On Wed, Dec 23, 2020 at 09:58:52PM +0100, Alexandre Belloni wrote:
> On 22/12/2020 15:41:41+0100, Andrew Lunn wrote:
> > > Yes the register based injection/extration is not going to be fast, but
> > > the FDMA and its driver is being sent later as separate series to keep
> > > the size of this review down.
> >
> > FDMA?
> >
> > I need a bit more background here, just to make use this should be a
> > pure switchdev driver and not a DSA driver.
> >
>
> I don't think this should be a DSA driver. As for Ocelot, the CPU
> port is not a MAC and in that use case, this would be like a top of the
> rack switch with traffic going to the CPU port being mostly used for
> managmement (dhcp, stp, etc...) as opposed to being used to forward
> traffic to another interface, like WAN or wifi.
>
> However, I would think there will be cases where the internal CPU is not
> use and instead use ths switch in a DSA setting, very much like what is
> done for Felix with regards to Ocelot.
>From what i have heard so far, it does seem like a pure switchdev
driver is correct. So long as FDMA is not a standalone Ethernet
driver, but just a DMA engine incorporated into this driver, the
architecture looks correct.
I was asking because from the information that was available, it was
impossible to say what the correct architecture should be.
Andrew
