[Bug 292293] Accurate ECN on the loopback interface causes an infinite ACK loop.

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=292293

--- Comment #3 from Richard Scheffenegger <[email protected]> ---
(In reply to adrik from comment #1)
Looks like the tcpdump  got captured around 1h17m after the onset of this
issue...

Do you have the binary trace? TCP dump only decodes the full 3 ACE bits in the
header after the commit 55dc320; 

ACE should start of with an ACE value of 5 (decoded as eE if the tcpdump
version supports decoding it), or 6 (CE mark on the SYN). A loopback port
shouldn't ever mark packets with the CE flag (unless introduced by a iptables
rule; older binaries of iptables/NAT may be using the AE bit - but that was
fixed with
https://reviews.freebsd.org/D43172.

Presumably you can recreate the effect at will? Do you have a minimalistic code
to induce this behavior - or are two instances of nc sufficient already?

-- 
You are receiving this mail because:
You are the assignee for the bug.