ok, if the only changes are cpu related, then they can be the same id, the s3c24xx code identifies the cpu for itself Hmm, but we have different phones, which also can have different hw revs iirc, that is handled by the revision tag being passed through fluffs: It's about EZX phones. They use all PXA27x fluffs: The differences are for keyboard,gpio, etc.. iirc, the mess that is mach-pxa selects cpu at compile time if the hardware is different, then that is a good reason for unique machine IDs especially if the difference is not detectable at runtime * stefan_schmidt gets the feeling that pxa is the worst mach type in arm-linux * fluffs has had to dig in it for a customer, eww one machine per kernel is not desperately nice fluffs: ok, E2 has no TS for example, other has TS only, TS and flip keyboard... fluffs: I have some hope that the earlier models can be used with a multimachine kernel the pxa needs work to get it to compile that way iirc, bits are still #ifdef hell it is worse than digging in hh.org's cvs fluffs: It seems the best way is to work out all the differences and what is runtime detectable that sounds like a good plan ok, thanks