Hi all,
I'm using LS2085A board. In my default dpl file contains 8 SFP port support , It's taking more time for initializing the 8 ports.
For my setup only 4 SFP support is enough for that i did some changes in dpl file. According to that changes only 4 SFP is enabled and time consumption also reduced.
But performance is reducing, before that i got 5.4 gbps from each SFP port after changes i got only around 3.5 gbps from each port.
I don't know how to change the dpl file without performance losing.
If anyone know about these kindly help me on this.
Thanks regards,
-Naveen
解決済! 解決策の投稿を見る。
Hello Naveen,
You reduced DPIO number in your dpl configuration file, probably this is the cause of your problem. One DPIO is assigned for each online CPU, in your Kernel log you could find that dpni.2, dpni.3 and dpni.4 could not get affine DPIO. DPIO objects is used to perform hardware enqueue and dequeue, and hardware buffer management, the interface between GPP cores and QBMAN software portals.
Please try whether the performance problem would be resolved if you still configure 8 DPIOs in your dpl file.
Have a great day,
Yiping
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------
Naveen, where did you get your default dpl file from? I have an LS2085A as well, and the QorIQ Linux SDK default dpl only configured one interface.
Hello Naveen,
Would you please provide your dpl file and Kernel boot log?
Customizing dpl file should not affect the networking performance, actually during the testing, we often delete DPNI definition and use restool to dynamically create DPNIs and assign to DPRC.
Have a great day,
Yiping
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------
Hi Yiping,
Here i have attached default dpl file, changed dpl, default kernel log and dpl changed kernel log.
Default dpl file :
For accessing single port i'm getting around 9.4 gbps,
For accessing four port i'm getting around 5.4 gbps from each port.
Changed dpl file:
For accessing single port i'm getting around 9.4 gbps,
For accessing four port i'm getting around 3.5 gbps from each port.
I don't know why the performance is losing, If i did anything wrong means please correct me.
Thanks regards,
-Naveen
Hello Naveen,
You reduced DPIO number in your dpl configuration file, probably this is the cause of your problem. One DPIO is assigned for each online CPU, in your Kernel log you could find that dpni.2, dpni.3 and dpni.4 could not get affine DPIO. DPIO objects is used to perform hardware enqueue and dequeue, and hardware buffer management, the interface between GPP cores and QBMAN software portals.
Please try whether the performance problem would be resolved if you still configure 8 DPIOs in your dpl file.
Have a great day,
Yiping
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------
Hi Yiping,
Thanks for your valuable and timely response,
I could changed the dpl file with 8 dpio. Now i got performance like whatever i got before.
And i'm having one more doubt:
In my test setup i'm having two boards(Ls2085a) with 4-SFP ports, I connected the both boards using 4 SFP optical fiber cables,
For accessing single port i got 9.4 gbps but for accessing 4 ports i got 5.4 gpps from each port,
May i know why it got reduced when i try to access all the 4 ports?
Again Thanks for your great support.
Thanks regards,
-Naveen