netdev - Re: [PATCH net-next] tcp: Fix CWV being too strict on thin streams

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: <1442933196.29850.94.camel@edumazet-glaptop2.roam.corp.google.com>
Date:	Tue, 22 Sep 2015 07:46:36 -0700
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Bendik Rønning Opstad <bro.devel@...il.com>
Cc:	Neal Cardwell <ncardwell@...gle.com>,
	"David S. Miller" <davem@...emloft.net>,
	Alexey Kuznetsov <kuznet@....inr.ac.ru>,
	James Morris <jmorris@...ei.org>,
	Hideaki YOSHIFUJI <yoshfuji@...ux-ipv6.org>,
	Patrick McHardy <kaber@...sh.net>,
	Netdev <netdev@...r.kernel.org>,
	Eric Dumazet <edumazet@...gle.com>,
	Andreas Petlund <apetlund@...ula.no>,
	Carsten Griwodz <griff@...ula.no>,
	Jonas Markussen <jonassm@....uio.no>,
	Kenneth Klette Jonassen <kennetkl@....uio.no>,
	Yuchung Cheng <ycheng@...gle.com>
Subject: Re: [PATCH net-next] tcp: Fix CWV being too strict on thin streams

On Tue, 2015-09-22 at 16:29 +0200, Bendik Rønning Opstad wrote:
> > Thanks for this report!
> 
> Thank you for answering!
> 
> > When you say "CWND is reduced due to loss", are you talking about RTO
> > or Fast Recovery? Do you have any traces you can share that illustrate
> > this issue?
> 
> The problem is not related to loss recovery, but only occurs in congestion
> avoidance state. This is because tcp_is_cwnd_limited(), called from the
> congestion controls ".cong_avoid" hook, will only use tp->is_cwnd_limited when
> in congestion avoidance.
> 
> I've made two traces to compare the effect with and without the patch:
> http://heim.ifi.uio.no/bendiko/traces/CWV_PATCH/
> 
> To ensure that the results are comparable, the traces are produced with an extra
> patch applied (shown below) such that both connections are in congestion
> avoidance from the beginning.
> 
> > Have you verified that this patch fixes the issue you identified? I
> > think the fix may be in the wrong place for that scenario, or at least
> > incomplete.
> 
> Yes, we have tested this and verified that the patch solves the issue in the
> test scenarios we've run.
> 
> As always (in networking) it can be difficult to reproduce the issue in a
> consistent manner, so I will describe the setup we have used.
> 


Ahem.

packetdrill can make this in one script, as you can exactly control the
packets that the 'remote' peer would answer.

No need for complex setup. You should try it, and as a bonus we could
easily reproduce the problem and check the fix.

Let see if we can cook a packetdrill scenario.



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html