Jean, Greg: * Jean Delvare <khali at linux-fr.org> [2004-08-25 10:17:18 +0100]: > > > Stuff like this probably isn't appropriate for kernel.org (Greg?) > > > > Why not? It looks useful to me. Care to send me a patch adding > > this to the main kernel tree? Later today, sure. > I agree it sounds very useful for testing, and the driver isn't that big. Just > make in clear in the help text that the regular guy doesn't need this. I'll make it impossible to build this driver into the kernel... it's only useful as a module anyway. Would it be appropriate to mark it experimental as well? > I think that this driver would be even more useful if it was possible to > "load" a register map into it. I guess it should be possible to have a sysfs > interface, much like we have for the eeprom driver, but writable. That way we > could test almost anything in the driver, including the > detection/identification step, and possibly even simulate temperature changes > and the like. Testing would look like: > 1* Load i2c-stub. > 2* Write (binary) register map. > 3* Load chip driver. > 4* Test the driver and possibly change registers values on the fly (dd or some > perl should do). > > An alternative interface file could be one to which we'd write "address value" > pairs. I don't know which is easier to implement or more convenient. Thoughts? Um, i2cset works fine for this. BTW: note how there are independent arrays for data/byte and data/word commands. Depending on if/how a chip driver mixes them, you'll need to be careful. Regards, -- Mark M. Hoffman mhoffman at lightlink.com
