Re: [PATCH v14 5/5] cpufreq: Only disable boost during cpu online when using frequency tables

[Mario Limonciello <mario.limonciello@amd.com>]

On 6/25/2024 01:30, Viresh Kumar wrote:
> On 24-06-24, 16:34, Mario Limonciello wrote:
>> The behavior introduced in commit f37a4d6b4a2c ("cpufreq: Fix per-policy
>> boost behavior on SoCs using cpufreq_boost_set_sw()") sets up the boost
>> policy incorrectly when boost has been enabled by the platform firmware
>> initially even if a driver sets the policy up.
>>
>> This is because there are no discrete entries in the frequency table.
>> Update that code to only run when a frequency table is present.
>>
>> Fixes: f37a4d6b4a2c ("cpufreq: Fix per-policy boost behavior on SoCs using cpufreq_boost_set_sw()")
>> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
>> ---
>> Cc: Sibi Sankar <quic_sibis@quicinc.com>
>> Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
>> Cc: Viresh Kumar <viresh.kumar@linaro.org>
>> Cc: Dhruva Gole <d-gole@ti.com>
>> Cc: Yipeng Zou <zouyipeng@huawei.com>
>> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
>>   drivers/cpufreq/cpufreq.c | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
>> index 1fdabb660231..32c119614710 100644
>> --- a/drivers/cpufreq/cpufreq.c
>> +++ b/drivers/cpufreq/cpufreq.c
>> @@ -1430,7 +1430,8 @@ static int cpufreq_online(unsigned int cpu)
>>   		}
>>   
>>   		/* Let the per-policy boost flag mirror the cpufreq_driver boost during init */
>> -		policy->boost_enabled = cpufreq_boost_enabled() && policy_has_boost_freq(policy);
>> +		if (policy->freq_table)
>> +			policy->boost_enabled = cpufreq_boost_enabled() && policy_has_boost_freq(policy);
> 
> I am not sure if I understand the problem properly here. Can you
> please explain a bit in detail ?
> 

The core issue is that there are drivers that have boost functionality 
but don't have a frequency table.  As pointed out by Gautham there are 
also drivers that have a frequency table but don't advertise boost 
pstates (CPUFREQ_BOOST_FREQ isn't set on any frequency).

So what happens is the driver may have choosen a policy to have boost 
enabled but when cpufreq_online() is called it gets "marked" disabled 
from this check introduced in f37a4d6b4a2c even though it's previously 
enabled.