Re: [PATCH v9 01/14] dt-bindings: media: qcom,glymur-iris: Add glymur video codec

Hi Rob,

On 7/21/2026 7:50 PM, Rob Herring wrote:
> On Mon, Jul 20, 2026 at 8:00 PM Vishnu Reddy
> <[email protected]> wrote:
>>
>> On 7/16/2026 10:24 PM, Rob Herring wrote:
>>> On Thu, Jul 16, 2026 at 08:37:17AM +0530, Vishnu Reddy wrote:
>>>> On 7/15/2026 7:52 PM, [email protected] wrote:
>>>>> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
>>>>> - [Low] Unnecessary `#address-cells` and `#size-cells` properties on a node without addressable children.
>>>>> --
>>>>>
>>>>> --- Patch [1]: [PATCH v9 01/14] dt-bindings: media: qcom,glymur-iris: Add glymur video codec ---
>>>>> commit 8ca344fa9395fcf1d13bd1ea93de8bee21a5c176
>>>>> Author: Vishnu Reddy <[email protected]>
>>>>>
>>>>> dt-bindings: media: qcom,glymur-iris: Add glymur video codec
>>>>>
>>>>> This commit adds the device tree binding for the Qualcomm Glymur Iris video
>>>>> codec. It introduces a new dual-core architecture requiring additional
>>>>> power domains and clocks, and updates the maxItems limits in the common
>>>>> venus schema to accommodate these changes.
>>>>>
>>>>>> diff --git a/Documentation/devicetree/bindings/media/qcom,glymur-iris.yaml b/Documentation/devicetree/bindings/media/qcom,glymur-iris.yaml
>>>>>> new file mode 100644
>>>>>> --- /dev/null
>>>>>> +++ b/Documentation/devicetree/bindings/media/qcom,glymur-iris.yaml
>>>>>> @@ -0,0 +1,255 @@
>>>>> [ ... ]
>>>>>> +  interconnect-names:
>>>>>> +    items:
>>>>>> +      - const: cpu-cfg
>>>>>> +      - const: video-mem
>>>>>> +
>>>>>> +  '#address-cells':
>>>>>> +    const: 2
>>>>>> +
>>>>>> +  '#size-cells':
>>>>>> +    const: 2
>>>>> [Severity: Low]
>>>>> Are these properties necessary here? The schema strictly enforces const: 2
>>>>> for #address-cells and #size-cells at the root node, but none of the child
>>>>> nodes (non-pixel, pixel, firmware) define a reg property, and the parent
>>>>> does not define a ranges property.
>>>>>
>>>>> Will this cause the DeviceTree compiler (dtc) to emit an
>>>>> avoid_unnecessary_addr_size warning when compiling device trees using this
>>>>> node?
>>>>>
>>>>>> +
>>>>>> +  non-pixel:
>>>>>> +    type: object
>>>>> [ ... ]
>>>>>> +        resets = <&gcc_video_axi0c_clk_ares>,
>>>>>> +                 <&gcc_video_axi0_clk_ares>,
>>>>>> +                 <&videocc_mvs0c_freerun_clk_ares>,
>>>>>> +                 <&videocc_mvs0_freerun_clk_ares>,
>>>>>> +                 <&gcc_video_axi1_clk_ares>,
>>>>>> +                 <&videocc_mvs1_freerun_clk_ares>;
>>>>>> +        reset-names = "core_bus",
>>>>>> +                      "vcodec0_bus",
>>>>>> +                      "core",
>>>>>> +                      "vcodec0_core",
>>>>>> +                      "vcodec1_bus",
>>>>>> +                      "vcodec1_core";
>>>>>> +
>>>>>> +        #address-cells = <2>;
>>>>>> +        #size-cells = <2>;
>>>>> [Severity: Low]
>>>>> Does the example need #address-cells and #size-cells? It demonstrates their
>>>>> use without any child nodes containing reg properties, which directly causes a
>>>>> dtc warning during dt_binding_check validation.
>>>> Without #address-cells and #size-cells, getting the below kernel warning prints.
>>>> [    6.953615] Missing '#address-cells' in /soc@0/video-codec@aa00000
>>>> [    6.968288] WARNING: drivers/of/base.c:115 at of_bus_n_addr_cells+0xa4/0xf4
>>>>
>>>> [    7.440909] Missing '#size-cells' in /soc@0/video-codec@aa00000
>>>> [    7.447044] WARNING: drivers/of/base.c:142 at of_bus_n_size_cells+0xa4/0xf8
>>> Then the kernel is wrong.
>> Hi Rob,
>>
>> Looking at the iris DT node from the link below:
>> https://lore.kernel.org/all/[email protected]/
>>
>> The iris is the parent node and non_pixel is the child node, which does not
>> have a reg property, but does have a memory-region reference containing
>> iommu-addresses. During of_translate_dma_region, #address-cells and
>> #size-cells are required from the parent node for address translation.
> iommu-addresses in the reserved memory node would use
> #address-cells/#size-cells from that node.

Currently, linux kernel does not use #address-cells/#size-cells from the
reserved memory node. Changing it something like below can make it to use
from that node as you suggested.

diff --git a/drivers/iommu/of_iommu.c b/drivers/iommu/of_iommu.c
index a18bb60f6f3d..e0d3f3728afd 100644
--- a/drivers/iommu/of_iommu.c
+++ b/drivers/iommu/of_iommu.c
@@ -251,7 +251,7 @@ void of_iommu_get_resv_regions(struct device *dev, struct
list_head *list)
    if (of_dma_is_coherent(dev->of_node))
        prot |= IOMMU_CACHE;

-    maps = of_translate_dma_region(np, maps, &iova, &length);
+    maps = of_translate_dma_region(it.node, maps, &iova, &length);
    if (length == 0) {
       
Kindly let me know your comment on this.

> IMO, you shouldn't even have 'pixel' and 'non-pixel' nodes. Their
> contents belong in the parent.
>
> Rob