Documentation/vhost-user-ovs-dpdk

From QEMU
Revision as of 09:17, 18 July 2016 by Mcoquelin (talk | contribs) (Update to align with DPDK v16.07, tested with Fedora 24 host and guests)

Vhost-user with OVS/DPDK as backend

The purpose of this document is to describe the steps needed to setup a development machine for vhost-user testing. All the information here is taken from various sources available online.

The goal is to connect guests' virtio-net devices having vhost-user backend to OVS dpdkvhostuser ports and be able to run any kind of network traffic between them.

Prerequisites

  • Ensure you have hugepages enabled in kernel command line:
 default_hugepagesz=2M hugepagesz=2M hugepages=2048
  • Install kernel development packages:
yum install kernel-devel

Build from sources

DPDK build

  • Get the sources
git clone git://dpdk.org/dpdk
cd dpdk
  • Edit the configuration file:
diff --git a/config/common_base b/config/common_base
index 7830535..91b2d29 100644
--- a/config/common_base
+++ b/config/common_base
@@ -67,7 +67,7 @@ CONFIG_RTE_ARCH_STRICT_ALIGN=n
 #
 # Compile to share library
 #
-CONFIG_RTE_BUILD_SHARED_LIB=n
+CONFIG_RTE_BUILD_SHARED_LIB=y
 
 #
 # Use newest code breaking previous ABI
  • Build it
export RTE_SDK=<dpdk-dir>
export RTE_TARGET=x86_64-native-linuxapp-gcc
make -j<cpu num> install T=$RTE_TARGET DESTDIR=install EXTRA_CFLAGS='-g'
cp -f install/lib/lib* /lib64/
export DPDK_BUILD_DIR=x86_64-native-linuxapp-gcc    #sometimes is "build", depending on your environment and make command parameters

OVS build

  • Get the sources:
 git clone https://github.com/openvswitch/ovs
 cd ovs
  • With DPDK 16.07, you will need an out-of-tree patch as of 07/18/2016:
 wget https://patchwork.ozlabs.org/patch/647253/mbox/ -O dpdk-16.07-api-update.mbox
 git am -3 dpdk-16.07-api-update.mbox
  • Build it:
./boot.sh
./configure --prefix=/usr --sysconfdir=/etc --localstatedir=/var --with-dpdk=$RTE_SDK/$DPDK_BUILD_DIR  --disable-ssl --with-debug  CFLAGS='-g'
make install -j <cpu num>

QEMU build

  • Get the sources:
git clone git://git.qemu.org/qemu.git
cd qemu
  • Build it:
mkdir bin
cd bin
../configure --target-list=x86_64-softmmu --enable-debug --extra-cflags='-g'
make -j <cpu num>

Setup

System setup

  • Prepare directories
mkdir -p /var/run/openvswitch
mount -t hugetlbfs -o pagesize=2048k none /dev/hugepages
  • Load kernel modules
modprobe openvswitch

OVS setup

  • Clean the environment
killall ovsdb-server ovs-vswitchd
rm -f /var/run/openvswitch/vhost-user*
rm -f /etc/openvswitch/conf.db
  • Start database server
export DB_SOCK=/var/run/openvswitch/db.sock
ovsdb-tool create /etc/openvswitch/conf.db /usr/share/openvswitch/vswitch.ovsschema
ovsdb-server --remote=punix:$DB_SOCK --remote=db:Open_vSwitch,Open_vSwitch,manager_options --pidfile --detach
  • Start OVS
ovs-vsctl --no-wait init
ovs-vsctl --no-wait set Open_vSwitch . other_config:dpdk-lcore-mask=0xf
ovs-vsctl --no-wait set Open_vSwitch . other_config:dpdk-socket-mem=1024
ovs-vsctl --no-wait set Open_vSwitch . other_config:dpdk-init=true
ovs-vswitchd unix:$DB_SOCK --pidfile --detach --log-file=/var/log/openvswitch/ovs-vswitchd.log
  • Configure the bridge
ovs-vsctl add-br ovsbr0 -- set bridge ovsbr0 datapath_type=netdev
ovs-vsctl add-port ovsbr0 vhost-user1 -- set Interface vhost-user1 type=dpdkvhostuser
ovs-vsctl add-port ovsbr0 vhost-user2 -- set Interface vhost-user2 type=dpdkvhostuser

QEMU setup

  • Start 'n' VMs
cd <qemu-dir>/bin/x86_64-softmmu/
qemu-system-x86_64 -enable-kvm -m 1024 -smp 2 \
    -chardev socket,id=char0,path=/var/run/openvswitch/vhost-user<n> \
    -netdev type=vhost-user,id=mynet1,chardev=char0,vhostforce \
    -device virtio-net-pci,netdev=mynet1,mac=52:54:00:02:d9:0<n> \
    -object memory-backend-file,id=mem,size=1024M,mem-path=/dev/hugepages,share=on \
    -numa node,memdev=mem -mem-prealloc \
    -net user,hostfwd=tcp::1002<n>-:22 -net nic \
    /path/to/img
  • VM setup
 ssh root@localhost -p 1002<n>
 ifconfig <eth interface> 192.168.100.<n>

Enabling multi-queue

Enable multi-queue

  • OVS setup
In OVS 2.5.0 or older version: (It only supports setting up same number of rx queues for all of PMD netdevs.)
    ovs-vsctl set Open_vSwitch . other_config:n-dpdk-rxqs=<queues_nr, the same as QEMU>
    ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=<cpu mask for rx queues, say 0xff00>
In post OVS 2.5.0:
    ovs-vsctl set Interface vhost-user1 options:n_rxq=<queues_nr, the same as QEMU>
    ovs-vsctl set Interface vhost-user2 options:n_rxq=<queues_nr, the same as QEMU>
    ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=<cpu mask for rx queues, say 0xff00>
  • QEMU command line modification
-netdev type=vhost-user,id=mynet1,chardev=char0,vhostforce,queues=<queues_nr>
-device virtio-net-pci,netdev=mynet1,mac=52:54:00:02:d9:0<n>,mq=on,vectors=<2 + 2 * queues_nr>
  • In VM
ethtool -L eth0 combined <queues_nr>

Notes

  • I want to keep it as simple as possible. If you see steps that can be skipped or unneeded configuration feel free to update the document.