linux-kernel - Re: [PATCH] cpufreq: qcom-hw: Add support for per-core-dcvs

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: <20220224043007.2s3jrwork4zxxlsq@vireshk-i7>
Date:   Thu, 24 Feb 2022 10:00:07 +0530
From:   Viresh Kumar <viresh.kumar@...aro.org>
To:     Bjorn Andersson <bjorn.andersson@...aro.org>
Cc:     "Rafael J. Wysocki" <rafael@...nel.org>,
        linux-arm-msm@...r.kernel.org, linux-pm@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] cpufreq: qcom-hw: Add support for per-core-dcvs

On 09-02-22, 21:01, Bjorn Andersson wrote:
> The OSM and EPSS hardware controls the frequency of each cluster in the
> system based on requests from the OS and various limiting factors, such
> as input from LMH.
> 
> In most systems the vote from the OS is done using a single register per
> cluster, but some systems are configured to instead take one request per
> core.  In this configuration a set of consecutive registers are used for
> the OS to request the frequency of each of the cores within the cluster.
> The information is then aggregated in the hardware and the frequency for
> the cluster is determined.
> 
> As the current implementation ends up only requesting a frequency for
> the first core in each cluster and only the vote of non-idle cores are
> considered it's often the case that the cluster will be clocked (much)
> lower than expected.
> 
> It's possible that there are benefits of performing the per-core
> requests from the OS, but more investigation of the outcome is needed
> before introducing such support. As such this patch extends the request
> for the cluster to be written to all the cores.
> 
> The weight of the policy's related_cpus is used to determine how many
> cores, and hence consecutive registers, each cluster has.
> 
> The OS is not permitted to disable the per-core dcvs feature.
> 
> Signed-off-by: Bjorn Andersson <bjorn.andersson@...aro.org>
> ---
>  drivers/cpufreq/qcom-cpufreq-hw.c | 18 ++++++++++++++++++
>  1 file changed, 18 insertions(+)

Applied. Thanks.

-- 
viresh