[PATCH v2 7/7] nvme_fc: add dev_loss_tmo timeout and remoteport resume support

[James Smart]

On 10/11/2017 3:29 AM, Sagi Grimberg wrote:
> OK, I think I finally understand the existence of the two tmos.
> But may I ask, what happens if we only have ctrl_loss_tmo support?
>
> If I understand correctly, it would behave exactly as expected as the
> controller will periodically attempt reconnect until ctrl_loss_tmo fires
> and then it is deleted. With dev_loss_tmo, once the lower between
> ctrl_loss_tmo and dev_loss_tmo expires, the controller is deleted.
> If my understanding is correct, this just adds a level of confusion
> to setup procedure doesn't it?

agree with your statements. And no, it doesn't add confusion. You are 
ignoring the whole history coming from the SCSI side that is expected 
and well-known. If anything, adding the per-controller ctrl_loss_tmo 
that has different values, making fc-nvme behave differently than scsi 
on the same port (running scsi and fc concurrently perhaps), is what 
adds confusion.

-- james