| Message ID | 20220930182212.209804-1-krzysztof.kozlowski@linaro.org |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [1/3] arm64: dts: qcom: sdm630: fix UART1 pin bias | expand |
Hi, On Sat, Oct 1, 2022 at 2:58 AM Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> wrote: > > > I would also note that convention on Qualcomm SoCs that I've worked on > > was that bias shouldn't be specified in the SoC dtsi file and should > > be left to board files. This is talked a bit about in a previous email > > thread [1]. > > Uh, that makes a lot of sense. It is almost always a property of a board. Right, though it can make sense to have a "default" in the SoC sometimes. For instance, for i2c you almost always want external pullups so you can tune them to the speed/trace lengths. Thus having a default in the SoC file to disable i2c pullups would make a lot of sense. The problem is the ugly / non-obvious "delete-property" we need to put in the board.dts file if we ever need to override the SoC's pull. :( I actually remember this not being a problem in Rockchip SoCs. I guess it's because they end up having an extra level of indirection. I guess there's no great way to do that for Qualcomm without changing the bindings. > > That being said, it does look like this was the intention of the > > original commit, so thus: > > > > Reviewed-by: Douglas Anderson <dianders@chromium.org> > > Thanks. > > I can also drop the property entirely to match existing behavior (not > the intention). Hopefully someone who cares about this board can test and let you know either way. > > [1] https://lore.kernel.org/lkml/CAD=FV=VUL4GmjaibAMhKNdpEso_Hg_R=XeMaqah1LSj_9-Ce4Q@mail.gmail.com/
On 03/10/2022 17:29, Doug Anderson wrote: > Hi, > > On Sat, Oct 1, 2022 at 2:58 AM Krzysztof Kozlowski > <krzysztof.kozlowski@linaro.org> wrote: >> >>> I would also note that convention on Qualcomm SoCs that I've worked on >>> was that bias shouldn't be specified in the SoC dtsi file and should >>> be left to board files. This is talked a bit about in a previous email >>> thread [1]. >> >> Uh, that makes a lot of sense. It is almost always a property of a board. > > Right, though it can make sense to have a "default" in the SoC > sometimes. If the default is safe, then could be. But it is still causing a risk of developer just forgetting to configure the configs for his board. Bringup of DTS should be a conscious decision, not just "copy and hope it works", therefore recommendation is to configure per-board properties in board. Even if it means duplication. The same was for board-provided clocks or aliases. > For instance, for i2c you almost always want external > pullups so you can tune them to the speed/trace lengths. Thus having a > default in the SoC file to disable i2c pullups would make a lot of > sense. The problem is the ugly / non-obvious "delete-property" we need > to put in the board.dts file if we ever need to override the SoC's > pull. :( Which might not help in reducing amount of code... Best regards, Krzysztof
diff --git a/arch/arm64/boot/dts/qcom/sdm630.dtsi b/arch/arm64/boot/dts/qcom/sdm630.dtsi index b51b85f583e5..e119060ac56c 100644 --- a/arch/arm64/boot/dts/qcom/sdm630.dtsi +++ b/arch/arm64/boot/dts/qcom/sdm630.dtsi @@ -779,7 +779,7 @@ rx-cts-rts { pins = "gpio17", "gpio18", "gpio19"; function = "gpio"; drive-strength = <2>; - bias-no-pull; + bias-disable; }; };
There is no "bias-no-pull" property. Assume intentions were disabling bias. Fixes: b190fb010664 ("arm64: dts: qcom: sdm630: Add sdm630 dts file") Cc: <stable@vger.kernel.org> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> --- Not tested on hardware. --- arch/arm64/boot/dts/qcom/sdm630.dtsi | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)