On Wed, Aug 14, 2019 at 4:28 PM André Roth <neolynx@xxxxxxxxx> wrote: > > > Hi Adam, > > > The Logic PD Torpedo dm3730 boots with a few messages that concern me: > > > > [ 2.824188] omap2_set_init_voltage: unable to find boot up OPP for > > vdd_mpu_iva > > [ 2.831481] omap2_set_init_voltage: unable to set vdd_mpu_iva > > [ 2.837280] omap2_set_init_voltage: unable to find boot up OPP for vdd_core > > [ 2.844329] omap2_set_init_voltage: unable to set vdd_core > > I noticed the same on my dm3730 and found that the order in which TWL, > SmartReflex and the omap init functions are called was wrong. > > See my patch set here: > https://marc.info/?l=linux-omap&m=156462291906717&w=2 > > With this I was able to get rid of half of those messages... > > > I noticed that abb_mpu_iva has a bunch of OPP's, but it's not clear to > > me how or what should power this and/or how the regulator is is > > related (if at all) > > It seems to be more complex than I assumed. In short, proper Dynamic > voltage and frequency scaling (DVFS) is required, which is currently not implemented. There happened to be a patch pushed yesterday that I didn't see: https://patchwork.kernel.org/patch/11094653/ It seems to fix the issue I was seeing. > > > The torpedo is using the TWL4030 PMIC, and I would expect those > > voltages to be acceptable to the twl4030 > > I found one patch [1] which enables SmartReflex on the twl4030, which > this dm3730 ran without RAM errors at 1GHz. > > However, I have learned that these patches are not enough, and would > reduce the chips lifetime without DVFS active. What makes DVFS different from what cpufreq does by changing the operating voltage and frequency? adam > > Kind Regards, > > André > > [1] https://marc.info/?l=linux-omap&m=156462292206718&w=2
