From: <a.o...@bl...> - 2004-01-22 14:26:43
|
On Thu, Jan 22, 2004 at 01:44:20PM +0100, Free The Cube wrote: > Michael Steil wrote: > >** SI/gamepad driver > >"Free the Cube" has announced yesterday that he was working on a > >SI/gamepad driver - great. Don't we need an independent driver for the > >SI bus first, that exports an API for the bus, and gamepad drivers > >later, that plug into the SI API? > > The GCPAD driver from GCLIB uses direct memory access to the mapped area > where are pad status, it should be possible to do like this > (temporarily)... Then we can modify this first implementation to use the > SI layer ! (i need confirmation) > IMHO, you're better off going with the bus-layer -> device-driver architecture, initially. That way, the bus driver handles things like probing for and maintaining a list of existing devices, binding device drivers to devices, handling PM and hotplug events, etc, while device drivers handle stuff like enabling the device, accessing configuration space and communicating with the drive. Don't get me wrong here - if you feel more comfortable starting with your approach, feel free to do so. We can always move over later on :) Arthur -- Linux is a true multitasking system. Are you? |