Hey Martyn, > I haven't got any hardware old enough to hand to play with how ISA devices get > bound into the device model. Neither do I! I studied the code in drivers/base/isa.c to understand how it is being done. More specifically, take a look at isa_register_driver() and isa_bus_match() and you will see the similarities with this patch :) > I was thinking that we may be able to do vme-$bus_id.$device_id, but this > would require the $device_id to be unique per bus, which isn't an issue on > buses such as PCI (or I guess USB) where such enumeration is built into the > spec. With the exception of VME64x, which I guess we should really be > treating as a special case (I guess most cards in actual use don't follow > this standard), we are much closer to the ISA bus. Exactly. This is great for VME64x but unfortunately makes the rest of the cases a bit constrained. > I assume that $driver_name-$bus_id.$device_id is roughly in line with how ISA > would do this (IIRC, ther would only ever be one ISA bus on a board, so the > bus entry wouldn't be needed, but we clearly need this)? Yup. (Most of this code was taken from drivers/base/isa.c) Thanks! -- /manohar _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
