Re: [PATCH v4 1/2] dt-bindings: media: i2c: Add IMX296 CMOS sensor binding

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Sakari,

On Thu, Oct 31, 2019 at 07:08:37PM +0200, Sakari Ailus wrote:
> On Thu, Oct 31, 2019 at 06:54:44PM +0200, Laurent Pinchart wrote:
> > On Thu, Oct 31, 2019 at 04:28:17PM +0200, Sakari Ailus wrote:
> >> On Thu, Oct 31, 2019 at 04:11:41PM +0200, Laurent Pinchart wrote:
> >>> On Thu, Oct 31, 2019 at 07:15:12PM +0530, Manivannan Sadhasivam wrote:
> >>>> On Thu, Oct 31, 2019 at 03:15:38PM +0200, Laurent Pinchart wrote:
> >>>>> On Wed, Oct 30, 2019 at 03:19:01PM +0530, Manivannan Sadhasivam wrote:
> >>>>>> Add YAML devicetree binding for IMX296 CMOS image sensor. Let's also
> >>>>>> add MAINTAINERS entry for the binding and driver.
> >>>>>> 
> >>>>>> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> >>>>>> ---
> >>>>>>  .../devicetree/bindings/media/i2c/imx296.yaml | 94 +++++++++++++++++++
> >>>>>>  MAINTAINERS                                   |  8 ++
> >>>>>>  2 files changed, 102 insertions(+)
> >>>>>>  create mode 100644 Documentation/devicetree/bindings/media/i2c/imx296.yaml
> >>>>>> 
> >>>>>> diff --git a/Documentation/devicetree/bindings/media/i2c/imx296.yaml b/Documentation/devicetree/bindings/media/i2c/imx296.yaml
> >>>>>> new file mode 100644
> >>>>>> index 000000000000..c04ec2203268
> >>>>>> --- /dev/null
> >>>>>> +++ b/Documentation/devicetree/bindings/media/i2c/imx296.yaml
> >>>>>> @@ -0,0 +1,94 @@
> >>>>>> +# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
> >>>>>> +%YAML 1.2
> >>>>>> +---
> >>>>>> +$id: http://devicetree.org/schemas/media/i2c/imx296.yaml#
> >>>>>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> >>>>>> +
> >>>>>> +title: Sony IMX296 1/2.8-Inch CMOS Image Sensor
> >>>>>> +
> >>>>>> +maintainers:
> >>>>>> +  - Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> >>>>>> +
> >>>>>> +description: |-
> >>>>>> +  The Sony IMX296 is a 1/2.9-Inch active pixel type CMOS Solid-state image
> >>>>>> +  sensor with square pixel array and 1.58 M effective pixels. This chip
> >>>>>> +  features a global shutter with variable charge-integration time. It is
> >>>>>> +  programmable through I2C and 4-wire interfaces. The sensor output is
> >>>>>> +  available via CSI-2 serial data output (1 Lane).
> >>>>>> +
> >>>>>> +properties:
> >>>>>> +  compatible:
> >>>>>> +    const: sony,imx296
> >>>>>> +
> >>>>>> +  reg:
> >>>>>> +    maxItems: 1
> >>>>>> +
> >>>>>> +  clocks:
> >>>>>> +    maxItems: 1
> >>>>>> +
> >>>>>> +  clock-names:
> >>>>>> +    description:
> >>>>>> +      Input clock for the sensor.
> >>>>>> +    items:
> >>>>>> +      - const: mclk
> >>>>> 
> >>>>> The pin is named INCK, let's name the clock accordingly.
> >>>> 
> >>>> Okay, I thought generic names are preferred here!
> >>>>  
> >>>>>> +  clock-frequency:
> >>>>>> +    description:
> >>>>>> +      Frequency of the mclk clock in Hertz.
> >>>>> 
> >>>>> This shouldn't be needed, you can retrieve the clock frequency at
> >>>>> runtime from the clock source.
> >>>> 
> >>>> Unless the clock source is a fixed one! What if the clock source comes from
> >>>> SoC? We need to set the rate, right?
> >>> 
> >>> In that case, if you want to hardcode the clock in DT, the preferred way
> >>> is to use the assigned-clock-rates property. Otherwise, if the driver
> >>> requires a specific clock frequency, it's better to hardcode it in the
> >>> driver itself. In this specific case, I think assigned-clock-rates is
> >>> best as the device can support three different clock frequencies.
> >> 
> >> Just note that if ACPI support is added to the sensor driver, you'll need
> >> the clock-frequency property again, for that's the only way how the driver
> >> will get the clock frequency.
> > 
> > Why is so ? Why can't we implement of assigned-clock-rates for ACPI ?
> 
> ACPI doesn't deal with clocks as such. So there's also no ACPI defined way
> to access clocks specifically, including the frequency --- instead the
> clock is controlled by an AML methods which implement power on and off
> sequences for the device.

It's a shortcoming of ACPI, which should be addressed at the ACPI level.
We shouldn't polute the DT bindings with a clock-frequency property for
this reason.

-- 
Regards,

Laurent Pinchart

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel