Attendees: jtl@, kp@, thj@, rscheff@, tuexen@ = Discussion = * kp@ asks about status of TCP test suite * decision to move tests to base system so we can align behavior changes with corresponding changes to tests * tool (packetdrill) will still live in a port, similar to things like scapy which also live in ports and are also used for tests * [[https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=252913 | PR 252913]]: tuexen@ got a tracefile and some information from Netflix; impacts 12, waiting for 13 to be released. * [[https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=238741 | PR 238741]]: Fixed and MFCd to 13. * [[https://reviews.freebsd.org/D23230 | D23230]]: ECN++, rscheff@ will investigate the sysctl interface. Still in progress. * [[https://reviews.freebsd.org/D28813 | D28813]]: MFCd. * [[https://reviews.freebsd.org/D28819 | D28819]]: MFCd. * [[https://reviews.freebsd.org/D28702 | D28702]]: Planned for main, mid March. Still in progress, waiting for 13 release. * [[https://reviews.freebsd.org/D28634 | D28634]]: Fix dup acks triggering a fast retransmission when we don't receive SACKs. Clarified the SACK negotiation. Needs reviews. * [[https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=253476 | PR 253476]]: ipfw injecting TCP segments without timestamps even if they are required. The result was that this should be fixed in ipfw. Other implementation might also do strict checks. There is a sysctl variable to enforce less strict checks. * kp@ checked and pf does not have the same bug * we consider this closed from the TCP side * [[https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=253848 | PR 253848]]: fixed by [[https://reviews.freebsd.org/D29083 | D29083]] * rscheff@ has something which works with PRR for ECN and congestion window reduction with !NewReno. No plans to MFC. Will wait for 13 to commit to main. * thj@ asks for TCP blackbox log tools. * Tools are already available in ports: * devel/bbparse * net/read_bbrlog * net/tcplog_dumper * There was a discussion at IETF/IRTF about TCP schema for Qlog * Discussion in our meeting about generic fields vs. fields needed for debugging * Lost retransmission detection: [[https://reviews.freebsd.org/D28931 | D28931]] is an idea for doing SACK loss recovery in the base stack. It is a simpler alternative to TLP and RACK. tuexen wonders if this should be an internet-draft? rscheff suggests it might have low interest at IETF. Next meeting: 25 March @ 1400 UTC