On Fri, Feb 10, 2012 at 08:38:40AM +0100, Jean Delvare wrote: > On Thu, 9 Feb 2012 18:13:35 -0800, Guenter Roeck wrote: > > On Thu, Feb 09, 2012 at 05:54:52PM -0500, Nikolaus Schulz wrote: > > > Unfortunately, the F75387 uses the very same registers for both modes. > > > The values are only interpreted differently, depending on the open vs > > > closed loop mode, either as RPM values, or raw PWM. > > > > Other chips do the same, or something similar. For example, the NCT6775F > > uses the same register for the target speed in one mode, and for the target > > temperature in another. [...] > In this specific case, what you want to do is decouple the register > cache from the actual hardware registers. In your per-device private > data structure, have room to store all the settings for all mode. This > means that a given hardware register can have 2 or more corresponding > members in the data structure, one for each meaning it can have > depending on the mode. [...] > When reading the register value from the chip, you write them to the > right struct member for the given mode, and leave the other untouched. > When values are changed by the user through sysfs, you store them in > the data structure. If they relate to the currently selected mode, you > write them to the chip immediately. If not, you keep them for later. > > When the user changes the mode through sysfs, you write that mode > change to the chip, and then automatically program the new mode using > the settings from the data structure. Very elegant and convincing. My thinking obviously has been too tied to the device specific details, thanks for your guidance. :) I will apply this to the f75375 driver. Nikolaus _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
