Re: [PATCH net-next v3 08/47] net: phylink: Support differing link speeds and interface speeds

[Sean Anderson]

On 7/18/22 1:28 PM, Andrew Lunn wrote:
>> > I am rather worried that we have drivers using ->speed today in their
>> > mac_config and we're redefining what that means in this patch.
>> 
>> Well, kind of. Previously, interface speed was defined to be link speed,
>> and both were just "speed". The MAC driver doesn't really care what the
>> link speed is if there is a phy, just how fast the phy interface mode
>> speed is.
> 
> I'm not sure that is true. At least for SGMII, the MAC is passed the
> line side speed, which can be 10, 100, or 1G. The PHY interface mode
> speed is fixed at 1G, since it is SGMII, but the MAC needs to know if
> it needs to repeat symbols because the line side speed is lower than
> the host side speed.

Right. In this case the phy interface speed is changing, so the MAC
needs to know about it. I suppose a more precise definition would be
something like

> The data rate of the media-independent interface between the MAC and
> the phy, without taking into account protocol overhead or flow control,
> but including encoding overhead.

as opposed to the link speed which is

> The data rate of the medium between the local device and the link
> partner, without ...

--Sean