[PATCH v2 03/18] ACPI: processor: thermal: Use __free(put_cpufreq_policy) for policy reference

[Zihuan Zhang]

在 2025/8/28 17:40, Rafael J. Wysocki 写道:
> On Wed, Aug 27, 2025 at 4:33 AM Zihuan Zhang <zhangzihuan at kylinos.cn> wrote:
>> Replace the manual cpufreq_cpu_put() with __free(put_cpufreq_policy)
>> annotation for policy references. This reduces the risk of reference
>> counting mistakes and aligns the code with the latest kernel style.
>>
>> No functional change intended.
>>
>> Signed-off-by: Zihuan Zhang <zhangzihuan at kylinos.cn>
>> ---
>>   drivers/acpi/processor_thermal.c | 12 +++---------
>>   1 file changed, 3 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/acpi/processor_thermal.c b/drivers/acpi/processor_thermal.c
>> index 1219adb11ab9..f99ed0812934 100644
>> --- a/drivers/acpi/processor_thermal.c
>> +++ b/drivers/acpi/processor_thermal.c
>> @@ -64,17 +64,13 @@ static int phys_package_first_cpu(int cpu)
>>
>>   static int cpu_has_cpufreq(unsigned int cpu)
>>   {
>> -       struct cpufreq_policy *policy;
>> +       struct cpufreq_policy *policy __free(put_cpufreq_policy);
>>
>>          if (!acpi_processor_cpufreq_init)
>>                  return 0;
>>
>>          policy = cpufreq_cpu_get(cpu);
>> -       if (policy) {
>> -               cpufreq_cpu_put(policy);
>> -               return 1;
>> -       }
>> -       return 0;
>> +       return !!policy;
> If you want to make this change, please also change the return type of
> the function to bool.
Thanks for pointing this out.
>>   }
>>
>>   static int cpufreq_get_max_state(unsigned int cpu)
>> @@ -95,7 +91,7 @@ static int cpufreq_get_cur_state(unsigned int cpu)
>>
>>   static int cpufreq_set_cur_state(unsigned int cpu, int state)
>>   {
>> -       struct cpufreq_policy *policy;
>> +       struct cpufreq_policy *policy __free(put_cpufreq_policy);
> This isn't correct AFAICS at least formally because the scope of the
> variable is the whole function, so it won't get out of scope at the
> point where you want cpufreq_cpu_put() to be called.
>
> The policy variable should be defined in the block following the "for"
> loop (and actually all of the local variables except for "i" can be
> defined there).


Sorry for the mistake — I did this correctly in other places, but forgot 
here.

> Or better still, please move that block to a separate function
> containing all of the requisite local variable definitions and call
> that function for each online CPU.


  In fact, I have realized that we cannot always use __free for cleanup 
directly.

The issue is that the release only happens at the end of the variable’s 
lifetime, while in some cases we want to drop the reference immediately 
after use.

To address this, I’m considering introducing a helper macro in 
include/linux/cpufreq.h that would make this more explicit and allow 
safe cleanup at the right point.


Before moving forward, I’d like to hear your opinion on this approach:

#define WITH_CPUFREQ_POLICY(cpu) \
for(struct cpufreq_policy *policy __free(put_cpufreq_policy) = \
     cpufreq_cpu_get(cpu);;)


Then we can use it for all code :

	WITH_CPUFREQ_POLICY(cpu) {
			if(!policy)
				return XXX; // error handing
			
			//code use policy here
		} // equal origin 'cpufreq_cpu_put' here
         ;;
        //left code

>>          struct acpi_processor *pr;
>>          unsigned long max_freq;
>>          int i, ret;
>> @@ -127,8 +123,6 @@ static int cpufreq_set_cur_state(unsigned int cpu, int state)
>>                  max_freq = (policy->cpuinfo.max_freq *
>>                              (100 - reduction_step(i) * cpufreq_thermal_reduction_pctg)) / 100;
>>
>> -               cpufreq_cpu_put(policy);
>> -
>>                  ret = freq_qos_update_request(&pr->thermal_req, max_freq);
>>                  if (ret < 0) {
>>                          pr_warn("Failed to update thermal freq constraint: CPU%d (%d)\n",
>> --