On 23-04-05, Florian Fainelli wrote: > On 4/5/2023 5:27 AM, Andrew Lunn wrote: > > On Wed, Apr 05, 2023 at 11:26:56AM +0200, Marco Felsch wrote: > > > Some phy's don't report the correct phy-id, e.g. the TJA1102 dual-port > > > report 0 for the 2nd port. To fix this a driver needs to supply the > > > phyid instead and tell the phy framework to not try to readout the > > > phyid. The latter case is done via the new 'phy_id_broken' flag which > > > tells the phy framework to skip phyid readout for the corresponding phy. > > > > In general, we try to avoid work around for broken hardware in the > > core. Please try to solve this within nxp-tja11xx.c. > > Agreed, and one way to solve working around broken PHY identification > registers is to provide them through the compatible string via > "ethernet-phyAAAA.BBBB". This forces the PHY library not to read from those > registers yet instantiate the PHY device and force it to bind to a certain > phy_driver. The nxp-tja11xx.c is a bit special in case of two-port devices since the 2nd port registers a 2nd phy device which is correct but don't have a dedicated compatible and so on. My 2nd idea here was to check if phy_id is !0 and in this case just use it. I went this way to make it a bit more explicit. Regards, Marco > -- > Florian >
