netdev - Re: [QUERY] tcp: remove dubious FIN exception from tcp_cwnd

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CANn89iLQdB5Qmd2U=mk6SsG0=GZqVYYP1YdoXqKxit4uRdPC=w@mail.gmail.com>
Date: Mon, 16 Jun 2025 07:31:16 -0700
From: Eric Dumazet <edumazet@...gle.com>
To: Alok Tiwari <alok.a.tiwari@...cle.com>
Cc: ncardwell@...gle.com, kuniyu@...zon.com, davem@...emloft.net, 
	dsahern@...nel.org, kuba@...nel.org, pabeni@...hat.com, horms@...nel.org, 
	netdev@...r.kernel.org, darren.kenny@...cle.com
Subject: Re: [QUERY] tcp: remove dubious FIN exception from tcp_cwnd_test()

On Mon, Jun 16, 2025 at 12:21 AM Alok Tiwari <alok.a.tiwari@...cle.com> wrote:
>
> Hi Eric,
>
> Thanks for the cleanup and RFC conformance improvements.
>
> However, we've observed a performance regression in our testing with
> Apachebench at higher concurrency levels. Specifically, throughput
> and HTTP tail latency in workloads.
> It seems the previous exception for FIN packets allowed graceful
> tear-down even when CWND was full, which is now delayed or stalled.
>
> We understand the motivation for the cleanup, but the change introduces
> practical impact for real-world workloads relying on timely
> FIN transmission.
>
> Would you be open to discussing potential mitigations?( not sure like
> selective FIN override, other approaches.)

Anything relying on being able to send more than the allowed budget
will break when rules are enforced.

You could try to add an opt-in feature inflating CWND at
shutdown(SHUT_WR) time, anything not in the fast path might be
considered.