diff mbox series

[v4,1/2] dt-bindings: ramoops: Inherit reserve memory property

Message ID 1674835252-31954-1-git-send-email-quic_mojha@quicinc.com (mailing list archive)
State Handled Elsewhere
Headers show
Series [v4,1/2] dt-bindings: ramoops: Inherit reserve memory property | expand

Commit Message

Mukesh Ojha Jan. 27, 2023, 4 p.m. UTC
The reserved memory region for ramoops is assumed to be at a
fixed and known location when read from the devicetree. This
is not desirable in an environment where it is preferred the
region to be dynamically allocated at runtime, as opposed to
being fixed at compile time.

So, update the ramoops binding by inheriting some reserve memory
property to allocate the ramoops region dynamically.

Cc: Kees Cook <keescook@chromium.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Guilherme G. Piccoli <gpiccoli@igalia.com>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
---
Changes in v4:
 - Addressed comment made by Krzysztof on ramoops node name.

Changes in v3:
 - Fixed yaml error and updated commit text as per comment.

Change in v2:
  - Added this patch as per changes going to be done in patch 3/3

 .../bindings/reserved-memory/ramoops.yaml          | 34 ++++++++++++++++++++--
 1 file changed, 32 insertions(+), 2 deletions(-)

Comments

Kees Cook Jan. 27, 2023, 7:18 p.m. UTC | #1
On Fri, Jan 27, 2023 at 09:30:51PM +0530, Mukesh Ojha wrote:
> The reserved memory region for ramoops is assumed to be at a
> fixed and known location when read from the devicetree. This
> is not desirable in an environment where it is preferred the
> region to be dynamically allocated at runtime, as opposed to
> being fixed at compile time.
> 
> So, update the ramoops binding by inheriting some reserve memory
> property to allocate the ramoops region dynamically.
> 
> Cc: Kees Cook <keescook@chromium.org>
> Cc: Tony Luck <tony.luck@intel.com>
> Cc: Guilherme G. Piccoli <gpiccoli@igalia.com>
> Cc: Rob Herring <robh+dt@kernel.org>
> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
> Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
> ---
> Changes in v4:
>  - Addressed comment made by Krzysztof on ramoops node name.

Thanks! Krzysztof, if this looks good to you, I can carry it in the
pstore tree.

-Kees

> 
> Changes in v3:
>  - Fixed yaml error and updated commit text as per comment.
> 
> Change in v2:
>   - Added this patch as per changes going to be done in patch 3/3
> 
>  .../bindings/reserved-memory/ramoops.yaml          | 34 ++++++++++++++++++++--
>  1 file changed, 32 insertions(+), 2 deletions(-)
> 
> diff --git a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> index 0391871..8741626 100644
> --- a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> +++ b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> @@ -10,7 +10,8 @@ description: |
>    ramoops provides persistent RAM storage for oops and panics, so they can be
>    recovered after a reboot. This is a child-node of "/reserved-memory", and
>    is named "ramoops" after the backend, rather than "pstore" which is the
> -  subsystem.
> +  subsystem. This region can be reserved both statically or dynamically by
> +  using appropriate property in device tree.
>  
>    Parts of this storage may be set aside for other persistent log buffers, such
>    as kernel log messages, or for optional ECC error-correction data.  The total
> @@ -112,7 +113,13 @@ unevaluatedProperties: false
>  
>  required:
>    - compatible
> -  - reg
> +
> +oneOf:
> +  - required:
> +      - reg
> +
> +  - required:
> +      - size
>  
>  anyOf:
>    - required: [record-size]
> @@ -142,3 +149,26 @@ examples:
>              };
>          };
>      };
> +
> +  - |
> +    / {
> +        compatible = "foo";
> +        model = "foo";
> +        #address-cells = <1>;
> +        #size-cells = <1>;
> +
> +        reserved-memory {
> +            #address-cells = <1>;
> +            #size-cells = <1>;
> +            ranges;
> +
> +            ramoops_region: ramoops {
> +                compatible = "ramoops";
> +                alloc-ranges = <0x00000000 0xffffffff>;
> +                size = <0x0 0x10000>;       /* 64kB */
> +                console-size = <0x8000>;    /* 32kB */
> +                record-size = <0x400>;      /*  1kB */
> +                ecc-size = <16>;
> +            };
> +        };
> +    };
> -- 
> 2.7.4
>
Krzysztof Kozlowski Jan. 27, 2023, 8:03 p.m. UTC | #2
On 27/01/2023 17:00, Mukesh Ojha wrote:
> The reserved memory region for ramoops is assumed to be at a
> fixed and known location when read from the devicetree. This
> is not desirable in an environment where it is preferred the
> region to be dynamically allocated at runtime, as opposed to
> being fixed at compile time.
> 
> So, update the ramoops binding by inheriting some reserve memory
> property to allocate the ramoops region dynamically.

Where is the update which adds "inheriting"?

> 
> Cc: Kees Cook <keescook@chromium.org>
> Cc: Tony Luck <tony.luck@intel.com>
> Cc: Guilherme G. Piccoli <gpiccoli@igalia.com>
> Cc: Rob Herring <robh+dt@kernel.org>
> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
> Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
> ---
> Changes in v4:
>  - Addressed comment made by Krzysztof on ramoops node name.
> 
> Changes in v3:
>  - Fixed yaml error and updated commit text as per comment.
> 
> Change in v2:
>   - Added this patch as per changes going to be done in patch 3/3
> 
>  .../bindings/reserved-memory/ramoops.yaml          | 34 ++++++++++++++++++++--
>  1 file changed, 32 insertions(+), 2 deletions(-)
> 
> diff --git a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> index 0391871..8741626 100644
> --- a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> +++ b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> @@ -10,7 +10,8 @@ description: |
>    ramoops provides persistent RAM storage for oops and panics, so they can be
>    recovered after a reboot. This is a child-node of "/reserved-memory", and
>    is named "ramoops" after the backend, rather than "pstore" which is the
> -  subsystem.
> +  subsystem. This region can be reserved both statically or dynamically by
> +  using appropriate property in device tree.
>  
>    Parts of this storage may be set aside for other persistent log buffers, such
>    as kernel log messages, or for optional ECC error-correction data.  The total
> @@ -112,7 +113,13 @@ unevaluatedProperties: false
>  
>  required:
>    - compatible
> -  - reg

This is okay, but:

> +
> +oneOf:
> +  - required:
> +      - reg
> +
> +  - required:
> +      - size

I now keep wondering - why do you need this?

>  
>  anyOf:
>    - required: [record-size]
> @@ -142,3 +149,26 @@ examples:
>              };
>          };
>      };
> +
> +  - |
> +    / {
> +        compatible = "foo";
> +        model = "foo";
> +        #address-cells = <1>;
> +        #size-cells = <1>;
> +
> +        reserved-memory {
> +            #address-cells = <1>;
> +            #size-cells = <1>;
> +            ranges;
> +
> +            ramoops_region: ramoops {
> +                compatible = "ramoops";
> +                alloc-ranges = <0x00000000 0xffffffff>;
> +                size = <0x0 0x10000>;       /* 64kB */
> +                console-size = <0x8000>;    /* 32kB */
> +                record-size = <0x400>;      /*  1kB */
> +                ecc-size = <16>;
> +            };
> +        };
> +    };

This example does not bring anything new for the ramoops. It's an
example for reserved-memory to show usage with alloc-ranges. There is
nothing useful here in terms of ramoops, so I think it should be dropped.

Best regards,
Krzysztof
Krzysztof Kozlowski Jan. 27, 2023, 8:04 p.m. UTC | #3
On 27/01/2023 17:00, Mukesh Ojha wrote:
> The reserved memory region for ramoops is assumed to be at a
> fixed and known location when read from the devicetree. This
> is not desirable in an environment where it is preferred the
> region to be dynamically allocated at runtime, as opposed to
> being fixed at compile time.
> 
> So, update the ramoops binding by inheriting some reserve memory
> property to allocate the ramoops region dynamically.
> 
> Cc: Kees Cook <keescook@chromium.org>
> Cc: Tony Luck <tony.luck@intel.com>
> Cc: Guilherme G. Piccoli <gpiccoli@igalia.com>
> Cc: Rob Herring <robh+dt@kernel.org>
> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>

Drop. There is no single need to store automated output of
get_maintainers.pl in the git log forever. It can be easily re-created
at any given time.

If you need it for your own patch management purposes, keep it under ---.

Additionally:
where is patch number 2?

Best regards,
Krzysztof
Rob Herring Jan. 27, 2023, 8:05 p.m. UTC | #4
On Fri, Jan 27, 2023 at 10:01 AM Mukesh Ojha <quic_mojha@quicinc.com> wrote:
>
> The reserved memory region for ramoops is assumed to be at a
> fixed and known location when read from the devicetree. This
> is not desirable in an environment where it is preferred the
> region to be dynamically allocated at runtime, as opposed to
> being fixed at compile time.

I don't see how this can work. How do you find the region after
rebooting? Or this is only used for the current boot? If so, provide a
userspace interface to configure it (or the existing module params).

The addition of 'size' type regions was primarily for large carveouts
that needed to be allocated before anything else. ramoops is not that.
It's 10s or 100s of KB at most.


> So, update the ramoops binding by inheriting some reserve memory
> property to allocate the ramoops region dynamically.
>
> Cc: Kees Cook <keescook@chromium.org>
> Cc: Tony Luck <tony.luck@intel.com>
> Cc: Guilherme G. Piccoli <gpiccoli@igalia.com>
> Cc: Rob Herring <robh+dt@kernel.org>
> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
> Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
> ---
> Changes in v4:
>  - Addressed comment made by Krzysztof on ramoops node name.
>
> Changes in v3:
>  - Fixed yaml error and updated commit text as per comment.
>
> Change in v2:
>   - Added this patch as per changes going to be done in patch 3/3
>
>  .../bindings/reserved-memory/ramoops.yaml          | 34 ++++++++++++++++++++--
>  1 file changed, 32 insertions(+), 2 deletions(-)
>
> diff --git a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> index 0391871..8741626 100644
> --- a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> +++ b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
> @@ -10,7 +10,8 @@ description: |
>    ramoops provides persistent RAM storage for oops and panics, so they can be
>    recovered after a reboot. This is a child-node of "/reserved-memory", and
>    is named "ramoops" after the backend, rather than "pstore" which is the
> -  subsystem.
> +  subsystem. This region can be reserved both statically or dynamically by
> +  using appropriate property in device tree.
>
>    Parts of this storage may be set aside for other persistent log buffers, such
>    as kernel log messages, or for optional ECC error-correction data.  The total
> @@ -112,7 +113,13 @@ unevaluatedProperties: false
>
>  required:
>    - compatible
> -  - reg
> +
> +oneOf:
> +  - required:
> +      - reg
> +
> +  - required:
> +      - size
>
>  anyOf:
>    - required: [record-size]
> @@ -142,3 +149,26 @@ examples:
>              };
>          };
>      };
> +
> +  - |
> +    / {

You can't have multiple root node examples. Check the example dts, the
examples will be merged together.

> +        compatible = "foo";
> +        model = "foo";
> +        #address-cells = <1>;
> +        #size-cells = <1>;
> +
> +        reserved-memory {
> +            #address-cells = <1>;
> +            #size-cells = <1>;
> +            ranges;
> +
> +            ramoops_region: ramoops {
> +                compatible = "ramoops";
> +                alloc-ranges = <0x00000000 0xffffffff>;
> +                size = <0x0 0x10000>;       /* 64kB */
> +                console-size = <0x8000>;    /* 32kB */
> +                record-size = <0x400>;      /*  1kB */
> +                ecc-size = <16>;
> +            };
> +        };
> +    };
> --
> 2.7.4
>
Mukesh Ojha Jan. 30, 2023, 5:14 a.m. UTC | #5
On 1/28/2023 1:34 AM, Krzysztof Kozlowski wrote:
> On 27/01/2023 17:00, Mukesh Ojha wrote:
>> The reserved memory region for ramoops is assumed to be at a
>> fixed and known location when read from the devicetree. This
>> is not desirable in an environment where it is preferred the
>> region to be dynamically allocated at runtime, as opposed to
>> being fixed at compile time.
>>
>> So, update the ramoops binding by inheriting some reserve memory
>> property to allocate the ramoops region dynamically.
>>
>> Cc: Kees Cook <keescook@chromium.org>
>> Cc: Tony Luck <tony.luck@intel.com>
>> Cc: Guilherme G. Piccoli <gpiccoli@igalia.com>
>> Cc: Rob Herring <robh+dt@kernel.org>
>> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
> 
> Drop. There is no single need to store automated output of
> get_maintainers.pl in the git log forever. It can be easily re-created
> at any given time.
> 
> If you need it for your own patch management purposes, keep it under ---.
> 

Will drop, thanks.

> Additionally:
> where is patch number 2?

It was merged with 2/2 as it was not making sense to keep it as separate 
patch

-Mukesh

> 
> Best regards,
> Krzysztof
>
Mukesh Ojha Jan. 30, 2023, 8:13 a.m. UTC | #6
On 1/28/2023 1:33 AM, Krzysztof Kozlowski wrote:
> On 27/01/2023 17:00, Mukesh Ojha wrote:
>> The reserved memory region for ramoops is assumed to be at a
>> fixed and known location when read from the devicetree. This
>> is not desirable in an environment where it is preferred the
>> region to be dynamically allocated at runtime, as opposed to
>> being fixed at compile time.
>>
>> So, update the ramoops binding by inheriting some reserve memory
>> property to allocate the ramoops region dynamically.
> 
> Where is the update which adds "inheriting"?

By inheriting, i meant using reserve memory properties..

Probably rephrase above as.

"dt-bindings: ramoops: Support dynamic ramoops region allocation

The reserved memory region for ramoops is assumed to be at a
fixed and known location when read from the devicetree. This
is not desirable in an environment where it is preferred the
region to be dynamically allocated at runtime, as opposed to
being fixed at compile time. This can be done with minor update
in ramoops binding as it inherit reserve memory property
(.yaml) in the binding.

Dynamic region could be used by providing size(region size) and
alloc-ranges(allowed ddr region to allocate the size from) instead
of mentioning regs"

Does it sound reasonable ?

-Mukesh
> 
>>
>> Cc: Kees Cook <keescook@chromium.org>
>> Cc: Tony Luck <tony.luck@intel.com>
>> Cc: Guilherme G. Piccoli <gpiccoli@igalia.com>
>> Cc: Rob Herring <robh+dt@kernel.org>
>> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
>> Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
>> ---
>> Changes in v4:
>>   - Addressed comment made by Krzysztof on ramoops node name.
>>
>> Changes in v3:
>>   - Fixed yaml error and updated commit text as per comment.
>>
>> Change in v2:
>>    - Added this patch as per changes going to be done in patch 3/3
>>
>>   .../bindings/reserved-memory/ramoops.yaml          | 34 ++++++++++++++++++++--
>>   1 file changed, 32 insertions(+), 2 deletions(-)
>>
>> diff --git a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
>> index 0391871..8741626 100644
>> --- a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
>> +++ b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
>> @@ -10,7 +10,8 @@ description: |
>>     ramoops provides persistent RAM storage for oops and panics, so they can be
>>     recovered after a reboot. This is a child-node of "/reserved-memory", and
>>     is named "ramoops" after the backend, rather than "pstore" which is the
>> -  subsystem.
>> +  subsystem. This region can be reserved both statically or dynamically by
>> +  using appropriate property in device tree.
>>   
>>     Parts of this storage may be set aside for other persistent log buffers, such
>>     as kernel log messages, or for optional ECC error-correction data.  The total
>> @@ -112,7 +113,13 @@ unevaluatedProperties: false
>>   
>>   required:
>>     - compatible
>> -  - reg
> 
> This is okay, but:
> 
>> +
>> +oneOf:
>> +  - required:
>> +      - reg
>> +
>> +  - required:
>> +      - size
> 
> I now keep wondering - why do you need this?

This should be same as..
Documentation/devicetree/bindings/reserved-memory/reserved-memory.yaml

> 
>>   
>>   anyOf:
>>     - required: [record-size]
>> @@ -142,3 +149,26 @@ examples:
>>               };
>>           };
>>       };
>> +
>> +  - |
>> +    / {
>> +        compatible = "foo";
>> +        model = "foo";
>> +        #address-cells = <1>;
>> +        #size-cells = <1>;
>> +
>> +        reserved-memory {
>> +            #address-cells = <1>;
>> +            #size-cells = <1>;
>> +            ranges;
>> +
>> +            ramoops_region: ramoops {
>> +                compatible = "ramoops";
>> +                alloc-ranges = <0x00000000 0xffffffff>;
>> +                size = <0x0 0x10000>;       /* 64kB */
>> +                console-size = <0x8000>;    /* 32kB */
>> +                record-size = <0x400>;      /*  1kB */
>> +                ecc-size = <16>;
>> +            };
>> +        };
>> +    };
> 
> This example does not bring anything new for the ramoops. It's an
> example for reserved-memory to show usage with alloc-ranges. There is
> nothing useful here in terms of ramoops, so I think it should be dropped.
> 

will drop this..

> Best regards,
> Krzysztof
>
Mukesh Ojha Jan. 30, 2023, 8:32 a.m. UTC | #7
On 1/28/2023 1:35 AM, Rob Herring wrote:
> On Fri, Jan 27, 2023 at 10:01 AM Mukesh Ojha <quic_mojha@quicinc.com> wrote:
>>
>> The reserved memory region for ramoops is assumed to be at a
>> fixed and known location when read from the devicetree. This
>> is not desirable in an environment where it is preferred the
>> region to be dynamically allocated at runtime, as opposed to
>> being fixed at compile time.
> 
> I don't see how this can work. How do you find the region after
> rebooting? Or this is only used for the current boot? If so, provide a
> userspace interface to configure it (or the existing module params).
> 

Client can do of_reserved_mem_lookup() on this ramoops node to get 
region addresses and it is for regular boots and it is not for
current boot.

Basically, here we are trying to support system which does not 
completely support warm reset(content could be inconsistent across 
boots) and has their own mechanism to dump ramoops region on crash.

Also, If it is required, to override the existing ramoops region of 
current boot(from static to dynamic), would provide user space interface 
as well.

> The addition of 'size' type regions was primarily for large carveouts
> that needed to be allocated before anything else. ramoops is not that.
> It's 10s or 100s of KB at most.

I agree..

Is this the only issue ?

I thought if it is inheriting reserve-memory binding, it can be used if 
it is not breaking anything what ramoops already supports..
you can have a look at 2/2 patch.


-Mukesh

> 
> 
>> So, update the ramoops binding by inheriting some reserve memory
>> property to allocate the ramoops region dynamically.
>>
>> Cc: Kees Cook <keescook@chromium.org>
>> Cc: Tony Luck <tony.luck@intel.com>
>> Cc: Guilherme G. Piccoli <gpiccoli@igalia.com>
>> Cc: Rob Herring <robh+dt@kernel.org>
>> Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>
>> Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
>> ---
>> Changes in v4:
>>   - Addressed comment made by Krzysztof on ramoops node name.
>>
>> Changes in v3:
>>   - Fixed yaml error and updated commit text as per comment.
>>
>> Change in v2:
>>    - Added this patch as per changes going to be done in patch 3/3
>>
>>   .../bindings/reserved-memory/ramoops.yaml          | 34 ++++++++++++++++++++--
>>   1 file changed, 32 insertions(+), 2 deletions(-)
>>
>> diff --git a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
>> index 0391871..8741626 100644
>> --- a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
>> +++ b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
>> @@ -10,7 +10,8 @@ description: |
>>     ramoops provides persistent RAM storage for oops and panics, so they can be
>>     recovered after a reboot. This is a child-node of "/reserved-memory", and
>>     is named "ramoops" after the backend, rather than "pstore" which is the
>> -  subsystem.
>> +  subsystem. This region can be reserved both statically or dynamically by
>> +  using appropriate property in device tree.
>>
>>     Parts of this storage may be set aside for other persistent log buffers, such
>>     as kernel log messages, or for optional ECC error-correction data.  The total
>> @@ -112,7 +113,13 @@ unevaluatedProperties: false
>>
>>   required:
>>     - compatible
>> -  - reg
>> +
>> +oneOf:
>> +  - required:
>> +      - reg
>> +
>> +  - required:
>> +      - size
>>
>>   anyOf:
>>     - required: [record-size]
>> @@ -142,3 +149,26 @@ examples:
>>               };
>>           };
>>       };
>> +
>> +  - |
>> +    / {
> 
> You can't have multiple root node examples. Check the example dts, the
> examples will be merged together.

Sure, will remove this..

> 
>> +        compatible = "foo";
>> +        model = "foo";
>> +        #address-cells = <1>;
>> +        #size-cells = <1>;
>> +
>> +        reserved-memory {
>> +            #address-cells = <1>;
>> +            #size-cells = <1>;
>> +            ranges;
>> +
>> +            ramoops_region: ramoops {
>> +                compatible = "ramoops";
>> +                alloc-ranges = <0x00000000 0xffffffff>;
>> +                size = <0x0 0x10000>;       /* 64kB */
>> +                console-size = <0x8000>;    /* 32kB */
>> +                record-size = <0x400>;      /*  1kB */
>> +                ecc-size = <16>;
>> +            };
>> +        };
>> +    };
>> --
>> 2.7.4
>>
diff mbox series

Patch

diff --git a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
index 0391871..8741626 100644
--- a/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
+++ b/Documentation/devicetree/bindings/reserved-memory/ramoops.yaml
@@ -10,7 +10,8 @@  description: |
   ramoops provides persistent RAM storage for oops and panics, so they can be
   recovered after a reboot. This is a child-node of "/reserved-memory", and
   is named "ramoops" after the backend, rather than "pstore" which is the
-  subsystem.
+  subsystem. This region can be reserved both statically or dynamically by
+  using appropriate property in device tree.
 
   Parts of this storage may be set aside for other persistent log buffers, such
   as kernel log messages, or for optional ECC error-correction data.  The total
@@ -112,7 +113,13 @@  unevaluatedProperties: false
 
 required:
   - compatible
-  - reg
+
+oneOf:
+  - required:
+      - reg
+
+  - required:
+      - size
 
 anyOf:
   - required: [record-size]
@@ -142,3 +149,26 @@  examples:
             };
         };
     };
+
+  - |
+    / {
+        compatible = "foo";
+        model = "foo";
+        #address-cells = <1>;
+        #size-cells = <1>;
+
+        reserved-memory {
+            #address-cells = <1>;
+            #size-cells = <1>;
+            ranges;
+
+            ramoops_region: ramoops {
+                compatible = "ramoops";
+                alloc-ranges = <0x00000000 0xffffffff>;
+                size = <0x0 0x10000>;       /* 64kB */
+                console-size = <0x8000>;    /* 32kB */
+                record-size = <0x400>;      /*  1kB */
+                ecc-size = <16>;
+            };
+        };
+    };