On Tue, Jun 01, 2021 at 10:29:48AM +0200, Václav Kubernát wrote: > Hello, > [ short reply - I am on the road. I'll look into your replies in more detail later ] > I have tested your patches. It seems like the driver is now a lot more > usable, thanks. What seems to be still unusable for me, is RPM mode. > The chip creates an effort to set the RPM, but it almost never > stabilizes. I think the reason is mainly because the driver doesn't > work with registers like "window" and "pwm rate of change". PWM mode > seems to work fine. > Agreed; I found the same problem. I actually tried to implement support for an attribute to support the window registers, but could not find working settings. I only played with window registers, though, so we'll probably need attributes for both window mode and pwm rate of change. > I think some of the patches in my patch series on the chip are now > obsolete. However, I do think at least fanX_div should be implemented. > When testing with Sunon PF36281BX-000U-S99 (its max RPM is 23000), the > default divisor is not enough (one bit of change equals to about 500 > RPM). The only way to change the divisor right now, is to set the > target RPM or min RPM. > Agreed. > There is also the regmap patch, but I've implemented that one mainly > because it made the driver a bit easier to debug. > Only reason for me to omit that was because it was buggy (FWIW, I didn't know either that the defalt caching mode is "do not cache", and that all the 'volatile' declarations were useless). Guenter
