On Tue, 17 Aug 2021 at 11:32, Maulik Shah [off-list ref] wrote:

Hi Daniel/Rafael,

 >> would it make sense to convert the governors into modules

i am not aware if this was not pursued further due to any issue.

Do you see any concerns to allow existing governors compiled as loadable
module?
if not i can work on same and post. please do let me know your thoughts
on this.

i have CCed Saravana and Todd for awareness.

I suggest you re-submit a new patch and put some arguments of why this
is useful for us in the commit message. Then we can discuss that
instead, rather than this old patch.

Kind regards
Uffe

Thanks,
Maulik

On 10/10/2020 1:56 AM, Daniel Lezcano wrote:

quoted

Hi Rafael,

On 22/09/2020 19:27, Rafael J. Wysocki wrote:

quoted

Hi Lina,

On Tue, Sep 22, 2020 at 6:12 PM Lina Iyer [off-list ref] wrote:

quoted

Hi Rafael,

On Tue, Sep 22 2020 at 10:00 -0600, Rafael J. Wysocki wrote:

quoted

Sorry for the delay.

On Wed, Sep 9, 2020 at 12:15 AM Lina Iyer [off-list ref] wrote:

quoted

Commit 83788c0caed3 ("cpuidle: remove unused exports") removed
capability of registering cpuidle governors, which was unused at that
time. By exporting the symbol, let's allow platform specific modules to
register cpuidle governors and use cpuidle_governor_latency_req() to get
the QoS for the CPU.

Which platform-specific modules may want to do that and why?

We are planning a custom cpuidle governor for QCOM SoCs. With Android,
the idea is to make them loadable modules so they can be in a separate
partition.

Well, the $subject patch is not applicable without a mainline user
requiring this, so it needs to be posted along with that user.

Putting apart the custom cpuidle governor mentioned above, would it make
sense to convert the governors into modules ? It is pointless to have
all of them compiled in, especially with distros doing make
allmodconfig, no?

--
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation