Re: [PATCH v1 3/3] cpufreq: Make cpufreq_frequency_table_verify() internal

[Zihuan Zhang <zhangzihuan@kylinos.cn>]

在 2025/9/4 12:45, Viresh Kumar 写道:
> On 04-09-25, 11:22, Zihuan Zhang wrote:
>> diff --git a/drivers/cpufreq/sh-cpufreq.c b/drivers/cpufreq/sh-cpufreq.c
>> index 642ddb9ea217..ee3fd1e71b90 100644
>> --- a/drivers/cpufreq/sh-cpufreq.c
>> +++ b/drivers/cpufreq/sh-cpufreq.c
>> @@ -90,10 +90,8 @@ static int sh_cpufreq_verify(struct cpufreq_policy_data *policy)
>>   {
>>   	struct clk *cpuclk = &per_cpu(sh_cpuclk, policy->cpu);
>>   
>> -	if (policy->freq_table)
>> -		return cpufreq_frequency_table_verify(policy);
>> -
>> -	cpufreq_verify_within_cpu_limits(policy);
>> +	if (!cpufreq_generic_frequency_table_verify(policy))
>> +		return 0;
>>   
>>   	policy->min = (clk_round_rate(cpuclk, 1) + 500) / 1000;
>>   	policy->max = (clk_round_rate(cpuclk, ~0UL) + 500) / 1000;
>> diff --git a/drivers/cpufreq/virtual-cpufreq.c b/drivers/cpufreq/virtual-cpufreq.c
>> index 6ffa16d239b2..2498f40cd57e 100644
>> --- a/drivers/cpufreq/virtual-cpufreq.c
>> +++ b/drivers/cpufreq/virtual-cpufreq.c
>> @@ -249,10 +249,7 @@ static int virt_cpufreq_offline(struct cpufreq_policy *policy)
>>   
>>   static int virt_cpufreq_verify_policy(struct cpufreq_policy_data *policy)
>>   {
>> -	if (policy->freq_table)
>> -		return cpufreq_frequency_table_verify(policy);
>> -
>> -	cpufreq_verify_within_cpu_limits(policy);
>> +	cpufreq_generic_frequency_table_verify(policy);
>>   	return 0;
>>   }
> You ended up changing the logic of both these files and it isn't the
> same anymore. Earlier if freq_table was there and
> cpufreq_frequency_table_verify() failed, we used to return, but not
> anymore.
>
> And we don't return the error anymore for virtual driver.


Thanks for pointing this out, Viresh.

You are correct — with the current changes, if
cpufreq_generic_frequency_table_verify() fails, we no longer return
an error in these drivers. For drivers that lack a frequency table,
they can still operate, which is why I wasn’t sure whether returning
an error is strictly necessary.

I would appreciate your advice here: should we preserve the old
behavior and return an error on failure, or is it acceptable for
drivers without a table to continue without it?


>