nvme: Delete lports in _nvmet_target_cleanup

[Mr-Bossman]

Delete the lports in _nvmet_target_cleanup, which is called when when the test ends. Additionally, ignore file not found errors when deleting the lports, in _cleanup_nvmet as they might have already been deleted by _nvmet_target_cleanup.

This will also fix the issue of common/nvme: line 186: echo: write error: No such file or directory being printed after the test has finished which will happen when running tests with NVMET_TRTYPES=fc ./check tests/nvme/*

[kawasaki]

It is still not clear for me why the change is required. My understanding is that,

• most of nvme group test cases call _setup_nvmet()
• _setup_nvmet() calls "_register_test_cleanup _cleanup_nvmet"
• then _cleanup_nvmet() is called for most of nvme group test cases regardless whether the test case passed or failed
• _cleanup_nvmet() deletes lport ( and calls stop_soft_rdma )

Then _nvmet_target_cleanup() should not need to delete lport.

I guess you needed this change to avoid some failures. What kind of failure do you observe?

[Mr-Bossman]

It is still not clear for me why the change is required. My understanding is that,

• most of nvme group test cases call _setup_nvmet()
• _setup_nvmet() calls "_register_test_cleanup _cleanup_nvmet"
• then _cleanup_nvmet() is called for most of nvme group test cases regardless whether the test case passed or failed
• _cleanup_nvmet() deletes lport ( and calls stop_soft_rdma )

Then _nvmet_target_cleanup() should not need to delete lport.

I guess you needed this change to avoid some failures. What kind of failure do you observe?

For some reason when _cleanup_nvmet calls _nvme_fcloop_del_lport this error happens common/nvme: line 186: echo: write error: No such file or directory moving _nvme_fcloop_del_lport to _nvmet_target_cleanup fixes this issue.

IMO _nvmet_target_cleanup should cleanup as much as possible and not leave stuff etxra stuff for _cleanup_nvmet to do which is why i also added stop_soft_rdma.

[kawasaki]

Thanks, I'm interested in the failure. May I know how to recreate the failure? I would like to take a closer look in it.

At the blktests master branch tip, the line 186 of common/nvme is not in _nvme_fcloop_del_lport(), but in _nvme_fcloop_del_tport(). When you see the failure, line 186 is in _nvme_fcloop_del_lport(), right?

k
IMO, we should have less duplication between _cleanup_nvmet() and _nvmet_target_cleanup(). Then I would like to fully understand why the duplication is required.

[Mr-Bossman]

This happens on Centos 10, same with #263. sorry for the lack of updates, I have been a bit busy.