[Bug 265588] [TCP] - tcp send a retransmission identical sequence number packet with different payload

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=265588

--- Comment #22 from Richard Scheffenegger <rscheff@freebsd.org> ---
(In reply to Keyong from comment #21)

Do you know, what devices are on the path between your fbds NFS server and the
Linux client?

As you can see, the ACKs sent by the Linux client are correct:

#41 ack 109097
#44 ack 111993
#45 ack 114889
#49 ack 129369 
#52 ack 129369 sack 132265-135161
#53 ack 129369 sack 132265-139505
#54 NFSv4 READ, ack 129369 sack 132265-139505

On the FreeBSD server, we observe:

#54 ack 111993
#56 ack 129369 sack 132265-135161
#58 ack 129369 sack 132265-139505
#59 ack 109097
#60 ack 114889
#61 ack 129369 
#64 NFSv4 READ, ack 129369 sack 132265-139505

So there is significant reordering of the ACKs going on here, with the bug
being triggered by the (server) #61 ACK.

I wonder if this reordering is due to a middlebox, or perhaps some Queue
reordering inside of the linux box, expediting "critical" ACKs (with SACK) over
normal ACKs...

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