2147136_en-US

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

2147136_en-US

2147136_en-US

reduced Tx network throughput speeds when writing large files over NFS

When running  iperf3 -s over the 10Gbase-t ethernet interface (using TQMLS1046A devkit) the benchmark shows 7.34 Gbaud (Giga bits per second) when the devkit sends (Tx) large data files (note that network path between the two systems is configured  with an MTU size of 9000 bytes for Jumbo frame). When TQMLS1046A is setup to receive the same workload (large files over NFS) on the Rx path of the ethernet interface we get the following error:

[ 1068.551554] fsl_dpaa_mac 1af2000.ethernet enet_exp: Err FD status = 0x00080000

AuroreL_0-1754439131315.png

My interpretation of this might be misconfiguration of the FSL DPAA ethernet drivers in the kernel config but I don't know where to start.

Re: reduced Tx network throughput speeds when writing large files over NFSTesting the changes and will provide an update on whether this is sufficient to resolve our issue.Re: reduced Tx network throughput speeds when writing large files over NFS

Please configure the following option in Linux Kernel configuration file.

FSL_DPAA_ETH_JUMBO_FRAME - "Optimize for jumbo frames"
Optimizes the DPAA1 Ethernet driver throughput for large frames termination traffic (For example, 4K and above).
Using this option in combination with small frames increases significantly the driver's memory footprint and may even deplete the system memory. Also, the skb truesize is altered and messages from the stack that warn against this are bypassed.

Tags (1)
No ratings
Version history
Last update:
‎11-21-2025 07:53 AM
Updated by: