[PATCH v5] dt-bindings: net: can: convert ifi_canfd to DT schema

[PATCH v5] dt-bindings: net: can: convert ifi_canfd to DT schema

[Quchaosheng]

Convert the IFI CAN-FD controller binding from free-form text to DT
schema, and reference the common can-controller.yaml schema.

The driver matches on the "ifi,canfd-1.0" compatible and obtains its
interrupt through platform_get_irq(), so interrupts is required.

The example uses the GIC and irq header defines instead of the raw
numbers from the original text, which keeps the values unchanged
(GIC_SPI is 0, IRQ_TYPE_NONE is 0 and 1 is IRQ_TYPE_EDGE_RISING). The
node name is changed from "canfd" to "can" so that it matches the
$nodename pattern required by can-controller.yaml.

---
v2:
 - Correct the description: this is a soft IP block typically synthesised
   into an FPGA or CPLD, not a controller found on Renesas SoCs.
   Wording now follows the driver's Kconfig entry.

v3:
 - Use unevaluatedProperties: false instead of additionalProperties: false.
   As can-controller.yaml is referenced through allOf, additionalProperties
   would have rejected properties defined by that common schema, such as
   termination-gpios and termination-ohms.

v4:
 - Use defines instead of raw numbers in the example, and rename the
   example node from canfd to can. No functional change.

v5:
 - Re-send the patch on its own. The v4 posting accidentally included
   unrelated changes to the xilinx-xadc binding and did not remove
   ifi_canfd.txt. Only the intended change is present now.

Signed-off-by: Quchaosheng <quchaosheng000406@163.com>
---
 .../bindings/net/can/ifi,canfd.yaml           | 46 +++++++++++++++++++
 .../devicetree/bindings/net/can/ifi_canfd.txt | 15 ------
 2 files changed, 46 insertions(+), 15 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/net/can/ifi,canfd.yaml
 delete mode 100644 Documentation/devicetree/bindings/net/can/ifi_canfd.txt

diff --git a/Documentation/devicetree/bindings/net/can/ifi,canfd.yaml b/Documentation/devicetree/bindings/net/can/ifi,canfd.yaml
new file mode 100644
index 000000000..63f982e73
--- /dev/null
+++ b/Documentation/devicetree/bindings/net/can/ifi,canfd.yaml
@@ -0,0 +1,46 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/net/can/ifi,canfd.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: IFI CAN-FD Controller
+
+maintainers:
+  - Quchaosheng <quchaosheng000406@163.com>
+
+description:
+  The I/F/I CAN-FD soft IP block connected to the platform bus. It is most
+  often synthesised into an FPGA or CPLD.
+
+allOf:
+  - $ref: can-controller.yaml#
+
+properties:
+  compatible:
+    const: ifi,canfd-1.0
+
+  reg:
+    maxItems: 1
+
+  interrupts:
+    maxItems: 1
+
+required:
+  - compatible
+  - reg
+  - interrupts
+
+unevaluatedProperties: false
+
+examples:
+  - |
+    #include <dt-bindings/interrupt-controller/arm-gic.h>
+    #include <dt-bindings/interrupt-controller/irq.h>
+
+    can@ff220000 {
+        compatible = "ifi,canfd-1.0";
+        reg = <0xff220000 0x1000>;
+        interrupts = <GIC_SPI 43 IRQ_TYPE_NONE>;
+    };
+...
diff --git a/Documentation/devicetree/bindings/net/can/ifi_canfd.txt b/Documentation/devicetree/bindings/net/can/ifi_canfd.txt
deleted file mode 100644
index 20ea5c70a..000000000
--- a/Documentation/devicetree/bindings/net/can/ifi_canfd.txt
+++ /dev/null
@@ -1,15 +0,0 @@
-IFI CANFD controller
---------------------
-
-Required properties:
-  - compatible: Should be "ifi,canfd-1.0"
-  - reg: Should contain CAN controller registers location and length
-  - interrupts: Should contain IRQ line for the CAN controller
-
-Example:
-
-	canfd0: canfd@ff220000 {
-		compatible = "ifi,canfd-1.0";
-		reg = <0xff220000 0x00001000>;
-		interrupts = <0 43 0>;
-	};
-- 
2.43.0

Re: [PATCH v5] dt-bindings: net: can: convert ifi_canfd to DT schema

[Krzysztof Kozlowski]

On 14/09/2026 04:21, Quchaosheng wrote:
> Convert the IFI CAN-FD controller binding from free-form text to DT
> schema, and reference the common can-controller.yaml schema.
> 
> The driver matches on the "ifi,canfd-1.0" compatible and obtains its
> interrupt through platform_get_irq(), so interrupts is required.
> 
> The example uses the GIC and irq header defines instead of the raw
> numbers from the original text, which keeps the values unchanged
> (GIC_SPI is 0, IRQ_TYPE_NONE is 0 and 1 is IRQ_TYPE_EDGE_RISING). The
> node name is changed from "canfd" to "can" so that it matches the
> $nodename pattern required by can-controller.yaml.
> 
> ---
> v2:
>  - Correct the description: this is a soft IP block typically synthesised
>    into an FPGA or CPLD, not a controller found on Renesas SoCs.
>    Wording now follows the driver's Kconfig entry.
> 
> v3:
>  - Use unevaluatedProperties: false instead of additionalProperties: false.
>    As can-controller.yaml is referenced through allOf, additionalProperties
>    would have rejected properties defined by that common schema, such as
>    termination-gpios and termination-ohms.
> 
> v4:
>  - Use defines instead of raw numbers in the example, and rename the
>    example node from canfd to can. No functional change.
> 
> v5:
>  - Re-send the patch on its own. The v4 posting accidentally included
>    unrelated changes to the xilinx-xadc binding and did not remove
>    ifi_canfd.txt. Only the intended change is present now.

You send this way too many times. Please slow down.

Best regards,
Krzysztof

Re: [PATCH v5] dt-bindings: net: can: convert ifi_canfd to DT schema

[Quchaosheng]

Hi Krzysztof,

You are right, and sorry for the noise. I have stopped sending. Nothing is
queued, and I will not repost the three CAN bindings unless you ask me to.

What I got wrong, and what is fixed before anything goes out again:

1. Threading. I attached new versions to older threads and unrelated
   patchsets. That is exactly what submitting-patches.rst warns about: it
   buried the patches and made Patchwork show several versions of one file.
   The next posting is one series, in its own thread, one version only.

2. spi-peripheral-props. Dropped, as you asked. The binding now references
   only can-controller.yaml, and dt_binding_check passes.

3. clock-frequency. Dropped, and clocks stays required. You were right that
   it is ACPI-only; I had argued from the driver code instead of reading the
   git history.

4. Commit messages rewritten as short declarative sentences that describe
   only what changed in the binding.

5. One thing I should have said from the start: these patches were prepared
   with an LLM coding assistant. Per
   Documentation/process/coding-assistants.rst they need an
   "Assisted-by: LLM" tag, and mine did not carry one. It is now on every
   patch. You told me not to paste LLM text, and I should also have
   disclosed that the text was LLM-assisted. Not doing that was the worst
   part of this, not the wording.

The bindings are otherwise unchanged. holt, mpc5xxx-mscan and ifi,canfd are
ready as a single clean v7 series. Tell me when you want it and I will post
it in one new thread.

Thanks for the pointers.

Quchaosheng

Re: [PATCH v5] dt-bindings: net: can: convert ifi_canfd to DT schema

[Conor]

On 17 September 2026 11:20:46 GMT+02:00, Quchaosheng <quchaosheng000406@163.com> wrote:
>Hi Krzysztof,
>
>You are right, and sorry for the noise. I have stopped sending. Nothing is
>queued, and I will not repost the three CAN bindings unless you ask me to.
>
>What I got wrong, and what is fixed before anything goes out again:
>
>1. Threading. I attached new versions to older threads and unrelated
>   patchsets. That is exactly what submitting-patches.rst warns about: it
>   buried the patches and made Patchwork show several versions of one file.
>   The next posting is one series, in its own thread, one version only.
>
>2. spi-peripheral-props. Dropped, as you asked. The binding now references
>   only can-controller.yaml, and dt_binding_check passes.
>
>3. clock-frequency. Dropped, and clocks stays required. You were right that
>   it is ACPI-only; I had argued from the driver code instead of reading the
>   git history.
>
>4. Commit messages rewritten as short declarative sentences that describe
>   only what changed in the binding.
>
>5. One thing I should have said from the start: these patches were prepared
>   with an LLM coding assistant. Per
>   Documentation/process/coding-assistants.rst they need an
>   "Assisted-by: LLM" tag, and mine did not carry one. It is now on every
>   patch. You told me not to paste LLM text, and I should also have

And yet, this looks exactly like LLM pasted text.
That's an insult to all who genuinely engage with the development process, who knows if you even understand what's been said here, and are just blindly pasting back and forth.
If we wanted to have a discussion with an LLM, we can do that ourselves. We don't need you in the loop for that.

Thanks,
Conor.

>   disclosed that the text was LLM-assisted. Not doing that was the worst
>   part of this, not the wording.
>
>The bindings are otherwise unchanged. holt, mpc5xxx-mscan and ifi,canfd are
>ready as a single clean v7 series. Tell me when you want it and I will post
>it in one new thread.
>
>Thanks for the pointers.
>
>Quchaosheng
>

Re: [PATCH v5] dt-bindings: net: can: convert ifi_canfd to DT schema

[Quchaosheng]

Conor,

You're right about that mail, and about the one before it. I was
relaying my replies through a tool instead of writing them.

I'm holding the three CAN bindings. Nothing is queued, and I won't post
a v7 unless Krzysztof or you ask for it. I won't keep going in this
thread either.

Quchaosheng

Re: [PATCH v5] dt-bindings: net: can: convert ifi_canfd to DT schema

[Quchaosheng]

Krzysztof,

On 09-15 you asked me to slow down and I said the three CAN bindings
would stay put until you asked for them. One thing has changed since:
Rob reviewed the v6 postings on 09-21. holt,hi311x and ifi,canfd-1.0 came
back with his Reviewed-by; fsl,mpc5xxx-mscan came back with four comments,
all of which are addressed now.

What changed in mscan, so you can judge it without opening the patch:

 - clock-names is an ordered list of ipg, ips, sys, ref, mclk, and clocks
   is minItems 5. fsl,mpc5121-mscan requires both.
 - The fsl,mpc5121-mscan if/then that only restated the enum is gone; the
   per-compatible defaults moved into the property description.
 - fsl,mscan-clock-divider has a minimum of 1, the driver divides by it.
 - One example instead of two.

If that is enough, may I send it as v7 - one series, one thread, one
version? If you would rather it wait, say so and it stays where it is.

Quchaosheng