Received-SPF: pass (google.com: domain of linux-kernel+bounces-84521-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Message-ID: <feb78f0d-1759-9478-f945-1f48107f6992@quicinc.com>
Date: Wed, 28 Feb 2024 10:37:36 +0530
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.11.0
Subject: Re: [PATCH 0/2] Fix per-policy boost behavior
Content-Language: en-US
To: Jie Zhan <zhanjie9@hisilicon.com>
CC: <sudeep.holla@arm.com>, <cristian.marussi@arm.com>,
        <linux-kernel@vger.kernel.org>, <linux-arm-msm@vger.kernel.org>,
        <quic_rgottimu@quicinc.com>, <linux-arm-kernel@lists.infradead.org>,
        <asahi@lists.linux.dev>, <linux-pm@vger.kernel.org>,
        <dietmar.eggemann@arm.com>, <marcan@marcan.st>, <sven@svenpeter.dev>,
        <alyssa@rosenzweig.io>, <rafael@kernel.org>, <viresh.kumar@linaro.org>,
        <xuwei5@hisilicon.com>
References: <20240227165309.620422-1-quic_sibis@quicinc.com>
 <91118802-eb8c-6225-3610-05e16270b3c4@hisilicon.com>
From: Sibi Sankar <quic_sibis@quicinc.com>
In-Reply-To: <91118802-eb8c-6225-3610-05e16270b3c4@hisilicon.com>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 8bit


On 2/28/24 07:36, Jie Zhan wrote:
> Hi Sibi,
> 
> Thanks for pointing this issue out.
> 
> However, I can't clearly see how the existing code fails.
> 
> cpufreq_frequency_table_cpuinfo() checks cpufreq_boost_enabled(),
> and that should be already set in cpufreq_boost_trigger_state() before
> calling cpufreq_boost_set_sw(), so presumably cpufreq_boost_set_sw()
> is supposed to work as expected.
> 
> Can you explain this a bit further?

In the existing code, per-policy flags doesn't have any impact i.e.
if cpufreq_driver boost is enabled and one or more of the per-policy
boost is disabled, the cpufreq driver will behave as if boost is
enabled. The second issue was just book keeping, meaning some drivers
enable boost by default, however the per-policy boost flags are set
as disabled during boot.

-Sibi

> 
> Cheers,
> Jie
> 
> On 28/02/2024 00:53, Sibi Sankar wrote:
>> Fix per-policy boost behavior by incorporating per-policy boost flag
>> in the policy->max calculation and setting the correct per-policy
>> boost_enabled value on devices that use cpufreq_enable_boost_support().
>>
>> Logs reported-by Dietmar Eggemann [1]:
>>
>> [1] 
>> https://lore.kernel.org/lkml/265e5f2c-9b45-420f-89b1-44369aeb8418@arm.com/

you can also have a look at ^^ thread for more info.

>>
>> Sibi Sankar (2):
>>    cpufreq: Fix per-policy boost behavior on SoCs using
>>      cpufreq_boost_set_sw
>>    cpufreq: apple-soc: Align per-policy and global boost flags
>>
>>   drivers/cpufreq/apple-soc-cpufreq.c |  1 +
>>   drivers/cpufreq/cpufreq.c           | 15 +++++++++------
>>   drivers/cpufreq/freq_table.c        |  2 +-
>>   3 files changed, 11 insertions(+), 7 deletions(-)
>>
>