netdev - Re: One way TCP bottleneck over 6 x Gbit tecl aggregated link

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]

Date:	Sun, 22 Mar 2015 09:54:54 -0700
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Wolfgang Rosner <wrosner@...net.de>
Cc:	netdev@...r.kernel.org
Subject: Re: One way TCP bottleneck over 6 x Gbit tecl aggregated link

On Sun, 2015-03-22 at 16:14 +0100, Wolfgang Rosner wrote:

> Agree. 
> This is why I considered it as a bug.
> 
> But just a sideline, not my main problem.
> And maybe it is even related to my build environment.
> Or to the fact that I call it with full path. or whatever....
> 
> I don't know whether this repo you pointed me
> https://git.kernel.org/cgit/linux/kernel/git/shemminger/iproute2.git/  
> is a bleeding edge developer fork, that is expected to show such fluctuations, 
> or a should-be-stable mainstream one, where such behaviur is not supposed to 
> occur. 
> 
> So if for you, there is reason to worry on this issue, we can try to debug 
> into it. Let me know, and give me instructions for tests, if required.
> 
> 
> 
> 
> 
> However, from my perspective, I'd prefer to return to my real problem:
> 	"One way TCP bottleneck over 6 x Gbit tecl aggregated link"
> 
> If you got lost in the flood of netdev list (I just subscribed, and am really 
> overwhelmed...), I don't mind to send you the info again and even remove the 
> superfluous lines manually for you - see below.
> 
> Do these figure tell you a story?
> Any hint where to look for a bottleneck?
> Presumaly related to some per-TCP-connection ressource?
> 

I see nothing wrong but a single retransmit.


You might be bitten by tcp_metric cache.

check /proc/sys/net/ipv4/tcp_no_metrics_save

(set it to 0, and flush tcp metric cache :

echo 0 >/proc/sys/net/ipv4/tcp_no_metrics_save
ip tcp_metrics flush


Also, not clear why you use jumbo frames with 1Gbit NIC (do not do that,
really), and it seems TSO is disabled on your NIC(s) ????


for dev in eth0 eth1 eth2 eth3 ....
do
 ethtool -k $dev
done



--
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