Re: [PATCH v2] arm64: dts: qcom: glymur: add coresight nodes

[Jie Gan <jie.gan@oss.qualcomm.com>]

On 3/23/2026 10:03 PM, Bjorn Andersson wrote:
> On Mon, Mar 23, 2026 at 09:27:41PM +0800, Jie Gan wrote:
>>
>>
>> On 3/23/2026 9:02 PM, Konrad Dybcio wrote:
>>> On 3/23/26 1:30 PM, Jie Gan wrote:
>>>>
>>>>
>>>> On 3/23/2026 7:05 PM, Konrad Dybcio wrote:
>>>>> On 3/18/26 12:42 PM, Jie Gan wrote:
>>>>>> Add CoreSight nodes to enable trace paths like TPDM->ETF/STM->ETF.
>>>>>> These devices are part of the AOSS, CDSP, QDSS, PCIe5, TraceNoc and
>>>>>> some small subsystems, such as GCC, IPCC, PMU and so on.
>>>>>>
>>>>>> Signed-off-by: Jie Gan <jie.gan@oss.qualcomm.com>
>>>>>> ---
>>>>>> Changes in V2:
>>>>>> 1. removed two cti devices due to GFX block is down
>>>>>
>>>>> i.e. "because GPU is not yet enabled"?
>>>>
>>>> Yeah, these CTI devices have clock issue for enabling due to the GPU block is not yet enabled.
>>>
>>> Do they need the GPU to be online, or a clock from GPU_CC, or
>>> maybe something else?
>>
>> We need a specific debug clock inside the GPU block. The debug clock only
>> can be enabled while GPU is online.
> 
> What happens once GPU has been delivered, but for some reason is
> inactive and we try to use this CTI device?

We will check these devices again once GPU is available. We also will 
cross check with AOP team if there is a clock enable/disable requirement 
with the specific debug clock.

> 
>> Also needs AOP to support enable/disable
>> the debug clock.
>>
> 
> When you hit such dependencies, please write and contribute the patch
> (in this case, you can do it in a separate patch/series).
> 

noted. The capability to enable/disable debug clocks is supported by 
mproc dt nodes(aoss_qmp). It will send message to AOP core to enable 
debug clocks by configuring a specific register in AOP via mailbox.

Thanks,
Jie

> Regards,
> Bjorn
> 
>>>
>>>>>>       - cti@11c42000
>>>>>>       - cti@11c4b000
>>>>>> 2. changes two TPDM devices to static:
>>>>>>       - tpdm-cdsp-cmsr
>>>>>>       - tpdm-cdsp-cmsr2
>>>>>
>>>>> They were TPDM instances in v1. What's the reason for the change?
>>>>
>>>> These TPDMs havent clock source for accessing registers. We only need enable its ports to output trace data. So I have changed them to static-TPDM compatible.
>>>
>>> The registers are clearly physically there. Are you saying that we
>>> (currently?) can't enable the clock required to access them? Or is
>>> there a design defect that's preventing us from doing so?
>>
>> It's about hardware design. Some of the TPDM devices are designed as static,
>> means we dont need access the register of the device for configuring. The
>> trace data of the static TPDM is enabled by default, we only need enable the
>> port of the connected TPDA device for receiving the data.
>>
>> I have tested these devices with Jtag attached, so I didnt observe issue
>> about these new devices in the CDSP block. (Jtag will provide debug
>> capability for all debug devices) Also cross-checked with hardware team for
>> confirming these devices are working as static.
>>
>> Thanks,
>> Jie
>>
>>>
>>> Konrad
>>>
>>