On Fri, 2017-05-05 at 09:18 +0800, Shawn Guo wrote: > On Thu, May 04, 2017 at 04:34:14PM +0200, Marek Vasut wrote: > > If you model the > > power distribution correctly, the OPP hackery can be removed.
> The OPP hackery can be removed even without reg_arm/reg_soc modeling. > That's why we can do hackery dropping and reg_arm/reg_soc modeling in > separate patches. > > @Leonard, if someday we support 'LDO bypass' mode in upstream kernel, > the OPP hackery needs to be back in some way even with reg_arm/reg_soc > modeling in place, right? Or will we have a better way to ensure SW1A > rail can always feed a correct voltage directly to reg_arm®_soc? Maybe? Or maybe the cpufreq driver could detect this situation and handle it internally. In the vendor tree this the cpufreq driver has a special fsl,arm-soc-shared property for this anyway. I posted another RFC at upstreaming ldo-bypass recently but it did not handle this particular case of a shared input rail. It can be handled separately. See: https://lkml.org/lkml/2017/3/22/640 But getting that series to an acceptable state might take a long time. -- Regards, Leonard
