[PATCH 0/4] clk: Provide option to unprepare unused clocks at late init

Ulf Hansson ulf.hansson at linaro.org
Thu Jan 24 15:41:48 EST 2013


Hi Mike,

On 24 January 2013 19:49, Mike Turquette <mturquette at linaro.org> wrote:
> Quoting Ulf Hansson (2013-01-24 08:45:52)
>> From: Ulf Hansson <ulf.hansson at linaro.org>
>>
>> The disable_unused sequence executed at late init, is already handling the
>> fast unused ungated clocks to be gated. This patchset extends this sequence to
>> include the slow unused prepared clocks to be unprepared.
>>
>> The default behavior will not change in this patchset. To unprepare unused
>> clocks during the disable_unused sequence, the clk_hw needs to implement
>> the new optional callback, is_prepared.
>>
>> The motivation for this patchset is to save power. Clocks that is from
>> bootloaders prepared|enabled, but not used should be unprepared|disabled.
>>
>> Patch 4 is specific for ux500, which implements the is_prepared callback
>> for it's PRMCU clocks.
>>
>> Note that patch 1-3 has been sent earlier, but since a proof of concept
>> patch could be useful, as patch 4 is, the hole patchset is resent.
>>
>
> Hi Ulf,
>
> I had some minor discussion in patch #1.  Otherwise patches #2 & #3 look
> good to me.
>
> Patch #4 is a bit interesting.  There is nothing wrong with it per se,
> but the general layering of the prcmu power stuff is the opposite of how
> I had envisioned things.  Your .prepare callback is making calls to some
> pm_qos api but a quick grep did not reveal any definitions for those
> functions, just some stubs in include/linux/mfd/dbx500-prcmu.h.
>
> Do these functions scale voltage?  I had figured that some day
> per-device pm qos apis would call the clk api, not the other way around.

You are right, these scale voltages. I would like to keep the code as
is, just as a reminder that some devices using some clocks will
requires these kind of operations.
On our internal STE development track this is how it is used but we a
are moving towards newer APIs.

So I am definitely sharing the same thought as you do and my plan is
to remove this code along the road, once drivers is adapted properly.

>
> Anyways that is just food for thought and no reason to block patch #4.
>
> Regards,
> Mike
>
>> Ulf Hansson (4):
>>   clk: Introduce optional is_prepared callback
>>   clk: Unprepare the unused prepared slow clocks at late init
>>   clk: Introduce optional unprepare_unused callback
>>   clk: ux500: Support is_prepared callback for clk-prcmu
>>
>>  drivers/clk/clk.c             |   53 ++++++++++++++++
>>  drivers/clk/ux500/clk-prcmu.c |  134 ++++++++++++++++++++++++-----------------
>>  include/linux/clk-provider.h  |   11 ++++
>>  3 files changed, 144 insertions(+), 54 deletions(-)
>>
>> --
>> 1.7.10

Thanks for the review comments!

Kind regards
Ulf Hansson



More information about the linux-arm-kernel mailing list