[PATCHv2 0/7] nvmet: debugfs support

[Sagi Grimberg]

On 21/03/2024 8:55, Hannes Reinecke wrote:
> On 3/21/24 00:07, Sagi Grimberg wrote:
>>
>>
>> On 20/03/2024 16:40, Hannes Reinecke wrote:
>>> Hi all,
>>>
>>> taking up the original patchset for nvmet debugfs
>>> I've improved upon that submission by modifying
>>> the layout:
>>>
>>> /dev/kernel/debug/nvmet
>>>    <subsysnqn>
>>>      ctrl<cntlid>
>>>        port
>>>        state
>>>        hostnqn
>>>        kato
>>>        queue<qnum>
>>>          host_traddr
>>
>> Why is host_traddr per queue?
>
> Blame TCP. We're having one connection per queue, and as such the
> peer address is a per-queue thingie.
> And it really felt daft to look for queue 0 to get the peer address
> of the controller.
> Unless we declare TCP the odd man out, and store the peer address
> per controller. Certainly would make sense for FC and loop.

It will never be different, I say just take it from queue 0...