Hi Chris, On Mon, 08 Jul 2013 17:18:46 -0400, Chris Pattenden wrote: > I'm working to configure lm_sensors and have notice a few strange issues. > > I'm working on a TI 8148 on custom hardware. > Linux 2.6.37 > lm_sensors lm_sensors-3.3.4 > > i2cdetect -l shows the following: > > i2c-2 i2c OMAP I2C adapter I2C adapter > i2c-4 i2c OMAP I2C adapter I2C adapter > > > > I have two TI ADC128D818 ADCs (I cobbled together my own driver while > waiting for support -- I see that someone else requested the same driver > last week :-). They are: > adc128d818-i2c-4-1d > adc128d818-i2c-4-37 > > > "sensors --bus-list" shows the following: > bus "i2c-4" "OMAP I2C adapter" > Why bus i2c-2 isn't shown, I don't know. That's likely an issue. No it's not. sensors --bus-list only lists the relevant buses - the ones with at least one hwmon chip. There's no chip on i2c-2, so it's not listed. > I started using the following sensors.conf: > chip "adc128d818-i2c-4-1d" > label in0 "Test" > > However, I was told that the "bus was unidentified". > I used the following sensors.conf, hoping to fix the problem: > bus "i2c-4" "OMAP I2C adapter" > chip "adc128d818-i2c-4-1d" > label in0 "Test" > However, none in0 was not renamed as expected -- it printed as "in0". > > It turns out that sensors_substitute_busses() incorrectly renumbers > adc128d818-i2c-4-1d to adc128d818-i2c-2-1d and then can't find the > labels. The problem here is that your i2c bus driver is evil and attributes the same name to two different buses. It should not do that. This is a bug in the i2c-omap driver. It should give each bus a distinct name, for example by including the base I/O address in it. > If I use the following sensors.conf (or comment out the call to > sensors_substitute_busses()), sensors works as expected: > chip "adc128d818-i2c-*-1d" > label in0 "Test" Yes, this works because you don't actually need a bus statement in your case. It is only needed if the same I2C address is used by the same chip type on two or more I2C buses. In your case the bus type and the I2C address are sufficient to uniquely identify each chip. > Am I doing something wrong or is there something strange going on with > the bus renaming. What is the point of the bus renaming? As explained in the manual page, the I2C bus numbers are not guaranteed to be stable across reboots, while I2C bus names are. But of course the whole logic assumes unique I2C bus names, and you don't have that on your system at the moment. -- Jean Delvare _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
