On Tue, 21 Oct 2008 15:43:23 -0700, Darrick J. Wong wrote: > On Tue, Oct 21, 2008 at 02:21:46PM +0200, Bernhard Rieder wrote: > > Hi Everybody, > > > > I am currently working on an ADT746[78] driver which will be finished soon. > > > > However, I am not sure how to react to errors. Which values do I return in case > > the chip reports errors (i.e. temp=-64 is an error value for the temperature reading, > > -63 is ok). Are there somewhere error values defined which can used for this? I do > > not think usinf -64 is a good choice, because the chip has two modes for temperature > > and an error would be reported as -128 in the other mode. Additionally, -64 could be > > interpreted as a real value (at least in sibiria when it is a cold winter day). > > > > Are there somewhere constants like SENSORS_VOLTAGE_INVALID, SENSORS_TEMP_INVALID, ... > > which I missed? I think common error values would be really beneficial ... > > Hm... I think the lm85/adt7468 have similar registers. Oddly, I was > working on a patch to the lm85 driver to add 7468 support. I suspect we > should collaborate. :) The lm85 driver is already pretty big. I'd only add support for chips which are 100% compatible with an already supported chip to it. If the ADT7468 and ADT7467 have specific properties (missing or extra inputs, different scaling factors, extra features...) I'd rather have a new driver for them (which might as well be derived from lm85.c, but then please make sure you start from the very last version as I fixed it and cleaned it up extensively in the past few months.) > Also, a driver for the adt7467 is (mostly?) implemented in > drivers/macintosh/therm_adt746x.c, though it relies on finding the > device via some sort of openfirmware calls, which isn't appropriate for > PCs. The macintosh driver also doesn't implement the standard sysfs interface and I don't think it supports voltage inputs. I could never convince the powerpc people to drop their custom interface in favor of ours. Maybe with the advent of new-style I2C drivers and if we implement an in-kernel interface for temperature value access, they will reconsider... -- Jean Delvare
