Re: [PATCH 1/4] PM / EM: and devices to Energy Model

From: Dietmar Eggemann <dietmar.eggemann@arm.com>
To: Lukasz Luba <lukasz.luba@arm.com>, Quentin Perret <qperret@google.com>
Cc: nm@ti.com, juri.lelli@redhat.com, peterz@infradead.org,
	viresh.kumar@linaro.org, dri-devel@lists.freedesktop.org,
	bjorn.andersson@linaro.org, bsegall@google.com,
	alyssa.rosenzweig@collabora.com, festevam@gmail.com,
	Morten.Rasmussen@arm.com, robh@kernel.org,
	amit.kucheria@verdurent.com, vincent.guittot@linaro.org,
	khilman@kernel.org, agross@kernel.org, daniel.lezcano@linaro.org,
	steven.price@arm.com, cw00.choi@samsung.com, mingo@redhat.com,
	linux-imx@nxp.com, rui.zhang@intel.com, kernel-team@android.com,
	mgorman@suse.de, daniel@ffwll.ch, linux-pm@vger.kernel.org,
	linux-arm-msm@vger.kernel.org, s.hauer@pengutronix.de,
	rostedt@goodmis.org, linux-mediatek@lists.infradead.org,
	matthias.bgg@gmail.com, Chris.Redpath@arm.com,
	linux-omap@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
	airlied@linux.ie, javi.merino@arm.com,
	tomeu.vizoso@collabora.com, sboyd@kernel.org,
	shawnguo@kernel.org, rjw@rjwysocki.net,
	linux-kernel@vger.kernel.org, b.zolnierkie@samsung.com,
	kernel@pengutronix.de, sudeep.holla@arm.com,
	ionela.voinescu@arm.com
Subject: Re: [PATCH 1/4] PM / EM: and devices to Energy Model
Date: Tue, 21 Jan 2020 10:10:16 +0100	[thread overview]
Message-ID: <c9910c74-ab47-0c78-a1c7-4f3978e1dd09@arm.com> (raw)
In-Reply-To: <b02da0ed-9e0b-36db-9813-daa334cbf2ba@arm.com>

On 20/01/2020 19:38, Lukasz Luba wrote:
> 
> 
> On 1/20/20 6:27 PM, Dietmar Eggemann wrote:
>> On 20/01/2020 16:09, Quentin Perret wrote:
>>> Hey Lukasz,
>>>
>>> On Monday 20 Jan 2020 at 14:52:07 (+0000), Lukasz Luba wrote:
>>>> On 1/17/20 10:54 AM, Quentin Perret wrote:

[...]

>> It's true that we need the policy->cpus cpumask only for cpu devices and
>> we have it available when we call em_register_perf_domain()
>> [scmi-cpufreq.c driver] or the OPP wrapper dev_pm_opp_of_register_em()
>> [e.g. cpufreq-dt.c driver].
>>
>> And we shouldn't make EM code dependent on OPP.
>>
>> But can't we add 'struct cpumask *mask' as an additional argument to
>> both which can be set to NULL for (devfreq) devices?
>>
>> We can check in em_register_perf_domain() that we got a valid cpumask
>> for a cpu device and ignore it for (devfreq) devices.
>>
> 
> I think we could avoid this additional argument 'cpumask'. I have
> checked the cpufreq_cpu_get function, which should do be good for this:
> 
> ---------->8-------------------------
> static int _get_sharing_cpus(struct device *cpu_dev, struct cpumask *span)
> {
>         struct cpufreq_policy *policy;
> 
>         policy = cpufreq_cpu_get(cpu_dev->id);
>         if (policy) {
>                 cpumask_copy(span, policy->cpus);
>                 cpufreq_cpu_put(policy);
>                 return 0;
>         } else {
>                 return -EINVAL;
>         }
> }
> --------------------------8<-------------------------------
> 
> It would be a replacement for:
> ret = dev_pm_opp_get_sharing_cpus(dev, span);

True. But then we hard-code that a CPU device performance domain can
only be a frequency domain (which is true today).

The task scheduler (build_perf_domains()) and thermal are already using
cpufreq_cpu_get() to access the cpufreq policy. Now the EM framework
would too for CPU devices. I assume that could work with a couple of
adaptations in Documentation/power/energy-model.rst.

BTW, there is a similar interface cpufreq_get_policy() in cpufreq.c
which is used less often?

_______________________________________________
Linux-mediatek mailing list
Linux-mediatek@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-mediatek