Hot swap of  SFPs 10G/1G current is not support on Intel x553 controller with dpdk 21.11.3 


CPU  Intel(R) Atom(TM) CPU C3558R @ 2.40GHz

x553 ports bond to dpdk inserted with 10G and 1G SFPs 

# lspci 


07:00.0 Ethernet controller: Intel Corporation Ethernet Connection X553 10 GbE SFP+ (rev 11)

07:00.1 Ethernet controller: Intel Corporation Ethernet Connection X553 10 GbE SFP+ (rev 11)


# /usr/share/dpdk/tools/dpdk-devbind.py --status


Network devices using DPDK-compatible driver

============================================


0000:07:00.0 'Ethernet Connection X553 10 GbE SFP+ 15c4' drv=vfio-pci unused=

0000:07:00.1 'Ethernet Connection X553 10 GbE SFP+ 15c4' drv=vfio-pci unused=

 

 et7   Driver PMDPort  HWaddr cc:1a:a3:ff:ea:e9 MTU 9236 

        Speed 1,000Mbps Link UP  Duplex FULL Autoneg ON  

        RX Queues: 1   RX Queue size: 4096 

        TX Queues: 1   TX Queue Size: 1024 

    Inc/RX packets: 22,327       bytes: 23,925,237     

        dropped: 0          

    Out/TX packets: 22,352       bytes: 23,924,291     

        dropped: 0          


et6   Driver PMDPort  HWaddr cc:1a:a3:ff:ea:e9 MTU 9236 

        Speed 10,000Mbps Link UP  Duplex FULL Autoneg ON  

        RX Queues: 1   RX Queue size: 4096 

        TX Queues: 1   TX Queue Size: 1024 

    Inc/RX packets: 2,798,567,431   bytes: 4,186,656,722,403  

        dropped: 0          

    Out/TX packets: 523,462,096    bytes: 783,087,633,300   

        dropped: 28,382       

 

Swap the SFPs with the cable between port6 and port 7.

 

et7   Driver PMDPort  HWaddr cc:1a:a3:ff:ea:e9 MTU 9236 

        Speed UNKNOWN  Link DOWN Duplex HALF Autoneg ON  

        RX Queues: 1   RX Queue size: 4096 

        TX Queues: 1   TX Queue Size: 1024 

    Inc/RX packets: 22,327       bytes: 23,925,237     

        dropped: 0          

    Out/TX packets: 22,362       bytes: 23,924,891     

        dropped: 0          

et7   Driver PMDPort  HWaddr cc:1a:a3:ff:ea:e9 MTU 9236 

        Speed UNKNOWN  Link DOWN Duplex HALF Autoneg ON  

        RX Queues: 1   RX Queue size: 4096 

        TX Queues: 1   TX Queue Size: 1024 

    Inc/RX packets: 22,327       bytes: 23,925,237     

        dropped: 0          

    Out/TX packets: 22,362       bytes: 23,924,891     

        dropped: 0          

 

Same experiment with Linux ixgbe kernel driver and on hotswap on SFPs both link and speed were correctly reported. Logs below:

 


I saw some patches in the archives posted in 2021 which did not make it to the upstream release that support these features by Stephen.


https://mails.dpdk.org/archives/dev/2021-December/230965.html


Is there a plan to support this feature dpdk like linux kernel ixgbe driver ?


Thanks,

Prasanna