152. vhost async data-path robust with cbdma test plan

152.1. Description

This document provides the test plan for testing Vhost asynchronous data path robust with CBDMA driver.

CBDMA is a kind of DMA engine, Vhost asynchronous data path leverages DMA devices to offload memory copies from the CPU and it is implemented in an asynchronous way. As a result, large packet copy can be accelerated by the DMA engine, and vhost can free CPU cycles for higher level functions.

Asynchronous data path is enabled per tx/rx queue, and users need to specify the DMA device used by the tx/rx queue. Each tx/rx queue only supports to use one DMA device, but one DMA device can be shared among multiple tx/rx queues of different vhostpmd ports.

Two PMD parameters are added: - dmas: specify the used DMA device for a tx/rx queue (Default: no queues enable asynchronous data path) - dma-ring-size: DMA ring size. (Default: 4096).

Here is an example: –vdev ‘eth_vhost0,iface=./s0,dmas=[txq0@0000:00.01.0;rxq0@0000:00.01.1],dma-ring-size=4096’

152.2. Test case

152.2.1. Common steps

  1. Bind 1 NIC port and CBDMA devices to vfio-pci:

    <dpdk dir># ./usertools/dpdk-devbind.py -b vfio-pci <DUT port pci device id>
    <dpdk dir># ./usertools/dpdk-devbind.py -b vfio-pci <DUT port DMA device id>
    
    For example, Bind 1 NIC port and 2 CBDMA devices::
    ./usertools/dpdk-devbind.py -b vfio-pci 0000:00:18.0
    ./usertools/dpdk-devbind.py -b vfio-pci 0000:00:04.0,0000:00:04.1
    
  2. Send imix packets [64,1518] to NIC by traffic generator:

    The TCP imix packets include packet size [64, 128, 256, 512, 1024, 1518], and the format of packet is as follows.
    +-------------+-------------+-------------+-------------+
    | MAC         | MAC         | IPV4        | IPV4        |
    | Src address | Dst address | Src address | Dst address |
    |-------------|-------------|-------------|-------------|
    | Random MAC  | Virtio mac  | Random IP   | Random IP   |
    +-------------+-------------+-------------+-------------+
    All the packets in this test plan use the Virtio mac: 00:11:22:33:44:10.
    

152.2.2. Test Case 1: PVP virtio-user quit test

This case is designed to test if virtio-user can quit normally regardless of whether the back-end stop sending packets.

  1. Bind 1 NIC port and 1 CBDMA devices to vfio-pci as common step 1.

  2. Launch vhost by below command:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 10-18 --file-prefix=vhost \
    -a 0000:18:00.1 -a 0000:00:04.0 \
    --vdev 'net_vhost0,iface=./vhost_net0,queues=1,dmas=[txq0@0000:00:04.0;rxq0@0000:00:04.0]' \
    --iova=va -- -i --nb-cores=1 --txq=1 --rxq=1 --txd=1024 --rxd=1024
    testpmd> set fwd mac
    testpmd> start
    
  3. Launch virtio-user with inorder mergeable path:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 2-6 --no-pci --file-prefix=virtio \
    --vdev=net_virtio_user0,mac=00:11:22:33:44:10,path=./vhost_net0,mrg_rxbuf=1,in_order=1,queues=1 \
    -- -i --nb-cores=1 --txq=1 --rxq=1 --txd=1024 --rxd=1024
    testpmd> set fwd csum
    testpmd> start
    
  4. Send TCP imix packets [64,1518] from packet generator as common step2.

  5. Quit virtio-user and relaunch virtio-user as step 3 while sending packets from packet generator.

  6. Stop vhost port, then quit virtio-user and reluanch virtio-user as step 3 while sending packets from packet generator.

  7. Stop sending packets from packet generator, then quit virtio-user and vhost.

152.2.3. Test Case 2: PVP vhost-user quit test

This case is designed to test if vhost-user can quit normally regardless of whether the back-end stop sending packets.

  1. Bind 1 NIC port and 1 CBDMA devices to vfio-pci as common step 1.

  2. Launch vhost by below command:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 10-18 --file-prefix=vhost \
    -a 0000:18:00.1 -a 0000:00:04.0 \
    --vdev 'net_vhost0,iface=./vhost_net0,queues=1,client=1,dmas=[txq0@0000:00:04.0;rxq0@0000:00:04.0]' \
    --iova=va -- -i --nb-cores=1 --txq=1 --rxq=1 --txd=1024 --rxd=1024
    testpmd> set fwd mac
    testpmd> start
    
  3. Launch virtio-user with inorder mergeable path:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 2-6 --no-pci --file-prefix=virtio \
    --vdev=net_virtio_user0,mac=00:11:22:33:44:10,path=./vhost_net0,mrg_rxbuf=1,in_order=1,queues=1,server=1 \
    -- -i --nb-cores=1 --txq=1 --rxq=1 --txd=1024 --rxd=1024
    testpmd> set fwd csum
    testpmd> start
    
  4. Send TCP imix packets [64,1518] from packet generator as common step2.

  5. Quit vhost-user and relaunch vhost-user as step 2 while sending packets from packet generator.

  6. Stop sending packets from packet generator, then quit vhost-user and virtio-user.

152.2.4. Test Case 3: PVP vhost async test with redundant device parameters

This case is designed to test if vhostpmd can work normally when binding and using redundant device parameters.

  1. Bind 1 NIC port and 4 CBDMA devices to vfio-pci as common step 1.

  2. Launch vhost by below command:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 10-18 --file-prefix=vhost \
    -a 0000:18:00.1 -a 0000:00:04.0 -a 0000:00:04.1 -a 0000:00:04.2 -a 0000:00:04.3 \
    --vdev 'net_vhost0,iface=./vhost_net0,queues=1,client=1,dmas=[txq0@0000:00:04.1;rxq0@0000:00:04.1]' \
    --iova=va -- -i --nb-cores=1 --txq=1 --rxq=1 --txd=1024 --rxd=1024
    testpmd> set fwd mac
    testpmd> start
    
  3. Launch virtio-user with inorder mergeable path:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 2-6 --no-pci --file-prefix=virtio \
    --vdev=net_virtio_user0,mac=00:11:22:33:44:10,path=./vhost_net0,mrg_rxbuf=1,in_order=1,queues=1,server=1 \
    -- -i --nb-cores=1 --txq=1 --rxq=1 --txd=1024 --rxd=1024
    testpmd> set fwd csum
    testpmd> start
    
  4. Send imix packets [64,1518] from packet generator as common step2, check the throughput.

152.2.5. Test Case 4: Loopback vhost async test with each queue using 2 DMA devices

Since each tx/rx queue only supports to use one DMA device, this case is designed to test if vhostpmd can work normally when each queue using 2 DMA devices.

  1. Bind 3 CBDMA devices to vfio-pci as common step 1.

  2. Launch vhost by below command:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 10-18 --file-prefix=vhost \
    -a 0000:00:04.0 -a 0000:00:04.1 -a 0000:00:04.2 \
    --vdev 'net_vhost0,iface=./vhost_net0,queues=2,client=1,dmas=[txq0@0000:00:04.0;txq0@0000:00:04.1;rxq0@0000:00:04.1;rxq0@0000:00:04.2]' \
    --iova=va -- -i --nb-cores=1 --txq=2 --rxq=2 --txd=1024 --rxd=1024
    testpmd> set fwd mac
    
  3. Launch virtio-user with inorder mergeable path:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 2-6 --no-pci --file-prefix=virtio \
    --vdev=net_virtio_user0,mac=00:11:22:33:44:10,path=./vhost_net0,mrg_rxbuf=1,in_order=1,queues=2,server=1 \
    -- -i --nb-cores=1 --txq=2 --rxq=2 --txd=1024 --rxd=1024
    testpmd> set fwd csum
    testpmd> start
    
  4. Send packets from vhost-user testpmd, check the throughput:

    testpmd>set txpkts 1024
    testpmd>start tx_first 32
    testpmd>show port stats all
    

152.2.6. Test Case 5: Loopback vhost async test with dmas parameters out of order

This case is designed to test if vhostpmd can work normally when dmas parameters out of order.

  1. Bind 2 CBDMA devices to vfio-pci as common step 1.

  2. Launch vhost by below command:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 10-18 --file-prefix=vhost \
    -a 0000:00:04.0 -a 0000:00:04.1 \
    --vdev 'net_vhost0,iface=./vhost_net0,queues=4,client=1,dmas=[rxq3@0000:00:04.1;txq0@0000:00:04.0;rxq1@0000:00:04.0;txq2@0000:00:04.1]' \
    --iova=va -- -i --nb-cores=1 --txq=4 --rxq=4 --txd=1024 --rxd=1024
    testpmd> set fwd mac
    
  3. Launch virtio-user with inorder mergeable path:

    <dpdk dir># ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -n 4 -l 2-6 --no-pci --file-prefix=virtio \
    --vdev=net_virtio_user0,mac=00:11:22:33:44:10,path=./vhost_net0,mrg_rxbuf=1,in_order=1,queues=4,server=1 \
    -- -i --nb-cores=1 --txq=4 --rxq=4 --txd=1024 --rxd=1024
    testpmd> set fwd csum
    testpmd> start
    
  4. Send packets from vhost-user testpmd, check the throughput:

    testpmd>set txpkts 1024
    testpmd>start tx_first 32
    testpmd>show port stats all
    

152.2.7. Test Case 6: VM2VM split and packed ring mergeable path with cbdma enable and server mode

This case tests split and packed ring with cbdma can work normally when the front-end change from virtio-net to virtio-pmd.

  1. Bind 16 CBDMA channels to vfio-pci, as common step 1.

  2. Launch the testpmd with 2 vhost ports below commands:

    ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -l 1-5 -n 4 --file-prefix=vhost \
    -a 0000:00:04.0 -a 0000:00:04.1 -a 0000:00:04.2 -a 0000:00:04.3 -a 0000:00:04.4 -a 0000:00:04.5 -a 0000:00:04.6 -a 0000:00:04.7 \
    -a 0000:80:04.0 -a 0000:80:04.1 -a 0000:80:04.2 -a 0000:80:04.3 -a 0000:80:04.4 -a 0000:80:04.5 -a 0000:80:04.6 -a 0000:80:04.7 \
    --vdev 'net_vhost0,iface=vhost-net0,client=1,queues=8,dmas=[txq0@0000:00:04.0;txq1@0000:00:04.1;txq2@0000:00:04.2;txq3@0000:00:04.3;txq4@0000:00:04.4;txq5@0000:00:04.1;rxq2@0000:00:04.2;rxq3@0000:00:04.3;rxq4@0000:00:04.4;rxq5@0000:00:04.5;rxq6@0000:00:04.6;rxq7@0000:00:04.7]' \
    --vdev 'net_vhost1,iface=vhost-net1,client=1,queues=8,dmas=[txq0@0000:80:04.0;txq1@0000:80:04.1;txq2@0000:80:04.2;txq3@0000:80:04.3;txq4@0000:80:04.4;txq5@0000:80:04.1;rxq2@0000:80:04.2;rxq3@0000:80:04.3;rxq4@0000:80:04.4;rxq5@0000:80:04.5;rxq6@0000:80:04.6;rxq7@0000:80:04.7]' \
    -- -i --nb-cores=4 --txd=1024 --rxd=1024 --rxq=8 --txq=8
    testpmd> start
    
  3. Launch VM1 and VM2:

    taskset -c 6-16 qemu-system-x86_64 -name vm1 -enable-kvm -cpu host -smp 9 -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/vm1_qga0.sock,server,nowait,id=vm1_qga0 -device virtio-serial \
    -device virtserialport,chardev=vm1_qga0,name=org.qemu.guest_agent.2 -daemonize \
    -monitor unix:/tmp/vm1_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=./vhost-net0,server \
    -netdev type=vhost-user,id=netdev0,chardev=char0,vhostforce,queues=8 \
    -device virtio-net-pci,netdev=netdev0,mac=52:54:00:00:00:01,disable-modern=false,mrg_rxbuf=off,mq=on,vectors=40,csum=on,guest_csum=on,host_tso4=on,guest_tso4=on,guest_ecn=on,guest_ufo=on,host_ufo=on -vnc :10
    
    taskset -c 17-27 qemu-system-x86_64 -name vm2 -enable-kvm -cpu host -smp 9 -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:127.0.0.1:6003-:22 \
    -chardev socket,id=char0,path=./vhost-net1,server \
    -netdev type=vhost-user,id=netdev0,chardev=char0,vhostforce,queues=8 \
    -device virtio-net-pci,netdev=netdev0,mac=52:54:00:00:00:02,disable-modern=false,mrg_rxbuf=off,mq=on,vectors=40,csum=on,guest_csum=on,host_tso4=on,guest_tso4=on,guest_ecn=on,guest_ufo=on,host_ufo=on,packed=on -vnc :12
    
  4. On VM1, set virtio device IP and run arp protocal:

    ethtool -L ens5 combined 8
    ifconfig ens5 1.1.1.2
    arp -s 1.1.1.8 52:54:00:00:00:02
    
  5. On VM2, set virtio device IP and run arp protocal:

    ethtool -L ens5 combined 8
    ifconfig ens5 1.1.1.8
    arp -s 1.1.1.2 52:54:00:00:00:01
    
  6. Scp 1MB file form VM1 to VM2:

    Under VM1, run: `scp <xxx> root@1.1.1.8:/`   <xxx> is the file name
    
  7. Check the iperf performance between two VMs by below commands:

    Under VM1, run: `iperf -s -i 1`
    Under VM2, run: `iperf -c 1.1.1.2 -i 1 -t 60`
    
  8. On VM1 and VM2, bind virtio device with vfio-pci driver:

    modprobe vfio
    modprobe vfio-pci
    echo 1 > /sys/module/vfio/parameters/enable_unsafe_noiommu_mode
    ./usertools/dpdk-devbind.py --force --bind=vfio-pci 0000:00:05.0
    
  9. Launch testpmd in VM1:

    ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -c 0x3 -n 4 -- -i --tx-offloads=0x00 --enable-hw-vlan-strip --txq=8 --rxq=8 --txd=1024 --rxd=1024
    testpmd> set mac fwd
    testpmd> start
    
  10. Launch testpmd in VM2 and send imix pkts, check imix packets can looped between two VMs for 1 mins:

    ./x86_64-native-linuxapp-gcc/app/dpdk-testpmd -c 0x3 -n 4 -- -i --tx-offloads=0x00 --enable-hw-vlan-strip --txq=8 --rxq=8 --txd=1024 --rxd=1024
    testpmd> set mac fwd
    testpmd> set txpkts 64,256,512
    testpmd> start tx_first 32
    testpmd> show port stats all
    
  11. Rerun step 4-10.