not use mcopy when transmit from bsd to dpdk
recommit
ff_traffic.tx info should save the sent packetssend_burst() send several pkts which maybe composed by several mbufs, use rte_pktmbuf_pkt_len() to get actual length.
F-Stack's kni can work on DPDK 18.11 LTS now.
modify packet_dispatcher to support response package direct.
Add support for getsockopt TCP_INFO
Fix the IP/TCP/UDP csum error of txmode offloads, refer #317.
Revert "DPDK:upgrade to 18.11.0 LTS."This reverts commit 8850115bd87ac2e168ca2ab33159528818d45fe8.
Format code.
DPDK:upgrade to 18.11.0 LTS.
CONTAINER(DOCKER): support aggregation model while running in container with OVS-DPDK, refer to #298.
Setting FF_DPDK with MACHINE_CPUARCH
Update Makefile for reserving FF_DPDK for x86The former patch, which add compiler option for arm64, removes the FF_DPDK for x86. That's a mistake.
Add arch arm64 compiler options
fix static building path of FF_DPDK
add `base_virtaddr` config.
fix for issue 261
add "idle_sleep" to reduce CPU usage when no pkts incomming.
fix a traffic bug
add traffic tool.
F-Stack: disable kni module by default.Some omissions.
F-Stack: disable kni module by default.This commit adds a macro `FF_KNI` to determine whether to enable kni.In Makefile, uncomment the `FF_KNI=1`, kni will be compiled into f-stack.
Config: if proc-id is not configured, use 0.
F-stack:fix bug, if ff_veth_softc malloc faild, memory should no be read.
make use of numa when setting up rx/tx queues
12345678