Re: [PATCH] dt-bindings: clock: agilex: mark mpu_l2ram_clk deprecated

[Chan, Genevieve]

On 4/8/2026 3:49 pm, Krzysztof Kozlowski wrote:
> [You don't often get email from krzk@xxxxxxxxxx. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
>
> On Mon, Jul 27, 2026 at 05:49:26PM +0800, genevieve.chan@xxxxxxxxxx wrote:
> > From: Genevieve Chan <genevieve.chan@xxxxxxxxxx>
> >
> > Preliminary documentation referred to mpu_l2ram_clk, but that clock is
> > now mpu_ccu_clk and is already represented by AGILEX_MPU_CCU_CLK.
> > Keep AGILEX_MPU_L2RAM_CLK as a legacy ABI-preserved ID (do not use in
> > new DTs) and mark clock ID 42 as reserved/unused so it is not reused,
> > keeping all existing clock IDs stable for DT ABI compatibility.
> >
> > Fixes: 6b3c59780ed3 ("dt-bindings: documentation: add clock bindings information for Agilex")
> > Cc: stable@xxxxxxxxxxxxxxx # 5.8+
> > Signed-off-by: Genevieve Chan <genevieve.chan@xxxxxxxxxx>
> > ---
> >   include/dt-bindings/clock/agilex-clock.h | 2 ++
> >   1 file changed, 2 insertions(+)
> >
> > diff --git a/include/dt-bindings/clock/agilex-clock.h b/include/dt-bindings/clock/agilex-clock.h
> > index 06feca07e08e..de7a58a06651 100644
> > --- a/include/dt-bindings/clock/agilex-clock.h
> > +++ b/include/dt-bindings/clock/agilex-clock.h
> > @@ -44,6 +44,7 @@
> >
> >   /* Gate clocks */
> >   #define AGILEX_MPU_CLK                       30
> > +/* Legacy ABI ID; do not use in new DTs, use AGILEX_MPU_CCU_CLK */
> >   #define AGILEX_MPU_L2RAM_CLK         31
> >   #define AGILEX_MPU_PERIPH_CLK                32
> >   #define AGILEX_L4_MAIN_CLK           33
> > @@ -55,6 +56,7 @@
> >   #define AGILEX_CS_TIMER_CLK          39
> >   #define AGILEX_S2F_USER0_CLK         40
> >   #define AGILEX_EMAC0_CLK             41
> > +/* 42 - reserved/unused */
>
> How IDs can be reserved? These are implemented by drivers (abstract IDs
> not used by hardware), so how driver could reserve it?
>
> Best regards,
> Krzysztof
You are right. These IDs are DT clock specifier indices (provider one-cell ABI), not hardware IDs. The clk driver registers clocks into hws[id]; unused indices remain -ENOENT.

ID 42 was never defined in the original Agilex binding (EMAC0=41, EMAC1=43). The comment was placed to only documents that historical hole so we do not renumber later IDs. It is not a hardware reservation — it is for ABI hygiene.

Happy to reword to something clearer, e.g. /* 42 never assigned; do not renumber */, or drop the comment if preferred.