[PATCH v5 04/10] coresight: Appropriately disable programming clocks

[Suzuki K Poulose]

On 29/07/2025 12:31, Mark Brown wrote:
> On Mon, Jul 28, 2025 at 05:45:04PM +0100, Mark Brown wrote:
>> On Thu, Jul 24, 2025 at 04:22:34PM +0100, Leo Yan wrote:
>>
>> Previously we would return NULL for any error (which isn't super great
>> for deferred probe but never mind).
>>
>>> +	pclk = devm_clk_get_enabled(dev, "apb_pclk");
>>> +	if (IS_ERR(pclk))
>>> +		pclk = devm_clk_get_enabled(dev, "apb");
>>
>> ...
>>
>>>   	return pclk;
>>>   }
>>
>> Now we pass errors back to the caller, making missing clocks fatal.
> 
> Thinking about this some more I think for compatiblity these clocks need
> to be treated as optional - that's what the original code was
> effectively doing, and I can imagine this isn't the only SoC which has
> (hopefully) always on clocks and didn't wire things up in DT.

You're right. The static components (funnels, replicators) don't have
APB programming interface and hence no clocks. That said, may be the
"is amba device" check could be used to enforce the presence of a clock.

I will let Leo sort this out

Suzuki