linux-kernel - Re: [Patch 3/5] ACPI: CPPC: support updating epp, auto_sel and {min|max

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <b2bd3258-51bd-462a-ae29-71f1d6f823f3@nvidia.com>
Date: Fri, 14 Mar 2025 18:41:28 +0530
From: Sumit Gupta <sumitg@...dia.com>
To: Pierre Gondois <pierre.gondois@....com>, <rafael@...nel.org>,
	<viresh.kumar@...aro.org>, <lenb@...nel.org>, <robert.moore@...el.com>,
	<corbet@....net>, <linux-pm@...r.kernel.org>, <linux-acpi@...r.kernel.org>,
	<linux-doc@...r.kernel.org>, <acpica-devel@...ts.linux.dev>,
	<linux-kernel@...r.kernel.org>
CC: <linux-tegra@...r.kernel.org>, <treding@...dia.com>,
	<jonathanh@...dia.com>, <sashal@...dia.com>, <vsethi@...dia.com>,
	<ksitaraman@...dia.com>, <sanjayc@...dia.com>, <bbasu@...dia.com>, "Sumit
 Gupta" <sumitg@...dia.com>
Subject: Re: [Patch 3/5] ACPI: CPPC: support updating epp, auto_sel and
 {min|max_perf} from sysfs



On 24/02/25 15:54, Pierre Gondois wrote:
> External email: Use caution opening links or attachments
> 
> 
> Hello Sumit,
> 
> On 2/11/25 11:37, Sumit Gupta wrote:
>> Add support to update the CPC registers used for Autonomous
>> Performance Level Selection from acpi_cppc sysfs store nodes.
>> Registers supported for updation are:
>> - Engergy Performance Preference (EPP): energy_perf
>> - Autonomous Selection: auto_sel
>> - Maximum Performance: max_perf
>> - Minimum Performance: min_perf
>>
>> Also, enable show nodes to read of the following CPC registers:
>> - Performance Limited: perf_limited
>> - Autonomous Activity Window: auto_activity_window
>>
>> Signed-off-by: Sumit Gupta <sumitg@...dia.com>
>> ---
>>   drivers/acpi/cppc_acpi.c | 191 ++++++++++++++++++++++++++++++++++++---
>>   include/acpi/cppc_acpi.h |   5 +
>>   2 files changed, 183 insertions(+), 13 deletions(-)
>>
>> diff --git a/drivers/acpi/cppc_acpi.c b/drivers/acpi/cppc_acpi.c
>> index cc2bf958e84f..c60ad66ece85 100644
>> --- a/drivers/acpi/cppc_acpi.c
>> +++ b/drivers/acpi/cppc_acpi.c
> 
> [...]
> 
>>   sysfs_cppc_data(cppc_get_perf_caps, cppc_perf_caps, highest_perf, ro);
>>   sysfs_cppc_data(cppc_get_perf_caps, cppc_perf_caps, lowest_perf, ro);
>>   sysfs_cppc_data(cppc_get_perf_caps, cppc_perf_caps, nominal_perf, ro);
>> @@ -177,9 +304,16 @@ sysfs_cppc_data(cppc_get_perf_caps, 
>> cppc_perf_caps, lowest_nonlinear_perf, ro);
>>   sysfs_cppc_data(cppc_get_perf_caps, cppc_perf_caps, guaranteed_perf, 
>> ro);
>>   sysfs_cppc_data(cppc_get_perf_caps, cppc_perf_caps, lowest_freq, ro);
>>   sysfs_cppc_data(cppc_get_perf_caps, cppc_perf_caps, nominal_freq, ro);
>> +sysfs_cppc_data(cppc_get_perf_caps, cppc_perf_caps, auto_sel, rw);
>>
>>   sysfs_cppc_data(cppc_get_perf_fb_ctrs, cppc_perf_fb_ctrs, 
>> reference_perf, ro);
>>   sysfs_cppc_data(cppc_get_perf_fb_ctrs, cppc_perf_fb_ctrs, 
>> wraparound_time, ro);
>> +sysfs_cppc_data(cppc_get_perf_fb_ctrs, cppc_perf_fb_ctrs, 
>> perf_limited, ro);
>> +
>> +sysfs_cppc_data(cppc_get_perf_ctrls, cppc_perf_ctrls, min_perf, rw);
>> +sysfs_cppc_data(cppc_get_perf_ctrls, cppc_perf_ctrls, max_perf, rw);
> 
> IIUC, this means that users can modify the min/max performance levels of 
> the CPU
> without having the cpufreq framework notified. Meaning that if a user 
> modifies these
> levels, the frequency selection will be done using the initial min/max 
> performance
> level.
> I think it would be better not allow users to modifies these values 
> directly. Reliying
> on existing scaling_min_freq/scaling_max_freq files would be better IMO.
> 
> Regards,
> Pierre


Hi Pierre,

Sorry to get back late on this.

Providing the option to modify these registers from sysfs will help to
easily tune for Power and Performance.
When min/max_perf are updated or auto_sel is enabled from sysfs, the
policy min/max can also be updated with those values.
When auto_sel is disabled for a CPU from sysfs, then its policy min/max
can be updated back to the nominal_perf and lowest_non_linear_perf.
Was thinking of doing this in v2 after conclusion of [1].

[1] 
https://lore.kernel.org/lkml/f0f1b31b-a0fc-4d21-8b79-c896833dae35@nvidia.com/

Best Regards,
Sumit Gupta