> > That would be great, do we have such a list anywhere already? > > Not exactly. However, such a list could be easily regenerated from the > confidence values stored in sensors-detect. Chips to which we never > grant a confidence value greater than (say) 5 are bad candidates to > autoloading. A closer look should let us refine the list. > > If you are interested, I may provide a comprehensive list of I2C > client drivers and chips stating the "autoloadability" for each chip. > Since I know the various hardware monitoring chips and sensors-detect > very well, I will probably come to a decent candidate list faster than > anyone else. If you want a similar document for ISA chips and > Super-I/O chips, just let me know. On second though, we don't need this, do we? The hotplug events will be generated by the kernel, but on the kernel side we have no idea what the client chips are (which is why we need the whole detection stuff in the first place). The detection code will remain in user-space and won't necessarily be very different from what it was. All we will need to do it convert the confidence value into a boolean balue (autoload a client driver or not). -- Jean Delvare http://khali.linux-fr.org/
