On Fri, Jul 22, 2022 at 12:44:17PM +0100, Russell King (Oracle) wrote: > Given that this managed property was introduced for mvneta, mvneta's > implementation of it is the best reference we have to work out what > the intentions of it were beyond the commit text. > > With in-band mode enabled, mvneta makes use of a fixed-link PHY, and > updates the fixed-link PHY with the status from its GMAC block (which > is the combined PCS+MAC). > > So, when in-band mode is specified, the results from SGMII or 1000base-X > negotiation are read from the MAC side of the link, pushed into the > fixed-PHY, which then are reflected back into the driver via the usual > phylib adjust_link(). ... and I should have said that this is exactly why in-band mode is treated as a fixed-link, even though it's nothing of the sort. It makes use of the infrastructure that was present at the time (fixed-phy) to implement this feature of reading the link status from the PCS/MAC end of the link. It may not have been the best design, but it's an evolved design based on the code that was available and what people thought at the time (and pre-dates my involvement with mvneta in mainline.) -- RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
