On Tue, 2011-02-08 at 13:03 -0500, Ian Dobson wrote: > > -------------------------------------------------- > From: "Guenter Roeck" <guenter.roeck@xxxxxxxxxxxx> > Sent: Tuesday, February 08, 2011 6:10 PM > To: <lm-sensors@xxxxxxxxxxxxxx> > Subject: tempX_source sysfs attribute needed ? > > > Hi all, > > > > in the recent months, I have seen two instances where a sysfs attribute > > identifying the source associated with a temperature sensor would have > > been > > helpful. > > > > One is max6639. For this chip, the source of the second temperature > > channel > > can be configured to be local or remote. > > > > The others are W83677HG-B, NCT6775F, and NCT6776F. Those chips have four > > to six > > temperature channels with a configurable source. > > > > Current approach is to stick with whatever is configured by the BIOS for > > the Nuvoton > > chips, and to use the local source for the max6639 driver. For the Nuvoton > > chips, > > my prototype driver for NCT6775F and NCT6776F reports the temperature > > source > > in tempX_label. > > > > Possible solutions might be: > > - Stick with the current situation, ie do nothing. > > - Use module parameters. Doesn't really work well for max6639 since it > > affects > > all instances of the driver, and if the driver is built into the kernel. > > - Use platform data. Might work for max6639, but not for the Nuvoton > > chips. > > - Define a new sysfs attribute to make the temperature source > > configurable, > > and to report the current selection. > > > > Any thoughts/comments ? > > > > Thanks, > > Guenter > > > > _______________________________________________ > > lm-sensors mailing list > > lm-sensors@xxxxxxxxxxxxxx > > http://lists.lm-sensors.org/mailman/listinfo/lm-sensors > > Hi Guenter, > > I actually think it's a good idea (tempX_label), I've already seen that > sensors can use this information if you don't have a label defined in > sensors3.conf. It actually makes is alot easier to configure the > sensors3.conf file correctly. I'm not sure if _label is the correct name, it > really describes the source of the signal. > tempX_label is a textual description, so I think it does make sense for this driver. > Also with the nuv chips what happens if you change the source for a > temperature input that's used by smartfan? The BIOS should know what's best > and playing with the fan control inputs could cause all sorts of problems. > During my hacking in the nct6776f driver I managed to screw up the smartfan > control so much that the temperature setpoint for one of the fans was 90Âc > and the fan stopped. > Yes, that is always a risk. I don't really know if tempX_source adds any real value - that is why I was asking for feedback. Ultimately the information is very board specific, so providing tempX_label for the Nuvoton chips and using - if ever needed - platform data for the max6639 may well be good enough. Thanks, Guenter _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
