Re: [PATCH blktests 00/10] nvme: make NVME_TARGET_CONTROL work with remote targets

[Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>]

On Nov 04, 2026 / 17:44, Daniel Wagner wrote:
> Running the nvme fabrics tests against a remote target with
> NVME_TARGET_CONTROL doesn't work in current upstream:
> 
> - contrib/nvme_target_control.py configures the target with nvmetcli
>   through a REST API which only exists in an nvmetcli branch that was
>   never merged, and its nvmetcli client crashes with configshell-fb 2.0.
>   The target is never set up and the tests fail to connect.
> 
> - With a target set up by other means, the tests pass without testing
>   anything: _nvme_ns_ready waits for the namespaces in the local nvmet
>   configfs, which doesn't exist for a remote target. _nvme_connect_subsys
>   returns an error after every connect, the tests skip their I/O and the
>   disconnect, and the harness reports leftover controllers.
> 
> - The tests which use more than one port configure the ports in the
>   local nvmet configfs, and add-ports is a stub. The authentication
>   tests change the host keys there too.

(snip)

Daniel, thank you for the series. I walked through the patches and all look
valuable. I applied first six patches. I made minor comments on the two patches
out of the left four patches. Please see if they make sense for you.