On Sat, Feb 26, 2011 at 05:43:49AM -0500, Jonathan Cameron wrote:
> On 02/25/11 21:32, Guenter Roeck wrote:
> > Hi Jonathan,
> >
> > On Fri, 2011-02-25 at 15:42 -0500, Jonathan Cameron wrote:
> >> One trivial formatting nitpick.
> >> Dug out datasheet for this one..
> >> I'm a little unclear how value from there map to those stored in here...
> > [ ... ]
> >
> >>> +static struct pmbus_driver_info max34440_info[] = {
> >>> + [max34440] = {
> >>> + .pages = 14,
> >>> + .direct[PSC_VOLTAGE_IN] = true,
> >>> + .direct[PSC_VOLTAGE_OUT] = true,
> >>> + .direct[PSC_TEMPERATURE] = true,
> >>> + .direct[PSC_CURRENT_OUT] = true,
> >>> + .m[PSC_VOLTAGE_IN] = 1,
> >>> + .b[PSC_VOLTAGE_IN] = 0,
> >>> + .R[PSC_VOLTAGE_IN] = 3,
> >>> + .m[PSC_VOLTAGE_OUT] = 1,
> >>> + .b[PSC_VOLTAGE_OUT] = 0,
> >>> + .R[PSC_VOLTAGE_OUT] = 3,
> >>> + .m[PSC_CURRENT_OUT] = 1,
> >>> + .b[PSC_CURRENT_OUT] = 0,
> >>> + .R[PSC_CURRENT_OUT] = 3,
> >> Table 3 of datasheet says R for current is 0... I may be missing
> >> something though!
> >
> > At least this one has a relatively simple reason ;).
> >
> > The base unit (to which R is applied to) in the datasheet is mV and mA.
> > The base unit used by max8688 and max16064 is V and A. Chips in linear
> > mode also use V and A. I had the choice of either adjusting the code to
> > expect mV and mA for all chips, or to expect V and A for all chips. I
> > decided to use the latter for consistency, and to adjust R in the core
> > driver to the units expected to be reported in the sysfs attributes. For
> > max34440/max34441, the adjustment ends up being 0 either way (3 + (-3)
> > in one direction and (-3) + 3 in the other).
> This reason actually occurred to me whilst cycling home but thanks for
> clarifying this.
> >
> > Another option would be to not do any sysfs related adjustments in the
> > core driver, but to do all adjustments via platform data instead. I
> > decided against that because the code to handle linear mode (which
> > expects V and A) would no longer match the code to handle direct mode,
> > which might cause even more confusion.
> What you have works fine for me though you'll have to keep a close eye
> on any drivers submitted by others as this will be easy to get wrong
> as people will blindly fill tables in from datasheets.
>
... though they should really test the code before submitting it.
But you are definitely right.
> Acked-by: Jonathan Cameron <jic23@xxxxxxxxx>
Thanks!
Guenter
_______________________________________________
lm-sensors mailing list
lm-sensors@xxxxxxxxxxxxxx
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
