RE: [PATCH v2 1/3] dt-bindings: thermal: samsung: Add tmu-name and sensor-index-ranges properties
From: 손신 <hidden>
Date: 2025-09-05 08:47:45
Also in:
linux-arm-kernel, linux-pm, linux-samsung-soc, lkml
Hello Krzysztof Kozlowski,
-----Original Message----- From: Krzysztof Kozlowski [mailto:krzk@kernel.org] Sent: Thursday, September 4, 2025 5:00 PM To: Shin Son <redacted> Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>; Rafael J . Wysocki [off-list ref]; Daniel Lezcano [off-list ref]; Zhang Rui [off-list ref]; Lukasz Luba [off-list ref]; Rob Herring [off-list ref]; Conor Dooley [off-list ref]; Alim Akhtar [off-list ref]; linux-pm@vger.kernel.org; linux-samsung- soc@vger.kernel.org; devicetree@vger.kernel.org; linux-arm- kernel@lists.infradead.org; linux-kernel@vger.kernel.org Subject: Re: [PATCH v2 1/3] dt-bindings: thermal: samsung: Add tmu-name and sensor-index-ranges properties On Wed, Sep 03, 2025 at 04:36:32PM +0900, Shin Son wrote:quoted
The exynosautov920 TMU requires per-sensor interrupt enablement for its critical trip points. Add a DT property to the Samsung thermal bindings to support this requirement:That's pretty redundant sentence.
I'll remove this sentence.
quoted
- **samsung,hw-sensor-indices**: Defines the sensors currently mapped to the TMU hardware. Indices not listed are absent or fused offDon't write here any code, but concise prose dxescribing hardware. If sensors are fused out, you certainly can read their status from efuse, no?
"fused out" was a wrong expression — sensors are not indicated in any register. Sorry for the confusion. The hardware does not provide a bitmask of present sensors. Therefore, the DT must explicitly list which indices belong to this TMU instance. Additionally, I'll rephrase this sentence to clearly describe the TMU hardware only.
This is really vague description of hardware. I don't understand why you are changing sensor-cells, why older variants of tmu gets now cells=1 (missing constraints?). Why older variants also get that property for sensors? It does not make sense there, because they have one-to-one mapping between TMU and sensor.
Older variants should be fixed to 0, but my patch mistakenly opened it with an enum so that 1 was also allowed there, I'll fix this. I'll also restrict the sensor indices property to v920 only.
quoted
Additionally, add myself to the bindings' maintainers list, as I plan to actively work on the exynosautov920 TMU support and handle further updates in this area. Signed-off-by: Shin Son <redacted> --- .../bindings/thermal/samsung,exynos-thermal.yaml | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) diff --git a/Documentation/devicetree/bindings/thermal/samsung,exynos-thermal.yam l b/Documentation/devicetree/bindings/thermal/samsung,exynos-thermal.yam l index 29a08b0729ee..abd89902d33a 100644 --- a/Documentation/devicetree/bindings/thermal/samsung,exynos-thermal.yam l+++ b/Documentation/devicetree/bindings/thermal/samsung,exynos-thermal +++ .yaml@@ -8,6 +8,7 @@ title: Samsung Exynos SoC Thermal Management Unit(TMU) maintainers: - Krzysztof Kozlowski [off-list ref] + - Shin Son [off-list ref] description: | For multi-instance tmu each instance should have an alias correctly numbered @@ -27,6 +28,7 @@ properties: - samsung,exynos5420-tmu-ext-triminfo - samsung,exynos5433-tmu - samsung,exynos7-tmu + - samsung,exynosautov920-tmu clocks: minItems: 1@@ -62,11 +64,22 @@ properties: minItems: 1 '#thermal-sensor-cells': - const: 0 + enum: + - 0 + - 1 vtmu-supply: description: The regulator node supplying voltage to TMU. + samsung,hw-sensor-indices: + description: |Drop |
I'll drop this.
quoted
+ List of hardware sensor indices that are physically present andusablequoted
+ in this TMU instance. Indices not listed are either unmapped orunused.quoted
+ $ref: /schemas/types.yaml#/definitions/uint32-array + minItems: 1 + maxItems: 16 + uniqueItems: trueBest regards, Krzysztof
Thanks. Best regards, Shin Son