[lm-sensors] [PATCH v2 2/2] hwmon: (lm63) Add support for LM96163

[Guenter Roeck]

On Mon, 2011-11-21 at 11:10 -0500, Jean Delvare wrote:
> Hi Guenter,
> 
> On Mon, 21 Nov 2011 05:48:25 -0800, Guenter Roeck wrote:
> > On Mon, Nov 21, 2011 at 08:07:07AM -0500, Jean Delvare wrote:
> > > On Mon, 21 Nov 2011 10:43:39 +0100, Jean Delvare wrote:
> > > > What about bit 3 (USF) in this enhanced configuration register? When
> > > > set, it would affect the way we encode and decode _max and _crit
> > > > temperature limits, right?
> > > 
> > > BTW, if you agree but don't have time to work on this, I would be
> > > perfectly fine with a check at probe time that would complain (and
> > > eventually bail out) if the way the chip is configured is not properly
> > > supported by the driver. If anyone needs the feature, it can be added
> > > later.
> > > 
> > Gives me something to do if I need some distraction ;). It will need
> > some thinking, so I may choose the complain option for now and submit
> > a patch later. Bailing out seems a bit harsh - as are temperatures
> > above 127 C anyway.
> 
> Temperatures above 127°C are frequent in high-end graphics cards. I
> would never want any of these in my own machines, but gamers do.
> 
Ok, guess that makes sense. Still a bit harsh, though.

> I am more surprised by the implementation. I'd have expected a single
> configuration but to change the range of all temperature-related
> registers, not just the high and critical limits. Other chips do this.
> 
There is a new set of registers reporting unsigned values for the
external temperature. Of course, that makes the implementation even more
tricky.

> > Do you remember why you did not add support for the remote temperature
> > offset registers ?
> 
> Probably because nobody ever asked for it. Also, originally we did not
> have a standard sysfs file name for this feature, so it was often left
> out of driver submissions. For some drivers it was added later when
> a standard name was defined, for others it was not. Feel free to add it
> if you want. I don't expect any difficulty.
> 
No problem, just wondering. I may add it just to simplify testing the
unsigned temperature configuration.

Thanks,
Guenter