Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 16 Nov 2022 14:55:04 +0100
From:   Marijn Suijten <marijn.suijten@somainline.org>
To:     =?utf-8?B?TsOtY29sYXMgRi4gUi4gQS4=?= Prado 
        <nfraprado@collabora.com>
Cc:     Rob Herring <robh+dt@kernel.org>,
        AngeloGioacchino Del Regno 
        <angelogioacchino.delregno@collabora.com>, kernel@collabora.com,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
        Matthias Brugger <matthias.bgg@gmail.com>,
        devicetree@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
        linux-kernel@vger.kernel.org, linux-mediatek@lists.infradead.org
Subject: Re: [PATCH] docs: dt: writing-schema: Document usage of CHECK_DTBS
 make flag
Message-ID: <20221116135504.mdmgm6ce2cynt5yt@SoMainline.org>
References: <20221102214300.309347-1-nfraprado@collabora.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20221102214300.309347-1-nfraprado@collabora.com>
Precedence: bulk

Hi N?colas,

On 2022-11-02 17:43:00, N?colas F. R. A. Prado wrote:
> It is possible to run checks on a Devicetree by passing the CHECK_DTBS
> flag when building. This is a useful shortcut to the dtbs_check make
> target since it avoids checking unrelated Devicetrees, which can take
> some time and is unnecessary if no bindings were modified. Document it.
> 
> Signed-off-by: N?colas F. R. A. Prado <nfraprado@collabora.com>
> 
> ---
> 
>  Documentation/devicetree/bindings/writing-schema.rst | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/Documentation/devicetree/bindings/writing-schema.rst b/Documentation/devicetree/bindings/writing-schema.rst
> index 4a381d20f2b4..55ad556472b4 100644
> --- a/Documentation/devicetree/bindings/writing-schema.rst
> +++ b/Documentation/devicetree/bindings/writing-schema.rst
> @@ -167,6 +167,13 @@ setting the ``DT_SCHEMA_FILES`` variable to a specific schema file or pattern.
>      make dt_binding_check DT_SCHEMA_FILES=/gpio/
>      make dtbs_check DT_SCHEMA_FILES=trivial-devices.yaml
>  
> +Note that ``make dtbs_check`` will validate every DT source file that is
> +enabled. When making changes to a DT but not to the bindings, a possible
> +shortcut to validate only the DT in question is to explicitly build it with
> +the ``CHECK_DTBS`` flag enabled. For example::
> +
> +    make CHECK_DTBS=y mediatek/mt8192-evb.dtb

I have a bit of trouble getting this to work on a _clean_ out directory
(perhaps this should have been reported at the original patch, I had
always been using Dmitry's version [1] which didn't suffer from this
problem).

Consider running with the following:

    rm out -r
    make ARCH=arm64 O=out defconfig
    make ARCH=arm64 O=out CHECK_DTBS=y qcom/sm8450-sony-xperia-nagara-pdx223.dtb

After compiling preliminaries, it exits with:

    make[3]: *** No rule to make target 'arch/arm64/boot/dts/qcom/sm8450-sony-xperia-nagara-pdx223.dtb'.  Stop.
    make[2]: *** [../scripts/Makefile.build:500: arch/arm64/boot/dts/qcom] Error 2
    make[1]: *** [/kernel/Makefile:1460: qcom/sm8450-sony-xperia-nagara-pdx223.dtb] Error 2
    make[1]: Leaving directory '/kernel/out'
    make: *** [Makefile:231: __sub-make] Error 2

However, if I lint all DTBs first by running `dtbs_check`, it seems the
schema preliminaries are built:

      LINT    Documentation/devicetree/bindings
      CHKDT   Documentation/devicetree/bindings/processed-schema.json
    ... bunch of warnings
      SCHEMA  Documentation/devicetree/bindings/processed-schema.json

And here I ctrl+c the build so that it doesn't run DTC_CHK over every
dts.  If I now re-run the original command on my .dtb of choice, it
completes successfully with the warnings that I expect.  Is the logic
behind `CHECK_DTBS=y` simply missing a step to make sure SCHEMA is built
and up-to-date?

Aside from not working in a clean output directly, could this imply
schema changes (edits in Documentation/devicetree/bindings) _are not_
propagated when running with `CHECK_DTBS=y?

At the same time running this command twice results in no output the
second time around, supposedly because the dtb has "already been built".
Is that also something we can improve?

[1]: https://lore.kernel.org/linux-arm-msm/20220623144357.297252-1-dmitry.baryshkov@linaro.org/

Thanks!
- Marijn