netdev - Re: [PATCH net v2] tcp_cubic: fix incorrect HyStart round start detection

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

Message-ID: <CADVnQy=3xz2_gjnM1vifjPJ_2TpT55mc=Oh7hO_omAAi9P6fxw@mail.gmail.com>
Date: Thu, 16 Jan 2025 09:42:32 -0500
From: Neal Cardwell <ncardwell@...gle.com>
To: Jason Xing <kerneljasonxing@...il.com>
Cc: Mahdi Arghavani <ma.arghavani@...oo.com>, 
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>, "edumazet@...gle.com" <edumazet@...gle.com>, 
	"haibo.zhang@...go.ac.nz" <haibo.zhang@...go.ac.nz>, 
	"david.eyers@...go.ac.nz" <david.eyers@...go.ac.nz>, "abbas.arghavani@....se" <abbas.arghavani@....se>
Subject: Re: [PATCH net v2] tcp_cubic: fix incorrect HyStart round start detection

On Thu, Jan 16, 2025 at 6:40 AM Jason Xing <kerneljasonxing@...il.com> wrote:
>
> On Thu, Jan 16, 2025 at 5:49 PM Mahdi Arghavani <ma.arghavani@...oo.com> wrote:
> >
> > Hi Jason,
> >
> > I will explain this using a test conducted on my local testbed. Imagine a client and a server connected through two Linux software routers. In this setup, the minimum RTT is 150 ms, the bottleneck bandwidth is 50 Mbps, and the bottleneck buffer size is 1 BDP, calculated as (50M / 1514 / 8) * 0.150 = 619 packets.
> >
> > I conducted the test twice, transferring data from the server to the client for 1.5 seconds:
> >
> > TEST 1) With the patch applied: HyStart stopped the exponential growth of cwnd when cwnd = 632 and the bottleneck link was saturated (632 > 619).
> >
> >
> > TEST 2) Without the patch applied: HyStart stopped the exponential growth of cwnd when cwnd = 516 and the bottleneck link was not yet saturated (516 < 619). This resulted in 300 KB less delivered data compared to the first test.
>
> Thanks for sharing these numbers. I would suggest in the v3 adding the
> above description in the commit message. No need to send v3 until the
> maintainers of TCP (Eric and Neal) give further suggestions :)
>
> Feel free to add my reviewed-by tag in the next version:
> Reviewed-by: Jason Xing <kerneljasonxing@...il.com>
>
> Thanks,
> Jason

Mahdi, a few quick questions about your test logs, beforePatch.log and
afterPatch.log:

+ What is moRTT? Is that ca->curr_rtt? It would be great to share the
debug patch you used, so we know for certain how to interpret each
column in the debug output.

+ Are both HYSTART-DELAY and HYSTART-ACK-TRAIN enabled for both of those tests?

thanks,
neal