On 5/30/2024 1:43 AM, Russell King (Oracle) wrote: > On Wed, May 29, 2024 at 04:28:16PM +0200, Andrew Lunn wrote: >>> Qualcomm ethernet HW supports 2.5G speed in overclocked SGMII mode. >>> we internally term it as OCSGMII. >> >> So it still does SGMII inband signalling? Not 2500BaseX signalling? It >> is not actually compatible with 2500BaseX? >> >>> End goal of these patches is to enable SGMII with 2.5G speed support. >>> The patch in these series enabled up SGMII with 2.5 for cases where we >>> don't have external phy. ( mac-to-mac connectivity) >> >> So the other end needs to be an over clocked SGMII MAC, not 2500BaseX? >> >>> The new patch posted extends this for the case when the MAC has an >>> external phy connected. ( hence we are advertising fr 2.5G speed by adding >>> 2500BASEX as supported interface in phylink) >> >> And i assume it does not actually work against a true 2500BaseX >> device, because it is doing SGMII inband signalling? > > I really hope the hardware isn't using any SGMII inband signalling > at 2.5G speeds. Other devices explicitly state that SGMII inband > signalling while operating the elevated 2.5G speed is *not* supported! Do we mean SGMII autoneg by SGMII inband signalling? Our hardware doesn't support autoneg for 2.5G mode in SGMII. >
