On 8/22/23 01:38, Daniel Wagner wrote:

Addressed Sagi's feedback and dropped the explicit --blkdev=device argument from
the _nvmet_target_setup calls, as it is the default.



original cover letter:

Introduce helpers to setup nvmet targets. This is spin off from the refactoring
patches and the allowed_host patches [1].

Sagi suggested to record all resources allocated by nvmet_target_setup and then
later clean them up in nvmet_target_cleanup. I opted to figure out in
nvmet_target_cleanup what was allocated via the newly introdcuded _get_nvmet_ports
helper. The reason being, Hannes told me offline that he would like to add ANA
tests which will add some more ports to the subsystem. I hope with this
the code is more future proof.

BTW, while looking at this I saw that the passthru code is using the awkward
return value port when calling nvmet_passthru_target_setup. It seems some
more refactoring is in order...

[1] https://lore.kernel.org/linux-nvme/5h333eqhtw252sjw6axjewlb5bbb5ze7awekczxe3kie2lnhw6@manyer42khct/ (local)


changes

v3:
  - rebased/retested
  - use the default with _nvmet_target_setup

v2:
  - drop local subsys variable in passthru tests
  - do not use port as handle in passthru tests
  - free port after unregistering from subsys
  - https://lore.kernel.org/linux-nvme/20230818141537.22332-1-dwagner@suse.de/ (local)

v1:
  - https://lore.kernel.org/linux-nvme/20230818095744.24619-1-dwagner@suse.de/ (local)

Daniel Wagner (3):
   nvme/{033,034,035,036}: use default subsysnqn variable directly
   nvme/{033,034,035,036,37}: drop port handle between passthru target
     setup and cleanup
   nvme: introduce nvmet_target_{setup/cleanup} common code

I really appreciate you doing this cleanup, it was long due.

Thanks a lot, for this series :-

Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>

-ck