Re: [PATCH 1/2] cpufreq: tegra186: Fix initial frequency

[Jon Hunter <jonathanh-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>]

On 14/07/2020 04:46, Viresh Kumar wrote:
> On 13-07-20, 17:37, Jon Hunter wrote:
>>
>> On 13/07/2020 04:25, Viresh Kumar wrote:
>>> On 12-07-20, 11:06, Jon Hunter wrote:
>>>> Commit 6cc3d0e9a097 ("cpufreq: tegra186: add
>>>> CPUFREQ_NEED_INITIAL_FREQ_CHECK flag") fixed CPUFREQ support for
>>>> Tegra186 but as a consequence the following warnings are now seen on
>>>> boot ...
>>>>
>>>>  cpufreq: cpufreq_online: CPU0: Running at unlisted freq: 0 KHz
>>>>  cpufreq: cpufreq_online: CPU0: Unlisted initial frequency changed to: 2035200 KHz
>>>>  cpufreq: cpufreq_online: CPU1: Running at unlisted freq: 0 KHz
>>>>  cpufreq: cpufreq_online: CPU1: Unlisted initial frequency changed to: 2035200 KHz
>>>>  cpufreq: cpufreq_online: CPU2: Running at unlisted freq: 0 KHz
>>>>  cpufreq: cpufreq_online: CPU2: Unlisted initial frequency changed to: 2035200 KHz
>>>>  cpufreq: cpufreq_online: CPU3: Running at unlisted freq: 0 KHz
>>>>  cpufreq: cpufreq_online: CPU3: Unlisted initial frequency changed to: 2035200 KHz
>>>>  cpufreq: cpufreq_online: CPU4: Running at unlisted freq: 0 KHz
>>>>  cpufreq: cpufreq_online: CPU4: Unlisted initial frequency changed to: 2035200 KHz
>>>>  cpufreq: cpufreq_online: CPU5: Running at unlisted freq: 0 KHz
>>>>  cpufreq: cpufreq_online: CPU5: Unlisted initial frequency changed to: 2035200 KHz
>>>>
>>>> Although we could fix this by adding a 'get' operator for the Tegra186
>>>> CPUFREQ driver, there is really little point because the CPUFREQ on
>>>> Tegra186 is set by writing a value stored in the frequency table to a
>>>> register and we just need to set the initial frequency.
>>>
>>> The hardware still runs at the frequency requested by cpufreq core here, right ?
>>
>> Yes.
>>
>>> It is better to provide the get() callback as it is also used to show the
>>> current frequency in userspace.
>>
>> I looked at that and I saw that if the get() callback is not provided,
>> the current frequency showed by userspace is policy->cur. For this
>> device, policy->cur is accurate and so if we added the get() callback we
>> essentially just going to return policy->cur. Therefore, given that we
>> already know policy->cur, I did not see the point in adding a device
>> specific handler to do the same thing.
> 
> The get() callback is supposed to read the frequency from hardware and
> return it, no cached value here. policy->cur may end up being wrong in
> case there is a bug.

OK, I can add a get callback. However, there are a few other drivers
that set the current frequency in the init() and don't implement a get()
callback ...

drivers/cpufreq/pasemi-cpufreq.c
drivers/cpufreq/ppc_cbe_cpufreq.c
drivers/cpufreq/intel_pstate.c

Jon

-- 
nvpublic