Stefan Roese wrote:
On Tuesday 08 January 2008, Andy Fleming wrote:
I think that if you use an Ethernet and need to differentiate between MII and RMII they should be mutually exclusive. You either define CONFIG_RMII OR CONFIG_MII but not both. If you have a PHY on the chip, then you do not define any of the two. MII uses more pins and maybe different pins, so a port for a CPU should differentiate.
I'm a little late to this, but I feel I should make sure people are aware that CONFIG_MII and CONFIG_RMII are not, in this case, two different options for the same concept. CONFIG_MII is the option you enable to allow MII Management operations to be done. It should *probably* be called CONFIG_MDIO.
Yes, this would be good. Anyone interested in fixing up a patch for this?
I think it's a noble idea, but probably a can of worms in practice. The use of the term 'MII' is rampant, for example we have 'mii' commands, and 'bbmii' commands. These both really concern MDIO, but I doubt we want to change them.
The real culprit is RMII. As Andy has pointed out, RMII is a data plane alternative to MII, but shares the MDIO control plane. RMII is a hardware configuration, and as such is 'set and forget' and highly implementation-specific. Having a global CONFIG_RMII makes no sense. If anything, CONFIG_RMII should be changed to stuff like CONFIG_TSEC_RMII_2 etc. (a quickly made-up example with no grounding in reality), where the hardware binding is obvious.
regards, Ben