On Aug 10, 2023 / 13:13, Daniel Wagner wrote:
> I've updated the series accoring the feedback. Also added the mentioned
> _nvmet_target_{setup|cleanup} helpers which reduced a lot of code.
> Roughly 350 lines code less after the refactoring.
>
> Maybe it's possible to refactor even more, e.g. this snippet
>
>
> local nvmedev
> nvmedev=$(_find_nvme_dev "${def_subsysnqn}")
> cat "/sys/block/${nvmedev}n1/uuid"
> cat "/sys/block/${nvmedev}n1/wwid"
>
> could be moved into _nvme_connect_subsys(). Though there are quiet a few tests
> which want the nvmedev later on for something like
>
> _run_fio_verify_io --size="${nvme_img_size}" \
> --filename="/dev/${nvmedev}n1"
>
> So we could return the nvmedev from _nvme_connect_subsys() but I don't know if
> this a good idea.
IMO, it is a good idea to make _nvme_connect_subsys() return the device. The
similar function _nvmet_passthru_target_connect() does that, so it is another
small goodness to have consistency between the two.
> FWIW, it would also fix the current problem we face with
> nvme/047 which seems to lack the second _find_nvme_dev() call.
I posted the fix patch for the nvme/047 problem reflecting your comments. I hope
that fix settled before further refactoring.
It is a fun to see the much of the boiler plates go away with the series :)
Thanks. I provided two minor comments on the 5th patch and 10th patch. Other
than that, this series looks good to me. Also I did another trial run, and
saw no regression. Good.
