controller deletion consolidation

[Sagi Grimberg]

> Hi all,
> 
> this series consolidates a lot of the boilerplate code in controller
> deletion.  I did this when applying the FC patches that finally made
> it look exactly like RDMA.
> 
> This will probably conflict a bit with Sagis consolidation series, but
> in the end should make it easier.

Definitely a step in the right direction. Looks good.

> Two notes:
> 
>   - do we need the cancellation of reset_work in FC?  If so we probably
>     want it in RDMA and loop as well.

We probably do.

>   - should RDMA really queue delete_work manually in
>     nvme_rdma_reconnect_or_remove instead of going through the state
>     machine?  That seems like a bug to me.
> 

I had this fix from the centralization patches, I can easily extract it
out and send it asap.