Re: [PATCH v2 3/9] dt-bindings: clock: ipq5332: drop the few nss clocks definition

[Dmitry Baryshkov]

On Wed, 22 Nov 2023 at 12:19, Kathiravan Thirumoorthy
<quic_kathirav@xxxxxxxxxxx> wrote:
>
>
>
> On 11/22/2023 3:42 PM, Krzysztof Kozlowski wrote:
> > On 22/11/2023 11:08, Kathiravan Thirumoorthy wrote:
> >>
> >>
> >> On 11/21/2023 8:36 PM, Krzysztof Kozlowski wrote:
> >>> On 21/11/2023 15:30, Kathiravan Thirumoorthy wrote:
> >>>> In commit 0dd3f263c810 ("clk: qcom: ipq5332: enable few nssnoc clocks in
> >>>
> >>> Where is this commit coming from?
> >>>
> >>>> driver probe"), gcc_snoc_nssnoc_clk, gcc_snoc_nssnoc_1_clk,
> >>>> gcc_nssnoc_nsscc_clk are enabled in driver probe to keep it always-on.
> >>>
> >>> Implementation can change and for example bring back these clocks. Are
> >>> you going to change bindings? No, drop the patch.
> >>>
> >>> Bindings should be dropped only in a few rare cases like clocks not
> >>> available for OS or bugs.
> >>
> >> Thanks Krzysztof. Will drop this patch in V3.
> >>
> >> One more question to understand further. In IPQ SoCs there are bunch of
> >> coresight / QDSS clocks but coresight framework doesn't handle all
> >> clocks. Those clocks are enabled in bootloader stage itself. In such
> >> case, should I drop the clocks from both binding and driver or only from
> >> driver?
> >
> > That's not really the reason to drop them at all. Neither from driver,
> > nor from bindings. You should not rely on bootloader handling your clocks
>
>
> Thanks, lets say if those clocks are not needed at all by OS since QDSS
> is not used and needed only for the boot loaders to access the
> corresponding address space, in such case what can be done? I
> understand, at first those clocks should not have been added to the driver.

First, what is QDSS? Yet another acronym?

Second, if they are not used now, they can get used later.

-- 
With best wishes
Dmitry