The implementation of this and the target side looks sane to me. I'll wait for the opinion of the NVMe/TCP on the usefulness of this, and for the networking maintainer on an ACK for patch 1.