22October2020 [FreeBSD WIKI]

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
Attendees:  rrs@, rscheff@, tuexen@

====== Discussion ======
  * tuexen@ discussed how to mitigate blackhole detection problems
    * tuexen@ has a review to ensure we don't do blackhole detection once we are at the minimum value [[https://reviews.freebsd.org/D24308|D24308]]; committed
    * Discussion on what should trigger blackhole detection -- full size? Or, even smaller messages which are above our min MTU? For at least IPv6, min MTU makes sense.
    * Otherwise, tuexen@'s plan remains unchanged.
    * BBR and RACK needs some modifcation
  * Discussion of [[https://reviews.freebsd.org/D24237|D24237]]: rscheff@ changed both the in-order and out-of-order data path so we call sorwakeup_locked() [when necessary] after the SACK blocks are updated. Still ongoing. rrs@ reviewed. Discussion continues. rrs@ accepted.
  * Can the TCP_STATS framework be turned on and sample no connections? Yes. thj@ will do.
  * [[https://reviews.freebsd.org/D18892|D18892]]: PRR, rrs@ will review.
  * [[https://reviews.freebsd.org/D18985|D18985]]: Upcoming: Rescue retransmission
  * [[https://reviews.freebsd.org/D21117|D21117]]: Upcoming: Reduce counter space
  * [[https://reviews.freebsd.org/D6611|D6611]], [[https://reviews.freebsd.org/D6612|D6612]], [[https://reviews.freebsd.org/D6656|D6656]]: Get gallatin@ involved. What needs to be done to also support IPv6? Can't we use some unused MPROTO flag?
  * [[https://reviews.freebsd.org/D8253|D8253]]: rscheff@ will look into it. Partially covered in [[https://reviews.freebsd.org/D26808|D26808]].
  * [[https://reviews.freebsd.org/D10894|D10894]]: possibly use hystart++ instead. rrs@ will look into this later this year.
  * [[https://reviews.freebsd.org/D26807|D26807]]: Refactoring. accepted.
  * [[https://reviews.freebsd.org/D26808|D26808]]: Cubic RTO reaction. accepted.
  * [[https://reviews.freebsd.org/D26897|D26897]]: Update iotctl() for TCP to not access socket buffers on listeners. Also report an error?
  * The current TCP stack is not compliant with [[https://tools.ietf.org/html/rfc7323|RFC7323]] when handling segments without timestamps when support has been negotiated or vice versa. Reported in [[https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=250499|PR 250499]]. tuexen@ will put up a fix to improve the compliance.
Next meeting: 5 November @ 1400 UTC