[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20180405180423.GC17495@lunn.ch>
Date: Thu, 5 Apr 2018 20:04:23 +0200
From: Andrew Lunn <andrew@...n.ch>
To: Ran Shalit <ranshalit@...il.com>
Cc: netdev@...r.kernel.org
Subject: Re: marvell switch
On Thu, Apr 05, 2018 at 05:26:37PM +0000, Ran Shalit wrote:
> בתאריך יום ה׳, 5 באפר׳ 2018, 19:09, מאת Andrew Lunn <andrew@...n.ch>:
>
> > > Is there a wiki which explains switch configuration ?
> >
> > Nope. The whole idea is that they behave like normal linux
> > interfaces. So there is no need to document them. You already know how
> > to use them.
> >
> > > Is it possible to open socket and send/recieve on switch ports (lan0
> > > for example) ?
> >
> > Sure. It is as normal interface. Give is an IP address and then do
> > TCP/IP as usual.
> >
>
> There is something I don't fully understand.
> I thought that ports in switch which are not connected to the manage cpu
> are only configured by cpu and then can coomunicate with each other.
>
> What does it mean to open socket in such port (not the cpu port) and send
> data (ethernet frames) ? Does it mean that the cpu port is sending data
> directly to that port ?
The CPU port is configured to use {E}DSA tagging. Frames ingressing on
the CPU port contain an extra header. That header tells the switch
which port to send the frame out of. The switch can also send frames
received on a port out the CPU port to the host, again, with a
header. The host can tell from the header which port the frame
ingressed.
Using this, we can make ports look like normal Linux interfaces.
Andrew
Powered by blists - more mailing lists