[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Tue, 12 Dec 2017 14:01:27 -0800
From: Stephen Hemminger <stephen@...workplumber.org>
To: Ed Swierk <eswierk@...portsystems.com>
Cc: Marcelo Ricardo Leitner <marcelo.leitner@...il.com>,
Dan Williams <dcbw@...hat.com>,
netdev <netdev@...r.kernel.org>,
Benjamin Warren <ben@...portsystems.com>,
Keith Holleman <holleman@...portsystems.com>
Subject: Re: [PATCH] veth: Optionally pad packets to minimum Ethernet length
On Tue, 12 Dec 2017 11:00:38 -0800
Ed Swierk <eswierk@...portsystems.com> wrote:
> On Tue, Dec 12, 2017 at 10:34 AM, Stephen Hemminger
> <stephen@...workplumber.org> wrote:
> > Why not add to netdevsim rather than cluttering up a normal driver
> > with test support. We just pulled a bunch of test stuff out of dummy
> > for the same reason.
>
> My test setup to trigger an openvswitch conntrack issue
> (https://marc.info/?l=linux-netdev&m=151309548725627) involves a lot
> of moving parts:
>
> [netns-a: vetha1] - [vetha0] - [ovsbr0] - [vethb0] - [netns-b: vethb1]
>
> with nc client and server in netns-a and -b, and tweaks like turning
> off tcp_timestamps to make sure the packets in the TCP stream are
> small enough to reproduce the problem. A simpler, less fragile test
> setup would be valuable, especially if it ends up as an automated
> regression test.
>
> Could netdevsim be useful for that? Are there any existing tests
> producing TCP traffic that might serve as an example?
>
> --Ed
Maybe add a netem impairment that does padding?
Powered by blists - more mailing lists