Re: [PATCH v5 1/5] dt-bindings: arm: coresight-tnoc: Drop arm,primecell to bind on platform bus
From: Jie Gan
Date: Thu Jul 09 2026 - 04:42:28 EST
On 7/9/2026 3:59 PM, Krzysztof Kozlowski wrote:
On Tue, Jul 07, 2026 at 05:08:49PM +0800, Jie Gan wrote:
The TNOC compatible previously required the two-string AMBA form
"qcom,coresight-tnoc", "arm,primecell", which forces the device onto the
AMBA bus.
Change the compatible to a single "qcom,coresight-tnoc" string with no
"arm,primecell" entry, so the device is created on the platform bus and
bound by the platform driver through its compatible string.
I asked already: this is not a valid reason. References to Linux
structures are not correct here. If this is your reason, then answer is:
fix drivers, by renaming or doing whatever is necessary to platform bus
to behave like amba bus.
The issue is that we are not allowed to introduce a DT property to bypass the AMBA bus validation, which reads the Component ID (CID) registers. Besides, I cannot fix this in the TraceNoC AMBA driver because the driver is only invoked after amba_match() succeeds. Since the device fails the AMBA match, the TraceNoC AMBA driver is never reached.
The solution proposed by the CoreSight maintainer, Suzuki, is to define a dedicated compatible string for the Aggregator TraceNoC device and have it probed by the existing platform driver. If the AMBA bus cannot identify the device as a valid AMBA device, we should not fake the hardware description to make it appear as one.
Thanks,
Jie
Drop the custom select block as well. It was only needed to avoid
matching the "arm,primecell" string on unrelated nodes; with the single
"qcom,coresight-tnoc" compatible the schema's auto-generated select
already matches the right nodes.
Why are you describing the diff? It's obvious.
Update the example node to use the standalone compatible form.
And here.
You got comments at v4, nothing improveed. Your commit msg tells useless
things like what you did or you want different Linux behavior.
Best regards,
Krzysztof