Jumbo frames work poorly on FreeBSD out of the box. There are a variety of opinions on this issue, but the status quo leaves users in the position to find this out the hard way.
A definition of done for this issue:
Jumbo frames work better out of the box
*** Bug 210901 has been marked as a duplicate of this bug. ***
See also https://reviews.freebsd.org/rS344817
See also https://lists.freebsd.org/pipermail/freebsd-net/2020-January/055164.html
See also https://gist.github.com/freqlabs/eba9b755f17a223260246becfbb150a1
We should mark drivers that handle 9k MTU and those that don't and exhibit this behaviour ... mark them in their man pages. I often buy hardware based on reading the FreeBSD man page for the hardware --- and were I again going to attempt jumbo packets on my network, I would want this in the man pages.
- This issue appears to be a a non-(bug)-specific issue, suitable as a parent/meta issue. Re-classify accordingly, incl switching blockers to dependencies.
- Reporter is committer, assign accordingly (coordinator/champion)