Message-ID: <097dbefb-f679-40da-bce4-d6e09dc4cf08@oss.qualcomm.com>
Date: Mon, 22 Dec 2025 14:58:47 +0800
From: Jie Gan <jie.gan@....qualcomm.com>
To: Suzuki K Poulose <suzuki.poulose@....com>,
        Krzysztof Kozlowski <krzk@...nel.org>, Rob Herring <robh@...nel.org>
Cc: Mike Leach <mike.leach@...aro.org>, James Clark <james.clark@...aro.org>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Krzysztof Kozlowski <krzk+dt@...nel.org>,
        Conor Dooley
 <conor+dt@...nel.org>,
        Tingwei Zhang <tingwei.zhang@....qualcomm.com>,
        Mao Jinlong <jinlong.mao@....qualcomm.com>,
        Bjorn Andersson <andersson@...nel.org>,
        Konrad Dybcio <konradybcio@...nel.org>, coresight@...ts.linaro.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        linux-arm-msm@...r.kernel.org, devicetree@...r.kernel.org
Subject: Re: [PATCH v8 5/8] dt-bindings: arm: add an interrupt property for
 Coresight CTCU



On 12/19/2025 5:54 PM, Suzuki K Poulose wrote:
> On 19/12/2025 02:05, Jie Gan wrote:
>>
>>
>> On 12/19/2025 7:19 AM, Suzuki K Poulose wrote:
>>> On 18/12/2025 10:17, Krzysztof Kozlowski wrote:
>>>> On 12/12/2025 02:12, Jie Gan wrote:
>>>>>
>>>>>
>>>>> On 12/11/2025 9:37 PM, Rob Herring wrote:
>>>>>> On Thu, Dec 11, 2025 at 02:10:44PM +0800, Jie Gan wrote:
>>>>>>> Add an interrupt property to CTCU device. The interrupt will be 
>>>>>>> triggered
>>>>>>> when the data size in the ETR buffer exceeds the threshold of the
>>>>>>> BYTECNTRVAL register. Programming a threshold in the BYTECNTRVAL 
>>>>>>> register
>>>>>>> of CTCU device will enable the interrupt.
>>>>>>>
>>>>>>> Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
>>>>>>> Reviewed-by: Mike Leach <mike.leach@...aro.org>
>>>>>>> Signed-off-by: Jie Gan <jie.gan@....qualcomm.com>
>>>>>>> ---
>>>>>>>    .../devicetree/bindings/arm/qcom,coresight-ctcu.yaml    | 17 + 
>>>>>>> + + ++++++++++++++
>>>>>>>    1 file changed, 17 insertions(+)
>>>>>>>
>>>>>>> diff --git a/Documentation/devicetree/bindings/arm/ 
>>>>>>> qcom,coresight- ctcu.yaml b/Documentation/devicetree/bindings/ 
>>>>>>> arm/qcom,coresight- ctcu.yaml
>>>>>>> index c969c16c21ef..90f88cc6cd3e 100644
>>>>>>> --- a/Documentation/devicetree/bindings/arm/qcom,coresight-ctcu.yaml
>>>>>>> +++ b/Documentation/devicetree/bindings/arm/qcom,coresight-ctcu.yaml
>>>>>>> @@ -39,6 +39,16 @@ properties:
>>>>>>>        items:
>>>>>>>          - const: apb
>>>>>>> +  interrupts:
>>>>>>> +    items:
>>>>>>> +      - description: Byte cntr interrupt for the first etr device
>>>>>>> +      - description: Byte cntr interrupt for the second etr device
>>>
>>> This is really vague. How do you define first vs second ? Probe order ?
>>> No way. This must be the "port" number to which the ETR is connected
>>> to the CTCU. IIUC, there is a config area for each ETR (e.g., trace id
>>> filter) connected to the CTCU. I was under the assumption that they
>>> are identified as "ports" (input ports). I don't really understand how
>>> this interrupt mapping works now. Please explain it clearly.
>>>
>>
>> Sorry for the misunderstanding.
>>
>> Each ETR device should have its own interrupt line and an IRQ register 
>> within the CTCU device, as defined by the specification. In existing 
>> projects, the maximum supported number of ETR devices is 2.
>>
>> Each interrupt is directly mapped to a specific ETR device, for example:
>> tmc@...0 → interrupt line 0
>> tmc@...1 → interrupt line 1
>>
>> The suggestion to identify devices by ‘ports’ is much clearer than my 
>> previous explanation, as it explicitly shows which device is connected 
>> to which port.
> 
> Thanks for confirming.
> 
>>
>>>>>>> +
>>>>>>> +  interrupt-names:
>>>>>>> +    items:
>>>>>>> +      - const: etrirq0
>>>>>>> +      - const: etrirq1
>>>>>>
>>>>>> Names are kind of pointless when it is just foo<index>.
>>>>>
>>>>> Hi Rob,
>>>>>
>>>>> I was naming them as etr0/etr1. Are these names acceptable?
>>>>
>>>> Obviously irq is redundant, but how does etr0 solves the problem of
>>>> calling it foo0?
>>>>
>>>> I don't think you really read Rob's comment.
>>>>
>>>>> The interrupts are assigned exclusively to a specific ETR device.
>>>>>
>>>>> But Suzuki is concerned that this might cause confusion because the 
>>>>> ETR
>>>>> device is named randomly in the driver. Suzuki suggested using 
>>>>> ‘port-0’
>>>>> and ‘port-1’ and would also like to hear your feedback on these names.
>>>>
>>>> There is no confusion here. Writing bindings luckily clarifies this 
>>>> what
>>>> the indices in the array mean.
>>>
>>> The point is there are "n" interrupts. Question is, could there be more
>>> devices(ETRs) connected to the CTCU than "n".
>>>
>>> e.g., Lets CTCU can control upto 4 ETRs and on a particular system, the
>>>
>>> TMC-ETR0 -> CTCU-Port0
>>>
>>> TMC-ETR1 -> CTCU-Port2
>>> TMC-ETR2 -> CTCU-Port3
>>>
>>> Now, how many interrupts are described in the DT ? How do we map which
>>> interrupts correspond to the CTCU-Portn. (Finding the TMC-ETRx back
>>> from the port is possible, with the topology).
>>>
>>
>> Got your point and it's much clearer.
>>
>>> This is what I raised in the previous version. Again, happy to hear
>>> if there is a standard way to describe the interrupts.
>>>
>>> Suzuki
>>>
>>>
>>>>
>>>>>
>>>>> Usually, the probe sequence follows the order of the addresses. In our
>>>>> specification, ‘ETR0’ is always probed before ‘ETR1’ because its 
>>>>> address
>>>>> is lower.
>>>>
>>>> How is this even relevant? You are answering to something completely
>>>> different, so I don't think you really tried to understand review.
>>>>
>>
>> My previous explanation was definitely unclear. As Suzuki suggested, 
>> mapping the interrupt to the port number (to identify the relevant 
>> device based on topology) makes sense and provides a much easier way 
>> to understand the relationship between the interrupt and the ETR device.
>>
>> So with the suggestion, here is the new description about the interrupts:
>>
>>    interrupts:
>>      items:
>>        - description: Interrupt for the ETR device connected to in-port0.
>>        - description: Interrupt for the ETR device connected to in-port1.
>>
>>   interrupt-names:
>>      items:
>>       - const: port0
>>       - const: port1
> 
> Which brings us back to the question I posted in the previous version. 
> Do we really need a "name" or are there other ways to define, a sparse
> list of interrupts ?
> 

Each interrupt is dedicated to a specific ETR device. While we can 
retrieve the list of interrupts using of_irq_get, we cannot guarantee 
that the obtained interrupt corresponds to the correct ETR device?
I believe it would be better to have an interrupt-name property, so we 
can assign the name in the data structure and retrieve the interrupt by 
its name, ensuring it maps to the correct ETR device.

Thanks,
Jie

Powered by blists - more mailing lists