Re: [RFC PATCH 0/4] net: dsa: link aggregation support

[Andrew Lunn <andrew@lunn.ch>]

On Tue, Oct 27, 2020 at 04:05:30PM +0100, Marek Behun wrote:
> When I first read about port trunking in the Peridot documentation, I
> immediately thought that this could be used to transparently offload
> that which is called Bonding in Linux...
> 
> Is this what you want to eventually do?
> 
> BTW, I thought about using port trunking to solve the multi-CPU DSA
> issue as well. On Turris Omnia we have 2 switch ports connected to the
> CPU. So I could trunk these 2 swtich ports, and on the other side
> create a bonding interface from eth0 and eth1.
> 
> Andrew, what do you think about this? Is this something that can be
> done? Or is it too complicated?
 
Hi Marek

trunking is something i've looked at once, but never had time to work
on. There are three different use cases i thought of:

1) trunk user ports, with team/bonding controlling it
2) trunk DSA ports, i.e. the ports between switches in a D in DSA setup
3) trunk CPU ports.

What Tobias is implementing here is 1). This seems like a good first
step.

I'm not sure 3) is even possible. Or it might depend on the switch
generation. The 6352 for example, the CPU Dest field is a port
number. It does not appear to allow for a trunk. 6390 moved this
register, but as far as i know, it did not add trunk support.  It
might be possible to have multiple SoC interfaces sending frames to
the Switch using DSA tags, but i don't see a way to have the switch
send frames to the SoC using multiple ports.

     Andrew