[PATCH 3/4] nvmet: prevent max_qid changes for discovered subsystems

[Max Gurtovoy]

On 26/09/2025 9:58, Daniel Wagner wrote:
> On Fri, Sep 26, 2025 at 01:09:22AM +0300, Max Gurtovoy wrote:
>> I'm not an expert in the blktests but it seems like we should do inside the
>> set_qid_max() something like:
>>
>> _nvmet_target_cleanup
>> _nvmet_target_setup --blkdev file (Add here --qid_max "${qid_max}" option)
>> nvmf_check_queue_count "${subsysnqn}" "${qid_max}" || return 1
>> _nvmf_wait_for_state "${subsysnqn}" "live" || return 1
>>
>> I expect the above to work with and without my patches..
> My patch was just a quick hack to see if it works. Updating
> __nvmet_target_setup to handle the qid_max argument is the right
> approach.

Thanks for trying this.


>   
>> If possible - lets try it also with tcp/rdma transports
> You are proposing a patch which has never run blktests, as it clearly
> would regress nvme/048. I don't see myself here to do your work.

This was not my intention :)

I've sent a patch series to blktests with the needed fixes.

I've tested nvme/048 with RDMA/TCP transports.