netdev - 802.11ad symmetry

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>] [day] [month] [year] [list]

Date:	Wed, 27 Jan 2016 10:33:41 +0000
From:	Samsonov Dmitry <dmitry.samsonov@...p.mail.ru>
To:	"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: 802.11ad symmetry

Hello!
bonding-devel@...ts.sourceforge.net is not accepting e-mails, so I assume this list is the correct destination.
Is there a way to make LACP bonding between a linux box and a switch to be symmetric?
As I understand, currently there is no way to guarantee, that hash, generated for two packet which differs only by direction (packet from connection with same MAC or IP or IP:PORT) would get in the same port on the switch and on the linux box.
Imagine that we have linux box with 2 CPUs and bond from two interfaces - eth0 (cpu node 0) and eth1 (cpu node 1). And there is an application thread which handle connection with client. Now it is possible, that incoming packet for this connection would end up in eth0 (cpu node 0) and outgoing packets would end up in eth1 (cpu node 1) which will lead to inter CPU communication.
So, is there a way to guarantee symmetric choice of bonded ports?

Best regards,
Dmitry Samsonov
Odnoklassniki.ru
Mail.ru Group
Lead System Administrator
e-mail: dmitry.samsonov@...oklassniki.ru