdk *pd)
 {
        uint16_t portid = pd->portid;
 static void nic_stats_display(struct pcap_dpdk *pd)
 {
        uint16_t portid = pd->portid;


@@ -364,12 +382,12 @@ static int pcap_dpdk_stats(pcap_t *p, struct pcap_stat *ps)
        struct pcap_dpdk *pd = p->priv;
        calculate_timestamp(&(pd->ts_helper), &(pd->curr_ts));
        rte_eth_stats_get(pd->portid,&(pd->curr_stats));
        struct pcap_dpdk *pd = p->priv;
        calculate_timestamp(&(pd->ts_helper), &(pd->curr_ts));
        rte_eth_stats_get(pd->portid,&(pd->curr_stats));

-
-       ps->ps_recv = pd->curr_stats.ipackets;
-       ps->ps_drop = pd->curr_stats.ierrors;
-       ps->ps_drop += pd->bpf_drop;
-       ps->ps_ifdrop = pd->curr_stats.imissed;
-       
+       if (ps){
+               ps->ps_recv = pd->curr_stats.ipackets;
+               ps->ps_drop = pd->curr_stats.ierrors;
+               ps->ps_drop += pd->bpf_drop;
+               ps->ps_ifdrop = pd->curr_stats.imissed;
+       }

        uint64_t delta_pkt = pd->curr_stats.ipackets - pd->prev_stats.ipackets;
        struct timeval delta_tm;
        timersub(&(pd->curr_ts),&(pd->prev_ts), &delta_tm);
        uint64_t delta_pkt = pd->curr_stats.ipackets - pd->prev_stats.ipackets;
        struct timeval delta_tm;
        timersub(&(pd->curr_ts),&(pd->prev_ts), &delta_tm);


@@ -384,36 +402,21 @@ static int pcap_dpdk_stats(pcap_t *p, struct pcap_stat *ps)
        return 0;
 }
 
        return 0;
 }
 

-static int pcap_dpdk_setnonblock(pcap_t *p, int fd _U_){
-       pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
-               errno, "dpdk error: setnonblock not support");
+static int pcap_dpdk_setnonblock(pcap_t *p, int nonblock){
+       struct pcap_dpdk *pd = (struct pcap_dpdk*)(p->priv);
+       pd->nonblock = nonblock;

        return 0;
 }
 
 static int pcap_dpdk_getnonblock(pcap_t *p){
        return 0;
 }
 
 static int pcap_dpdk_getnonblock(pcap_t *p){

-       pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
-               errno, "dpdk error: getnonblock not support");
-       return 0;
+       struct pcap_dpdk *pd = (struct pcap_dpdk*)(p->priv);
+       return pd->nonblock;

 }
 }

-

 static int check_link_status(uint16_t portid, struct rte_eth_link *plink)
 {
 static int check_link_status(uint16_t portid, struct rte_eth_link *plink)
 {

-       uint8_t count = 0;
-       int is_port_up = 0;
-       int max_check_time = 50;
-       int check_interval = 100; // 100ms
-       for (count = 0; count <= max_check_time; count++) {
-               memset(plink, 0, sizeof(struct rte_eth_link));
-               rte_eth_link_get_nowait(portid, plink);
-               if (plink->link_status == ETH_LINK_UP)
-               {
-                       is_port_up = 1;
-                       break;
-               }else{
-                       rte_delay_ms(check_interval);
-               }
-       }
-       return is_port_up;
+       // wait up to 9 seconds to get link status
+       rte_eth_link_get(portid, plink);
+       return plink->link_status == ETH_LINK_UP;

 }
 static void eth_addr_str(struct ether_addr *addrp, char* mac_str, int len)
 {
 }
 static void eth_addr_str(struct ether_addr *addrp, char* mac_str, int len)
 {


@@ -458,6 +461,9 @@ static uint16_t portid_by_device(char * device)
                }
        }
        ret_ul = strtoul(&(device[prefix_len]), &pEnd, 10);
                }
        }
        ret_ul = strtoul(&(device[prefix_len]), &pEnd, 10);

+       if (pEnd == &(device[prefix_len]) || *pEnd != '\0'){
+               return ret;
+       }

        // too large for portid
        if (ret_ul >= DPDK_PORTID_MAX){ 
                return ret;
        // too large for portid
        if (ret_ul >= DPDK_PORTID_MAX){ 
                return ret;


@@ -477,11 +483,11 @@ static int parse_dpdk_cfg(char* dpdk_cfg,char** dargv)
        // find first non space char
        // The last opt is NULL
        for (i=0;dpdk_cfg[i] && cnt<DPDK_ARGC_MAX-1;i++){
        // find first non space char
        // The last opt is NULL
        for (i=0;dpdk_cfg[i] && cnt<DPDK_ARGC_MAX-1;i++){

-               if (skip_space && dpdk_cfg[i]!=0x20){ // not space
+               if (skip_space && dpdk_cfg[i]!=' '){ // not space

                        skip_space=!skip_space; // skip normal char
                        dargv[cnt++] = dpdk_cfg+i;
                }
                        skip_space=!skip_space; // skip normal char
                        dargv[cnt++] = dpdk_cfg+i;
                }

-               if (!skip_space && dpdk_cfg[i]==0x20){ // fint a space
+               if (!skip_space && dpdk_cfg[i]==' '){ // fint a space

                        dpdk_cfg[i]=0x00; // end of this opt
                        skip_space=!skip_space; // skip space char
                }
                        dpdk_cfg[i]=0x00; // end of this opt
                        skip_space=!skip_space; // skip space char
                }


@@ -491,18 +497,28 @@ static int parse_dpdk_cfg(char* dpdk_cfg,char** dargv)
 }
 
 // only called once
 }
 
 // only called once

-static int dpdk_pre_init()
+static int dpdk_pre_init(char * ebuf)

 {
        int dargv_cnt=0;
        char *dargv[DPDK_ARGC_MAX];
        char *ptr_dpdk_cfg = NULL;
 {
        int dargv_cnt=0;
        char *dargv[DPDK_ARGC_MAX];
        char *ptr_dpdk_cfg = NULL;

-       int ret = -1; //default is error
+       int ret = PCAP_ERROR; 

        // globale var
        if (is_dpdk_pre_inited)
        {
                // already inited
                return 0;
        }
        // globale var
        if (is_dpdk_pre_inited)
        {
                // already inited
                return 0;
        }

+       // check for root permission
+       if( geteuid() != 0)
+       {
+               RTE_LOG(ERR, USER1, "%s\n", DPDK_ERR_PERM_MSG);
+               pcap_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
+                           errno, "dpdk error: %s",
+                           DPDK_ERR_PERM_MSG);
+               ret = PCAP_ERROR_PERM_DENIED;
+               return ret;     
+       }

        // init EAL
        ptr_dpdk_cfg = getenv(DPDK_CFG_ENV_NAME);
        // set default log level to debug
        // init EAL
        ptr_dpdk_cfg = getenv(DPDK_CFG_ENV_NAME);
        // set default log level to debug


@@ -539,7 +555,7 @@ static int pcap_dpdk_activate(pcap_t *p)
        struct rte_eth_link link;
        do{
                //init EAL
        struct rte_eth_link link;
        do{
                //init EAL

-               ret = dpdk_pre_init();
+               ret = dpdk_pre_init(p->errbuf);

                if (ret < 0)
                {
                        pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
                if (ret < 0)
                {
                        pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,


@@ -572,19 +588,12 @@ static int pcap_dpdk_activate(pcap_t *p)
                        pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
                            errno, "dpdk error: portid is invalid. device %s",
                            p->opt.device);
                        pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
                            errno, "dpdk error: portid is invalid. device %s",
                            p->opt.device);

-                       ret = PCAP_ERROR;
+                       ret = PCAP_ERROR_NO_SUCH_DEVICE;

                        break;
                }
 
                        break;
                }
 

-               if (portid >= nb_ports)
-               {
-                       pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
-                           errno, "dpdk error: portid(%u) is larger than nb_ports(%u)",
-                           portid, nb_ports);
-                       ret = PCAP_ERROR;
-                       break;
-               }

                pd->portid = portid;
                pd->portid = portid;

+

                if (p->snapshot <= 0 || p->snapshot > MAXIMUM_SNAPLEN)
                {
                        p->snapshot = MAXIMUM_SNAPLEN;
                if (p->snapshot <= 0 || p->snapshot > MAXIMUM_SNAPLEN)
                {
                        p->snapshot = MAXIMUM_SNAPLEN;


@@ -682,15 +691,17 @@ static int pcap_dpdk_activate(pcap_t *p)
                        ret = PCAP_ERROR;
                        break;
                }
                        ret = PCAP_ERROR;
                        break;
                }

-               // set promisc mode
-               pd->must_clear_promisc=1;
-               rte_eth_promiscuous_enable(portid);
+               // set promiscuous mode
+               if (p->opt.promisc){
+                       pd->must_clear_promisc=1;
+                       rte_eth_promiscuous_enable(portid);
+               }

                // check link status
                is_port_up = check_link_status(portid, &link);
                if (!is_port_up){
                        pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
                            errno, "dpdk error: link is down, port=%u",portid);
                // check link status
                is_port_up = check_link_status(portid, &link);
                if (!is_port_up){
                        pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
                            errno, "dpdk error: link is down, port=%u",portid);

-                       ret = PCAP_ERROR;
+                       ret = PCAP_ERROR_IFACE_NOT_UP;

                        break;
                }
                // reset statistics
                        break;
                }
                // reset statistics


@@ -708,18 +719,23 @@ static int pcap_dpdk_activate(pcap_t *p)
                p->selectable_fd = p->fd;
                p->read_op = pcap_dpdk_dispatch;
                p->inject_op = pcap_dpdk_inject;
                p->selectable_fd = p->fd;
                p->read_op = pcap_dpdk_dispatch;
                p->inject_op = pcap_dpdk_inject;

-               p->setfilter_op = pcap_dpdk_setfilter;
+               // using pcap_filter currently, though DPDK provides their own BPF function. Because DPDK BPF needs load a ELF file as a filter.
+               p->setfilter_op = install_bpf_program;

                p->setdirection_op = NULL;
                p->set_datalink_op = NULL;
                p->getnonblock_op = pcap_dpdk_getnonblock;
                p->setnonblock_op = pcap_dpdk_setnonblock;
                p->stats_op = pcap_dpdk_stats;
                p->cleanup_op = pcap_dpdk_close;
                p->setdirection_op = NULL;
                p->set_datalink_op = NULL;
                p->getnonblock_op = pcap_dpdk_getnonblock;
                p->setnonblock_op = pcap_dpdk_setnonblock;
                p->stats_op = pcap_dpdk_stats;
                p->cleanup_op = pcap_dpdk_close;

-               p->breakloop_op = pcap_dpdk_breakloop;
+               p->breakloop_op = pcap_breakloop_common;
+               // set default timeout
+               pd->required_select_timeout.tv_sec = 0;
+               pd->required_select_timeout.tv_usec = DPDK_DEF_MIN_SLEEP_MS*1000;
+               p->required_select_timeout = &pd->required_select_timeout;

                ret = 0; // OK
        }while(0);
 
                ret = 0; // OK
        }while(0);
 

-       if (ret == PCAP_ERROR)
+       if (ret <= PCAP_ERROR) // all kinds of error code

        {
                pcap_cleanup_live_common(p);
        }else{
        {
                pcap_cleanup_live_common(p);
        }else{


@@ -754,14 +770,14 @@ pcap_t * pcap_dpdk_create(const char *device, char *ebuf, int *is_ours)
 int pcap_dpdk_findalldevs(pcap_if_list_t *devlistp, char *ebuf)
 {
        int ret=0;
 int pcap_dpdk_findalldevs(pcap_if_list_t *devlistp, char *ebuf)
 {
        int ret=0;

-       int nb_ports = 0;
+       unsigned int nb_ports = 0;

        char dpdk_name[DPDK_DEV_NAME_MAX];
        char dpdk_desc[DPDK_DEV_DESC_MAX];
        struct ether_addr eth_addr;
        char mac_addr[DPDK_MAC_ADDR_SIZE];
        char pci_addr[DPDK_PCI_ADDR_SIZE];
        do{
        char dpdk_name[DPDK_DEV_NAME_MAX];
        char dpdk_desc[DPDK_DEV_DESC_MAX];
        struct ether_addr eth_addr;
        char mac_addr[DPDK_MAC_ADDR_SIZE];
        char pci_addr[DPDK_PCI_ADDR_SIZE];
        do{

-               ret = dpdk_pre_init();
+               ret = dpdk_pre_init(ebuf);

                if (ret < 0)
                {
                        pcap_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,
                if (ret < 0)
                {
                        pcap_fmt_errmsg_for_errno(ebuf, PCAP_ERRBUF_SIZE,


@@ -777,17 +793,55 @@ int pcap_dpdk_findalldevs(pcap_if_list_t *devlistp, char *ebuf)
                        ret = PCAP_ERROR;
                        break;
                }
                        ret = PCAP_ERROR;
                        break;
                }

-               for (int i=0; i<nb_ports; i++){
-                       pcap_snprintf(dpdk_name,DPDK_DEV_NAME_MAX-1,"dpdk:%d",i);
+               for (unsigned int i=0; i<nb_ports; i++){
+                       pcap_snprintf(dpdk_name, DPDK_DEV_NAME_MAX-1,
+                           "%s%u", DPDK_PREFIX, i);

                        // mac addr 
                        rte_eth_macaddr_get(i, &eth_addr);
                        eth_addr_str(&eth_addr,mac_addr,DPDK_MAC_ADDR_SIZE);    
                        // PCI addr
                        rte_eth_dev_get_name_by_port(i,pci_addr);
                        pcap_snprintf(dpdk_desc,DPDK_DEV_DESC_MAX-1,"%s %s, MAC:%s, PCI:%s", DPDK_DESC, dpdk_name, mac_addr, pci_addr);
                        // mac addr 
                        rte_eth_macaddr_get(i, &eth_addr);
                        eth_addr_str(&eth_addr,mac_addr,DPDK_MAC_ADDR_SIZE);    
                        // PCI addr
                        rte_eth_dev_get_name_by_port(i,pci_addr);
                        pcap_snprintf(dpdk_desc,DPDK_DEV_DESC_MAX-1,"%s %s, MAC:%s, PCI:%s", DPDK_DESC, dpdk_name, mac_addr, pci_addr);

-                       // continue add all dev, even error happens
-                       add_dev(devlistp, dpdk_name, 0, dpdk_desc, ebuf);
+                       if (add_dev(devlistp, dpdk_name, 0, dpdk_desc, ebuf)==NULL){
+                               ret = PCAP_ERROR;
+                               break;
+                       }

                }       
        }while(0);      
        return ret;
 }
                }       
        }while(0);      
        return ret;
 }

+
+#ifdef DPDK_ONLY
+/*
+ * This libpcap build supports only DPDK, not regular network interfaces.
+ */
+
+/*
+ * There are no regular interfaces, just DPDK interfaces.
+ */
+int
+pcap_platform_finddevs(pcap_if_list_t *devlistp _U_, char *errbuf)
+{
+       return (0);
+}
+
+/*
+ * Attempts to open a regular interface fail.
+ */
+pcap_t *
+pcap_create_interface(const char *device, char *errbuf)
+{
+       pcap_snprintf(errbuf, PCAP_ERRBUF_SIZE,
+           "This version of libpcap only supports DPDK");
+       return NULL;
+}
+
+/*
+ * Libpcap version string.
+ */
+const char *
+pcap_lib_version(void)
+{
+       return (PCAP_VERSION_STRING " (DPDK-only)");
+}
+#endif