[PATCH] nvme: fix multiple ctrl removal scheduling

[Sagi Grimberg]

> cancel_work_sync seems to be a good idea

In fact, I think we should have it independently of
the state discussion, it healthy to make sure nothing runs
concurrently with remove.

> but I would still consider
> new state to be a better alternative as it seems to make overal design
> simpler.

I'm still not convinced on why adding a new state is making things
simpler.

> 
> Anyway, once we can decide which way to go, I would be happy to give
> this a test spin.

Keith, Christoph? can you please explain again why adding a new
state is better?

> You just forgot to remove the now reduntant WARN_ON
> in untested one but of course this was just untested and for
> demonstration only.

Obviously.