211. Vswitch sample test with vhost async data path test plan

211.1. Description

Vswitch sample can leverage IOAT to accelerate vhost async data-path from dpdk 20.11. This plan test vhost DMA operation callbacks for CBDMAV PMD and vhost async data-path in vhost sample. Also compare the performance between CPU copy, CBDMA copy and Sync copy. CPU copy means vhost enqueue w/o cbdma channel; CBDMA copy needs vhost enqueue with cbdma channel using parameter ‘-dmas’; Sync copy needs vhost enqueue with cbdma channel, but threshold ( can be adjusted by change value of f.async_threshold in dpdk code) is larger than forwarding packet length.

211.2. CBDMA performance indicator

(1)CPU copy vs. sync copy delta < 10% for 64B packet size (2)CBDMA copy vs sync copy delta > 5% for 1518 packet size

211.3. Prerequisites

Modify the testpmd code as following:

--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -29,7 +29,7 @@
 #include "main.h"

 #ifndef MAX_QUEUES
-#define MAX_QUEUES 128
+#define MAX_QUEUES 512
 #endif

 /* the maximum number of external ports supported */

211.4. Test Case1: PVP performance check with CBDMA channel using vhost async driver

  1. Adjust dpdk code as below:

    --- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -1343,7 +1343,7 @@ new_device(int vid)

if (async_vhost_driver) {
                f.async_inorder = 1;
-               f.async_threshold = 256;
+               f.async_threshold = 1518;
                return rte_vhost_async_channel_register(vid, VIRTIO_RXQ,
                        f.intval, &channel_ops);
}

  2. Bind physical port to vfio-pci and CBDMA channel to igb_uio.

  3. On host, launch dpdk-vhost by below command:

    ./dpdk-vhost -c 0x1c000000 -n 4 -- \
-p 0x1 --mergeable 1 --vm2vm 1 --dma-type ioat --stats 1 --socket-file /tmp/vhost-net --dmas [txd0@00:04.0]

  4. Launch virtio-user with testpmd:

    ./dpdk-testpmd -l 29-30 -n 4 --no-pci --file-prefix=testpmd0 \
--vdev=net_virtio_user0,mac=00:11:22:33:44:10,path=/tmp/vhost-net,queues=1 -- -i --rxq=1 --txq=1 --txd=1024 --rxd=1024 --nb-cores=1

  5. Start pkts from virtio-user side to let vswitch know the mac addr:

    testpmd>set fwd mac #if use io fwd ,ixia can't receive packets
testpmd>start tx_first
testpmd>stop
testpmd>start #start to forward

  6. Inject different length pkts (packets length=64,512,1024,1518) seperately with VLAN_id=1000 and dest_mac=virtio_mac_addresss ( specific in above cmd with 00:11:22:33:44:10) to NIC using packet generator, record performance number.

  7. Adjust vswitch code as below, re-test step 3-6 and record perfromance of different packet length:

    --- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -1343,7 +1343,7 @@ new_device(int vid)

if (async_vhost_driver) {
                f.async_inorder = 1;
-               f.async_threshold = 256;
+               f.async_threshold = 0;
                return rte_vhost_async_channel_register(vid, VIRTIO_RXQ,
                        f.intval, &channel_ops);
}

  8. Re-test step 3-6 except to launch dpdk-vhost by below command, record perfromance of different packet length:

    ./dpdk-vhost -c 0x1c000000 -n 4 -- -p 0x1 --mergeable 1 --vm2vm 1 --stats 1 --socket-file /tmp/vhost-net

  9. Compare performance, check below two performance indicator:

    (1)CPU copy vs. sync copy delta < 10% for 64B packet size
(2)CBDMA copy vs sync copy delta > 5% for 1518 packet size

211.5. Test Case2: PV test with multiple CBDMA channels using vhost async driver

  1. Bind one physical ports to vfio-pci and two CBDMA channels to igb_uio.

  2. On host, launch dpdk-vhost by below command:

    ./dpdk-vhost -l 26-28 -n 4 -- \
-p 0x1 --mergeable 1 --vm2vm 1 --dma-type ioat --stats 1 --socket-file /tmp/vhost-net0 --socket-file /tmp/vhost-net1 --dmas [txd0@00:04.0,txd1@00:04.1] --client

  3. launch two virtio-user ports:

    ./dpdk-testpmd -l 29-30 -n 4 --no-pci --file-prefix=testpmd0 \
--vdev=net_virtio_user0,mac=00:11:22:33:44:10,path=/tmp/vhost-net0,queues=1,server=1 -- -i --rxq=1 --txq=1 --txd=1024 --rxd=1024 --nb-cores=1

./dpdk-testpmd -l 31-32 -n 4 --no-pci --file-prefix=testpmd1 \
--vdev=net_virtio_user0,mac=00:11:22:33:44:11,path=/tmp/vhost-net1,queues=1,server=1 -- -i --rxq=1 --txq=1 --txd=1024 --rxd=1024 --nb-cores=1

  4. Start pkts from two virtio-user side individually to let vswitch know the mac addr:

    testpmd0>start tx_first
testpmd1>start tx_first

  5. Inject 1518B packets with VLAN_id=1000 and dest_mac=virtio_mac_addresss (00:11:22:33:44:10 and 00:11:22:33:44:11) to NIC using packet generator,check two virtio-user ports can receive according packets.

  6. Stop dpdk-vhost side and relaunch it with same cmd as step2.

7.Start pkts from two virtio-user side individually to let vswitch know the mac addr, check two virtio-user ports still can receive according packets and get same throughput as step5:

testpmd0>stop
testpmd0>start tx_first
testpmd1>stop
testpmd1>start tx_first

211.6. Test Case3: VM2VM performance test with two CBDMA channels using vhost async driver

1.Bind one physical ports to vfio-pci and two CBDMA channels to igb_uio.

  1. On host, launch dpdk-vhost by below command:

    ./dpdk-vhost -l 26-28 -n 4 -- -p 0x1 --mergeable 1 --vm2vm 1 --dma-type ioat \
--socket-file /tmp/vhost-net0 --socket-file /tmp/vhost-net1 --dmas [txd0@00:04.0,txd1@00:04.1]

  2. Launch virtio-user:

    ./dpdk-testpmd -l 29-30 -n 4 --no-pci --file-prefix=testpmd0 \
--vdev=net_virtio_user0,mac=00:11:22:33:44:10,path=/tmp/vhost-net0,queues=1 -- -i --rxq=1 --txq=1 --txd=1024 --rxd=1024 --nb-cores=1

./dpdk-testpmd -l 31-32 -n 4 --no-pci --file-prefix=testpmd1 \
--vdev=net_virtio_user0,mac=00:11:22:33:44:11,path=/tmp/vhost-net0,queues=1 -- -i --rxq=1 --txq=1 --txd=1024 --rxd=1024 --nb-cores=1

  3. Start pkts from two virtio-user sides, record performance number with txpkts=256 and 2000 from testpmd1 seperately:

    testpmd0>set fwd mac
testpmd0>start tx_first
testpmd0>stop
testpmd0>set eth-peer 0 00:11:22:33:44:11
testpmd0>start

testpmd1>set fwd mac
testpmd1>set eth-peer 0 00:11:22:33:44:10
testpmd1>set txpkts 2000
testpmd1>start tx_first
testpmd1>show port stats all
testpmd1>stop
testpmd1>set txpkts 256
testpmd1>start tx_first
testpmd1>show port stats all

  4. Re-test step 2-4 except to launch dpdk-vhost by below command, record perfromance:

    ./dpdk-vhost -l 27-28 -n 4 -- -p 0x1 --mergeable 1 --vm2vm 1 --socket-file /tmp/vhost-net0 --socket-file /tmp/vhost-net1

  5. Compare perfromance number, when txpkts=2000, first test has better performance; while txpkts=256, the second test has better performance.

211.7. Test Case4: VM2VM test with 2 vhost device using vhost async driver

  1. Bind one physical ports to vfio-pci and two CBDMA channels to igb_uio.

  2. On host, launch dpdk-vhost by below command:

    ./dpdk-vhost -l 26-28 -n 4 -- -p 0x1 --mergeable 1 --vm2vm 1 --dma-type ioat \
--socket-file /tmp/vhost-net0 --socket-file /tmp/vhost-net1 --dmas [txd0@00:04.0,txd1@00:04.1]

  3. Start VM0:

    /home/qemu-install/qemu-4.2.1/bin/qemu-system-x86_64 -name vm1 -enable-kvm -cpu host -smp 4 -m 4096 \
-object memory-backend-file,id=mem,size=4096M,mem-path=/mnt/huge,share=on \
-numa node,memdev=mem -mem-prealloc -drive file=/home/osimg/ubuntu20-04.img  \
-chardev socket,path=/tmp/vm2_qga0.sock,server,nowait,id=vm2_qga0 -device virtio-serial \
-device virtserialport,chardev=vm2_qga0,name=org.qemu.guest_agent.2 -daemonize \
-monitor unix:/tmp/vm2_monitor.sock,server,nowait -device e1000,netdev=nttsip1 \
-netdev user,id=nttsip1,hostfwd=tcp:127.0.0.1:6002-:22 \
-chardev socket,id=char0,path=/tmp/vhost-net0 \
-netdev type=vhost-user,id=netdev0,chardev=char0,vhostforce \
-device virtio-net-pci,netdev=netdev0,mac=52:54:00:00:00:01,disable-modern=true,mrg_rxbuf=off,csum=on,guest_csum=on,host_tso4=on,guest_tso4=on,guest_ecn=on -vnc :10

  4. Start VM1:

    /home/qemu-install/qemu-4.2.1/bin/qemu-system-x86_64 -name vm2 -enable-kvm -cpu host -smp 4 -m 4096 \
-object memory-backend-file,id=mem,size=4096M,mem-path=/mnt/huge,share=on \
-numa node,memdev=mem -mem-prealloc -drive file=/home/osimg/ubuntu20-04-2.img  \
-chardev socket,path=/tmp/vm2_qga0.sock,server,nowait,id=vm2_qga0 -device virtio-serial \
-device virtserialport,chardev=vm2_qga0,name=org.qemu.guest_agent.2 -daemonize \
-monitor unix:/tmp/vm2_monitor.sock,server,nowait -device e1000,netdev=nttsip1 \
-netdev user,id=nttsip1,hostfwd=tcp:10.67.119.61:6003-:22 \
-chardev socket,id=char0,path=/tmp/vhost-net1 \
-netdev type=vhost-user,id=netdev0,chardev=char0,vhostforce \
-device virtio-net-pci,netdev=netdev0,mac=52:54:00:00:00:02,disable-modern=true,mrg_rxbuf=off,csum=on,guest_csum=on,host_tso4=on,guest_tso4=on,guest_ecn=on -vnc :12

  5. Bind virtio port to vfio-pci in both two VMs:

    modprobe vfio enable_unsafe_noiommu_mode=1
modprobe vfio-pci
echo 1 > /sys/module/vfio/parameters/enable_unsafe_noiommu_mode
./usertools/dpdk-devbind.py --bind=vfio-pci 00:05.0

  6. Start testpmd in VMs seperately:

    ./dpdk-testpmd -l 1-2 -n 4 -- -i --rxq=1 --txq=1 --nb-cores=1 --txd=1024 --rxd=1024

  7. Start pkts from two virtio-pmd, record performance number with txpkts=256 and 2000 from testpmd1 seperately:

    testpmd0>set fwd mac
testpmd0>start tx_first
testpmd0>stop
testpmd0>set eth-peer 0 52:54:00:00:00:02
testpmd0>start

testpmd1>set fwd mac
testpmd1>set eth-peer 0 52:54:00:00:00:01
testpmd1>set txpkts 2000
testpmd1>start tx_first
testpmd1>show port stats all
testpmd1>stop
testpmd1>set txpkts 256
testpmd1>start tx_first
testpmd1>show port stats all

  8. Inject traffic with VLAN_id=1000 and dest_mac=virtio_mac_addresss (52:54:00:00:00:02 and 52:54:00:00:00:02) to NIC using packet generator, check two virtio-pmd can receive according packets.

  9. Quit two testpmd in two VMs, bind virtio-pmd port to virtio-pci,then bind port back to vfio-pci, rerun below cmd 50 times:

    ./usertools/dpdk-devbind.py -u 00:05.0
./usertools/dpdk-devbind.py --bind=virtio-pci 00:05.0
./usertools/dpdk-devbind.py --bind=vfio-pci 00:05.0

  10. Rerun step 6-8，check vhost can stable work.