> Flashrom has really nice generic routines like probe_jedec. They work > perfectly as long as flashrom does not have to care about the bus > protocol it should use to talk to the chip. > Even current external flashers (nic3com, satasii) work fine because > flashrom does not have to care about the bus protocol there either. > However, any external flasher supporting multiple bus protocols (and > such flashers are being worked on) really wants to know which protocol > to use. Unfortunately, chip_read* and chip_write* don't know about the > bus protocol.
If you're referring to my work, yes the protocol is designed so to be able to support many bustypes, but i have no real hardware for anything else than parallel. And then, i realized a missing feature in my protocol, it cannot tell the flasher of the bustype to be used. It's only an optional command away though. > Adding struct flashchip * as parameter to chip_read* and chip_write* > would solve the issue quite nicely and improve our abstraction at the > same time. > Of course, this means every function calling chip_read* and chip_write* > has to pass struct flashchip * to them. Ergo, we need struct flashchip * > as parameter in the whole call chain. > > An alternative would be using global variables for current flash chip > properties, but that is madness during probe where we want to check for > multiple chips. I think that having a simple global current_flashchip pointer would be simpler in many ways: - it would only need a single assign to operation to the loop at probe_flash to support it - it wouldnt be necessary to pass unnecessary parameters to chip_* functions ( most dont need this info - atleast atm ;P ), making for a smaller binary -- urjaman -- coreboot mailing list: [email protected] http://www.coreboot.org/mailman/listinfo/coreboot
