Re: [PATCH 2/3] dt-bindings: spmi: Add SmartDV SPMI controller

From: VAMSHI GAJJELA

Date: Sun Nov 15 2026 - 11:43:54 EST


On Tue, Nov 3, 2026 at 11:47 PM Conor Dooley <conor@xxxxxxxxxx> wrote:
>
> On Tue, Nov 03, 2026 at 09:56:06AM +0530, vamshi gajjela wrote:
> > From: Vamshi Gajjela <vamshigajjela@xxxxxxxxxx>
> >
> > Document the device tree bindings for the SmartDV IP Solutions MIPI SPMI
> > master controller ("smartdv,spmi").
> >
> > This controller IP is integrated in Google Tensor SoCs starting with the
> > G5 generation (codename: laguna, canonically written as "lga"), so add
> > "google,lga-spmi" with a fallback to "smartdv,spmi".
> >
> > Signed-off-by: Vamshi Gajjela <vamshigajjela@xxxxxxxxxx>
> > ---
> > .../bindings/spmi/smartdv,spmi.yaml | 126 ++++++++++++++++++
> > MAINTAINERS | 6 +
> > 2 files changed, 132 insertions(+)
> > create mode 100644 Documentation/devicetree/bindings/spmi/smartdv,spmi.yaml
> >
> > diff --git a/Documentation/devicetree/bindings/spmi/smartdv,spmi.yaml b/Documentation/devicetree/bindings/spmi/smartdv,spmi.yaml
> > new file mode 100644
> > index 000000000000..e0512e60d932
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/spmi/smartdv,spmi.yaml
> > @@ -0,0 +1,126 @@
> > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/spmi/smartdv,spmi.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > +
> > +title: SmartDV IP Solutions SPMI Controller
> > +
> > +maintainers:
> > + - Vamshi Gajjela <vamshigajjela@xxxxxxxxxx>
> > +
> > +description:
> > + SmartDV IP Solutions MIPI System Power Management Interface (SPMI) master
> > + controller supporting MIPI SPMI Specification versions 1.0 and 2.0.
> > +
> > +allOf:
> > + - $ref: spmi.yaml#
> > +
> > +properties:
> > + compatible:
> > + oneOf:
> > + - items:
> > + - const: google,lga-spmi
>
> > + - const: smartdv,spmi
> > + - const: smartdv,spmi
>
> I hate compatibles for purchasable IP like this, particularly when
> unversioned as they maybe become effectively useless when a new revision
> releases. The standalone smartdv,spmi compatible has to go at least,
> soc-specific compatibles are mandatory.

Agreed. In v2 I will drop "smartdv,spmi" (both standalone and as a
fallback) and use only the SoC-specific "google,lga-spmi" compatible,
renaming the binding to google,lga-spmi.yaml and dropping patch 1/3.

>
> > +
> > + reg:
> > + maxItems: 1
> > +
> > + clocks:
> > + description:
> > + Core clock to the controller from which SPMI SCL is derived. Optional if
> > + the controller uses a free-running clock and clock-frequency is specified.
> > + maxItems: 1
> > +
> > + resets:
> > + maxItems: 1
> > +
> > + clock-frequency:
> > + description:
> > + Input core clock frequency in Hz when the controller uses a free-running
> > + clock not described in clocks.
> > +
> > + smartdv,master-arb-enable:
> > + $ref: /schemas/types.yaml#/definitions/uint32
> > + enum: [0, 1]
> > + default: 0
> > + description:
> > + Enable master bus arbitration phase (0 = disable, 1 = enable).
>
> What aspect of board design governs the decision here?
> Why can it not be determined from the soc specific compatible?

These parameters do not vary between boards on LGA (this applies
to all smartdv,* properties below). In v2 I will drop all
of them from the binding and move the LGA controller configuration into
of_device_id match data in the driver keyed off "google,lga-spmi"

>
> > +
> > + smartdv,master-id:
> > + $ref: /schemas/types.yaml#/definitions/uint32
> > + minimum: 0
> > + maximum: 3
> > + default: 0
> > + description:
> > + Master ID for the controller on the SPMI bus.
>
> What aspect of board design governs the decision here?
> Why can it not be determined from the soc specific compatible?
>
>
> > +
> > + smartdv,spmi-version:
> > + $ref: /schemas/types.yaml#/definitions/uint32
> > + enum: [0, 1]
> > + default: 0
> > + description:
> > + MIPI SPMI specification version used for transfers (0 for MIPI SPMI v1.0,
> > + 1 for MIPI SPMI v2.0).
>
> What aspect of board design governs the decision here?
> Why can it not be determined from the soc specific compatible?
>
>
> > +
> > + smartdv,arb-retry-limit:
> > + $ref: /schemas/types.yaml#/definitions/uint32
> > + minimum: 0
> > + maximum: 255
> > + default: 32
> > + description:
> > + Maximum number of times the master retries a command when losing bus
> > + arbitration.
>
> What aspect of board design governs the decision here?
> Why can it not be determined from the soc specific compatible?
>
>
> > +
> > + smartdv,prescaler:
> > + $ref: /schemas/types.yaml#/definitions/uint32
> > + minimum: 0
> > + maximum: 255
> > + default: 0
> > + description:
> > + Clock divider value to generate SPMI SCL from the input core clock. Writing
> > + 1 divides the system clock by 2; more generally, a non-zero value n
> > + divides the system clock by 2 * n.
>
> What aspect of board design governs the decision here?
> Why can it not be determined from the soc specific compatible?
>
>
> > +
> > + smartdv,read-latency:
> > + $ref: /schemas/types.yaml#/definitions/uint32
> > + minimum: 0
> > + maximum: 255
> > + default: 32
> > + description:
> > + Number of SPMI SCL clock cycles to pause during bus turnaround on read
> > + commands to allow the target device to fetch data.
>
> What aspect of board design governs the decision here?
> Why can it not be determined from the soc specific compatible?
>
>
> > +
> > + smartdv,arb-scl-delay:
> > + $ref: /schemas/types.yaml#/definitions/uint32
> > + minimum: 0
> > + maximum: 3
> > + default: 0
> > + description:
> > + Delay between arbitration and Sequence Start Condition (SSC) when the
> > + controller is the Bus Owner Master (BOM).
>
> What aspect of board design governs the decision here?
> Why can it not be determined from the soc specific compatible?
>
>
> > +
> > +required:
> > + - compatible
> > + - reg
>
> I find it hard to believe that this is the set of required properties.

Agreed. In v2 I will make clocks and resets required alongside compatible
and reg, and drop the remaining optional properties.

Thanks,
Vamshi Gajjela