Scheffenegger, Richard Richard.Scheffenegger at netapp.com
Thu Oct 18 15:16:03 UTC 2018


Just wanted to test the waters, if there is any appetite to approve a patch that prevents cwnd to collapse down to 1 after loss recovery, under certain circumstance.

When flightsize becomes very small, and cwnd ends up at 1x MSS, the continuation of the session is gated with a 50:50 change by the clients delayed ACK timeout.

The change in RFC6582 is to limit cwnd to only collapse down to 2x MSS, which is a straight-forward 1-line change in the modular CC methods.

That change allows FreeBSD to claim conformance with RFC6582.

Any comments?

Richard Scheffenegger

More information about the freebsd-transport mailing list