search

amzn / amzn-drivers

Official AWS drivers repository for Elastic Network Adapter (ENA) and Elastic Fabric Adapter (EFA)
444 stars 166 forks source link

[Support]: coalescing features #305

Open likewise opened 2 weeks ago

likewise commented 2 weeks ago

Preliminary Actions

Driver Type

Linux kernel driver for Elastic Network Adapter (ENA)

Driver Tag/Commit

ena_linux_2.11.1

Custom Code

No

OS Platform and Distribution

different Linux kernel versions, custom and default

Support request

What interrupt coalescing options should be available with the latest ENA?
Was "adaptive TX" ever supported for the Linux kernel driver?

We have a 5.10 kernel that shows all the options available, and are confused, because the (current) driver source code does not seem to support these. Or are these implemented in-kernel rather than in-driver??

Coalesce parameters for ens7: Adaptive RX: off TX: off stats-block-usecs: 0 sample-interval: 0 pkt-rate-low: 0 pkt-rate-high: 0

rx-usecs: 0 rx-frames: 0 rx-usecs-irq: 0 rx-frames-irq: 0

tx-usecs: 0 tx-frames: 0 tx-usecs-irq: 0 tx-frames-irq: 0

rx-usecs-low: 0 rx-frame-low: 0 tx-usecs-low: 0 tx-frame-low: 0

rx-usecs-high: 0 rx-frame-high: 0 tx-usecs-high: 0 tx-frame-high: 0

Contact Details

leon@sidebranch.com

davidarinzon commented 2 weeks ago

Hi @likewise

Adaptive TX moderation was never supported, only Adaptive RX moderation is supported and is enabled by default in the driver. Though execution of ethtool may show all of these parameters, you should focus on rx-usecs, tx-usecs and Adaptive RX. ethtool outputs all the standard information it has, and each driver fills the fields that it supports, though ethtool still outputs all of the fields.

I do see that all of the values are 0 in your case, I assume you've loaded the github driver and have tried manipulating some of the values (Like disabling interrupt moderation, as suggested in our best practices for some use-cases)?

Would you like to share your use-case and what you're trying to achieve?

likewise commented 1 week ago

Use case is quick response time to RX for low traffic, either quick interrupts or no interrupts and busy polling I guess.

davidarinzon commented 1 week ago

Hi @likewise Please check the best practices page, there's a question about best latency there (Q: What are the optimal settings for achieving the best latency), I believe that the steps there will be helpful. Let us know otherwise.