RE: [Patch V2 3/4] spi: dt-bindings: Add Tegra TPM wait polling flag

[Krishna Yarlagadda]

> -----Original Message-----
> From: Rob Herring <robh+dt@xxxxxxxxxx>
> Sent: 04 February 2023 01:20
> To: Krishna Yarlagadda <kyarlagadda@xxxxxxxxxx>
> Cc: broonie@xxxxxxxxxx; peterhuewe@xxxxxx; jgg@xxxxxxxx;
> jarkko@xxxxxxxxxx; krzysztof.kozlowski+dt@xxxxxxxxxx; linux-
> spi@xxxxxxxxxxxxxxx; linux-tegra@xxxxxxxxxxxxxxx; linux-
> integrity@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> thierry.reding@xxxxxxxxx; Jonathan Hunter <jonathanh@xxxxxxxxxx>;
> Sowjanya Komatineni <skomatineni@xxxxxxxxxx>; Laxman Dewangan
> <ldewangan@xxxxxxxxxx>
> Subject: Re: [Patch V2 3/4] spi: dt-bindings: Add Tegra TPM wait polling flag
> 
> External email: Use caution opening links or attachments
> 
> 
> On Fri, Feb 3, 2023 at 7:02 AM Krishna Yarlagadda
> <kyarlagadda@xxxxxxxxxx> wrote:
> >
> > Add "nvidia,wait-polling" flag to enable TCG TIS hardware flow control.
> 
> Tell me something that the diff doesn't.
> 
> >
> > Signed-off-by: Krishna Yarlagadda <kyarlagadda@xxxxxxxxxx>
> > ---
> >  .../bindings/spi/nvidia,tegra210-quad-peripheral-props.yaml | 6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> > diff --git a/Documentation/devicetree/bindings/spi/nvidia,tegra210-quad-
> peripheral-props.yaml
> b/Documentation/devicetree/bindings/spi/nvidia,tegra210-quad-peripheral-
> props.yaml
> > index 2c3cada75339..19d2b30cadbf 100644
> > --- a/Documentation/devicetree/bindings/spi/nvidia,tegra210-quad-
> peripheral-props.yaml
> > +++ b/Documentation/devicetree/bindings/spi/nvidia,tegra210-quad-
> peripheral-props.yaml
> > @@ -29,4 +29,10 @@ properties:
> >      minimum: 0
> >      maximum: 255
> >
> > +  nvidia,wait-polling:
> > +    description:
> > +      Enable TPM wait polling feature for QSPI as specified in TCG PC Client
> > +      Specific TPM Interface Specification (TIS).
> > +    $ref: /schemas/types.yaml#/definitions/flag
> 
> Why do you need this flag when you have a compatible that also
> indicates you have a quirk.
> 
> If this a TPM feature, why is it enabled for every single SPI slave device?
> 
> If the fundamental issue is the controller only supports half-duplex,
> why can't you just check that from the driver? Can't the SPI subsystem
> tell you that the host controller is half-duplex? Though sometimes
> that may be board level property I suppose. If so, define the h/w
> quirk, not the driver mode in DT. Half-duplex is probably something
> everyone could use, not just Nvidia.
> 
> Please discuss this series internally with the folks you marked as
> maintainers. It has issues I'm sure they would have also pointed out.
> 
> Rob
QSPI is a multi-chip-select controller and HW wait polling is only for TPM
Both controller and device would need a flag/setting to identify support
for this feature. Using SPI controller flags and SPI device mode flags to
avoid device tree flags. Moved HW/alternate TPM flow control into existing
driver. No need of new compatible now.

KY