On Sat, 28 Jan 2012 10:42:19 -0800, Guenter Roeck wrote: > On Sat, Jan 28, 2012 at 01:15:36PM -0500, Jean Delvare wrote: > > The die temperature seems to be a simple linear extrapolation, this > > should actually be fairly easy to implement. > > > > The battery temperature is a typical thermistor computation. It is best > > done in user-space by a libsensors configuration statement (i.e. the > > driver exports a value in mV instead of m°C and sensors.conf takes care > > of the conversion.) > > > > So I don't see anything particularly frightening here. Do you? > > Not really; I didn't expect the implementation to be difficult, just understanding > the documentation. > > For the battery temperature, I would prefer to report a temperature if the chip > is configured for it (assuming that can be detected). Agreed. > At least so far we can not convert voltages to temperatures via sensors.conf, > or can we ? We can't change the channel type, no. There is a long standing ticket opened for this: http://www.lm-sensors.org/ticket/2258 but I'm still not sure what to do with it. What can be done is that drivers export temperatures in mV but they don't say so, and we do the conversion in sensors.conf. This is essentially an undocumented violation of our sysfs interface but several drivers are doing it already: vt1211, vt8231, pc87360. See etc/sensors.conf.eg for examples of conversion formula. I'm nor particularly happy with the situation, but I have no immediate solution to propose and the problem is rare enough that it doesn't prevent me from sleeping so far. If you or anyone else want to propose something or even implement it, feel free. -- Jean Delvare _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
