Re: [TEGRA194_CPUFREQ Patch 2/3] cpufreq: Add Tegra194 cpufreq driver

[Sumit Gupta <sumitg@nvidia.com>]

On 09/04/20 1:14 PM, Viresh Kumar wrote:
> External email: Use caution opening links or attachments
> 
> 
> On 08-04-20, 16:54, sumitg wrote:
>>
>>
>> On 08/04/20 11:23 AM, Viresh Kumar wrote:
>>> External email: Use caution opening links or attachments
>>>
>>>
>>> On 07-04-20, 23:48, sumitg wrote:
>>>> On 06/04/20 8:25 AM, Viresh Kumar wrote:
>>>>> On 05-04-20, 00:08, sumitg wrote:
>>>>>> On 26/03/20 5:20 PM, Viresh Kumar wrote:
>>>>>>> On 03-12-19, 23:02, Sumit Gupta wrote:
>>>>>>>> diff --git a/drivers/cpufreq/tegra194-cpufreq.c b/drivers/cpufreq/tegra194-cpufreq.c
>>>>>>>> +static unsigned int tegra194_get_speed_common(u32 cpu, u32 delay)
>>>>>>>> +{
>>>>>>>> +     struct read_counters_work read_counters_work;
>>>>>>>> +     struct tegra_cpu_ctr c;
>>>>>>>> +     u32 delta_refcnt;
>>>>>>>> +     u32 delta_ccnt;
>>>>>>>> +     u32 rate_mhz;
>>>>>>>> +
>>>>>>>> +     read_counters_work.c.cpu = cpu;
>>>>>>>> +     read_counters_work.c.delay = delay;
>>>>>>>> +     INIT_WORK_ONSTACK(&read_counters_work.work, tegra_read_counters);
> 
> Initialize the work only once from init routine.
> 
We are using "read_counters_work" as local variable. So every invocation 
the function will have its own copy of counters for corresponding cpu. 
That's why are doing INIT_WORK_ONSTACK here.

>>>>>>>> +     queue_work_on(cpu, read_counters_wq, &read_counters_work.work);
>>>>>>>> +     flush_work(&read_counters_work.work);
>>>>>>>
>>>>>>> Why can't this be done in current context ?
>>>>>>>
>>>>>> We used work queue instead of smp_call_function_single() to have long delay.
>>>>>
>>>>> Please explain completely, you have raised more questions than you
>>>>> answered :)
>>>>>
>>>>> Why do you want to have long delays ?
>>>>>
>>>> Long delay value is used to have the observation window long enough for
>>>> correctly reconstructing the CPU frequency considering noise.
>>>> In next patch version, changed delay value to 500us which in our tests is
>>>> considered reliable.
>>>
>>> I understand that you need to put a udelay() while reading the freq from
>>> hardware, that is fine, but why do you need a workqueue for that? Why can't you
>>> just read the values directly from the same context ?
>>>
>> The register to read frequency is per core and not accessible to other
>> cores. So, we have to execute the function remotely as the target core to
>> read frequency might be different from current.
>> The functions for that are smp_call_function_single or queue_work_on.
>> We used queue_work_on() to avoid long delay inside ipi interrupt context
>> with interrupts disabled.
> 
> Okay, I understand this now, finally :)
> 
> But if the interrupts are disabled during some call, won't workqueues face the
> same problem ?
> 
Yes, we are trying to minimize the case.

> --
> viresh
>