"Hong Xu" hongxu.cn@gmail.com wrote:
The point is I'm not sure all AT91 series will behave the same way. But since AVR32 has less cases, it's a good idea to use #else to handle AT91 variants. So at least till now, this patch makes sense.
When I said "let's stop the overengineering", I meant let's not attempt to handle cases that don't exist yet :-)
If there turns out to be lots of differences within chip families, we should probably consider handling it some other way than through adding lots of #ifdefs. For example by adding a macb_set_usrio() hook in an arch/platform/chip/whatever-specific file. But until we actually _need_ that, I think we should go for the simplest solution possible.
Note that AT91 boards that don't actually have ethernet, or use some other implementation, are not interesting in this context since they shouldn't enable the macb driver in the first place.
Haavard