Re: [PATCH v4 2/3] PM / EM: Expose perf domain struct

[Quentin Perret]

On Wednesday 15 May 2019 at 11:56:30 (+0200), Daniel Lezcano wrote:
> On 15/05/2019 11:17, Quentin Perret wrote:
> > Hi Daniel,
> > 
> > On Wednesday 15 May 2019 at 11:06:18 (+0200), Daniel Lezcano wrote:
> >> On 15/05/2019 10:23, Quentin Perret wrote:
> >>> In the current state, the perf_domain struct is fully defined only when
> >>> CONFIG_ENERGY_MODEL=y. Since we need to write code that compiles both
> >>> with or without that option in the thermal framework, make sure to
> >>> actually define the struct regardless of the config option. That allows
> >>> to avoid using stubbed accessor functions all the time in code paths
> >>> that use the EM.
> >>>
> >>> Acked-by: Viresh Kumar <viresh.kumar@xxxxxxxxxx>
> >>> Signed-off-by: Quentin Perret <quentin.perret@xxxxxxx>
> >>
> >> This patch implies the cpu cooling device can be set without the energy
> >> model.
> >>
> >> Isn't it possible to make a strong dependency for the cpu cooling device
> >> on the energy model option, add the energy model as default on arm arch
> >> and drop this patch?
> > 
> > Right, that should work too.
> > 
> >> After all, the cpu cooling is using the em framework.
> > 
> > The reason I did it that way is simply to keep things flexible. If you
> > don't compile in THERMAL_GOV_POWER_ALLOCATOR, you will never use the EM
> > for CPU thermal. So I thought it would be good to not mandate compiling
> > in ENERGY_MODEL in this case -- that should save a bit of space.
> > 
> > But TBH I don't have a strong opinion on this one, so if everybody
> > agrees it's fine to just make CPU_THERMAL depend on ENERGY_MODEL, I'm
> > happy to drop this patch and fix patch 3/3. That would indeed simplify
> > things a bit.
> 
> Ok in this case it will be better to drop the 2/3 and add a small series
> doing for the cpu_cooling.c
> 
> #ifdef CONFIG_THERMAL_GOV_POWER_ALLOCATOR
> 
> /* structure freq */
> 
> /* power2state */
> 
> /* state2power*/
> 
> /* getrequestedpower */
> 
> /* All functions needed for the above */
> 
> #endif
> 
> static struct thermal_cooling_device_ops cpufreq_cooling_ops = {
>         .get_max_state          = cpufreq_get_max_state,
>         .get_cur_state          = cpufreq_get_cur_state,
>         .set_cur_state          = cpufreq_set_cur_state,
> #ifdef CONFIG_THERMAL_GOV_POWER_ALLOCATOR
>         .get_requested_power    = cpufreq_get_requested_power,
>         .state2power            = cpufreq_state2power,
>         .power2state            = cpufreq_power2state,
> #endif
> };
> 
> So you don't have to care about ENERGY_MODEL to be set as
> THERMAL_GOV_POWER_ALLOCATOR depends on it.
> 
> I think the result for cpu_cooling.c will be even more cleaner with the
> em change.

OK, that works for me. I'll give it a go in v5.

Thanks !
Quentin