[PATCH 1/1] nvme-rdma: fix possible list_del corruption while freeing ctrl

[Sagi Grimberg]

> In case of concurrent calls to nvme_rdma_free_ctrl will occur,
> we may end up with a wrong and non-protected check for list emptiness.
> For current implementation, this is only for good practice.

.free_ctrl is called from a single call-site serialized by a refcount
so we don't have concurrent access by design, so I'm not sure I
understand why this makes sense.