Are dt-bindings headers part of the device tree spec?

Are dt-bindings headers part of the device tree spec?

[Sean Anderson]

Hi all,

I'm reviewing a clock driver [1], and the submitter has deviated from
the defines used by Linux. For example, where Linux might have

	#define CLOCK_FOOBAR 5

his driver might have

	#define CLK_FUBAR 6

Which means that both the device tree source and the resulting device
tree binary will be different.

As I understand it, we try to be compatible with Linux on these things.
However, it is unclear to me if include/dt-bindings is also part of
this, or just the things in Documentation/devicetree/bindings. And if it
is, do we need to have compatible sources, compatible binaries, or both?
In general, I think we should try to have the same headers as well, but
is it permitted to allow deviations with reasonable justification?

--Sean

[1] https://lore.kernel.org/u-boot/de6e75a083647dfeec3058dd4dcc0419b08e155c.1637285375.git.weijie.gao@mediatek.com/

Re: Are dt-bindings headers part of the device tree spec?

[Tom Rini]

[-- Attachment #1: Type: text/plain, Size: 1060 bytes --]

On Wed, Dec 15, 2021 at 10:53:26AM -0500, Sean Anderson wrote:
> 
> Hi all,
> 
> I'm reviewing a clock driver [1], and the submitter has deviated from
> the defines used by Linux. For example, where Linux might have
> 
> 	#define CLOCK_FOOBAR 5
> 
> his driver might have
> 
> 	#define CLK_FUBAR 6
> 
> Which means that both the device tree source and the resulting device
> tree binary will be different.
> 
> As I understand it, we try to be compatible with Linux on these things.
> However, it is unclear to me if include/dt-bindings is also part of
> this, or just the things in Documentation/devicetree/bindings. And if it
> is, do we need to have compatible sources, compatible binaries, or both?
> In general, I think we should try to have the same headers as well, but
> is it permitted to allow deviations with reasonable justification?
> 
> --Sean
> 
> [1] https://lore.kernel.org/u-boot/de6e75a083647dfeec3058dd4dcc0419b08e155c.1637285375.git.weijie.gao@mediatek.com/

Adding a few more people to the thread.

-- 
Tom

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 659 bytes --]

Re: Are dt-bindings headers part of the device tree spec?

[Simon Glass]

Hi Sean,

On Wed, 15 Dec 2021 at 08:53, Sean Anderson <seanga2@gmail.com> wrote:
>
>
> Hi all,
>
> I'm reviewing a clock driver [1], and the submitter has deviated from
> the defines used by Linux. For example, where Linux might have
>
>         #define CLOCK_FOOBAR 5
>
> his driver might have
>
>         #define CLK_FUBAR 6
>
> Which means that both the device tree source and the resulting device
> tree binary will be different.
>
> As I understand it, we try to be compatible with Linux on these things.
> However, it is unclear to me if include/dt-bindings is also part of
> this, or just the things in Documentation/devicetree/bindings. And if it
> is, do we need to have compatible sources, compatible binaries, or both?
> In general, I think we should try to have the same headers as well, but
> is it permitted to allow deviations with reasonable justification?

I think they should be the same, including the name and value. It gets
too confusing otherwise.

If there are conversions needed for U-Boot, perhaps they could be
handled by some #defines in the driver or another file in U-Boot.

Regards,
Simon

>
> --Sean
>
> [1] https://lore.kernel.org/u-boot/de6e75a083647dfeec3058dd4dcc0419b08e155c.1637285375.git.weijie.gao@mediatek.com/