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

From: <bugzilla-noreply_at_freebsd.org>
Date: Thu, 20 Oct 2022 11:47:47 UTC
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=265588

--- Comment #17 from Richard Scheffenegger <rscheff@freebsd.org> ---
1448 byte chunks between server and client align perfectly until server frame
#51 (5 chunks: 0xb23ee730dd    0xc2fd29b2eb    0xbe27bbf1be    0xb829e51fdc   
0xbc5b20682b) and client frame #48 (3 chunks 0xb23ee730dd    0xc2fd29b2eb   
0xbe27bbf1be).

The last two chunks of the TSO get dropped.

server #62 is the lost 4th chunk of #51.

And server #63 starts with chunk 5 of #51 but doesn't have the correct sequence
number!

Frame 66 repeats chunk 5 of #51, with the correct sequence number.


TSO frame 63 seems to send everything from snd.una for an entire window, but
the seq number is off by one segment (1448) to the left. However, 19 packets
during SACK loss recovery is unexpected too!

(In short, I feel like the TSO frame #63 shouldn't be there at all).

I wonder if this may be the issue https://reviews.freebsd.org/D36637 although
that's dealing with the traffic burst, not the erraneous sequence number...

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