[PATCH v25 00/20] nvme-tcp receive offloads

[Jakub Kicinski]

On Wed, 26 Jun 2024 18:21:54 +0300 Aurelien Aptel wrote:
> We have taken some time to review your documents and code and have had
> several internal discussions regarding the CI topic. We truly appreciate
> the benefits that a CI setup could bring. However, we believe that since
> this feature primarily relies on nvme-tcp, it might achieve better
> coverage and testing if integrated with blktest. Your design focuses on
> the netdev layer, which we don't think is sufficient.
> 
> blktests/nvme is designed to test the entire nvme upstream
> infrastructure including nvme-tcp that targets corner cases and bugs in
> on-going development.  Chaitanya, Shinichiro, Daniel and other
> developers are actively developing blktests and running these tests in
> timely manner on latest branch in linux-nvme repo and for-next branch in
> linux-block repo.
> 
> Again, we are open to provide NIC so that others can also test this
> feature on upstream kernel on our NIC to facilitate easier testing
> including distros, as long as they are testing this feature on upstream
> kernel. In this way we don't have to replicate the nvme-block storage
> stack infra/tools/tests in the framework that is focused on netdev
> development and yet achieve good coverage, what do you think?

I'm not sure we're on the same page. The ask is to run the tests on 
the netdev testing branch, at 12h cadence, and generate a simple JSON
file with results we can ingest into our reporting. Extra points to
reporting it to KCIDB. You mention "framework that is focused on
netdev", IDK what you mean.