On 10-05-19, 08:47, Andy Tang wrote: > + Viresh for help. > > > -----Original Message----- > > From: Daniel Lezcano <daniel.lezcano@xxxxxxxxxx> > > Sent: 2019年5月10日 15:17 > > To: Andy Tang <andy.tang@xxxxxxx>; Shawn Guo <shawnguo@xxxxxxxxxx> > > Cc: Leo Li <leoyang.li@xxxxxxx>; robh+dt@xxxxxxxxxx; > > mark.rutland@xxxxxxx; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; > > devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; > > linux-pm@xxxxxxxxxxxxxxx; rui.zhang@xxxxxxxxx; edubezval@xxxxxxxxx > > Subject: Re: [EXT] Re: [PATCH v6] arm64: dts: ls1088a: add one more thermal > > zone node > > > > Caution: EXT Email > > > > On 10/05/2019 05:40, Andy Tang wrote: > > >> -----Original Message----- > > >> From: Shawn Guo <shawnguo@xxxxxxxxxx> > > >> Sent: 2019年5月10日 11:14 > > >> To: Andy Tang <andy.tang@xxxxxxx> > > >> Cc: Leo Li <leoyang.li@xxxxxxx>; robh+dt@xxxxxxxxxx; > > >> mark.rutland@xxxxxxx; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; > > >> devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; > > >> linux-pm@xxxxxxxxxxxxxxx; daniel.lezcano@xxxxxxxxxx; > > >> rui.zhang@xxxxxxxxx; edubezval@xxxxxxxxx > > >> Subject: [EXT] Re: [PATCH v6] arm64: dts: ls1088a: add one more > > >> thermal zone node > > >> > > >> Caution: EXT Email > > >> > > >> On Tue, Apr 23, 2019 at 10:25:07AM +0800, Yuantian Tang wrote: > > >>> Ls1088a has 2 thermal sensors, core cluster and SoC platform. Core > > >>> cluster sensor is used to monitor the temperature of core and SoC > > >>> platform is for platform. The current dts only support the first sensor. > > >>> This patch adds the second sensor node to dts to enable it. > > >>> > > >>> Signed-off-by: Yuantian Tang <andy.tang@xxxxxxx> > > >>> --- > > >>> v6: > > >>> - add cooling device map to cpu0-7 in platform node. > > > I like to explain a little. I think it makes sense that multiple thermal zone > > map to same cooling device. > > > In this way, no matter which thermal zone raises a temp alarm, it can call > > cooling device to chill out. > > > I also asked cpufreq maintainer about the cooling map issue, he think it > > would be fine. Yes, you asked me and I said it should be okay. > > > I have tested and no issue found. > > > > > > Daniel, what's your thought? > > > > If there are multiple thermal zones, they will be managed by different > > instances of a thermal governor. Each instances will act on the shared cooling > > device and will collide in their decisions: > > > > - If the sensors are closed, their behavior will be similar regarding the > > temperature. The governors may take the same decision for the cooling > > device. But in such case having just one thermal zone managed is enough. > > > > - If the sensors are not closed, their behavior will be different regarding the > > temperature. The governors will take different decision regarding the cooling > > device (one will decrease the freq, other will increase the freq). > > > > As the thermal governors are not able to manage several thermal zones and > > there is one cooling device (the cpu cooling device), this setup won't work as > > expected IMO. > > > > The setup making sense is having a thermal zone per 'cluster' and a cooling > > device per 'cluster'. That means the platform has one clock line per 'cluster'. > > The thermal management happens in a self-contained thermal zone (one > > cooling device - one governor - one thermal zone). > > > > In the case of HMP, other combinations are possible to be optimal. But not sure how I missed the obvious, though I do remember thinking about this. So the problem is that the cpu_cooling driver will get requests in parallel to set different max frequencies and the last call will always win and may result in undesired outcome. Sorry about creating the confusion. -- viresh
