pkts | Pure java based pcap library capable of reading and writing | Learning library

It appears I'm just a bit stupid and really not used to webserver configuration, I missed the configuration of my server, it fell on the default nginx server, which is sadly only listening on port 80 on IPv4. Just another dumb mistake, warm thanks for @Petros, who set me om the right tracks.

Looks like it has either DPDK or NUMA backed page issue. The same is working with DPDK version 19.11 LTS and 20.11 LTS.

• DPDK application: rm /tmp/sock0; sudo ./build/l2fwd --legacy-mem -l 1-2 --no-pci --vdev=net_vhost0,iface=/tmp/sock0 --vdev=net_tap0 -m 1024 -- -p 3 -T 1 --no-mac-updating

• QEMU: taskset -c 4-9 qemu-system-x86_64 -cpu host -enable-kvm -m 1024 -smp 4,sockets=1,cores=4,threads=1 \ -object memory-backend-file,id=mem,size=1024M,mem-path=/mnt/huge,share=on \ -numa node,memdev=mem,nodeid=0 -mem-prealloc \ -name test \ -no-reboot \ -vnc none \ -nographic \ -net user,hostfwd=tcp::10023-:22 -net nic \ -chardev socket,id=charnet0,path=/tmp/sock0 \ -netdev type=vhost-user,chardev=charnet0,queues=1,id=hostnet0 \ -device virtio-net-pci,mq=on,vectors=18,netdev=hostnet0,id=net0,mac=fa:16:3e:52:30:73 \ -hda [disk name]

Once the VM is booted, you can login via ssh port 10023.

You can use load in order to access Raw bytes:

The reason for the error is because of the check for virtio PMD in function virtio_dev_configure file [dpdk root folder]drivers/net/virtio/virtio_ethdev.c. This can be due to the current Fstack enables RSS for better flow distribution over its port-queue pair.

There 2 possible solution to fix the problem is to

1. find the configuration parameter in f-stack.conf to disable RSS or
2. change the FSTACK port configuration logic not to use RSS (by editing code).

File: lib/ff_dpdk_if.c edit: line 627 from port_conf.rxmode.mq_mode = ETH_MQ_RX_RSS; to port_conf.rxmode.mq_mode = ETH_MQ_RX_NONE; and rebuild the fstack

Note: if you use physical NIC RSS is supported in most of cases. hence there will be no error there.

If the goal is to have packet transfer between Pktgen and testpmd that is connected by OVS-DPDK one has to use net_vhost and virtio_user pair.

DPDK API rte_ipv4_fragment_packet is used under both testpmd and DPDK example ip_fragementation. These are also included under the DPDK test suite which is run for each release too.

Based on the internal test and proper use of API for example Ip_fragementation the issue is not been able to reproduce. Hence the API leaking memory pool buffers are highly unlikely other than some special corner case (which is yet to be found).

Based on the code snippet analysis following could be the cause of mempool exhaust

1. fail to free direct buffer after fragmentation
2. fail to free one or more fragments from the indirect buffer when tx_burst fails.

[EDIT-1] based on the email update and comments, there is indeed no problem with DPDK API rte_ipv4_fragment_packet. The issue is from Application logic, with new behaviour as

1. DPDK BOND PMD leads to mempool exhaustion with current code snippet
2. DPDK BOND PMD has no issues with DPDK example with rte_ipv4_fragment_packet
3. DPDK i40e PMD has an issue with the current code snippet
4. DPDK i40e PMD has no issue with dpdk example rte_ipv4_fragment_packet

Hence the issue is with sample code snippet and usage and not DPDK API.

I don't think there's a way to do this. I also misread the output. Docker daemon was forwarding to port 53.

read udp 172.20.0.127:37928->10.99.0.1:53: i/o timeout

So turns out issues was indeed iptables. The default policy for the INPUT chain in this case is DROP. Since the container is now trying to reach the host system itself, its traffic will come through the input chain.

So I added 2 simple rules for now, that will allow all traffic in the INPUT chain from docker0 and docker_gwbridge.

There are multiple factors which affect performance for NIC PMD. Some of them are listed below

1. cpu core isolation to explicitly make user-space threads to sole user of CPU core time
2. Kernel watchdog timer callback reduction
3. disable Transparent Huge page (especially with 1GB)
4. firmware of NIC
5. DPDK version
6. vector code for RX-TX
7. PCIe lane (direct attach to CPU give higher performance than south bridge)
8. CPU clock frequency
9. DDIO ability of NIC
10. Traffic pattern (with RSS on RX-queue or FLow DIrector)
11. Resoruce Director for preventing cache posioning.

I highly recommend @Anuradha to check FM10K PMD capacity, BIOS, and using smap_affinity, isol_cpu, rcu_callback etc.

Note: I have been able to achieve 29 Mpps (64B) packets using single core and DPDK example skeleton with X710 NIC.