diff mbox series

[02/10] dt-bindings: bus: add device tree bindings for RIFSC

Message ID 20230705172759.1610753-3-gatien.chevallier@foss.st.com
State New
Headers show
Series Introduce STM32 Firewall framework | expand

Commit Message

Gatien CHEVALLIER July 5, 2023, 5:27 p.m. UTC
Document RIFSC (RIF security controller). RIFSC is a firewall controller
composed of different kinds of hardware resources.

Signed-off-by: Gatien Chevallier <gatien.chevallier@foss.st.com>
---
 .../bindings/bus/st,stm32-rifsc.yaml          | 101 ++++++++++++++++++
 1 file changed, 101 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml

Comments

Rob Herring July 5, 2023, 7:39 p.m. UTC | #1
On Wed, 05 Jul 2023 19:27:51 +0200, Gatien Chevallier wrote:
> Document RIFSC (RIF security controller). RIFSC is a firewall controller
> composed of different kinds of hardware resources.
> 
> Signed-off-by: Gatien Chevallier <gatien.chevallier@foss.st.com>
> ---
>  .../bindings/bus/st,stm32-rifsc.yaml          | 101 ++++++++++++++++++
>  1 file changed, 101 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
> 

My bot found errors running 'make DT_CHECKER_FLAGS=-m dt_binding_check'
on your patch (DT_CHECKER_FLAGS is new in v5.13):

yamllint warnings/errors:

dtschema/dtc warnings/errors:
/builds/robherring/dt-review-ci/linux/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml: title: 'STM32 Resource isolation framework security controller bindings' should not be valid under {'pattern': '([Bb]inding| [Ss]chema)'}
	hint: Everything is a binding/schema, no need to say it. Describe what hardware the binding is for.
	from schema $id: http://devicetree.org/meta-schemas/core.yaml#

doc reference errors (make refcheckdocs):

See https://patchwork.ozlabs.org/project/devicetree-bindings/patch/20230705172759.1610753-3-gatien.chevallier@foss.st.com

The base for the series is generally the latest rc1. A different dependency
should be noted in *this* patch.

If you already ran 'make dt_binding_check' and didn't see the above
error(s), then make sure 'yamllint' is installed and dt-schema is up to
date:

pip3 install dtschema --upgrade

Please check and re-submit after running the above command yourself. Note
that DT_SCHEMA_FILES can be set to your schema file to speed up checking
your schema. However, it must be unset to test all examples with your schema.
Krzysztof Kozlowski July 6, 2023, 6:28 a.m. UTC | #2
On 05/07/2023 19:27, Gatien Chevallier wrote:
> Document RIFSC (RIF security controller). RIFSC is a firewall controller
> composed of different kinds of hardware resources.
> 
> Signed-off-by: Gatien Chevallier <gatien.chevallier@foss.st.com>

A nit, subject: drop second/last, redundant "device tree bindings for".
The "dt-bindings" prefix is already stating that these are bindings. 4
words of your 6 word subject is meaningless...

> ---
>  .../bindings/bus/st,stm32-rifsc.yaml          | 101 ++++++++++++++++++
>  1 file changed, 101 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
> 
> diff --git a/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml b/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
> new file mode 100644
> index 000000000000..68d585ed369c
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml

Filename like compatible, unless you know list of compatibles will
grow... but then add them.

> @@ -0,0 +1,101 @@
> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> +%YAML 1.2
> +---
> +$id: http://devicetree.org/schemas/bus/st,stm32-rifsc.yaml#
> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> +
> +title: STM32 Resource isolation framework security controller bindings

Drop bindings

> +
> +maintainers:
> +  - Gatien Chevallier <gatien.chevallier@foss.st.com>
> +
> +description: |
> +  Resource isolation framework (RIF) is a comprehensive set of hardware blocks
> +  designed to enforce and manage isolation of STM32 hardware resources like
> +  memory and peripherals.
> +
> +  The RIFSC (RIF security controller) is composed of three sets of registers,
> +  each managing a specific set of hardware resources:
> +    - RISC registers associated with RISUP logic (resource isolation device unit
> +      for peripherals), assign all non-RIF aware peripherals to zero, one or
> +      any security domains (secure, privilege, compartment).
> +    - RIMC registers: associated with RIMU logic (resource isolation master
> +      unit), assign all non RIF-aware bus master to one security domain by
> +      setting secure, privileged and compartment information on the system bus.
> +      Alternatively, the RISUP logic controlling the device port access to a
> +      peripheral can assign target bus attributes to this peripheral master port
> +      (supported attribute: CID).
> +    - RISC registers associated with RISAL logic (resource isolation device unit
> +      for address space - Lite version), assign address space subregions to one
> +      security domains (secure, privilege, compartment).
> +
> +properties:
> +  compatible:
> +    const: st,stm32mp25-rifsc
> +
> +  reg:
> +    maxItems: 1
> +
> +  "#address-cells":
> +    const: 1
> +
> +  "#size-cells":
> +    const: 1
> +
> +  "#feature-domain-cells":
> +    const: 1
> +
> +  ranges: true
> +
> +  feature-domain-controller: true
> +
> +patternProperties:
> +  "^.*@[0-9a-f]+$":
> +    description: Peripherals
> +    type: object
> +    properties:
> +      feature-domains:
> +        minItems: 1
> +        maxItems: 2
> +        description:
> +          The first argument must always be a phandle that references to the
> +          firewall controller of the peripheral. The second can contain the
> +          platform specific firewall ID of the peripheral.

It does not make much sense to me to have hierarchy parent-child and via
phandle at the same time. You express the similar relationship twice.

> +
> +required:
> +  - compatible
> +  - reg
> +  - "#address-cells"
> +  - "#size-cells"
> +  - feature-domain-controller
> +  - "#feature-domain-cells"
> +  - ranges
> +
> +additionalProperties: false
> +
> +examples:
> +  - |
> +    // In this example, the usart2 device refers to rifsc as its domain
> +    // controller.
> +    // Access rights are verified before creating devices.
> +
> +    #include <dt-bindings/interrupt-controller/arm-gic.h>
> +
> +    rifsc: rifsc-bus@42080000 {
> +        compatible = "st,stm32mp25-rifsc";
> +        reg = <0x42080000 0x1000>;
> +        #address-cells = <1>;
> +        #size-cells = <1>;
> +        ranges;
> +        feature-domain-controller;
> +        #feature-domain-cells = <1>;
> +
> +        usart2: serial@400e0000 {
> +            compatible = "st,stm32h7-uart";
> +            reg = <0x400e0000 0x400>;
> +            interrupts = <GIC_SPI 115 IRQ_TYPE_LEVEL_HIGH>;
> +            clocks = <&ck_flexgen_08>;
> +            feature-domains = <&rifsc 32>;
> +            status = "disabled";

No status in the examples.

> +        };
> +    };

Best regards,
Krzysztof
Gatien CHEVALLIER July 6, 2023, 9:29 a.m. UTC | #3
Hello Krzysztof,

Firstly, I will correct the bindings error pointed by Rob's robot.
Obviously, I did not pass the bindings check the proper way or maybe I'm 
running an old version.

On 7/6/23 08:28, Krzysztof Kozlowski wrote:
> On 05/07/2023 19:27, Gatien Chevallier wrote:
>> Document RIFSC (RIF security controller). RIFSC is a firewall controller
>> composed of different kinds of hardware resources.
>>
>> Signed-off-by: Gatien Chevallier <gatien.chevallier@foss.st.com>
> 
> A nit, subject: drop second/last, redundant "device tree bindings for".
> The "dt-bindings" prefix is already stating that these are bindings. 4
> words of your 6 word subject is meaningless...

Ack, I will rephrase, it is indeed redundant

> 
>> ---
>>   .../bindings/bus/st,stm32-rifsc.yaml          | 101 ++++++++++++++++++
>>   1 file changed, 101 insertions(+)
>>   create mode 100644 Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
>>
>> diff --git a/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml b/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
>> new file mode 100644
>> index 000000000000..68d585ed369c
>> --- /dev/null
>> +++ b/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
> 
> Filename like compatible, unless you know list of compatibles will
> grow... but then add them.
> 
>> @@ -0,0 +1,101 @@
>> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
>> +%YAML 1.2
>> +---
>> +$id: http://devicetree.org/schemas/bus/st,stm32-rifsc.yaml#
>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
>> +
>> +title: STM32 Resource isolation framework security controller bindings
> 
> Drop bindings

Ack

> 
>> +
>> +maintainers:
>> +  - Gatien Chevallier <gatien.chevallier@foss.st.com>
>> +
>> +description: |
>> +  Resource isolation framework (RIF) is a comprehensive set of hardware blocks
>> +  designed to enforce and manage isolation of STM32 hardware resources like
>> +  memory and peripherals.
>> +
>> +  The RIFSC (RIF security controller) is composed of three sets of registers,
>> +  each managing a specific set of hardware resources:
>> +    - RISC registers associated with RISUP logic (resource isolation device unit
>> +      for peripherals), assign all non-RIF aware peripherals to zero, one or
>> +      any security domains (secure, privilege, compartment).
>> +    - RIMC registers: associated with RIMU logic (resource isolation master
>> +      unit), assign all non RIF-aware bus master to one security domain by
>> +      setting secure, privileged and compartment information on the system bus.
>> +      Alternatively, the RISUP logic controlling the device port access to a
>> +      peripheral can assign target bus attributes to this peripheral master port
>> +      (supported attribute: CID).
>> +    - RISC registers associated with RISAL logic (resource isolation device unit
>> +      for address space - Lite version), assign address space subregions to one
>> +      security domains (secure, privilege, compartment).
>> +
>> +properties:
>> +  compatible:
>> +    const: st,stm32mp25-rifsc
>> +
>> +  reg:
>> +    maxItems: 1
>> +
>> +  "#address-cells":
>> +    const: 1
>> +
>> +  "#size-cells":
>> +    const: 1
>> +
>> +  "#feature-domain-cells":
>> +    const: 1
>> +
>> +  ranges: true
>> +
>> +  feature-domain-controller: true
>> +
>> +patternProperties:
>> +  "^.*@[0-9a-f]+$":
>> +    description: Peripherals
>> +    type: object
>> +    properties:
>> +      feature-domains:
>> +        minItems: 1
>> +        maxItems: 2
>> +        description:
>> +          The first argument must always be a phandle that references to the
>> +          firewall controller of the peripheral. The second can contain the
>> +          platform specific firewall ID of the peripheral.
> 
> It does not make much sense to me to have hierarchy parent-child and via
> phandle at the same time. You express the similar relationship twice
Thank you for pointing this out.

About the parent-child relation:

The bus-like device tree architecture allows a bus-probe mechanism with 
which we want to check accesses of peripherals before probing their 
driver. This has several advantages:
-This bus architecture provides a clearer view of the hardware.
-No peripheral driver modifications as it is fully handled by the 
firewall drivers.
-Drivers for devices that aren't accessible will not even be probed => 
no probe fail.

It would be possible to manage this mechanism another way by handling 
probe deferrals in drivers. But it would mean modifying every driver 
with a check on ST firewall that we probe and some of them aren't from 
STMicroelectronics.

About the phandle relation:

I agree on the fact that this double expression of the relationship is 
redundant.

I've done it this way because there will be other nodes outside the 
RIFSC node that will need to reference it as their feature-domain 
controller. I kept the same information in the property to be coherent 
between all.

For nodes under the RIFSC, the phandle is indeed useless and could be 
removed, just to leave the firewall ID. And I'm inclined to do so. I 
just have one worry on the YAML binding files where I will have a 
pattern property in the RIFSC that will state something and maybe 
another description in the peripheral YAML files. What is your take on that?

> 
>> +
>> +required:
>> +  - compatible
>> +  - reg
>> +  - "#address-cells"
>> +  - "#size-cells"
>> +  - feature-domain-controller
>> +  - "#feature-domain-cells"
>> +  - ranges
>> +
>> +additionalProperties: false
>> +
>> +examples:
>> +  - |
>> +    // In this example, the usart2 device refers to rifsc as its domain
>> +    // controller.
>> +    // Access rights are verified before creating devices.
>> +
>> +    #include <dt-bindings/interrupt-controller/arm-gic.h>
>> +
>> +    rifsc: rifsc-bus@42080000 {
>> +        compatible = "st,stm32mp25-rifsc";
>> +        reg = <0x42080000 0x1000>;
>> +        #address-cells = <1>;
>> +        #size-cells = <1>;
>> +        ranges;
>> +        feature-domain-controller;
>> +        #feature-domain-cells = <1>;
>> +
>> +        usart2: serial@400e0000 {
>> +            compatible = "st,stm32h7-uart";
>> +            reg = <0x400e0000 0x400>;
>> +            interrupts = <GIC_SPI 115 IRQ_TYPE_LEVEL_HIGH>;
>> +            clocks = <&ck_flexgen_08>;
>> +            feature-domains = <&rifsc 32>;
>> +            status = "disabled";
> 
> No status in the examples.
> 
>> +        };
>> +    };
> 
> Best regards,
> Krzysztof
> 

Best regards,
Gatien
Gatien CHEVALLIER July 20, 2023, 2:58 p.m. UTC | #4
Hello Krzysztof,

On 7/6/23 11:29, Gatien CHEVALLIER wrote:
> Hello Krzysztof,
> 
> Firstly, I will correct the bindings error pointed by Rob's robot.
> Obviously, I did not pass the bindings check the proper way or maybe I'm 
> running an old version.
> 
> On 7/6/23 08:28, Krzysztof Kozlowski wrote:
>> On 05/07/2023 19:27, Gatien Chevallier wrote:
>>> Document RIFSC (RIF security controller). RIFSC is a firewall controller
>>> composed of different kinds of hardware resources.
>>>
>>> Signed-off-by: Gatien Chevallier <gatien.chevallier@foss.st.com>
>>
>> A nit, subject: drop second/last, redundant "device tree bindings for".
>> The "dt-bindings" prefix is already stating that these are bindings. 4
>> words of your 6 word subject is meaningless...
> 
> Ack, I will rephrase, it is indeed redundant
> 
>>
>>> ---
>>>   .../bindings/bus/st,stm32-rifsc.yaml          | 101 ++++++++++++++++++
>>>   1 file changed, 101 insertions(+)
>>>   create mode 100644 
>>> Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
>>>
>>> diff --git 
>>> a/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml 
>>> b/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
>>> new file mode 100644
>>> index 000000000000..68d585ed369c
>>> --- /dev/null
>>> +++ b/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
>>
>> Filename like compatible, unless you know list of compatibles will
>> grow... but then add them.
>>
>>> @@ -0,0 +1,101 @@
>>> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
>>> +%YAML 1.2
>>> +---
>>> +$id: http://devicetree.org/schemas/bus/st,stm32-rifsc.yaml#
>>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
>>> +
>>> +title: STM32 Resource isolation framework security controller bindings
>>
>> Drop bindings
> 
> Ack
> 
>>
>>> +
>>> +maintainers:
>>> +  - Gatien Chevallier <gatien.chevallier@foss.st.com>
>>> +
>>> +description: |
>>> +  Resource isolation framework (RIF) is a comprehensive set of 
>>> hardware blocks
>>> +  designed to enforce and manage isolation of STM32 hardware 
>>> resources like
>>> +  memory and peripherals.
>>> +
>>> +  The RIFSC (RIF security controller) is composed of three sets of 
>>> registers,
>>> +  each managing a specific set of hardware resources:
>>> +    - RISC registers associated with RISUP logic (resource isolation 
>>> device unit
>>> +      for peripherals), assign all non-RIF aware peripherals to 
>>> zero, one or
>>> +      any security domains (secure, privilege, compartment).
>>> +    - RIMC registers: associated with RIMU logic (resource isolation 
>>> master
>>> +      unit), assign all non RIF-aware bus master to one security 
>>> domain by
>>> +      setting secure, privileged and compartment information on the 
>>> system bus.
>>> +      Alternatively, the RISUP logic controlling the device port 
>>> access to a
>>> +      peripheral can assign target bus attributes to this peripheral 
>>> master port
>>> +      (supported attribute: CID).
>>> +    - RISC registers associated with RISAL logic (resource isolation 
>>> device unit
>>> +      for address space - Lite version), assign address space 
>>> subregions to one
>>> +      security domains (secure, privilege, compartment).
>>> +
>>> +properties:
>>> +  compatible:
>>> +    const: st,stm32mp25-rifsc
>>> +
>>> +  reg:
>>> +    maxItems: 1
>>> +
>>> +  "#address-cells":
>>> +    const: 1
>>> +
>>> +  "#size-cells":
>>> +    const: 1
>>> +
>>> +  "#feature-domain-cells":
>>> +    const: 1
>>> +
>>> +  ranges: true
>>> +
>>> +  feature-domain-controller: true
>>> +
>>> +patternProperties:
>>> +  "^.*@[0-9a-f]+$":
>>> +    description: Peripherals
>>> +    type: object
>>> +    properties:
>>> +      feature-domains:
>>> +        minItems: 1
>>> +        maxItems: 2
>>> +        description:
>>> +          The first argument must always be a phandle that 
>>> references to the
>>> +          firewall controller of the peripheral. The second can 
>>> contain the
>>> +          platform specific firewall ID of the peripheral.
>>
>> It does not make much sense to me to have hierarchy parent-child and via
>> phandle at the same time. You express the similar relationship twice
> Thank you for pointing this out.
> 
> About the parent-child relation:
> 
> The bus-like device tree architecture allows a bus-probe mechanism with 
> which we want to check accesses of peripherals before probing their 
> driver. This has several advantages:
> -This bus architecture provides a clearer view of the hardware.
> -No peripheral driver modifications as it is fully handled by the 
> firewall drivers.
> -Drivers for devices that aren't accessible will not even be probed => 
> no probe fail.
> 
> It would be possible to manage this mechanism another way by handling 
> probe deferrals in drivers. But it would mean modifying every driver 
> with a check on ST firewall that we probe and some of them aren't from 
> STMicroelectronics.
> 
> About the phandle relation:
> 
> I agree on the fact that this double expression of the relationship is 
> redundant.
> 
> I've done it this way because there will be other nodes outside the 
> RIFSC node that will need to reference it as their feature-domain 
> controller. I kept the same information in the property to be coherent 
> between all.
> 
> For nodes under the RIFSC, the phandle is indeed useless and could be 
> removed, just to leave the firewall ID. And I'm inclined to do so. I 
> just have one worry on the YAML binding files where I will have a 
> pattern property in the RIFSC that will state something and maybe 
> another description in the peripheral YAML files. What is your take on 
> that?
> 

Looking back at it, feature-domains is a phandle-array. I guess I can't
derogate to the following architecture:

items:
   - items:
       - description: A phandle
       - description: 1st arg cell
       - description: 2nd arg cell

can I?

Some devices' nodes that are not subnodes of the firewall controllers
will need the phandle reference. Should I keep the redundant information
then?

Best regards,
Gatien

>>
>>> +
>>> +required:
>>> +  - compatible
>>> +  - reg
>>> +  - "#address-cells"
>>> +  - "#size-cells"
>>> +  - feature-domain-controller
>>> +  - "#feature-domain-cells"
>>> +  - ranges
>>> +
>>> +additionalProperties: false
>>> +
>>> +examples:
>>> +  - |
>>> +    // In this example, the usart2 device refers to rifsc as its domain
>>> +    // controller.
>>> +    // Access rights are verified before creating devices.
>>> +
>>> +    #include <dt-bindings/interrupt-controller/arm-gic.h>
>>> +
>>> +    rifsc: rifsc-bus@42080000 {
>>> +        compatible = "st,stm32mp25-rifsc";
>>> +        reg = <0x42080000 0x1000>;
>>> +        #address-cells = <1>;
>>> +        #size-cells = <1>;
>>> +        ranges;
>>> +        feature-domain-controller;
>>> +        #feature-domain-cells = <1>;
>>> +
>>> +        usart2: serial@400e0000 {
>>> +            compatible = "st,stm32h7-uart";
>>> +            reg = <0x400e0000 0x400>;
>>> +            interrupts = <GIC_SPI 115 IRQ_TYPE_LEVEL_HIGH>;
>>> +            clocks = <&ck_flexgen_08>;
>>> +            feature-domains = <&rifsc 32>;
>>> +            status = "disabled";
>>
>> No status in the examples.
>>
>>> +        };
>>> +    };
>>
>> Best regards,
>> Krzysztof
>>
> 
> Best regards,
> Gatien
diff mbox series

Patch

diff --git a/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml b/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
new file mode 100644
index 000000000000..68d585ed369c
--- /dev/null
+++ b/Documentation/devicetree/bindings/bus/st,stm32-rifsc.yaml
@@ -0,0 +1,101 @@ 
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/bus/st,stm32-rifsc.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: STM32 Resource isolation framework security controller bindings
+
+maintainers:
+  - Gatien Chevallier <gatien.chevallier@foss.st.com>
+
+description: |
+  Resource isolation framework (RIF) is a comprehensive set of hardware blocks
+  designed to enforce and manage isolation of STM32 hardware resources like
+  memory and peripherals.
+
+  The RIFSC (RIF security controller) is composed of three sets of registers,
+  each managing a specific set of hardware resources:
+    - RISC registers associated with RISUP logic (resource isolation device unit
+      for peripherals), assign all non-RIF aware peripherals to zero, one or
+      any security domains (secure, privilege, compartment).
+    - RIMC registers: associated with RIMU logic (resource isolation master
+      unit), assign all non RIF-aware bus master to one security domain by
+      setting secure, privileged and compartment information on the system bus.
+      Alternatively, the RISUP logic controlling the device port access to a
+      peripheral can assign target bus attributes to this peripheral master port
+      (supported attribute: CID).
+    - RISC registers associated with RISAL logic (resource isolation device unit
+      for address space - Lite version), assign address space subregions to one
+      security domains (secure, privilege, compartment).
+
+properties:
+  compatible:
+    const: st,stm32mp25-rifsc
+
+  reg:
+    maxItems: 1
+
+  "#address-cells":
+    const: 1
+
+  "#size-cells":
+    const: 1
+
+  "#feature-domain-cells":
+    const: 1
+
+  ranges: true
+
+  feature-domain-controller: true
+
+patternProperties:
+  "^.*@[0-9a-f]+$":
+    description: Peripherals
+    type: object
+    properties:
+      feature-domains:
+        minItems: 1
+        maxItems: 2
+        description:
+          The first argument must always be a phandle that references to the
+          firewall controller of the peripheral. The second can contain the
+          platform specific firewall ID of the peripheral.
+
+required:
+  - compatible
+  - reg
+  - "#address-cells"
+  - "#size-cells"
+  - feature-domain-controller
+  - "#feature-domain-cells"
+  - ranges
+
+additionalProperties: false
+
+examples:
+  - |
+    // In this example, the usart2 device refers to rifsc as its domain
+    // controller.
+    // Access rights are verified before creating devices.
+
+    #include <dt-bindings/interrupt-controller/arm-gic.h>
+
+    rifsc: rifsc-bus@42080000 {
+        compatible = "st,stm32mp25-rifsc";
+        reg = <0x42080000 0x1000>;
+        #address-cells = <1>;
+        #size-cells = <1>;
+        ranges;
+        feature-domain-controller;
+        #feature-domain-cells = <1>;
+
+        usart2: serial@400e0000 {
+            compatible = "st,stm32h7-uart";
+            reg = <0x400e0000 0x400>;
+            interrupts = <GIC_SPI 115 IRQ_TYPE_LEVEL_HIGH>;
+            clocks = <&ck_flexgen_08>;
+            feature-domains = <&rifsc 32>;
+            status = "disabled";
+        };
+    };